They are doing Gen5 SSDs wrong. We don't need more speed, they should use half the pci lanes for the same speed as a Gen4 ssd. We could have more m.2 slots with the same amount of lanes used.
I still wonder who need more speed? I need a very high capacity SSD for back-up and archiving. They should stop restricting themselves to M.2. There's 2.5/3.5 inch using U.2 or the newer U.3. I need 16TB SSD.
Well it's so inaccurate that it's clearly not tied to die sense. It's probably a thermocouple/thermistor on the HS hot side with an assumed offset -- like a forehead thermometer.
@@concinnus I think you are reading it wrong. The reading on the meter is taking the temperature from the heatsink and not as much from the SSD. You would want to see a much higher temperature on that readout if the heatsink were working correctly.
@@concinnus yeah that absolutely makes sense - it is useful for checking how the heat transfer is working and, assuming that's working, you'd see a rough approximation of the temp lol
@@DizConnected Err no, if the heatsink is hotter than the SSD, it's just become the heat source, and the SSD is then the Heat Sink. Basic physics, heat flows from hottest part to coldest part.
The bad idea isn't the heatsinks, but that they made these Gen 5 SSD that require major cooling to just run without "exploding", especially as the majority of users are not tech savvy and having to cool the SSD will be a new thing.
Its dumb anyways, because the insane sequential speeds these offer are not ever going to be used by 99.99% of buyers. People need faster random reads, which is useful for near every task, not faster sequential which is only useful for moving massive files 50GB+ files, we've already had the ability to move big files plenty quick for years, like a Gen5 SSD could move a single 50GB file in say 5 seconds, while Gen4 can only do it in 6 seconds.
I was thinking the exact same thing. That thing puts out so much heat that literately everything including the massive fan only lasted around 10 minutes before getting to hot : ( Pretty soon, your computer will need a chiller and a dedicated 30 amp circuit just to run. It sure does seem like efficiency is loosing the battle here.
The problem is that M.2 ever became a mainstream thing in desktop. It was originally laptop format. Ind desktop the logic was to use Sata-express or U.2 directly. The M.2 drives, with their zero cooling and the way they are usually installed (under graphics card, in dead zone with no airflow between the GPU and CPU, etc.), its just all wrong. I personally adapted all the M.2 slots to U.2 connectors are replaced all this consumer crap with industrial U.2 drives and problem solved.
Especially with the already flaky/questionable reliability/longevity of SSDs that didn't even get hot. We were promised so much more reliability and longevity with SSDs over HDDS, but it really doesn't seem much or any better.
The fact the drive can't stay idle without running into temperature issues tells me that's where the problem is. How do you explain that with throttling, and back essentially to Gen 4 speeds, it is still at 80°C?
@@Azuraken Not all motherboards or M.2 slots have a heatsink though. Not all SSDs come with a heatsink either. This is fairly new territory for people, as 2.5" SSD's obviously never had heatsinks and a lot of gen 3 and gen 4 M.2 ones didnt need one either. Also M.2 will go into laptops and consoles that cant even have a heatsink.
@@Azuraken You're comparing apples to oranges. CPUs and GPUs consume massively more power even at idle than NVME drives do, so they've always needed a heatsink. NVME drives have not *required* heatsinks in the past the way these Gen 5 NVME drives do. The idea of an idle NVME drive being so hot as to risk tripping a thermal cutoff is very much a new one.
@@pollin522 GPUs didn't have heat sinks in the beginning either. But as technology progressed and more power was needed to power them they got hotter. And current SSDs require more power than early heat sink GPUs but in a much smaller package.
For normal home use I dont see the point of faster than Gen 3 yet~ We kind of need better ways to split PCIE lanes out so we can add NVME like SATA today, Gen 4/5 drives split out to use 1PCIE lane each.
For normal home user SATA-3 600MB/s is also enough. Just look around a real world speed test on the internet hdd vs sata ssd vs pcie gen3 ssd. Few games have noticable advantage and few workloads thats it.
@@maestrohuni agree with this statement, sata 3 is enough for pretty much everything that a gamer or a casual worker will need. I still use my 2,5" Intel 530 series SSD as boot drive, because there is literally a 1 second difference between that and my m.2 Samsung 970 Evo Plus.
@@maestrohun We will hit the point NVME is cheaper than SATA, amazon has Crucial P3 1TB Gen3 drives for £35 & P3 Plus 1TB Gen4 for £44. That's on sale but still cheaper and faster than SATA drives.
@@liaminwalesthat being said, the P3 is an awful SSD. It really doesn't matter which interface it uses if it just falls on its face during regular use (which the P3 does).
Consumers need more random read speed improvements, which Optane was, they dont need sequential improvements that are only useful for people moving hundreds of Gigabytes of a single file around, like an 8k video.
Make sure the controller chip on the M.2 board is making firm contact with the cooler surface. Usually the NAND chips are thicker and might prevent the cooler's surface reaching the controller chip, when applied with thinner thermal pad.
I'm wondering if the 2 coolers that looked quite decent (the red one and the dual heat pipe one) had old or too thick thermal pads. I'd retest them with something decent with a thickness of 0.5mm. Also, just an idea, but the red one might not have any fins under the plate since there is an lcd/oled there.
With all SSD they should really say what is the continues speed after like 30 minutes. After the cache and over heating is done. Cause showing only peak speed doesn't give you the real performance.
@@maestrohun What are you doing 12GB/s anyway? You have two PCIe Gen 5 SSDs and are just copying 20GB files from one to the other all day long? This is why peak sequential read/write is useless as a measurement of performance. This SSD will overheat while idling, causing the system to crash, so over an hour its performance is: 0MB/s...But for a few seconds it can show 12GB/s in the benchmark tool, so that looks cool... It's like saying this new Volkswagen can do 400KM/h. Yeah but is it safe, nice to drive, do the groceries and whatever fit in the boot? Can I actually use it?
