I must have a video of him rambling about how he would redesign the rails for overclocking and how side panels would become fan panels with a 380mm monster box fan built in specifically for ambient airflow and motherboard components.
While you have a good start, especially with the fluid dynamics of air heating up through the finstack. I think you are missing some of the key parameters at work here. You have to keep in mind that radial flow ("blower") fans have much higher pressure than an axial flow fan. So your "put a high pressure fan right in front of the intake" (6:56) is the blower fan itself. So the greater impedance to flow through the heatsink, and the much smaller opening is much less of a problem for them. Axial flow fans have higher airflow, but worse pressure. This is why you have so many exits for the air to go: the resistance to flow is a problem for them, so freeing it up is crucial to the design. I haven't run the numbers, but I strongly suspect a blower card can get away with a more restrictive heatsink that has more surface area than they do now. The reason you see really "deep" contra-rotating fans in servers in because the restriction is so high they need a lot of pressure to be able to move air through it. They could use radial flow "blower" fans, but there are not well optimized for the servers which intake air from the front instead of the side. Here's an example comparison of sanyo fans. I tried to pick two that are close together. The axial fan is more powerful. 9GV1212P1J011 axial flow fan 120*38mm, 12V, 3A, 64 dB. It has 224 CFM and 360 Pa of pressure. 9BFB12P2H003 blower fan 120 * 32mmw 12V, 2.3A, 62 dB. It has 56 CFM and 1250 Pa of pressure. Compared to the slightly larger and more powerful axial fan, the blower has 1/4 the CFM and 4x the pressure. So given the tradeoff between axial flow and radial flow fans, the reason why axial flow fans are better in this application is because while they have better airflow and worse pressure, you can take advantage of the better airflow more than you can take advantage of the better pressure. In other words: they are more efficient. You also have multiple axial fans, so there's more fan available than with a blower card which only has one fan. Putting multiple radial flow fans on a single card has its issues. Perhaps a better comparison is to take a blower card and compare it with an axial card with one fan disabled. Or you can compare multiple blower cards all stacked up with multiple axial ones: you will see the airflow out of the blower card is less affected by the restricted airflow at the input. Since most people have only one GPU, the input to the fan is much less restricted, another reason why axial fans are better here. Put another way: compare it with water cooling. If you have a pump with more pressure, you can put it with a more restrictive coldplate and get good performance. If you have a pump with less pressure and more flow rate, you will have issues getting it to work with a restrictive coldplate. Perhaps this is a bad example, since all WC pumps are WAY over-rated in pressure and in flow rate both, but the principle is the same. I watched on in the video and I have some more disagreements. For example, the reason why the blower heatsink is longer is to take advantage of the higher pressure. It's true that the efficiency of the cooling decreases the longer the air is in the heatsink, but as it travels across, a single particle of air can absorb more heat from the sink. So there's less air flowing through it, but each particle of air can move more heat away from your device. It would only be "literally the worst shape" if it had very low static pressure and high airflow like an axial fan. I think I'll stop editing my post here and just say: please talk to a heatsink designer. Not necessarily one for GPUs.
"please talk to a heatsink designer. Not necessarily one for GPUs." Why he says the heatsink is fine(well subjectively) its the fact that the fanblower actuates airflow in the most ineffiencient manner
If we assume identical heatsink area, you could probably argue that a blower card is removing twice as much heat per unit if air - I’d imagine thermal transfer is one of those things that has exponential decay, so having a heatsink 4-8x “deeper” than an axial card is only going to deliver 2-3x the heat dissipation per unit of air (thanks to the aforementioned heating of the air and the reduced thermal transfer that entails). So far, so good. Then we run into the fact that despite the pressure it produces, the typical blower is probably going to move 1/4-1/3 as much air as an equivalently loud axial fan (as your own example shows). So the blower card overall is moving probably 1/5 to 1/8 as much air as an axial card. 1/5*3 = 3/5ths the cooling performance of an axial card. The *extremely* rough numbers above show why blower cards are kinda shit. You can quibble about the specifics, but I think it’s difficult to argue with the overarching point.
It also depends on your case, blower cards are the best thing ever for mini-ITX builds since you can't afford for anymore heated air to cycle in your case than what already is being displaced by the CPU cooler.
amd radeon devision literally for years now: "we have a nice chip. a bit less efficient than the competition but it is capable and has a good price/performance ratio. how can we improve on that? i know let us choke it by putting a glorified hairdryer on top ot if and make sure it sounds like a jet engine the second it has to do more than render a still image!"
Hope you don't mind if I paste my comment, I'm on phone and don't fancy re typing. If you think BZ videos are long read this lol. 1) obviously the air is exhausted out the case, better for cpu and ram thermals. 2) if you have a 100mmx100mm fin stack and then put in a 100mm fan only 73.6% of the fin stack gets airflow from swept area, the corners and fan hub get nothing. Atleast with a blower 100% of the heatsink cross section is used. Example. Say you have a 2 slot 10.5in (266mm) card thats 100mm wide with its fans.... you only get 1 slot of heatsink 1 slot of fan. For a total fin stack of 100mm x12mm x266mm= 320K mm2. But only 176K mm2 (55%) is getting direct airflow through it. If you have a blower card thats also 2 slot with a 100mm fan thats 160mm heatstack length x100 x24mm thick so 384K mm2 of area. *Obviously* you need allot of pressure to push through 160mm length of fins but even if it was 120mm thats 288K mm2 of area. Unfortunately the biggest fans they make now at 85mm and the blade depth seems thin and I think that's the main cause, the 70mm fans in recent use seemed deeper like 20mm. Either way wouldn't be an issue if we had bigger slower blower fans. ALA PS5. 3) as mentioned allot of cards (like my r9 290 twin frozrs) have the air exhaust out the sides right against the motherboard and m.2, assuming it doesn't get stuck there by lower slots... it just gets recirculated so only the sides against the window is what you have to rely on and even that can get recirculated back under the fans. I'm not saying blower is better but we have 3-4 slot axial coolers like the morpheus so obviously axials aren't perfect either. I'd love to see a 3 or 4 slot 100mm high (above pcie slot hight like the 3090 is and older lightning/kingpin models) blowers, How about a quad slot which has an 80mm noctua fan blowing through it like allot of data centers (but less noise and airflow and more heatsink). They just need to put effort into them but they won't because gamers need thier plastic pointy shrouds with RGB lighting etc to make themselves feel validated about thier purchase. I also like the style of the gtx680 and vega56 and 5700xt stock coolers. Its a graphics card, its a tool like a screwdriver, so its clean and simple black box, it doesn't look like a power rangers lunchbox made out of knex. This is just my opinion and I test in cases, each to thier own. Casually waits for triple fan quad slot coolers because anti-sag sticks look good. /s Anti-lag is for noobs, anti sag is the new thing. Aunty sag is someone else. You're welcome. The rear io on allot of blower cards is less restrictive than the fin stack. The rear io being removed depending on model might drop you 1-2c at best. I have 4x reference gtx480s in sli, I have an interest in this. Edit and 1080ti fe and titanV. Safe to say there are other benefits.
@@tomstech4390 finally, someone who actually gets my frustration with blower style cooling. It's not the style that is the issue, rather it is the stagnated engineering process. Maybe it's a little niche, but i would not have mind buying a 3 slot size blower style cooler for my gpu. I really the blower style cooler because it reduces the point of failure for my PC, e.g if my front or rear panel fan decides bites the dust, the blower style cooler makes sure that there is still cold air flowing through the gpu.
On the Hardware Unboxed channel, people were talking about Harbor Boxed and I was wondering where it was coming from. They never mentioned the auto-generated CC and I would never click on it. So it got stuck in my head for days until a good samaritan posted a comment about the CC system. Now I always check the auto CC to have a good laugh.
While the thermodynamic principles you laid out were valid, I feel like there are a couple of counterpoints: 1) There is a difference in cooling efficiency between axial and blower coolers, as you have outlined, with the axial type having a marked advantage. I think the context really matters, though, in whether that is meaningful. For SIs and OEMs, sustaining 10% lower GPU performance at the same cost while remaining in-spec thermally is going to be worth the trade versus managing the heat placed awkwardly into the system. Enthusiast systems tend to have massive airflow relative to OEM builds that is capable of forcefully [and inelegantly IMO] dealing with that problem. Reference blower designs are for maximum compatibility as a part of a larger thermal system, not chasing the ultimate thermal efficiency in isolation. 2) A lot of the criticism leveled at blower designs comes from a crowd that gets stuck on the word 'throttling'. In a world with ever more fine-grained DVFS and opportunistic boost algorithms in play, the concept of throttling has become somewhat lost on a lot of hardware enthusiasts. It's not really a matter of one cooler design throttling and not another, rather how much thermal headroom (and other parameters) is present in each design under which the DVFS logic affords performance. This is a matter of optics: the lower performance achieved with a blower is not some indication of a failed design or of problematic operation, rather it is just a data point in the broader cost-benefit analysis of system design. I strongly agree with your assertion that case design is at the root of this problem. ATX is entirely unsuitable for what the modern enthusiast system has become. ATX is in no way optimal to deal with how much power modern CPUs and GPUs use, and the layout of the add-in cards couldn't have contended with a future of 350 W triple-slot behemoths. Assuming it has too much inertia to easily change, I think there is still room for a couple of design wins in the short term: With the reasonable death of NVLink/SLI outside of extremely niche systems, I hope that the majority of ATX motherboards ditch the second PCIE x8 slot and repeater chips in favor of better placement of the single X16 slot. Slot 1 tends to be needlessly close to the CPU and DIMMs and can easily be pushed down 1-2 slots for better clearance. Asus has already moved it down 1 with the Maximus XII Extreme and Apex, something many older boards had back in the northbridge days. This even leaves a better spot open above slot 1 for M.2 devices, rather than being roasted below a GPU finstack. Next, we have triple-slot GPUs, but not much clever use of the reverse side of the PCB. Why not have a triple-slot GPU where the reverse side is one of those slots? Imagine for example if the I/O (HDMI, DP) was mounted on the reverse, or spread some of the thermal load out, like mounting the memory packages and/or some of the VRM components on that side. I also think if multigpu is properly dead, bring on the 4 or even 5-slot coolers. Not many users install a second add-in card anyway, why not use that volume to further increase the efficiency of the GPU cooler (the single biggest source of heat in the system). I think the Mac Pro, of all products, does the best job of demonstrating what is possible without the constraints of ATX. In that design, you have the DIMMs on the opposite side of the motherboard, isolated from the CPU heat. You have the add-in cards done in the style of a 3U rack with high volume full-length coolers designed with very low fin density for low static pressure airflow penetration. The cooling concept of a server rack can be achieved without insane airflow and static pressure, it's just that the heat sink volume, fin densities and obstructions have to be scaled to compensate. ATX, by comparison, is full of dead end corners, right angle airflow, pointless close proximity of heat-generating components, and almost unavoidable obstructions. I want the motherboard vendors to be more brave with this stuff.
I think you have a lot of valid points, but moving away from ATX is going to take more than a brave motherboard vendor. PSUs and cases will also have to change at the same time. That requires multiple manufacturers to work together on a new concept that may or may not be successful. Or... a brand new standard to be created for the entire industry. I just don't see this kind of stuff happening any time soon.
@@prycenewberg3976 Agreed. The enthusiast PC platform, as opposed to the more locked-down and bespoke SI PC platforms, prides itself on modularity and backwards compatibility. Being so far on the open side of the closed-open platform spectrum makes change very difficult. For it to happen, a more radical design change would have to prove itself extremely valuable to the market-something customers are ready to line up and pay for. While a more modern standard could clearly do better than ATX, it probably couldn't be so significantly better that it would drive high demand for it immediately. Without that sort of incentive, it's going to be too much of a risk. Maybe it just takes running harder and harder into walls until we're actively looking everywhere for a solution to the headaches, and for now it's still tolerable. I think ATX12VO and the new 12-pin GPU power connector could be interesting in terms of their market penetration: how quickly will these standards find their way into enthusiast products? If we see rapid uptake, it might predict a more positive outlook for moving past ATX at some point.
To be fair the earlier attempt to improve ATX, BTX clung on much longer in OEMs after colossal failure in the DIY market. Really I think we need to take design queues from the server space for full size boards and more dual chamber cases for ITX which then will allow for more rear mounting of M.2 or memory.
I still remember how we made fun of the first cards with blowers: The dreaded GeForce FX 5000 series. Back when ATi was king of the hill. *sigh* I'm old
@@SteelSkin667 the 9500 Pro was the budget version, with 128b wide RAM, but a beefy 9700 or 9800 GPU. I ditched AMD for an FX 5900 XT, though, and an aftermarket cooler. My 9600 XT ran quiet, with a tiny fan, but the drivers were SO BAD. I couldn't take it.
The grill is usually part of the IO-shield. The IO-shield usually contains the mounting holes to screw your card onto your case. So, good luck getting rid of that grill. There are three reasons why one could opt for a blower card: 1. Your system is very small and you want to direct hot air out of your case right away 2. You're going to WC the card anyway, in which case it doesn't matter which thermal design you pick 3. You love the sound of fighter jets
Often (but not always) we will also see a 'too small' blower fan combined with a crappy cheap tiny heat sink, constricted exhaust orifice and cheap thermal paste. So a quadruple whammy of ineptness. RX 470/480 reference cards, I'm looking at you!
I have actually had great success overclocking blower cards in a interesting manner. I noticed that if ambient is really really low, open air cooling cards have a flaw that they make a heat bubble around themselves, while a blower card will push that air through its cooler and exhaust all the heated air to not get recirculated. I managed to get a blower rx 470 cooled to 0C under load (didnt show temps below zero, might have been less, north europe winters, took my rig outside), making for a insanely good overclocker without any LN2. Did not get the same results with a open air rx 480, which still managed to stay toasty in the same cool ambient air when running it over 250w, yielding minimal performance increase with voltage increase. I remember taking a non ln2 world record for the 470, dont know if it still stands.
Yep, I knew all this from back when I put a ATI Radeon X 1800 XT into my PC. It not only sounded like a hairdryer, it's cooling performance was also incredibly awful.
There's one great positive effect of blower fans. They keep the heat inside the GPU-case and blow it outside the tower. No hot air is blown on the mainboard like it happens with the axial fans. The fins are always build to blow the air to the mainboard and to the side of the tower. My mainboard for example is having the SSD under the GPU. That means, an axial fan is always blowing the hot air onto my SSD. I don't think that's a great idea. ;-) Anyway, of course axial fans are keeping the GPU cooler than radial fans, but it also has it faults. I'm sure there are some good reasons to better choose a radial blower (which usually sucks) instead of an axial fan. But hey, nice video and great presentation. I've enjoyed it and also learned some interesting things I didn't have in mind before. So, thanks for that.
Only a couple of days left until we will get reviews of the 3080 and a week or so later we should get 3090 reviews. We might just wait this short amount of time and see how the cooler design performs.
The blower portion has a MUCH shorter HS, so won't have the long HS problem, and the back fan is the much more efficient flow through design. I expect most cooling power at the flow through portion, but blower part will help prevent heat build up under the card. I am betting it works really well.
Seems like an effective design, I just worry about the longevity of the pull-orientation fan, because that motor and bearings will heat up a lot more than any fan installed in a push-orientation.
@@ydihtty I would say that it would last longer in my opinion. Without having the fans work against themselves by having a pcb to block the airflow. The fan can now spin slower to do the same amount of work because there isn't an pcb in the way. I have a feeling a lot of people's rooms are going to heat up because the hot air will now actually be exhausted more efficiently instead of getting stuck in a gpu loop.
there will be a downside that the Fan hanging off the back there will provide mediocre Cooling since it's nowhere near the source of the Heat. the Heatpipes that i'm sure they're using will help compensate for it just like we do with all Heatsinks, but it means that part of the Heatsink will be completely reliant on those Heatpipes to do anything at all and also means that only the surface area of like a few Millimeters around those Heatpipes will do anything at all, and the rest will just be weight for basically no performance result. if they really wanted to impress me, they'd have the other Fan also be able to move air through the PCB by cutting some holes. to make that really work you'd want to move the GPU die farther away from the PCI-E Bracket though, to keep the most effective Fan as the one directly above the Die. in short it's a better Reference design but one that's still pissing a way a lot of performance for little benefit.
For my 50th anniversary edition Rx 5700 xt (basically a binned reference card) I swapped out the original grill I/O bracket for a single slot bracket and cut off the metal bar covering the pcle slots on the case. I also undervolted, used thermal paste, put thermal pads under the back plate, and did the washer mod. My undervolt is 1800mhz at 930mv but the actual frequency sits at ~1750mhz instead of 1800mhz. At 30% fan speed, my temps are around ~77°C with mem at 88°C while in side of a mini itx case (Silverstone sg13) with basically no direct airflow to the gpu. The temps aren’t awful for being in an itx case and it’s actually not too loud. I don’t notice it too much but even after doing all this work to improve the temps, I think blowers still suck. I’m actually thinking about trying to water cool my gpu instead because my memory temps are too high. I don’t think a customer, especially a beginner or noobie customer has to put this much effort into fixing AMD’s crappy blower cooler. Well I honestly don’t really regret buying it. It still performs great and at the time of purchase there weren’t any other 2 slot models that would’ve fit in my case. Blower cards can still be effective depending on the itx case. Here’s a video by Optimum Tech th-cam.com/video/_dhOsTS6-C8/w-d-xo.html Edit: oh yeah the anniversary edition pcb is different from a regular reference card. I can confirm this too when I took apart my card. That the anniversary edition card has all the aluminum polymer capacitors populated on the back of the pcb which I assume is why it’s better at undervolting and overclocking This is a reddit post I found about it. It’s not by me www.reddit.com/r/Amd/comments/dm3uv2/5700xt_50th_anniversary_isnt_just_a_base/?
I don't think blower style cards get enough Love, they keep your internals much cooler specifically the ram, vrm, also power supply, as long as you have a AIO for the CPU they are great cards. I've had a few open-air coolers and they dramatically heated up my internals. But another good option is a AIO blower style card they are just very expensive, but they do great work at staying cool along with keeping the heat down inside the case.
If blowers are so bad, how come reference cards from AMD or NVidia come with blowers? Top engineers have decided that blowers are GOOD. Since blowers exhaust the hot air outside the PC, the air inside the PC case is cooler, leading to more efficiency in heat transfer.
The way I see it, blower style fans are good when you have bad overall case airflow, since they will discard the heat immediately. "Normal" style cards are better if you have good airflow since they blow a lot more hot air back into the pc, not only heating the GPU but all the other components as well, which isn't a problem if the case has good airflow
Limited viewpoint expressed on this video. Some benefits of blower cards are that they don't dump all the heat into the case affecting the whole system, works well with multiple cards, for smaller systems can be the better solution. Anything over 220w I found blower cards negative is noise and insufficient cooling. Vega FE at stock is a rather quiet blower card, rated at 225w. Vega and the rather power hungry if you try to push it makes that rather super quiet blower setup into a rather noisy one quick. Will be amusing when people are dumping 320w+, ASUS 400w, OCing 400w+ with the upcoming Ampere cards adverse side effects from that solution, probably should be a AIO for over 300w cards like AMD has done several times. AMD also has made some great cooling solution cards as well, well not many but they can do it right at times.
19:30 Yes, you're referring to laminar flow, a boundary layer which sounds great but ruins everything. Clean separation = undisturbed air cushion coating the fin stack
21:17 Blowers are perfect if you want to have like 3 or more cards running in parallel. If you went with open air cards, you'd need to have heaps of case airflow and space them out a lot.
I’ve got an RTX 2060 and an RTX 2060 KO on the same motherboard with one empty slot between them (the KO is two slots, the vanilla 2060 effectively 3). The 2060 is giving the KO some nice warm air and the KO’s running at 82C normally. If the 2060 blew most of its hot air out of the case, I think it would help. Maybe I should go without a case? Open-air test bench permanently?
Thank you for pointing this out because it's absolutely true, Blowers aren't bad coolers it's always their heat-sinks that are designed completely garbage. making them appear like they are absolutely terrible. they are much more efficient at exhausting heat out of the heat-ex-changer and the system Case, They are loud, yes, but if made properly they can provide superior cooling results then axial cooling that's popular with AIBs. It's just there hasn't been a proper blower heat-sink made. Thanks for reinforcing this idea Buildzoid, Every-time i bring this up to other-people they dismiss me as crazy. But it's a fact especially when you look at common sense at engineering and enterprise systems.