I wonder if there will be a difference when using something like Fan Control to ensure the fans are indeed running at 100%. Otherwise they might defailted to default 3-pin case fan setting
He could have just use a converter to power the fans off molex or sata so that they are always running at %100. I did that with some in my PC because motherboard use to run them too slow and fan control wasn't properly working with 3 pin fans.
3 pin fans don't have fan control beyond lowering their voltage. They have 12VDC, GND and tachometer pins, the last one being a signal that is generated by the fan to indicate the speed it is spinning at so the system knows whether a fan has failed or not.
There is one actually good m.2 cooler that worked well for me its the Thermalright HR-09 it has two heat pipes that go through an proper heatsink with fins. also If you want to have better temps remove any stickers from the m.2, coper or otherwise and replace with a nice thin thermal pad.
Plugging a fan onto the motherboard defaults to the BIOS fan curve for that header, which most likely goes by CPU temp o chipset temp which were idling during testing. You probably never run those fans at full speed (and probably from this comes your wonder why were they quiet).
For me, as content creator, these SSDs are necessary, especially when I render big music files, let's say like projects with 200 or more multitracks, sometimes even 400 tracks, if you have to mix a big orchestra, with different microphones in an opera, each actor has its own microphone. And all the instruments. A consumer SSD would throttle the whole time, that's why we use M2.nvme in Raid 0, with a realtime backup running in the background.
@@someguy4915 I'm already maxed out 128Gb, and already using SSD cache, but that is also limited. in my studio there is an Apple Pro, and this one also has problems with that many tracks. On many track are plugins loaded, these are recourses eaters. That means, when the system is saturated it uses HD' or in this case SSD's, that's why we love these gens 5 SSDs.
@@Bluelagoonstudios I guess with Apple you're limited with what they offer in terms of RAM yeah so then PCIe Gen 5 SSDs are the way to go. With Windows/Linux I'd say a 512GB RAM system isn't that much more expensive, assuming the content creation is a job so the RAM (or SSD) is just an investment to do your job faster and easier. But for any other situation, seeing these Gen 5 SSDs overheat while idling is ridiculous xD
Another thing to add into the custom waterlopp :) Seriously you showed that no cooling makes too hot to run and even shutting the system off. Many people without proper airflow over the motherborad wont be able to run this ( a " fancy" motherboard with SSD coolingpads is necessary). If the SSD doesnt fit onto the motherboard without proper cooling maybe we go back to mount them externally with riser cables or something. I dont need more speed if that much heat comes with it, Gen4 is fine by me...
The one with heat pipes needs a shroud so the air can effect entire heatsink, it blows sideways almost immediately. I doubt it will make too much of a difference but still basic underrstanding of airflow should be applied to design.
Intel sells high-performance CPUs without any cooling solutions included as well, but this SSD cooling stuff might be a new concept for some people so I hope they label it very clearly if it requires additional cooling.
Very nice (and fun) testing! One thing I want to point out is that with the motherboard built-in heatsink, you are testing it with the GPU idle and on an open bench, so it will likely hit near 75-80°C even idling when playing a game on the GPU (I'm seeing similar behavior on my X570S board with gen4), so if you load it up alongside the GPU, it will definitely throttle. IMO these things are still a generation or two worth of refinements away from general viability... even gen4 isnt quite low power enough to just "slap anywhere and it will be fine".
In my opinion, the problem is not in the heat sink, but in the design of the ssd. A plastic cap on the chips, plus paper/plastic labels, plus a thermal pad - these are very large losses in thermal conductivity. Even water cooling will not help, at least the difference will not be so significant. I'm pretty sure the ssd doesn't produce enough heat to be dissipated by the aluminum heat sink with fan.
It is fucking bananas how hot these controllers get (on pretty much all SSDs). What the hell. Problem also is the controllers are packaged in plastic and have no metal interface to wick the heat out properly. So for most users the nominal use case is going to be it running at half speed pretty much 99% of the time.
I don’t have an Ssd, but this method works for pendrives to fully reduce the heat: connecting to a Samsung OTG cable, and to set the phone to standby. Also works with cards inserted into card reader. After this, if you connect it to a Windows Pc, and if you copy some gb files such as videos within the card, into other folders. That speeds up the drive, no failure will happen. The copying sets up the real mb/s of the card. So if an sd card is advertised at 50 mb/s, the copying sets up its 4 mb/s copying speed for example 4 gb sd card. The setup dramatically reduces the heat.
The fact that these things exist to begin with means to me that there's a serious design flaw in Gen 5 M.2 SSDs they shouldn't get this hot to begin with.
Just basic physics, the same as how your CPU heats up if used without a heatsink. Don't forget we once had PCs without a heatsink. Then we had to start adding little heatsinks which over time had to be bigger and bigger as CPUs started consuming far more power.
Pretty crazy to think that we've barely had to think about temps in the past, and now you want the best and even actively-cooled devices cannot keep up
Good video again. I think, that sticker over the chips, might cause poor thermal transfer. And here is my idea, the manufacturer, can make those stickers like thermal pads style, with some thermal conductivity, or just remove them and change the type of labeling.
If you need a heatsink that badly just get a board with one already on it. Your case will proved the airflow on its own. If you think it looks cool go for it otherwise unless you are writing to the drive 24/7 you dont need one. Your OS on a regular SSD is perfectly fine.
I am curious as to what happens with temp when you remove the sticker. I have always removed the M.2 sticker for more direct contact to heatsinks, but became aware that some companies make the stickers infused with copper.
I actually bought a few of those tiny 20mm centrifugal fans. For cooling random electronics. They are better than nothing, but without any engineering in cooling fins it won't do much.
Ah the temp reading on that heat sink seemed to have found the kid in you ha ha. It's a nice touch for sure. By the way I used copper shims with thermal grizzly & a cut up back plate from an old GPU to cool mines & it does a great job!
These heatsinks seem made for the look. I'm sure the heat pipes aren't functional. They are meant to be made of pure copper, have a granular surface finish on the inside, and be under a precise-near vacuum where some water vapour is added. I doubt any of the functional parameters have been implemented there. Thanks for the video, and buy computer parts from reputable sources which know what they are selling and respect their customers...
Also for copper heat pipes to work properly there needs to be a temperature difference between evaporative and condensation zones. They seemed to be connected to top and bottom of the same aluminium extrusion which after the while will have equalized temperature distribution.
This is a big challenge. Mounting pressure, clearance, performance, and compatibility will all be difficult. I'm looking forward to seeing what you develop for Gen 5 ssd cooling.
The problem is that regardless of how much mass and air throughput you have, the weak link is the thermal tape, because it is highly inefficient at conducting heat away. Replace the thermal tape with paste on the top of the chips, get a nice bare metal surface on the graugear, and retest with something like kryonaut
All current PCIe 5 SSDs shipping today were designed to be used with a cooler. If you look at the documentation it clearly states use the drive with the motherboards heatsink. The drives shipping without a heatsink are for users that have motherboards that include a cooling solution.
I think first one is for Samsung SSDs, which is known for hot controllers. Its fan is right in usual place of controller. Samsung also have dedicated temperature sensor for controller. This Crucial likely just have one for memory.
You should use the better working coolers from you previous vids on the Gen 5- also, not all motherboards have this huge heatshield like gigabyte does - if you have other motherboards - you should check out how the built in solution works there.
You hardly need any airflow to cool a 10W load but you do need good thermal transfer from the load to the heatsink for it to have any chance of doing anything, which is likely where most cheap aftermarket solutions are failing.
I think the Graugear cooler is a good idea with bad execution. It definitely needs walls on the sides to guide the air across the heatsink! As it is, the air hits the first 20 maybe 30% of the heatsink and then disperses out and loses velocity. Would be interesting to see how it performs with paper walls on the sides for science. Also at these temps, maybe the paper sticker is hurting the temp performance a bit. It is probably not by much something like 1-2 C but still that may be the difference between throttle or not. The quality of the thermal pads provided is quite questionable as well, I know that if i get a ssd cooler i will use my own pads.
Would be "cool" to go deeper on how built-in motherboard m.2 coolers perform (including fully populated) and if/when it even makes sense to buy separate coolers.
To get better results, before using the heatsink, it is better to remove the paper sticker on the SSD so that the thermal pad can be placed directly on the memories and have a better heat exchange.
Yeah, I'll stick to my gen 4 nvme, thank you very much. I do not need another oven radiating from my PC case, especially during the hot days in the summer. Funny thing is even the motherboard cooling solution is not good enough, which is big hunk of aluminium with heat pipes! Yikes!
I have the "natural cooling disk" and mine never goes higher than 41 degrees (normally 2 degrees off of what HW info reads) Why is it so different? Do gen 5 m.2s just use loads more power?
Same here. I've used these coolers for years. My drives (gen 3 + 4) stay way cooler. I'm surprised at how much hotter gen 5 seems to be. Looking at my gen 4 drive now it's 14 degrees colder in idle with this cooler than how it's running in the video.
I would be very interested in seeing how these would perform with water-cooling in my ROG MAXIMUS Z690 EXTREME GLACIAL, which has a full coverage water block on the motherboard and a dedicated Alphacool NexXxoS Monsta 560mm Radiator with push pull and x8 Noctua NF-F12 iPPC 3000 PWM, Heavy Duty Cooling Fan, 4-Pin, 3000 RPM (120mm, Black) with a dedicated Aquacomputer ULTITUBE D5 200 PRO reservoir with D5 NEXT pump. Total overkill but it would be interesting to test.
There must be some kind of a problem with the interface, or the component which is generating heat. Memory chip gets hot, but the controller is the component which heats up more. Even mosfets run cooler as soon as you apply any sort of heatsink to them, and they run even cooler when you add a fan to the equation. So I think that the problem there must be where the heat is produced and/or sampled...
Hey Roman, be curious to see the results with the supplied motherboard gen 5 heatsinks , I have a x670e taichi amd a arous master z790 both using gen 4 at the moment so supplied heatsinks not utilised yet
There's not one, not two, but three seven segment dispays on this ssd cooler. If motherboard manufacturers are to be believed, this ssd cooler costs at least $600 to produce.
a. motherboard manufactures are not to believed for almost everything. b. I don't think they ever claim that adding 2 seven segment displays cost too much, it's just their shameless way to up-sale you expensive boards. c. If we the consumers would demand 7 segment displays on cheaper boards, they can integrate it to 150$ boards.
The test is flawed, you didn't take into account the room temperature + you didn't configure your motherboard to which fan speed to set the fans in the bios, its literally depend on the voltage your MB send ... instead you kept plug/unplug the fans which was super cringe. As well it looked like your setup was horrible with the graphic card literally throwing heat on the M2 area, you had to use some small and cheap graphic card or use the on board graphic card (if available), clearly not that monster you put there .. but that's another story.
The design on the top M.2 cooler for the Gigabyte AORUS Master X670E board is how they need to do things for Gen 5 SSDs - on my Gen 4 drives, it idles 3-5C cooler than the M.2s I have under the main plate area down below (which has 3 drives under it, to be fair) and under load, it's a huge difference, 10-15C or more cooler than the ones under the plate below the GPU slot. Would be interesting to see results with a Gen 5 drive, but I have neither the need nor the money to get one at the moment lul
This is the part when the testers/reviewers must realize that their open test bench not suitable to test it to the rest of the word that has a pc case and a significant airflow trough it.
But poor airflow over the motherboard is quite standard. Especially in most cases where the GPU basically blocks any significant airflow over the motherboard itself. And since most coolers he tested here have fans, airflow (or lack thereof) is not the issue here.
It did not show clearly, but was the label taken off before the heat sink was put on? Those stick on labels ack as thermo insulators, preventing heat transfer.