14:07 So those same people who think slower water increases thermal transfer also hold their breath during a jog to increase O2 saturation XD. Thanks for the easy fix on blower gpus!
At 19:30 you're talking about boundary layer, and thermodynamically that's correct. A turbulent boundary layer promotes better thermal transfer compared to laminar (all else being equal). Although the effect is less significant at the low flow speed and volume we're talking about (small PC fans). A lot of the time, the fancy impeller geometry and funky fin stack shapes manufacturers like to use has more to do with aesthetics and marketing than actual performance improvement (don't get me wrong, design features that promote turbulent boundary layer flow DOES help with thermal transfer, but at this scale, the same improvement can be easily be achieved just as well just by brute-forcing it with running a higher RPM fan)
Great video! But : "from my understanding of thermal physics blower heatsinks suck" They dont suck. They are just used in incorrect applications. Blowers are not meant to be "silent", they are meant to be efficient in relatively low temps. Just blow the shit out of long thermal connective area. They just introduce high speed air flow. That's it. It's the reason we have vacuum cleaners.
Blower cooler has one big advantage That if you case is too small or bad airflow design That the blower cooler wont effect and heat the other parts in case Beside this, blower cooler doesnt really good for every body
@@andruloni Oh I misunderstand then However still blower cooler is better for ITX Especially for SFF CASE ITX video card just too close to the cpu cooler Even I use the NCASE M1 still prefer blower cooler
In regards to the turbulent airflow thing! Here is the TLDR: Turbulent flow results in more mixing of the hot and cold air, which is how you transfer heat. Conversely, laminar flow, has much less mixing of the hot and cold air, basically causing a moving column of cool air in the finstack that never contributes to cooling.
It might help to explain the air temperature doesn’t rise linearly. Instead, the air will decrease the amount of heat it can soak up due to the smaller temperature differential, meaning the temperature will rise slower and slower. In the linear example shown in this video, heat transfer with minimal airflow would be reached if the air temperature is equal to the heatsink temperature exactly at the end of the heatsink. This would only be true if Tdelta doesn’t impact transfer efficacy. Ignoring radiation, Q=hA*Tdelta/d. So the higher Tdelta, the higher heat transfer.
Open air cards exhausting in the case increase the CPU temps while blower cards prevent that to the tune of around 8-9°C iirc. Cause the CPU cooler then has to deal with the GPU's hot air. This is assuming you're using a case and not an open bench setup. Radiator placement has similar effects as intake taking in cool air is better than it being exhaust as it then has to deal with other thermals in addition to its own. That's my understanding anyways. So I suppose you could say it's more about regulating other component thermals rather than being superior for the GPU itself.
The biggest advantage for blower cards is that they work much better with WARM and poorly ventilated cases. Otherwise, they SUCK. Open cards also have the benefit of heat pipes. One design change that I wish we see more often.... is holes in the PCB & backplate to allow air to exhaust more freely.
The only thing that I think you might have overlooked about blowers is their use in servers because for servers it is all about not recirculating any air. So for server cards they are almost exclusively blower style because perform better in that type of system (it also helps that noise is an after thought)
It's bugged me for a while that no GPUs are actually using the PCIe slot above the GPU, it seems like the perfect location to vent the hot air. Giving you the best of both worlds.
Just took Heat Transfer last semester. When I saw the new 3080 and 3090 designs I instantly turned into Michael Scott after seeing Toby return from Cost Rica. NOOOOOOOOO!!!!! GOD NOO! NYOOOOOOOOOO
Silverstone-style reverse ATX (even better would be smaller form factors) cases would work really well with exhaust GPU axial fans. Just have GPUs exhaust everything straight out of the top with no PSU shroud in the way. It'd solve the entire problem you mentioned where nobody thought about the heat output modern GPUs would turn out to have. But obviously, it'd need an industry-wide push so that's not happening.
One of the most relevant point of cooling is "Countercurrent exchange". If it is implemented, it is the best way to transport heat away. Full stop. Also very important is the speed and the volume of the air and the area of contact of the air and the fan. So I can not agree, blower are not all bad, bad blower are bad, but blower are not necessarily bad.
Vega 64 airboost oc here! Card is only cool after undervolting and fan curve optimization. Still noisy. Hits about 70ºc on games and have put a sensor on the exhaust. Temps can reach high 50s there, but the air goes out, not affecting the interior mean temperature.
It all comes down to the use case scenario. Airflow is not the only parameter, static pressure is another one. Blower has more static pressure, which is more suited for closed PC case such as workstation. Blower fan is not to be blamed, what makes it a bad cooler is the unoptimized cooler design. Blower is more universal to all types of PC case, at least it guarantees cooling in all use case.
Very good summary of the thermodynamics of GPU cooler designs. BZ is right that blowers were designed to compensate for poor airflow cases however he didn't mention that this is actually the case for server racks and other SFF PC cases. It's the reason why Quadros, Titans, and Tesla cards are all blowers.
We're talking about airflow vs. cooling surface area. Would you use an aluminum tube or would you use a brick of aluminum? Both have their advantages and disadvantages
Probably worth pointing out that cooling solutions like GPU blower fans are used a lot in servers for pretty much all of the reasons that make them annoying for desktop use cases (to the point where even the CPU's use these setups). Multiple fast fans draw cool air in through the front, passes over every component and exhausts out of the back. Noise isn't a concern but the depth of the chassis is.
Sounds like at the end you are describing the SFFPC deshrouding mod like the one on Optimum Tech channel. Remove fans of GPU heatsink, stick the heatsink to 2x 25mm case fan in exhaust at the bottom of the case, and there you have a full exhaust system for your GPU. The hardest part is getting the case / heatsink combo which fits tightly in the case if not using after market cooler.
What if you replaced the blower fan with like two 40mm fans blowing air the same direction? On one hand, you can have more heatsink. On the other, I can't imagine they would move nearly as much air, unless they're super thick 1U server fans.
triple fan 2070 super with a 200 mm front intake at the bottom of case (Phanteks Enthoo Pro) and a 140mm rear exhaust makes my gpu very happy got a 360mm EK aio at the top for the cpu one of the best cases under £100 only thing is I wish i had gone tempred glass but that would be more than I was willing to pay for a case at the time
I'd like to see a 4-slot blower card, with vertically (?) mounted fans in a push-pull configuration. One low-pressure fan blowing hot air out of the rear slots, one high-pressure fan pushing cool air into the card from near the front of the case. And possibly a high-pressure fan in the middle, but I don't know if that'd actually be useful. For four slots, standard/replaceable 92mm fans should be perfectly feasible.
Through my own experience with several different cards, some blower, some with dual axial, and some with triple axial fans, I feel that more of a mention needs to be made of the manufacture's shrouds. While yes, blower style coolers have to exhaust all of their heat out the back of the card AROUND the ports, far too many of the axial style cooler manufacturers design their cards with heat sink fins aligned vertically, only to then almost COMPLETELY OBSTRUCT the majority of them with the shrouding of the card, purely for esthetic reasons. Unless you have a dam decent amount of airflow through the case to overcome those obstructions and to prevent it recirculating the same hot air, axial coolers aren't much, if any better than blower coolers with the fan set to a high rpm. I currently have 5 1070's running full tilt right now, and the blower style are keeping up with the axial coolers just fine. If manufacturers turned the fins on axial cards to take advantage of the gaps in their shrouds, that would obviously change, but those 1070's are pretty clear evidence of the problem.
My imaginary 3 slot blower with heat pipes in addition to a vapor chamber beats all your coolers. Due to high airflow blowers are impervious to dust buildup and take the heat out of the case, which is very useful when your cpu is 280w itself.
@@ActuallyHardcoreOverclocking Still better than an open air cooler, in equal conditions that is. There is simply a higher chance that dust gets exhausted out of the case, than with an open air one. I think one of the main reasons why Nvidia abandoned blowers for consumer cards starting with the 20 series, is to upsell Quadros etc for serious people, who need to stack 4 of them into a case.
For a single consumer tower, blowers do indeed suck. HOWEVER, consider compute clusters / datacenters / servers, where dozens, hundreds, or thousands of components all need to be cooled. In these applications, airflow throughout the entire space must be carefully considered, and it becomes important to have dedicated "inlets" and "outlets" for each system. Ideally, cool air is delivered to the inlets, flows consistently through the relevant cooling components with minimal recirculation, then exits through exhausts (to AC units, etc). Blower fans are perfect for this - not only do they actively contribute to the pumping of air in one direction through the system, but they also minimize recirculation of hot air within the case itself (especially if many GPUs are installed right next to one another). In other words, blowers work well when ensuring a certain "flow pattern" is critical for maintaining efficient cooling of many tightly-packed, power-dense systems. (It turns out this is also a useful arrangement for managing dust with minimal maintenance). Note that enterprise compute cards exclusively use blower-style coolers: Quadro, RTX Server, Radeon Pro, etc - all use blowers! This lets you pack many of them into a server case right next to one another. Then, you can stack a bunch of those cases together in one rack. Assuming you deliver proper ventilation to and from the rack, you will still be able to get acceptable cooling on all cards in all systems (which would not be possible with open-air coolers!) On the other hand, open-air coolers work great for individual PCs, i.e. applications where there are 1-2 GPUs surrounded by lots of empty space. Open-air coolers are very "selfish" coolers from a systems perspective (they don't drive bulk airflow in any single direction and tend to throw exhaust heat towards all the other components), but the overall power density is low and ambient case temperature essentially agnostic to cooling performance. As long as there are some fans in the system, hot air will "just find its way out of the case" and it doesn't really matter where or how. Raw airflow volume (rather than direction / inlet / exhaust) is all that really matters, making open-air coolers great for individual PCs. In this use case, open-air is definitely the way to go! Blowers aren't nearly as efficient or quiet here, but that doesn't mean they don't still have their own applications!
Jayz tested the impact of the NVidia 3000 cards on the temperature in a case. so its funny to see that the blower part of the cooler help by moving a lot of hot air outside the case. so to compensate this when using an open card, we should have more fans to move the air out of the case faster. like mini fans below the card pushing hot air outside the case (like the top back fan) or other options like this one.
The blower will extract the maximum cooling potential out of a given amount of air if it's hitting equilibrium right at ejection, but will not maximimize the cooling potential of the fin stack. If blowers weren't way louder for moving the same amount of air, it might be a good idea ;). I think a blower will generate plenty of turbulence since the air exits tangentially. Static pressure isn't much of a concern with a blower, but of course you pay for that by sacrificing cfm anyway.
Good explanation! And regarding the typical exhaust on blower cards.,.. You hit the nail right on! :) I kinda agree with you. But blowers can easily push the air efficiently out of that short heatsink they normally use. You don't get such huge difference in temps over the HS. Usless you keep the fan very low speed. Forward curved blower fans are not really airlow efficient, but high static pressure efficient. static pressure is what is needed inside that "pressure" chamber. Especialy with the higher resistance expansion slot cover plate. They should really just make a big hole. I totally agree on that. Blowers are noisy because of the high static pressure from the high rpm. But because bad case airflow is not something rare, and especially with air cooled multi GPU setups, blowers advantage in that case, has been tried and proved.
I do not think it is the fault of the case designers either. With the way the pcie slots are arranged on the mainboard there are no good options to place the graphic card without using some crazy stuff like riser cables. Maybe it is time to change the mainboard layout completely, to allow graphic cards being mounted parallel to the backside of the case. This combined with holes inside the PCB like for the RTX 3080 would allow to blow the air out big time, not through such a small hole like you draw here. The question is then, were do you put the io for external devices like soundcards and such. Maybe to the side? This will limit the possibilities for fancy side glas panels, but the overall cooling performance would be worlds ahead of what we have now. I have to admit it is kind of hard to establish this kind of revolution to mainboard form factors and with the current development it is more likely that we reach efficient hardware than a revolution of the ATX Standard.
Even worse, most of the air pulled into a blower is expelled into directions other than the actual sink's. Plus, blowers accelerate air radially - meaning there's only a small amount of air that goes into the fin stack straight while everywhere else you get a lot of turbulence (=air resistance) due to the air needing to change directions so it will flow through the stack's channels. I mean, one could possibly design a blower with fin stacks in front and behind it, but that leaves the middle part of the card without an effective means of transferring its heat into either of them, barring the use of heat pipes - and these require a larger cooling solution, plus they impinge on airflow yet again as they'd need to be threaded through the fin stack to facilitate efficient heat transfer. Whichever way you look at it, blower designs range from bad to horribly inadequate, trending towards the latter.
Actually blowers have a place in the inverted case markets, (they would typically have a panel to take in air if you had a case with an upside down mount for the GPU) But also i was using a Titan X (maxwell) for 3 years that would max out at 74c (and i ran that thing pretty much 24/7 on BDO) in the FT05 (and a passively cooled 4690k on a megahalems for that matter) x2 180mm fans if im correct blowing all the hot air out, Only used the high fan speeds in the summer, I had to change the card i had before that because the gainward phantom would melt the top of my TJ08E, that cooler has the 3 fans... Under the heatsink, Blowing on a cold plate (that would be warm after 10 mins anyway hurrr) to grab that air out, I mean it looks lovely but the blower has much more of a use case compared to that thing.
Also blower cooler fans doesn't push air straight through the heatsink, it pushes it inside the card in all direction, which creates high pressure, which pushes the air through the heatsink, but also creates a lot of turbulance
Very good explanation. You did mention something that might be slightly wrong, depending how you look at it. If you look at a blower fan, the air does exit the full curcumference. The duct is shaped to guide the air evenly across the one opening at a much higher speed than it exits the fan blades. Edit: what do you think of ducting air from the axial fan heatsink to the outside of the case using a third and fourth slot?
I think the visualisation would be much improved if you use increasing distances between temperature steps. Like, right now, the distance between 20 and 30 degrees is the same as between 50 and 60 degrees, but if it becomes larger, that visualises the decrease in transfer rates.
Msi Air Boost Vega 56 have a bigger window and would recommended a cutting tool to make full use of the blower exhaust gate. Blower fan are Pressure fan to create high volume of air pressuring inside the chamber and as physics, the air will find the way out creating thru a wind tunnel, that said if part of the window is being block at exhaust, more heat will build up inside the fins than being push out fast enough. Not great for OC dues to lack of making it a triple slot or go old school to put in a blower bigger than the width of the card.
I was actually a little shocked at how bad mine was, given that the entire shroud was also made of metal. I think it was aluminum, but I sold the card and can't tell now. That fin stack inside only did half the cooling, and the outside got brutally hot as well. Curious how much of the temp issues were just related to the 7nm process. Like a ~70w ryzen 3600 is harder to cool than a ~90w 2600x.
I once removed the terrible blower fan from a GeForce 6800 and glued a good quality axial fan over the hole. It was actually way quieter than the original fan even though the air had to make a 90 degree turn to go through the heatsink. Those squirrel cage fans are noisy by default.
if blower style sucks , may I point out, that servers cool, by blowing out air out of the housing ? and they often have a multitude of GPU's in the case
Maybe you should make a 2nd part with case/laptop/console thermals. 14:55 same thing applies in the case. The air moving within your components is too hot to make a difference. Just having even one properly placed exhaust fan and one intake fan makes all the difference. It's also why just by removing a panel on your case you drop 20C while doing nothing to the fan situation, basically free cooling. Then there's consoles and laptops where almost anything you do will be bad to mediocre, because there isn't enough space to fit the air without very loud blowers.
I always wanted one of those gigabyte 680’s that had all those small 40mm fans for my collection. I check ebay all the time for it and can never find one.
I actually tested my reference RX 5700 XT without the rear IO bracket. I did not see any significant difference in temps on same fan curve, driver and same games/settings. Narrow slats design is often used in many intake/exhaust aplications like car front grille, AC, ventilation etc. It is there because something could get inside and damage or block the air vent. According the fluid mechanics, slats are pretty effiecient regarding air flow (90% intake and 92% exhaust compared to 100% unrestricted air flow). Nice video, though. Thanx
I did a LEARNEDING during this video. Knew blowers are not good coolers, but now I understand more details on why and I feel I've done a SCIENCE today thanks to you. Thanks for the video! :D :D :D
My friend built a new PC where the Asus 2070 Super video card had ~1mm clearance between its back plate and the Be Quiet! Dark Pro 4 air cooler. The processor ran hot because the video card transferred heat to the air cooler. I put in my modest 1050 Ti video card and the processor temperature dropped 7C. My friend gave me his air cooler and got an Arctic Freezer II AIO.
Who volunteers to write a successor to the ATX specification to fix some of the problems mentioned here? A recent fix for a related problem is the 12-V-connector proposed by Nvidia. I think the whole industry needs to standardize better on certain aspects as well to solve some of these legacy issues (good luck dealing with all of the diverse interests of the CPU, GPU, PSU and case vendors of the world).
It´s not as bad as you think, or else servers would not be as they are. Heat pipes really save the day in these designs. The problem is the tolerances and forcing all the air properly though the hood. Any gap, will let air through and reduce the pressure in the heatsink. So, I would say it´s not a bad design, and it can actually be a very efficient design. Just getting it right is difficult.
Blowers don't suck at all.
they blow
ba dum tss
They do both, technically.
They’re ambiblowsterous
they have to suck to blow :D
damn it, 12 hours late to that joke.
RE: PCIe was never made for 200W+ cards:
I would definitely watch a video of you rambling about how you think modern PCs should be laid out.
imagine a gpu parallel to the motherboard, like with a riser
buildzoid pls.
I must have a video of him rambling about how he would redesign the rails for overclocking and how side panels would become fan panels with a 380mm monster box fan built in specifically for ambient airflow and motherboard components.
@@Flaimbot so a vertically mounted GPU
pls
Buildzoind: you could not chose a worse design
Me: vacum insulated card to prevent GPU heat from heating up other stuff
dont give AMD ideas...
Congratulations, you’ve just invented a XFX THICC 3 series GPU :)
Sun in a vacuum of space: Am I a joke to you?
@@Loachie90 sun only sends out heat through radiation. Although the gpu will have minor radiation most of the energy will be kept
@@gingeritto All that was needed was Happy Meal plastic!
While you have a good start, especially with the fluid dynamics of air heating up through the finstack. I think you are missing some of the key parameters at work here.
You have to keep in mind that radial flow ("blower") fans have much higher pressure than an axial flow fan. So your "put a high pressure fan right in front of the intake" (6:56) is the blower fan itself. So the greater impedance to flow through the heatsink, and the much smaller opening is much less of a problem for them. Axial flow fans have higher airflow, but worse pressure. This is why you have so many exits for the air to go: the resistance to flow is a problem for them, so freeing it up is crucial to the design. I haven't run the numbers, but I strongly suspect a blower card can get away with a more restrictive heatsink that has more surface area than they do now.
The reason you see really "deep" contra-rotating fans in servers in because the restriction is so high they need a lot of pressure to be able to move air through it. They could use radial flow "blower" fans, but there are not well optimized for the servers which intake air from the front instead of the side.
Here's an example comparison of sanyo fans. I tried to pick two that are close together. The axial fan is more powerful.
9GV1212P1J011 axial flow fan 120*38mm, 12V, 3A, 64 dB. It has 224 CFM and 360 Pa of pressure.
9BFB12P2H003 blower fan 120 * 32mmw 12V, 2.3A, 62 dB. It has 56 CFM and 1250 Pa of pressure.
Compared to the slightly larger and more powerful axial fan, the blower has 1/4 the CFM and 4x the pressure.
So given the tradeoff between axial flow and radial flow fans, the reason why axial flow fans are better in this application is because while they have better airflow and worse pressure, you can take advantage of the better airflow more than you can take advantage of the better pressure. In other words: they are more efficient. You also have multiple axial fans, so there's more fan available than with a blower card which only has one fan. Putting multiple radial flow fans on a single card has its issues. Perhaps a better comparison is to take a blower card and compare it with an axial card with one fan disabled. Or you can compare multiple blower cards all stacked up with multiple axial ones: you will see the airflow out of the blower card is less affected by the restricted airflow at the input. Since most people have only one GPU, the input to the fan is much less restricted, another reason why axial fans are better here.