Silicon Teflon brush it on and leave it in the sun until you see the plastics appearing wet then clean kinda keep it wet looking put it together and use it , should have no throttle issues chip wise
I tried removing this heatsink (Natural cooling disk), when you turn on the fan, in the lower position of the fan's shaft is the fan control board, it causes heat and radiates heat out, actually this little fan is not needed. , circuit board display. The temperature display and the sensor are located quite far away, so the results may not be accurate compared to the crystal software
That little fan is adorable, you need to name it. Worlds hardest trying fan. Edit: Temp display is neato as well. Give me more stats, screens upon screens of information that I don't even need.
Couple of years ago I saw a video from LTT, that said that most SSDs actually like heat, because the hotter they are, the less damaging the SSD writes become.
The main issue is that he plugs into fan connector without setting it to run max RPM, we have no idea whats the rpm and it probably runs some low numbers not full speed how it supposed to be with these fans
i think we should be looking to the manufacturers with the questions about temps and why this is on the market, because i bet a few more min even on the mobo cooler it would've over heated
I added a Noctua NF-A4x blowing across the top to achieve 60/64 degrees sustained read and write on Samsung 980 Pro's. An additional -3 degrees with copper fins on top of MSI heat plate in slot M.2_1. Temperatures are NAND/Controller. Nothing could be done on slots M.2_2 & 3, (under GPU). On slots M.2_4 & 5 my results with Samsung 970 EVO were: stock solution - 57/61 C +Noctua NF-A4x blowing across the top - 54/61 C +Copper finned heat sink and new and slightly thicker thermal pads adjusted for different height between NAND and controller - 51/61 C In order of importance I feel are: properly contacting thermal pads, a copper finned heat sink and a fan moving stagnant air areas.
7:00, I think 3 minutes of max performance is adequate cooling capacity. At 12 GB/s, you can read the entire capacity of the drive in less than 3 minutes. So yes, while academically, it can't sustain max performance over extended period, I don't know if the test is relevant for any demanding, but real-life scenario. Of course, this is assuming your motherboard didn't come with a reasonable solution built-in.
I think it would have been interesting to test the time it takes for each cooler to reach their idle time from the max C. in real use scenarios, the sad isn't going 100% the whole time
For a better test, you also need to have it installed in a "real" system, meaning a closed case with case fans circulating air through it, which would help with cooling and show better values for a real scenario. These open test-benches don't really work for real use scenarios, but even so they give a good picture of what you can expect, and I believe in this example, that the onboard heatsink would perform even better with case fan through flow.
M.2 drives that need active cooling, should have been U.2 drives in the first place. We just need U.2 on consumer boards. There's no reason NOT to have U.2 over some SAS SFF on-board port, and bundle a couple adapter cables for a couple U.2 drives. There never should have been so many M.2 slots directly on the board anyway. That should have been specced out for M.2 to U.2 2.5" or other M.2 enclosures.
They are doing Gen5 SSDs wrong. We don't need more speed, they should use half the pci lanes for the same speed as a Gen4 ssd. We could have more m.2 slots with the same amount of lanes used.
maybe we coudl start stacking them like hard disk ... disks. For mor capacity.
@@Thisandthat8908not a bad idea
I still wonder who need more speed? I need a very high capacity SSD for back-up and archiving.
They should stop restricting themselves to M.2. There's 2.5/3.5 inch using U.2 or the newer U.3. I need 16TB SSD.
@@AlfaPro1337 100% i want an ssd nas with 2 to 4 16tb ssds. 2.5inch ssds is a formfactor sweetsport for me. i dont mind the sata and satapowerplug.
@@AlfaPro1337 U.2 is just a connector, but U.3 requires a pricey controller that can do SATA, SAS *and* NVMe...
I'm hating that watercooling SSDs might be a thing in the not too distant future
😂😂😂
not really.. i think the majority of people wants higher capacity M.2 and not speed (heater).. let the data centers have the faster ones.
It's already a thing Corsair has blocks for SSDs
@@heyitsmejm4792 I want both but bring on the 30TB M.2 drives!
You should see what wonders a solid copper block does on top of an SSD.
That thermometer on the cooler is pretty dang cool, I absolutely dig it. Shame the heatsink itself is essentially useless.
Well it's so inaccurate that it's clearly not tied to die sense. It's probably a thermocouple/thermistor on the HS hot side with an assumed offset -- like a forehead thermometer.
@@concinnus I think you are reading it wrong. The reading on the meter is taking the temperature from the heatsink and not as much from the SSD. You would want to see a much higher temperature on that readout if the heatsink were working correctly.
@@DizConnected That's...what I was saying? My point is, it's not an especially useful measurement.
@@concinnus yeah that absolutely makes sense - it is useful for checking how the heat transfer is working and, assuming that's working, you'd see a rough approximation of the temp lol
@@DizConnected Err no, if the heatsink is hotter than the SSD, it's just become the heat source, and the SSD is then the Heat Sink.
Basic physics, heat flows from hottest part to coldest part.
That thermal pad is the bottleneck in the first cooler. That thermal pad was made for chips that never get hotter than 40c.
yep, it'd be better with arctic silver.
sounds like phone chip type of thing as phone CPUs dont really go above 50C, likely just warning and shutting down if it somehow goes to like 60C
@@buggerlugz6753 Arctic silver is not a thermal pad. But 40W/mK graphite pads or even the 21W/mK would work wonders.
Am I the only one concerned about the sticker on top of the components being a bottleneck?
@@approximatelybored most stickers on those drives are actually foil stickers helping thermal transfer atleast on decent drives
The bad idea isn't the heatsinks, but that they made these Gen 5 SSD that require major cooling to just run without "exploding", especially as the majority of users are not tech savvy and having to cool the SSD will be a new thing.
Its dumb anyways, because the insane sequential speeds these offer are not ever going to be used by 99.99% of buyers. People need faster random reads, which is useful for near every task, not faster sequential which is only useful for moving massive files 50GB+ files, we've already had the ability to move big files plenty quick for years, like a Gen5 SSD could move a single 50GB file in say 5 seconds, while Gen4 can only do it in 6 seconds.