Put another way: compare it with water cooling. If you have a pump with more pressure, you can put it with a more restrictive coldplate and get good performance. If you have a pump with less pressure and more flow rate, you will have issues getting it to work with a restrictive coldplate. Perhaps this is a bad example, since all WC pumps are WAY over-rated in pressure and in flow rate both, but the principle is the same.
I watched on in the video and I have some more disagreements. For example, the reason why the blower heatsink is longer is to take advantage of the higher pressure. It's true that the efficiency of the cooling decreases the longer the air is in the heatsink, but as it travels across, a single particle of air can absorb more heat from the sink. So there's less air flowing through it, but each particle of air can move more heat away from your device. It would only be "literally the worst shape" if it had very low static pressure and high airflow like an axial fan. I think I'll stop editing my post here and just say: please talk to a heatsink designer. Not necessarily one for GPUs.
At the end of the blower heatsink the air is too hot to absorb more heat, amirite?
"please talk to a heatsink designer. Not necessarily one for GPUs." Why he says the heatsink is fine(well subjectively) its the fact that the fanblower actuates airflow in the most ineffiencient manner
I love blower GPUs they remove the heat from inside the case and also allow me to have 4 of them at the same time.
If we assume identical heatsink area, you could probably argue that a blower card is removing twice as much heat per unit if air - I’d imagine thermal transfer is one of those things that has exponential decay, so having a heatsink 4-8x “deeper” than an axial card is only going to deliver 2-3x the heat dissipation per unit of air (thanks to the aforementioned heating of the air and the reduced thermal transfer that entails). So far, so good. Then we run into the fact that despite the pressure it produces, the typical blower is probably going to move 1/4-1/3 as much air as an equivalently loud axial fan (as your own example shows). So the blower card overall is moving probably 1/5 to 1/8 as much air as an axial card. 1/5*3 = 3/5ths the cooling performance of an axial card.
The *extremely* rough numbers above show why blower cards are kinda shit. You can quibble about the specifics, but I think it’s difficult to argue with the overarching point.
It also depends on your case, blower cards are the best thing ever for mini-ITX builds since you can't afford for anymore heated air to cycle in your case than what already is being displaced by the CPU cooler.
amd radeon devision literally for years now:
"we have a nice chip. a bit less efficient than the competition but it is capable and has a good price/performance ratio. how can we improve on that?
i know let us choke it by putting a glorified hairdryer on top ot if and make sure it sounds like a jet engine the second it has to do more than render a still image!"
@Cleru Zinek don't forget those god awful drivers that kept getting even more broken for months
nVIDIA: how can we compete vs that? all these 1650,1660, 1660ti, 1660 super failed... wait i know! take the xx60's cards and lower the price!
@Cleru Zinek Vega cards are really power efficient at the time if your consider undervolting.
I should be working... but this video can definitely run in the background & I won't end up wasting 30 minutes not working... right?
That's seems reasonable 👍
Sure...
Just take a 25min break and move on.
There is an upside with blower coolers and that's the fact that the "spent" air goes out of the system.
But that's it.
Hope you don't mind if I paste my comment, I'm on phone and don't fancy re typing. If you think BZ videos are long read this lol.
1) obviously the air is exhausted out the case, better for cpu and ram thermals.
2) if you have a 100mmx100mm fin stack and then put in a 100mm fan only 73.6% of the fin stack gets airflow from swept area, the corners and fan hub get nothing. Atleast with a blower 100% of the heatsink cross section is used.
Example.
Say you have a 2 slot 10.5in (266mm) card thats 100mm wide with its fans.... you only get 1 slot of heatsink 1 slot of fan.
For a total fin stack of 100mm x12mm x266mm= 320K mm2. But only 176K mm2 (55%) is getting direct airflow through it.
If you have a blower card thats also 2 slot with a 100mm fan thats 160mm heatstack length x100 x24mm thick so 384K mm2 of area.
*Obviously* you need allot of pressure to push through 160mm length of fins but even if it was 120mm thats 288K mm2 of area.
Unfortunately the biggest fans they make now at 85mm and the blade depth seems thin and I think that's the main cause, the 70mm fans in recent use seemed deeper like 20mm.
Either way wouldn't be an issue if we had bigger slower blower fans. ALA PS5.
3) as mentioned allot of cards (like my r9 290 twin frozrs) have the air exhaust out the sides right against the motherboard and m.2, assuming it doesn't get stuck there by lower slots... it just gets recirculated so only the sides against the window is what you have to rely on and even that can get recirculated back under the fans.
I'm not saying blower is better but we have 3-4 slot axial coolers like the morpheus so obviously axials aren't perfect either.
I'd love to see a 3 or 4 slot 100mm high (above pcie slot hight like the 3090 is and older lightning/kingpin models) blowers, How about a quad slot which has an 80mm noctua fan blowing through it like allot of data centers (but less noise and airflow and more heatsink).
They just need to put effort into them but they won't because gamers need thier plastic pointy shrouds with RGB lighting etc to make themselves feel validated about thier purchase.
I also like the style of the gtx680 and vega56 and 5700xt stock coolers. Its a graphics card, its a tool like a screwdriver, so its clean and simple black box, it doesn't look like a power rangers lunchbox made out of knex.
This is just my opinion and I test in cases, each to thier own.
Casually waits for triple fan quad slot coolers because anti-sag sticks look good. /s
Anti-lag is for noobs, anti sag is the new thing. Aunty sag is someone else. You're welcome.
The rear io on allot of blower cards is less restrictive than the fin stack. The rear io being removed depending on model might drop you 1-2c at best.
I have 4x reference gtx480s in sli, I have an interest in this.
Edit and 1080ti fe and titanV.
Safe to say there are other benefits.
That's an enormous upside, though. You can put them in almost any case without worry, even in a prebuilt with only one or two stock fans.
@@tomstech4390 finally, someone who actually gets my frustration with blower style cooling. It's not the style that is the issue, rather it is the stagnated engineering process.
Maybe it's a little niche, but i would not have mind buying a 3 slot size blower style cooler for my gpu. I really the blower style cooler because it reduces the point of failure for my PC, e.g if my front or rear panel fan decides bites the dust, the blower style cooler makes sure that there is still cold air flowing through the gpu.
@@tomstech4390 that's big holy cow!
& here I was thinking the only upside to a Reference Blower Card is the Partners can't make a Worse Design
"Hello guys Bill Loyd here" gotta love closed captions
On the Hardware Unboxed channel, people were talking about Harbor Boxed and I was wondering where it was coming from. They never mentioned the auto-generated CC and I would never click on it. So it got stuck in my head for days until a good samaritan posted a comment about the CC system.
Now I always check the auto CC to have a good laugh.
While the thermodynamic principles you laid out were valid, I feel like there are a couple of counterpoints:
1) There is a difference in cooling efficiency between axial and blower coolers, as you have outlined, with the axial type having a marked advantage. I think the context really matters, though, in whether that is meaningful. For SIs and OEMs, sustaining 10% lower GPU performance at the same cost while remaining in-spec thermally is going to be worth the trade versus managing the heat placed awkwardly into the system. Enthusiast systems tend to have massive airflow relative to OEM builds that is capable of forcefully [and inelegantly IMO] dealing with that problem. Reference blower designs are for maximum compatibility as a part of a larger thermal system, not chasing the ultimate thermal efficiency in isolation.
2) A lot of the criticism leveled at blower designs comes from a crowd that gets stuck on the word 'throttling'. In a world with ever more fine-grained DVFS and opportunistic boost algorithms in play, the concept of throttling has become somewhat lost on a lot of hardware enthusiasts. It's not really a matter of one cooler design throttling and not another, rather how much thermal headroom (and other parameters) is present in each design under which the DVFS logic affords performance. This is a matter of optics: the lower performance achieved with a blower is not some indication of a failed design or of problematic operation, rather it is just a data point in the broader cost-benefit analysis of system design.
I strongly agree with your assertion that case design is at the root of this problem. ATX is entirely unsuitable for what the modern enthusiast system has become. ATX is in no way optimal to deal with how much power modern CPUs and GPUs use, and the layout of the add-in cards couldn't have contended with a future of 350 W triple-slot behemoths. Assuming it has too much inertia to easily change, I think there is still room for a couple of design wins in the short term:
With the reasonable death of NVLink/SLI outside of extremely niche systems, I hope that the majority of ATX motherboards ditch the second PCIE x8 slot and repeater chips in favor of better placement of the single X16 slot. Slot 1 tends to be needlessly close to the CPU and DIMMs and can easily be pushed down 1-2 slots for better clearance. Asus has already moved it down 1 with the Maximus XII Extreme and Apex, something many older boards had back in the northbridge days. This even leaves a better spot open above slot 1 for M.2 devices, rather than being roasted below a GPU finstack.
Next, we have triple-slot GPUs, but not much clever use of the reverse side of the PCB. Why not have a triple-slot GPU where the reverse side is one of those slots? Imagine for example if the I/O (HDMI, DP) was mounted on the reverse, or spread some of the thermal load out, like mounting the memory packages and/or some of the VRM components on that side. I also think if multigpu is properly dead, bring on the 4 or even 5-slot coolers. Not many users install a second add-in card anyway, why not use that volume to further increase the efficiency of the GPU cooler (the single biggest source of heat in the system).
I think the Mac Pro, of all products, does the best job of demonstrating what is possible without the constraints of ATX. In that design, you have the DIMMs on the opposite side of the motherboard, isolated from the CPU heat. You have the add-in cards done in the style of a 3U rack with high volume full-length coolers designed with very low fin density for low static pressure airflow penetration. The cooling concept of a server rack can be achieved without insane airflow and static pressure, it's just that the heat sink volume, fin densities and obstructions have to be scaled to compensate. ATX, by comparison, is full of dead end corners, right angle airflow, pointless close proximity of heat-generating components, and almost unavoidable obstructions.
I want the motherboard vendors to be more brave with this stuff.
I think you have a lot of valid points, but moving away from ATX is going to take more than a brave motherboard vendor. PSUs and cases will also have to change at the same time. That requires multiple manufacturers to work together on a new concept that may or may not be successful. Or... a brand new standard to be created for the entire industry. I just don't see this kind of stuff happening any time soon.
@@prycenewberg3976 Agreed. The enthusiast PC platform, as opposed to the more locked-down and bespoke SI PC platforms, prides itself on modularity and backwards compatibility. Being so far on the open side of the closed-open platform spectrum makes change very difficult. For it to happen, a more radical design change would have to prove itself extremely valuable to the market-something customers are ready to line up and pay for.
While a more modern standard could clearly do better than ATX, it probably couldn't be so significantly better that it would drive high demand for it immediately. Without that sort of incentive, it's going to be too much of a risk. Maybe it just takes running harder and harder into walls until we're actively looking everywhere for a solution to the headaches, and for now it's still tolerable.
I think ATX12VO and the new 12-pin GPU power connector could be interesting in terms of their market penetration: how quickly will these standards find their way into enthusiast products? If we see rapid uptake, it might predict a more positive outlook for moving past ATX at some point.
To be fair the earlier attempt to improve ATX, BTX clung on much longer in OEMs after colossal failure in the DIY market. Really I think we need to take design queues from the server space for full size boards and more dual chamber cases for ITX which then will allow for more rear mounting of M.2 or memory.
I still remember how we made fun of the first cards with blowers: The dreaded GeForce FX 5000 series.
Back when ATi was king of the hill. *sigh* I'm old
@@louisfriend9323 Do you mean 9700 Pro? that thing was revolutionary - the biggest step in graphics technology since the Voodoo.
@@SteelSkin667 the 9500 Pro was the budget version, with 128b wide RAM, but a beefy 9700 or 9800 GPU.
I ditched AMD for an FX 5900 XT, though, and an aftermarket cooler. My 9600 XT ran quiet, with a tiny fan, but the drivers were SO BAD. I couldn't take it.
🤔 I thought they are supposed to suck by design .
.❤️
xD
Nah man they blow.
@@badass6300 then I need it 😫
Nah, trust me, they really blow.
@@badass6300 homie needs it.
The grill is usually part of the IO-shield. The IO-shield usually contains the mounting holes to screw your card onto your case. So, good luck getting rid of that grill.
There are three reasons why one could opt for a blower card:
1. Your system is very small and you want to direct hot air out of your case right away
2. You're going to WC the card anyway, in which case it doesn't matter which thermal design you pick
3. You love the sound of fighter jets
Oxygen Not Included prepared me so well for understanding the thermodynamics of heatsinks
Often (but not always) we will also see a 'too small' blower fan combined with a crappy cheap tiny heat sink, constricted exhaust orifice and cheap thermal paste.
So a quadruple whammy of ineptness.
RX 470/480 reference cards, I'm looking at you!
I had a reference RX 480. Loudest, hottest card I've ever owned. What a crappy card.
"You could not choose a worse way to air cool a GPU."
AMD's deaf intern - "Hold my beer!"
MSI: ey guys i want to play too!
I have actually had great success overclocking blower cards in a interesting manner. I noticed that if ambient is really really low, open air cooling cards have a flaw that they make a heat bubble around themselves, while a blower card will push that air through its cooler and exhaust all the heated air to not get recirculated. I managed to get a blower rx 470 cooled to 0C under load (didnt show temps below zero, might have been less, north europe winters, took my rig outside), making for a insanely good overclocker without any LN2. Did not get the same results with a open air rx 480, which still managed to stay toasty in the same cool ambient air when running it over 250w, yielding minimal performance increase with voltage increase. I remember taking a non ln2 world record for the 470, dont know if it still stands.
Yep, I knew all this from back when I put a ATI Radeon X 1800 XT into my PC. It not only sounded like a hairdryer, it's cooling performance was also incredibly awful.
There's one great positive effect of blower fans. They keep the heat inside the GPU-case and blow it outside the tower. No hot air is blown on the mainboard like it happens with the axial fans. The fins are always build to blow the air to the mainboard and to the side of the tower. My mainboard for example is having the SSD under the GPU. That means, an axial fan is always blowing the hot air onto my SSD. I don't think that's a great idea. ;-)
Anyway, of course axial fans are keeping the GPU cooler than radial fans, but it also has it faults. I'm sure there are some good reasons to better choose a radial blower (which usually sucks) instead of an axial fan.
But hey, nice video and great presentation. I've enjoyed it and also learned some interesting things I didn't have in mind before. So, thanks for that.
I wish there was a blower design that worked because I like the idea of exhausting the GPU heat out without introducing it to the system.
Me: turns on captions.
captions: "Hi guys, Bill Lloyd here."
blower style should only be utilized when there is a strict airflow management... and they do great job with it!
only 5:00 in, so don't know if you covered it, but would love to know your thoughts on the 3090 FE design with the hybrid blower / axial cooler
Only a couple of days left until we will get reviews of the 3080 and a week or so later we should get 3090 reviews. We might just wait this short amount of time and see how the cooler design performs.
The blower portion has a MUCH shorter HS, so won't have the long HS problem, and the back fan is the much more efficient flow through design. I expect most cooling power at the flow through portion, but blower part will help prevent heat build up under the card. I am betting it works really well.
Seems like an effective design, I just worry about the longevity of the pull-orientation fan, because that motor and bearings will heat up a lot more than any fan installed in a push-orientation.
@@ydihtty I would say that it would last longer in my opinion.
Without having the fans work against themselves by having a pcb to block the airflow. The fan can now spin slower to do the same amount of work because there isn't an pcb in the way.
I have a feeling a lot of people's rooms are going to heat up because the hot air will now actually be exhausted more efficiently instead of getting stuck in a gpu loop.
there will be a downside that the Fan hanging off the back there will provide mediocre Cooling since it's nowhere near the source of the Heat.
the Heatpipes that i'm sure they're using will help compensate for it just like we do with all Heatsinks, but it means that part of the Heatsink will be completely reliant on those Heatpipes to do anything at all and also means that only the surface area of like a few Millimeters around those Heatpipes will do anything at all, and the rest will just be weight for basically no performance result.
if they really wanted to impress me, they'd have the other Fan also be able to move air through the PCB by cutting some holes.
to make that really work you'd want to move the GPU die farther away from the PCI-E Bracket though, to keep the most effective Fan as the one directly above the Die.
in short it's a better Reference design but one that's still pissing a way a lot of performance for little benefit.
Ah, thanks, was curious when new tech ASMR episode is out.
For my 50th anniversary edition Rx 5700 xt (basically a binned reference card) I swapped out the original grill I/O bracket for a single slot bracket and cut off the metal bar covering the pcle slots on the case. I also undervolted, used thermal paste, put thermal pads under the back plate, and did the washer mod. My undervolt is 1800mhz at 930mv but the actual frequency sits at ~1750mhz instead of 1800mhz. At 30% fan speed, my temps are around ~77°C with mem at 88°C while in side of a mini itx case (Silverstone sg13) with basically no direct airflow to the gpu. The temps aren’t awful for being in an itx case and it’s actually not too loud. I don’t notice it too much but even after doing all this work to improve the temps, I think blowers still suck. I’m actually thinking about trying to water cool my gpu instead because my memory temps are too high. I don’t think a customer, especially a beginner or noobie customer has to put this much effort into fixing AMD’s crappy blower cooler. Well I honestly don’t really regret buying it. It still performs great and at the time of purchase there weren’t any other 2 slot models that would’ve fit in my case. Blower cards can still be effective depending on the itx case. Here’s a video by Optimum Tech th-cam.com/video/_dhOsTS6-C8/w-d-xo.html
Edit: oh yeah the anniversary edition pcb is different from a regular reference card. I can confirm this too when I took apart my card. That the anniversary edition card has all the aluminum polymer capacitors populated on the back of the pcb which I assume is why it’s better at undervolting and overclocking
This is a reddit post I found about it. It’s not by me
www.reddit.com/r/Amd/comments/dm3uv2/5700xt_50th_anniversary_isnt_just_a_base/?
I don't think blower style cards get enough Love, they keep your internals much cooler specifically the ram, vrm, also power supply, as long as you have a AIO for the CPU they are great cards. I've had a few open-air coolers and they dramatically heated up my internals.
But another good option is a AIO blower style card they are just very expensive, but they do great work at staying cool along with keeping the heat down inside the case.
If blowers are so bad, how come reference cards from AMD or NVidia come with blowers? Top engineers have decided that blowers are GOOD.
Since blowers exhaust the hot air outside the PC, the air inside the PC case is cooler, leading to more efficiency in heat transfer.
I love blower style fans.
The design is very seamless and more logical for a natural airflow.
The way I see it, blower style fans are good when you have bad overall case airflow, since they will discard the heat immediately. "Normal" style cards are better if you have good airflow since they blow a lot more hot air back into the pc, not only heating the GPU but all the other components as well, which isn't a problem if the case has good airflow
Id love a vid about how the 3080 cooler works!
I love any technical talks that includes the word "sucks" ..... As an engineer myself it is one of my favorite descriptive terms ....
Every design has it's purpose, blower may be not best in your open bench build but its perfect in micro ATX case...
Limited viewpoint expressed on this video. Some benefits of blower cards are that they don't dump all the heat into the case affecting the whole system, works well with multiple cards, for smaller systems can be the better solution. Anything over 220w I found blower cards negative is noise and insufficient cooling. Vega FE at stock is a rather quiet blower card, rated at 225w. Vega and the rather power hungry if you try to push it makes that rather super quiet blower setup into a rather noisy one quick.
Will be amusing when people are dumping 320w+, ASUS 400w, OCing 400w+ with the upcoming Ampere cards adverse side effects from that solution, probably should be a AIO for over 300w cards like AMD has done several times.
AMD also has made some great cooling solution cards as well, well not many but they can do it right at times.
Rtx 3080 "new" flow through cooling design that combines both: allow me to introduce myself.
I did always suspect Nvidia was a fan of his videos!
About the sideways-type of GPU cooler, I've seen one - "gigabyte gtx 680 super overclock" was this type of card :)
Buildzoid needs to get his hands on one!
I have one 😁
19:30 Yes, you're referring to laminar flow, a boundary layer which sounds great but ruins everything. Clean separation = undisturbed air cushion coating the fin stack
21:17 Blowers are perfect if you want to have like 3 or more cards running in parallel. If you went with open air cards, you'd need to have heaps of case airflow and space them out a lot.