I was thinking the exact same thing. That thing puts out so much heat that literately everything including the massive fan only lasted around 10 minutes before getting to hot : ( Pretty soon, your computer will need a chiller and a dedicated 30 amp circuit just to run. It sure does seem like efficiency is loosing the battle here.
The problem is that M.2 ever became a mainstream thing in desktop. It was originally laptop format. Ind desktop the logic was to use Sata-express or U.2 directly. The M.2 drives, with their zero cooling and the way they are usually installed (under graphics card, in dead zone with no airflow between the GPU and CPU, etc.), its just all wrong. I personally adapted all the M.2 slots to U.2 connectors are replaced all this consumer crap with industrial U.2 drives and problem solved.
@@__aceofspadeswhat's the best m.2 then ?
Especially with the already flaky/questionable reliability/longevity of SSDs that didn't even get hot. We were promised so much more reliability and longevity with SSDs over HDDS, but it really doesn't seem much or any better.
You know what would really help? If there were m.2 adapter that propped the ssd sideways, I think it would open up cooling options and help cooling
why didn't they build it that way? i mean how stupid of them.
99% of ssds only have chips that get hot on 1 side.
The fact the drive can't stay idle without running into temperature issues tells me that's where the problem is.
How do you explain that with throttling, and back essentially to Gen 4 speeds, it is still at 80°C?
You should try idling your cpu and gpu without a heatsink
You talk about open bench system with no airflow
@@Azuraken Not all motherboards or M.2 slots have a heatsink though. Not all SSDs come with a heatsink either. This is fairly new territory for people, as 2.5" SSD's obviously never had heatsinks and a lot of gen 3 and gen 4 M.2 ones didnt need one either. Also M.2 will go into laptops and consoles that cant even have a heatsink.
@@Azuraken You're comparing apples to oranges. CPUs and GPUs consume massively more power even at idle than NVME drives do, so they've always needed a heatsink. NVME drives have not *required* heatsinks in the past the way these Gen 5 NVME drives do. The idea of an idle NVME drive being so hot as to risk tripping a thermal cutoff is very much a new one.
@@pollin522 GPUs didn't have heat sinks in the beginning either. But as technology progressed and more power was needed to power them they got hotter. And current SSDs require more power than early heat sink GPUs but in a much smaller package.
For normal home use I dont see the point of faster than Gen 3 yet~
We kind of need better ways to split PCIE lanes out so we can add NVME like SATA today, Gen 4/5 drives split out to use 1PCIE lane each.
For normal home user SATA-3 600MB/s is also enough. Just look around a real world speed test on the internet hdd vs sata ssd vs pcie gen3 ssd.
Few games have noticable advantage and few workloads thats it.
@@maestrohuni agree with this statement, sata 3 is enough for pretty much everything that a gamer or a casual worker will need. I still use my 2,5" Intel 530 series SSD as boot drive, because there is literally a 1 second difference between that and my m.2 Samsung 970 Evo Plus.
@@maestrohun We will hit the point NVME is cheaper than SATA, amazon has Crucial P3 1TB Gen3 drives for £35 & P3 Plus 1TB Gen4 for £44. That's on sale but still cheaper and faster than SATA drives.
@@liaminwalesthat being said, the P3 is an awful SSD. It really doesn't matter which interface it uses if it just falls on its face during regular use (which the P3 does).
Consumers need more random read speed improvements, which Optane was, they dont need sequential improvements that are only useful for people moving hundreds of Gigabytes of a single file around, like an 8k video.
Make sure the controller chip on the M.2 board is making firm contact with the cooler surface. Usually the NAND chips are thicker and might prevent the cooler's surface reaching the controller chip, when applied with thinner thermal pad.
great evolution. storage with moving parts -> no moving parts -> storage that gets so hot it needs moving parts
I'm wondering if the 2 coolers that looked quite decent (the red one and the dual heat pipe one) had old or too thick thermal pads. I'd retest them with something decent with a thickness of 0.5mm. Also, just an idea, but the red one might not have any fins under the plate since there is an lcd/oled there.
With all SSD they should really say what is the continues speed after like 30 minutes. After the cache and over heating is done. Cause showing only peak speed doesn't give you the real performance.
What are you doing with 12GB/sec speed for 1800 seconds on a 2000GB drive? :) Are you boring?
@@maestrohun What are you doing 12GB/s anyway? You have two PCIe Gen 5 SSDs and are just copying 20GB files from one to the other all day long?
This is why peak sequential read/write is useless as a measurement of performance. This SSD will overheat while idling, causing the system to crash, so over an hour its performance is: 0MB/s...But for a few seconds it can show 12GB/s in the benchmark tool, so that looks cool...
It's like saying this new Volkswagen can do 400KM/h. Yeah but is it safe, nice to drive, do the groceries and whatever fit in the boot? Can I actually use it?
@@maestrohun I download lots of p0rn 😁
I wonder if there will be a difference when using something like Fan Control to ensure the fans are indeed running at 100%. Otherwise they might defailted to default 3-pin case fan setting
Yeah, I think this was the case...
He could have just use a converter to power the fans off molex or sata so that they are always running at %100. I did that with some in my PC because motherboard use to run them too slow and fan control wasn't properly working with 3 pin fans.
3 pin fans don't have fan control beyond lowering their voltage. They have 12VDC, GND and tachometer pins, the last one being a signal that is generated by the fan to indicate the speed it is spinning at so the system knows whether a fan has failed or not.
The difference between running the fan and not running it was minimal, I don't think a faster fan would change that.
@@user-eg6rs1zp7lwell, if the fan is barely running, it won't make much of an impact...
There is one actually good m.2 cooler that worked well for me its the Thermalright HR-09 it has two heat pipes that go through an proper heatsink with fins. also If you want to have better temps remove any stickers from the m.2, coper or otherwise and replace with a nice thin thermal pad.
Plugging a fan onto the motherboard defaults to the BIOS fan curve for that header, which most likely goes by CPU temp o chipset temp which were idling during testing. You probably never run those fans at full speed (and probably from this comes your wonder why were they quiet).
Keep in mind those small fans usually have very short life anyway. So running them at 100% will kill them very fast
For me, as content creator, these SSDs are necessary, especially when I render big music files, let's say like projects with 200 or more multitracks, sometimes even 400 tracks, if you have to mix a big orchestra, with different microphones in an opera, each actor has its own microphone. And all the instruments. A consumer SSD would throttle the whole time, that's why we use M2.nvme in Raid 0, with a realtime backup running in the background.
Why not just run it in RAM at that point?
@@someguy4915 I'm already maxed out 128Gb, and already using SSD cache, but that is also limited. in my studio there is an Apple Pro, and this one also has problems with that many tracks. On many track are plugins loaded, these are recourses eaters. That means, when the system is saturated it uses HD' or in this case SSD's, that's why we love these gens 5 SSDs.
@@Bluelagoonstudios I guess with Apple you're limited with what they offer in terms of RAM yeah so then PCIe Gen 5 SSDs are the way to go.
With Windows/Linux I'd say a 512GB RAM system isn't that much more expensive, assuming the content creation is a job so the RAM (or SSD) is just an investment to do your job faster and easier.
But for any other situation, seeing these Gen 5 SSDs overheat while idling is ridiculous xD
Another thing to add into the custom waterlopp :) Seriously you showed that no cooling makes too hot to run and even shutting the system off. Many people without proper airflow over the motherborad wont be able to run this ( a " fancy" motherboard with SSD coolingpads is necessary). If the SSD doesnt fit onto the motherboard without proper cooling maybe we go back to mount them externally with riser cables or something. I dont need more speed if that much heat comes with it, Gen4 is fine by me...
The one with heat pipes needs a shroud so the air can effect entire heatsink, it blows sideways almost immediately.
I doubt it will make too much of a difference but still basic underrstanding of airflow should be applied to design.
EKWB had great NVMe heatsinks, simple yet effective design.
I like these type of videos. Why manufacturers sell these SSDs without a heatsink or a cooler is beyond me.
Intel sells high-performance CPUs without any cooling solutions included as well, but this SSD cooling stuff might be a new concept for some people so I hope they label it very clearly if it requires additional cooling.
Very nice (and fun) testing! One thing I want to point out is that with the motherboard built-in heatsink, you are testing it with the GPU idle and on an open bench, so it will likely hit near 75-80°C even idling when playing a game on the GPU (I'm seeing similar behavior on my X570S board with gen4), so if you load it up alongside the GPU, it will definitely throttle. IMO these things are still a generation or two worth of refinements away from general viability... even gen4 isnt quite low power enough to just "slap anywhere and it will be fine".
In my opinion, the problem is not in the heat sink, but in the design of the ssd. A plastic cap on the chips, plus paper/plastic labels, plus a thermal pad - these are very large losses in thermal conductivity. Even water cooling will not help, at least the difference will not be so significant. I'm pretty sure the ssd doesn't produce enough heat to be dissipated by the aluminum heat sink with fan.
I remove those labels on my SSD stick before adding on a heat sink. It works well as a passive one. No throttling that I could tell.
It is fucking bananas how hot these controllers get (on pretty much all SSDs). What the hell. Problem also is the controllers are packaged in plastic and have no metal interface to wick the heat out properly. So for most users the nominal use case is going to be it running at half speed pretty much 99% of the time.
AliExpress heat sinks - HOT 🔥
DerBauer saying “Aluminum” - COOL 😎
I can’t think of any situation where I needed anything more than a 980 Pro SSD.
I don’t have an Ssd, but this method works for pendrives to fully reduce the heat: connecting to a Samsung OTG cable, and to set the phone to standby. Also works with cards inserted into card reader. After this, if you connect it to a Windows Pc, and if you copy some gb files such as videos within the card, into other folders. That speeds up the drive, no failure will happen. The copying sets up the real mb/s of the card. So if an sd card is advertised at 50 mb/s, the copying sets up its 4 mb/s copying speed for example 4 gb sd card. The setup dramatically reduces the heat.
You know what I'd like to see? Liquid crystal thermometers being integrated into components. No power needed.
The fact that these things exist to begin with means to me that there's a serious design flaw in Gen 5 M.2 SSDs they shouldn't get this hot to begin with.
Just basic physics, the same as how your CPU heats up if used without a heatsink. Don't forget we once had PCs without a heatsink. Then we had to start adding little heatsinks which over time had to be bigger and bigger as CPUs started consuming far more power.
Pretty crazy to think that we've barely had to think about temps in the past, and now you want the best and even actively-cooled devices cannot keep up
Good video again. I think, that sticker over the chips, might cause poor thermal transfer. And here is my idea, the manufacturer, can make those stickers like thermal pads style, with some thermal conductivity, or just remove them and change the type of labeling.
I was Recently looking at Purchasing the Graugear Heatpipe Cooler...after this Review I'm glad I saved my Money and Passed on it.
der8auer's AliExpress test series are the best, even LTT can't come near the quality and knowledge.
If you need a heatsink that badly just get a board with one already on it. Your case will proved the airflow on its own. If you think it looks cool go for it otherwise unless you are writing to the drive 24/7 you dont need one. Your OS on a regular SSD is perfectly fine.
My thermaltake core p90 case is open. No airflow.
I am curious as to what happens with temp when you remove the sticker. I have always removed the M.2 sticker for more direct contact to heatsinks, but became aware that some companies make the stickers infused with copper.
I actually bought a few of those tiny 20mm centrifugal fans. For cooling random electronics. They are better than nothing, but without any engineering in cooling fins it won't do much.
Ah the temp reading on that heat sink seemed to have found the kid in you ha ha. It's a nice touch for sure. By the way I used copper shims with thermal grizzly & a cut up back plate from an old GPU to cool mines & it does a great job!
These heatsinks seem made for the look. I'm sure the heat pipes aren't functional. They are meant to be made of pure copper, have a granular surface finish on the inside, and be under a precise-near vacuum where some water vapour is added. I doubt any of the functional parameters have been implemented there.