I’ve got an RTX 2060 and an RTX 2060 KO on the same motherboard with one empty slot between them (the KO is two slots, the vanilla 2060 effectively 3). The 2060 is giving the KO some nice warm air and the KO’s running at 82C normally. If the 2060 blew most of its hot air out of the case, I think it would help.
Maybe I should go without a case? Open-air test bench permanently?
@@markiangooley if you don't have dust in the house maybe.
I have to shovel out dust bunnies semi monthly :(
Thank you for pointing this out because it's absolutely true, Blowers aren't bad coolers it's always their heat-sinks that are designed completely garbage. making them appear like they are absolutely terrible. they are much more efficient at exhausting heat out of the heat-ex-changer and the system Case, They are loud, yes, but if made properly they can provide superior cooling results then axial cooling that's popular with AIBs. It's just there hasn't been a proper blower heat-sink made. Thanks for reinforcing this idea Buildzoid, Every-time i bring this up to other-people they dismiss me as crazy. But it's a fact especially when you look at common sense at engineering and enterprise systems.
14:07 So those same people who think slower water increases thermal transfer also hold their breath during a jog to increase O2 saturation XD.
Thanks for the easy fix on blower gpus!
At 19:30 you're talking about boundary layer, and thermodynamically that's correct. A turbulent boundary layer promotes better thermal transfer compared to laminar (all else being equal). Although the effect is less significant at the low flow speed and volume we're talking about (small PC fans). A lot of the time, the fancy impeller geometry and funky fin stack shapes manufacturers like to use has more to do with aesthetics and marketing than actual performance improvement (don't get me wrong, design features that promote turbulent boundary layer flow DOES help with thermal transfer, but at this scale, the same improvement can be easily be achieved just as well just by brute-forcing it with running a higher RPM fan)
Great video! But : "from my understanding of thermal physics blower heatsinks suck" They dont suck. They are just used in incorrect applications. Blowers are not meant to be "silent", they are meant to be efficient in relatively low temps. Just blow the shit out of long thermal connective area. They just introduce high speed air flow. That's it. It's the reason we have vacuum cleaners.
Blower cooler has one big advantage
That if you case is too small or bad airflow design
That the blower cooler wont effect and heat the other parts in case
Beside this, blower cooler doesnt really good for every body
Well you just buy a small case that is all ventilation holes and that advantage is gone.
@@Mysteoa I just like ITX case
So tiny so powerful so cute
@@justbaka5946 That's what he said. mITX, very open case layout
@@andruloni Oh I misunderstand then
However still blower cooler is better for ITX
Especially for SFF CASE
ITX video card just too close to the cpu cooler
Even I use the NCASE M1 still prefer blower cooler
Hi guys Bill Lloyd here
(turn on the subtitles)
I honestly love my 1080ti FE in my tiny itx case. She runs pretty cool at 66c undervolted 1900@.912
In regards to the turbulent airflow thing! Here is the TLDR: Turbulent flow results in more mixing of the hot and cold air, which is how you transfer heat. Conversely, laminar flow, has much less mixing of the hot and cold air, basically causing a moving column of cool air in the finstack that never contributes to cooling.
grills, gaping holes, blowing, i knew this was gonna be a good video
It might help to explain the air temperature doesn’t rise linearly. Instead, the air will decrease the amount of heat it can soak up due to the smaller temperature differential, meaning the temperature will rise slower and slower.
In the linear example shown in this video, heat transfer with minimal airflow would be reached if the air temperature is equal to the heatsink temperature exactly at the end of the heatsink. This would only be true if Tdelta doesn’t impact transfer efficacy. Ignoring radiation, Q=hA*Tdelta/d. So the higher Tdelta, the higher heat transfer.
Gapping hole?!...Im in the wrong channel?! 😆😅
Open air cards exhausting in the case increase the CPU temps while blower cards prevent that to the tune of around 8-9°C iirc. Cause the CPU cooler then has to deal with the GPU's hot air. This is assuming you're using a case and not an open bench setup. Radiator placement has similar effects as intake taking in cool air is better than it being exhaust as it then has to deal with other thermals in addition to its own. That's my understanding anyways. So I suppose you could say it's more about regulating other component thermals rather than being superior for the GPU itself.
The biggest advantage for blower cards is that they work much better with WARM and poorly ventilated cases. Otherwise, they SUCK. Open cards also have the benefit of heat pipes.
One design change that I wish we see more often.... is holes in the PCB & backplate to allow air to exhaust more freely.
Me: watching my 2070 super blower black from evga...
My card: please love me!
The only thing that I think you might have overlooked about blowers is their use in servers because for servers it is all about not recirculating any air. So for server cards they are almost exclusively blower style because perform better in that type of system (it also helps that noise is an after thought)
It's bugged me for a while that no GPUs are actually using the PCIe slot above the GPU, it seems like the perfect location to vent the hot air. Giving you the best of both worlds.
I have a fix! Attach a Hoover to the exhaust port, of course. One must compensate for sucky designs with higher powered suckage!
@ 3:46
Why?
RF Shielding for FCC certification.
I clicked for the title. I stayed for the more air, more better.
Just took Heat Transfer last semester. When I saw the new 3080 and 3090 designs I instantly turned into Michael Scott after seeing Toby return from Cost Rica. NOOOOOOOOO!!!!! GOD NOO!
NYOOOOOOOOOO
Next up: Why blower heatsinks are great.
Silverstone-style reverse ATX (even better would be smaller form factors) cases would work really well with exhaust GPU axial fans. Just have GPUs exhaust everything straight out of the top with no PSU shroud in the way. It'd solve the entire problem you mentioned where nobody thought about the heat output modern GPUs would turn out to have. But obviously, it'd need an industry-wide push so that's not happening.
One of the most relevant point of cooling is "Countercurrent exchange".
If it is implemented, it is the best way to transport heat away. Full stop.
Also very important is the speed and the volume of the air and the area of contact of the air and the fan.
So I can not agree, blower are not all bad, bad blower are bad, but blower are not necessarily bad.
What? My vega reference cooler is fine!
You dare!!!
This joke isn't cool.
Vega 64 airboost oc here! Card is only cool after undervolting and fan curve optimization.
Still noisy. Hits about 70ºc on games and have put a sensor on the exhaust. Temps can reach high 50s there, but the air goes out, not affecting the interior mean temperature.
unironically true... but only with undervolt and custom fan curve
The "what" is because you are semi-deaf by this point because of it, right ? :D
It all comes down to the use case scenario. Airflow is not the only parameter, static pressure is another one. Blower has more static pressure, which is more suited for closed PC case such as workstation. Blower fan is not to be blamed, what makes it a bad cooler is the unoptimized cooler design. Blower is more universal to all types of PC case, at least it guarantees cooling in all use case.
Very good summary of the thermodynamics of GPU cooler designs. BZ is right that blowers were designed to compensate for poor airflow cases however he didn't mention that this is actually the case for server racks and other SFF PC cases. It's the reason why Quadros, Titans, and Tesla cards are all blowers.
racks normally don't even bother with the blower fans they just let the chassi fans RAM air into the cards fin stack directly.
You're a great artist Buildzoid.
We're talking about airflow vs. cooling surface area. Would you use an aluminum tube or would you use a brick of aluminum? Both have their advantages and disadvantages
Probably worth pointing out that cooling solutions like GPU blower fans are used a lot in servers for pretty much all of the reasons that make them annoying for desktop use cases (to the point where even the CPU's use these setups). Multiple fast fans draw cool air in through the front, passes over every component and exhausts out of the back. Noise isn't a concern but the depth of the chassis is.
Sounds like at the end you are describing the SFFPC deshrouding mod like the one on Optimum Tech channel. Remove fans of GPU heatsink, stick the heatsink to 2x 25mm case fan in exhaust at the bottom of the case, and there you have a full exhaust system for your GPU. The hardest part is getting the case / heatsink combo which fits tightly in the case if not using after market cooler.
What if you replaced the blower fan with like two 40mm fans blowing air the same direction?
On one hand, you can have more heatsink. On the other, I can't imagine they would move nearly as much air, unless they're super thick 1U server fans.
triple fan 2070 super with a 200 mm front intake at the bottom of case (Phanteks Enthoo Pro) and a 140mm rear exhaust makes my gpu very happy got a 360mm EK aio at the top for the cpu one of the best cases under £100 only thing is I wish i had gone tempred glass but that would be more than I was willing to pay for a case at the time
I'd like to see a 4-slot blower card, with vertically (?) mounted fans in a push-pull configuration. One low-pressure fan blowing hot air out of the rear slots, one high-pressure fan pushing cool air into the card from near the front of the case. And possibly a high-pressure fan in the middle, but I don't know if that'd actually be useful. For four slots, standard/replaceable 92mm fans should be perfectly feasible.
Through my own experience with several different cards, some blower, some with dual axial, and some with triple axial fans, I feel that more of a mention needs to be made of the manufacture's shrouds. While yes, blower style coolers have to exhaust all of their heat out the back of the card AROUND the ports, far too many of the axial style cooler manufacturers design their cards with heat sink fins aligned vertically, only to then almost COMPLETELY OBSTRUCT the majority of them with the shrouding of the card, purely for esthetic reasons. Unless you have a dam decent amount of airflow through the case to overcome those obstructions and to prevent it recirculating the same hot air, axial coolers aren't much, if any better than blower coolers with the fan set to a high rpm. I currently have 5 1070's running full tilt right now, and the blower style are keeping up with the axial coolers just fine. If manufacturers turned the fins on axial cards to take advantage of the gaps in their shrouds, that would obviously change, but those 1070's are pretty clear evidence of the problem.
My imaginary 3 slot blower with heat pipes in addition to a vapor chamber beats all your coolers. Due to high airflow blowers are impervious to dust buildup and take the heat out of the case, which is very useful when your cpu is 280w itself.
man I recently cleaned a blowered that pretty much had a wall of dust at at the front of the finstack.
@@ActuallyHardcoreOverclocking Still better than an open air cooler, in equal conditions that is. There is simply a higher chance that dust gets exhausted out of the case, than with an open air one. I think one of the main reasons why Nvidia abandoned blowers for consumer cards starting with the 20 series, is to upsell Quadros etc for serious people, who need to stack 4 of them into a case.
For a single consumer tower, blowers do indeed suck. HOWEVER, consider compute clusters / datacenters / servers, where dozens, hundreds, or thousands of components all need to be cooled. In these applications, airflow throughout the entire space must be carefully considered, and it becomes important to have dedicated "inlets" and "outlets" for each system. Ideally, cool air is delivered to the inlets, flows consistently through the relevant cooling components with minimal recirculation, then exits through exhausts (to AC units, etc). Blower fans are perfect for this - not only do they actively contribute to the pumping of air in one direction through the system, but they also minimize recirculation of hot air within the case itself (especially if many GPUs are installed right next to one another). In other words, blowers work well when ensuring a certain "flow pattern" is critical for maintaining efficient cooling of many tightly-packed, power-dense systems. (It turns out this is also a useful arrangement for managing dust with minimal maintenance).
Note that enterprise compute cards exclusively use blower-style coolers: Quadro, RTX Server, Radeon Pro, etc - all use blowers! This lets you pack many of them into a server case right next to one another. Then, you can stack a bunch of those cases together in one rack. Assuming you deliver proper ventilation to and from the rack, you will still be able to get acceptable cooling on all cards in all systems (which would not be possible with open-air coolers!)
On the other hand, open-air coolers work great for individual PCs, i.e. applications where there are 1-2 GPUs surrounded by lots of empty space. Open-air coolers are very "selfish" coolers from a systems perspective (they don't drive bulk airflow in any single direction and tend to throw exhaust heat towards all the other components), but the overall power density is low and ambient case temperature essentially agnostic to cooling performance. As long as there are some fans in the system, hot air will "just find its way out of the case" and it doesn't really matter where or how. Raw airflow volume (rather than direction / inlet / exhaust) is all that really matters, making open-air coolers great for individual PCs. In this use case, open-air is definitely the way to go! Blowers aren't nearly as efficient or quiet here, but that doesn't mean they don't still have their own applications!
Jayz tested the impact of the NVidia 3000 cards on the temperature in a case.
so its funny to see that the blower part of the cooler help by moving a lot of hot air outside the case.
so to compensate this when using an open card, we should have more fans to move the air out of the case faster.
like mini fans below the card pushing hot air outside the case (like the top back fan) or other options like this one.
The blower will extract the maximum cooling potential out of a given amount of air if it's hitting equilibrium right at ejection, but will not maximimize the cooling potential of the fin stack. If blowers weren't way louder for moving the same amount of air, it might be a good idea ;). I think a blower will generate plenty of turbulence since the air exits tangentially. Static pressure isn't much of a concern with a blower, but of course you pay for that by sacrificing cfm anyway.
Good explanation! And regarding the typical exhaust on blower cards.,.. You hit the nail right on! :)
I kinda agree with you. But blowers can easily push the air efficiently out of that short heatsink they normally use. You don't get such huge difference in temps over the HS. Usless you keep the fan very low speed. Forward curved blower fans are not really airlow efficient, but high static pressure efficient. static pressure is what is needed inside that "pressure" chamber. Especialy with the higher resistance expansion slot cover plate. They should really just make a big hole. I totally agree on that.
Blowers are noisy because of the high static pressure from the high rpm. But because bad case airflow is not something rare, and especially with air cooled multi GPU setups, blowers advantage in that case, has been tried and proved.
I do not think it is the fault of the case designers either. With the way the pcie slots are arranged on the mainboard there are no good options to place the graphic card without using some crazy stuff like riser cables. Maybe it is time to change the mainboard layout completely, to allow graphic cards being mounted parallel to the backside of the case. This combined with holes inside the PCB like for the RTX 3080 would allow to blow the air out big time, not through such a small hole like you draw here. The question is then, were do you put the io for external devices like soundcards and such. Maybe to the side? This will limit the possibilities for fancy side glas panels, but the overall cooling performance would be worlds ahead of what we have now.
I have to admit it is kind of hard to establish this kind of revolution to mainboard form factors and with the current development it is more likely that we reach efficient hardware than a revolution of the ATX Standard.
Even worse, most of the air pulled into a blower is expelled into directions other than the actual sink's. Plus, blowers accelerate air radially - meaning there's only a small amount of air that goes into the fin stack straight while everywhere else you get a lot of turbulence (=air resistance) due to the air needing to change directions so it will flow through the stack's channels.
I mean, one could possibly design a blower with fin stacks in front and behind it, but that leaves the middle part of the card without an effective means of transferring its heat into either of them, barring the use of heat pipes - and these require a larger cooling solution, plus they impinge on airflow yet again as they'd need to be threaded through the fin stack to facilitate efficient heat transfer.
Whichever way you look at it, blower designs range from bad to horribly inadequate, trending towards the latter.
Actually blowers have a place in the inverted case markets, (they would typically have a panel to take in air if you had a case with an upside down mount for the GPU)
But also i was using a Titan X (maxwell) for 3 years that would max out at 74c (and i ran that thing pretty much 24/7 on BDO) in the FT05 (and a passively cooled 4690k on a megahalems for that matter) x2 180mm fans if im correct blowing all the hot air out, Only used the high fan speeds in the summer,
I had to change the card i had before that because the gainward phantom would melt the top of my TJ08E, that cooler has the 3 fans... Under the heatsink, Blowing on a cold plate (that would be warm after 10 mins anyway hurrr) to grab that air out, I mean it looks lovely but the blower has much more of a use case compared to that thing.
Also blower cooler fans doesn't push air straight through the heatsink, it pushes it inside the card in all direction, which creates high pressure, which pushes the air through the heatsink, but also creates a lot of turbulance
Does no one see the irony in the title saying that something that blows sucks?
Like a Data Vac Blows instead of sucking???
Mega Maid?
Very good explanation.
You did mention something that might be slightly wrong, depending how you look at it. If you look at a blower fan, the air does exit the full curcumference. The duct is shaped to guide the air evenly across the one opening at a much higher speed than it exits the fan blades.
Edit: what do you think of ducting air from the axial fan heatsink to the outside of the case using a third and fourth slot?
I think the visualisation would be much improved if you use increasing distances between temperature steps.
Like, right now, the distance between 20 and 30 degrees is the same as between 50 and 60 degrees, but if it becomes larger, that visualises the decrease in transfer rates.
Most ventilators sucks by design! And then they blow...
I know: I'm fun at parties :D
Msi Air Boost Vega 56 have a bigger window and would recommended a cutting tool to make full use of the blower exhaust gate.
Blower fan are Pressure fan to create high volume of air pressuring inside the chamber and as physics, the air will find the way out creating thru a wind tunnel, that said if part of the window is being block at exhaust, more heat will build up inside the fins than being push out fast enough. Not great for OC dues to lack of making it a triple slot or go old school to put in a blower bigger than the width of the card.
I have to admit i only clicked because of your amazing paintskills Buildzoid :)
I was actually a little shocked at how bad mine was, given that the entire shroud was also made of metal. I think it was aluminum, but I sold the card and can't tell now. That fin stack inside only did half the cooling, and the outside got brutally hot as well. Curious how much of the temp issues were just related to the 7nm process. Like a ~70w ryzen 3600 is harder to cool than a ~90w 2600x.
19:30 you're thinking of the Boundary Layer and yes you are correct turbulent air reduces the size the boundary layer and improves thermal transfer.
Now I want to see BZ's improved case designs
I once removed the terrible blower fan from a GeForce 6800 and glued a good quality axial fan over the hole. It was actually way quieter than the original fan even though the air had to make a 90 degree turn to go through the heatsink.
Those squirrel cage fans are noisy by default.
Can't wait for you to take on the 3000 series cooling issues. The thing literally spews heat toward the CPU block.
if blower style sucks , may I point out, that servers cool, by blowing out air out of the housing ? and they often have a multitude of GPU's in the case
Maybe you should make a 2nd part with case/laptop/console thermals.
14:55 same thing applies in the case. The air moving within your components is too hot to make a difference. Just having even one properly placed exhaust fan and one intake fan makes all the difference. It's also why just by removing a panel on your case you drop 20C while doing nothing to the fan situation, basically free cooling.
Then there's consoles and laptops where almost anything you do will be bad to mediocre, because there isn't enough space to fit the air without very loud blowers.
I always wanted one of those gigabyte 680’s that had all those small 40mm fans for my collection. I check ebay all the time for it and can never find one.
I actually tested my reference RX 5700 XT without the rear IO bracket. I did not see any significant difference in temps on same fan curve, driver and same games/settings. Narrow slats design is often used in many intake/exhaust aplications like car front grille, AC, ventilation etc. It is there because something could get inside and damage or block the air vent. According the fluid mechanics, slats are pretty effiecient regarding air flow (90% intake and 92% exhaust compared to 100% unrestricted air flow). Nice video, though. Thanx
I did a LEARNEDING during this video. Knew blowers are not good coolers, but now I understand more details on why and I feel I've done a SCIENCE today thanks to you.
Thanks for the video! :D :D :D
if u take out the plastic cover and stock fans, replace them with high SP fans and they get as good as a AIO and really good OC
Loved the video I love listening to you especially when you rethink your presentations realtime Love this
My friend built a new PC where the Asus 2070 Super video card had ~1mm clearance between its back plate and the Be Quiet! Dark Pro 4 air cooler. The processor ran hot because the video card transferred heat to the air cooler. I put in my modest 1050 Ti video card and the processor temperature dropped 7C. My friend gave me his air cooler and got an Arctic Freezer II AIO.
Who volunteers to write a successor to the ATX specification to fix some of the problems mentioned here? A recent fix for a related problem is the 12-V-connector proposed by Nvidia. I think the whole industry needs to standardize better on certain aspects as well to solve some of these legacy issues (good luck dealing with all of the diverse interests of the CPU, GPU, PSU and case vendors of the world).
It´s not as bad as you think, or else servers would not be as they are. Heat pipes really save the day in these designs. The problem is the tolerances and forcing all the air properly though the hood.
Any gap, will let air through and reduce the pressure in the heatsink. So, I would say it´s not a bad design, and it can actually be a very efficient design. Just getting it right is difficult.