Thanks for the video, and buy computer parts from reputable sources which know what they are selling and respect their customers...
Also for copper heat pipes to work properly there needs to be a temperature difference between evaporative and condensation zones. They seemed to be connected to top and bottom of the same aluminium extrusion which after the while will have equalized temperature distribution.
This is a big challenge. Mounting pressure, clearance, performance, and compatibility will all be difficult. I'm looking forward to seeing what you develop for Gen 5 ssd cooling.
The problem is that regardless of how much mass and air throughput you have, the weak link is the thermal tape, because it is highly inefficient at conducting heat away. Replace the thermal tape with paste on the top of the chips, get a nice bare metal surface on the graugear, and retest with something like kryonaut
All current PCIe 5 SSDs shipping today were designed to be used with a cooler. If you look at the documentation it clearly states use the drive with the motherboards heatsink. The drives shipping without a heatsink are for users that have motherboards that include a cooling solution.
I’m calling it remote synthetic brain wave assistance , it emotes frequencies that can stimulate parts of your brain for focus memory and function
I think first one is for Samsung SSDs, which is known for hot controllers. Its fan is right in usual place of controller. Samsung also have dedicated temperature sensor for controller. This Crucial likely just have one for memory.
The Cat is the best bit of the video.
You should use the better working coolers from you previous vids on the Gen 5- also, not all motherboards have this huge heatshield like gigabyte does - if you have other motherboards - you should check out how the built in solution works there.
Ali Express is just going to start shipping Gen 5 ice cubes at this point.
You hardly need any airflow to cool a 10W load but you do need good thermal transfer from the load to the heatsink for it to have any chance of doing anything, which is likely where most cheap aftermarket solutions are failing.
Makes me appreciate my SATA SSDs, glued together against a PSU in an ITX case they did not overheat
just way slower though...so fail....gluing ssd's.....hahahaa
I think the Graugear cooler is a good idea with bad execution. It definitely needs walls on the sides to guide the air across the heatsink! As it is, the air hits the first 20 maybe 30% of the heatsink and then disperses out and loses velocity. Would be interesting to see how it performs with paper walls on the sides for science.
Also at these temps, maybe the paper sticker is hurting the temp performance a bit. It is probably not by much something like 1-2 C but still that may be the difference between throttle or not. The quality of the thermal pads provided is quite questionable as well, I know that if i get a ssd cooler i will use my own pads.
Would be "cool" to go deeper on how built-in motherboard m.2 coolers perform (including fully populated) and if/when it even makes sense to buy separate coolers.
To get better results, before using the heatsink, it is better to remove the paper sticker on the SSD so that the thermal pad can be placed directly on the memories and have a better heat exchange.
That is how you void the warranty. Like Samsung you leave it on and place the heat sink on top and I get 34 degrees Celsius.
6:06 looks like a badass stapler.
At this point I think Gen 6 SSD would have 8 pin for extra power. You know, because we don't have enough heat.
Yeah, I'll stick to my gen 4 nvme, thank you very much.
I do not need another oven radiating from my PC case, especially during the hot days in the summer.
Funny thing is even the motherboard cooling solution is not good enough, which is big hunk of aluminium with heat pipes! Yikes!
Would be interesting to test running a thick thermal pad on top of the drive to a rtx founders metal shroud. Should make a incredible heatsink.
I suspect the heat pipes are purely decorative, no working liquid inside.
Remove the thermal pads and apply some thermal paste.. you will see much much better results!
I was going to post this exact comment. Hope he reads it!
You have to refresh Crystal disk to get the right reading
I have the "natural cooling disk" and mine never goes higher than 41 degrees (normally 2 degrees off of what HW info reads)
Why is it so different? Do gen 5 m.2s just use loads more power?
Same here. I've used these coolers for years. My drives (gen 3 + 4) stay way cooler. I'm surprised at how much hotter gen 5 seems to be. Looking at my gen 4 drive now it's 14 degrees colder in idle with this cooler than how it's running in the video.
they consume nearly twice the power of a typical gen 4 drive. the Phison controller is to be blamed here (mostly).
Would be interesting if you retested a good M.2 cooler from the previous video's, like the Thermalright HR-09 PRO with the GEN 5 SSD.
yes!!
i bought this one after the review :)
really good cooler
I would be very interested in seeing how these would perform with water-cooling in my ROG MAXIMUS Z690 EXTREME GLACIAL, which has a full coverage water block on the motherboard and a dedicated Alphacool NexXxoS Monsta 560mm Radiator with push pull and x8 Noctua NF-F12 iPPC 3000 PWM, Heavy Duty Cooling Fan, 4-Pin, 3000 RPM (120mm, Black) with a dedicated Aquacomputer ULTITUBE D5 200 PRO reservoir with D5 NEXT pump. Total overkill but it would be interesting to test.
There must be some kind of a problem with the interface, or the component which is generating heat. Memory chip gets hot, but the controller is the component which heats up more. Even mosfets run cooler as soon as you apply any sort of heatsink to them, and they run even cooler when you add a fan to the equation. So I think that the problem there must be where the heat is produced and/or sampled...
It's all just chunks of copper or aluminium either way, just not overpriced, unlike the branded ones.
@2:29 Even the cat is screaming at how bad an idea it is to slot a gen 5 nvme SSD without a heatsink.....
Hey Roman, be curious to see the results with the supplied motherboard gen 5 heatsinks , I have a x670e taichi amd a arous master z790 both using gen 4 at the moment so supplied heatsinks not utilised yet
There's not one, not two, but three seven segment dispays on this ssd cooler. If motherboard manufacturers are to be believed, this ssd cooler costs at least $600 to produce.
a. motherboard manufactures are not to believed for almost everything.
b. I don't think they ever claim that adding 2 seven segment displays cost too much, it's just their shameless way to up-sale you expensive boards.
c. If we the consumers would demand 7 segment displays on cheaper boards, they can integrate it to 150$ boards.
The test is flawed, you didn't take into account the room temperature + you didn't configure your motherboard to which fan speed to set the fans in the bios, its literally depend on the voltage your MB send ... instead you kept plug/unplug the fans which was super cringe.
As well it looked like your setup was horrible with the graphic card literally throwing heat on the M2 area, you had to use some small and cheap graphic card or use the on board graphic card (if available), clearly not that monster you put there .. but that's another story.
My takeaway from this is water blocks.
The design on the top M.2 cooler for the Gigabyte AORUS Master X670E board is how they need to do things for Gen 5 SSDs - on my Gen 4 drives, it idles 3-5C cooler than the M.2s I have under the main plate area down below (which has 3 drives under it, to be fair) and under load, it's a huge difference, 10-15C or more cooler than the ones under the plate below the GPU slot.
Would be interesting to see results with a Gen 5 drive, but I have neither the need nor the money to get one at the moment lul
how well would the Gen 5 drives do in something like the ASUS ROG Hyper Expansion card like they include with their Maxims Hero mother boards?
It helps with my Apex motherboard with 3X intake 140 mm fans and one 140 mm fan blow from above.💨💨💨💨
Seems like a million dollar engineering issue.... for the one guy that can make this work with under 70c average under load. GGZ
The two tiny fans cooler cracked me up 😂
Gen.4 was really enough for anything.
Your cat is cool in every way! What a great co-host for your videos.
This is the part when the testers/reviewers must realize that their open test bench not suitable to test it to the rest of the word that has a pc case and a significant airflow trough it.
But poor airflow over the motherboard is quite standard. Especially in most cases where the GPU basically blocks any significant airflow over the motherboard itself.
And since most coolers he tested here have fans, airflow (or lack thereof) is not the issue here.
Stupid question perhaps, but how much of a difference would it make if you removed the stickers that are on the drive?
Just watercool it mate don't fuck with the warranty.
It did not show clearly, but was the label taken off before the heat sink was put on? Those stick on labels ack as thermo insulators, preventing heat transfer.
Silicon Teflon brush it on and leave it in the sun until you see the plastics appearing wet then clean kinda keep it wet looking put it together and use it , should have no throttle issues chip wise
I tried removing this heatsink (Natural cooling disk), when you turn on the fan, in the lower position of the fan's shaft is the fan control board, it causes heat and radiates heat out, actually this little fan is not needed. , circuit board display. The temperature display and the sensor are located quite far away, so the results may not be accurate compared to the crystal software
That little fan is adorable, you need to name it. Worlds hardest trying fan.
Edit: Temp display is neato as well. Give me more stats, screens upon screens of information that I don't even need.
Are these actually throttling due to HEAT, or are they throttling due to filling up its cache or memory?
Couple of years ago I saw a video from LTT, that said that most SSDs actually like heat, because the hotter they are, the less damaging the SSD writes become.
the NAND FLASH likes heat, The controller does not.
Bros for sure making a ssd cooler behind the scenes and these videos are just him testing the most efficient designs🤔🤔
6:05 that's really close to Gaimin Gladiators logo
I would try flipping positive and negative so the fan goes in reverse and see if that makes it cooler
Very interesting, will SSD cooling keep in mind for my next build.
Nice. Very informative. Thank you!!!
The main issue is that he plugs into fan connector without setting it to run max RPM,
we have no idea whats the rpm and it probably runs some low numbers not full speed how it supposed to be with these fans
i think we should be looking to the manufacturers with the questions about temps and why this is on the market, because i bet a few more min even on the mobo cooler it would've over heated
I’d like to see how the Alphacool pcie nvme waterblock performs
WOW! Gen 5 is hot hot hot! I guess Corsair had it right with the m.2 water cooling block. We made fun of it back then but now we need it.
Its the thermal pad what heat up and isolated it.
That is the reason it hit 80 celcius on all coolers.
I added a Noctua NF-A4x blowing across the top to achieve 60/64 degrees sustained read and write on Samsung 980 Pro's. An additional -3 degrees with copper fins on top of MSI heat plate in slot M.2_1.
Temperatures are NAND/Controller.
Nothing could be done on slots M.2_2 & 3, (under GPU). On slots M.2_4 & 5 my results with Samsung 970 EVO were:
stock solution - 57/61 C
+Noctua NF-A4x blowing across the top - 54/61 C
+Copper finned heat sink and new and slightly thicker thermal pads adjusted for different height between NAND and controller - 51/61 C
In order of importance I feel are: properly contacting thermal pads, a copper finned heat sink and a fan moving stagnant air areas.
9:00 Ah; reminds me of that time when a freshly installed no-brand LED strip went up in smoke and tripped my PSU.
I really do enjoy these daft little heatsinks
7:00, I think 3 minutes of max performance is adequate cooling capacity. At 12 GB/s, you can read the entire capacity of the drive in less than 3 minutes. So yes, while academically, it can't sustain max performance over extended period, I don't know if the test is relevant for any demanding, but real-life scenario. Of course, this is assuming your motherboard didn't come with a reasonable solution built-in.
I think it would have been interesting to test the time it takes for each cooler to reach their idle time from the max C. in real use scenarios, the sad isn't going 100% the whole time
For a better test, you also need to have it installed in a "real" system, meaning a closed case with case fans circulating air through it, which would help with cooling and show better values for a real scenario. These open test-benches don't really work for real use scenarios, but even so they give a good picture of what you can expect, and I believe in this example, that the onboard heatsink would perform even better with case fan through flow.
speak for yourself my sad is going 100% the whole time
Suddenly the corsair watercooled "block" for ssd's ain't just a joke.
what do u expect when u run the fans close to there minimum speed thats why even those small fans are so quiet
M.2 drives that need active cooling, should have been U.2 drives in the first place. We just need U.2 on consumer boards. There's no reason NOT to have U.2 over some SAS SFF on-board port, and bundle a couple adapter cables for a couple U.2 drives.
There never should have been so many M.2 slots directly on the board anyway. That should have been specced out for M.2 to U.2 2.5" or other M.2 enclosures.
Thank you. I have been wondering about SSD coolers.