I had a non-reference Voodoo2 years ago and it came with green anodized aluminum heatsinks factory fitted. they were pretty small but im sure they made a difference. it was a Gainward Dragon 3000
I got exactly this card yesterday. It generally works fine, but it does lock up in 3DMark 2000 on an Athlon XP@1500MHz. It gets crazy hot during the bench.
I found some 28x28x10mm heatsinks with 3M 8810 thermal tape on Amazon that are the perfect size for the Voodoo ICs. Just sharing with the community in case anyone else would like to add some passive cooling to their retro voodoo cards. Hope this helps.
I can understand it running cooler with the slower processor, there is less data to crunch with the slower processor than with the faster ones, so you do not get to utilize the card to its full potential. This was actually a great way of showing how a processor bottle neck can affect a video card, without having to see benchmark results :)
Like some other people mentioned, it doesn't measure exactly at the laser dot, but the laser is a sort of an indication where you are aiming. The measurement is just below the laser dot. I got a IR thermometer as well, and it usualy helps to move it around a bit around the area you want to measure until you hit the highest temperature. The closer to the object, the more accurate as it measures in a cone shape (closerby is a smaller cone, is a more accurate reading). My old 3S805C 1MB onboard chip of my 486 would hit 70ºC even when just 'idling' on the desktop (chips didn't have power saving modes back then, they always ran full speed all the time, no matter the load).
Yea I stopped using the laser. But the other issue that is doesn't measure all materials the same. For example a silver aluminium heatsink, it totally throws off the device.
I got a bit more expensive IR thermometer, and it can even do shiney surfaces (silver/alluminium heatsinks for example). Only glass throws it off. I never use the laser on the deveice, it drains battery too fast, potentionaly dangerous (for your or someone else's eyes), and the stuff I measure is always closeby so I can aim without a laser :)
Get a thermal probe from banggood, you can tape the probes onto the components to measure the temperature more easily. Kapton tape will hold under temperature, and they have a dual chsnnel model, too.
Phil, this kind of contactless thermometers are practical for some applications where area to measure is large and material emissivity is well known and adjusted for. While the latter *might* not be a problem, the distance that you measured from and the fact that laser is not really collimated with the actual sensor will seriously impair the measurement.
Yes I've mentioned this in the video. I won't use the laser the next time, people get too caught up over it. It's a simple case of searching for the "hottest spot".
Your thermometers infrared aperature and where the laser points where certainly not aligned correctly. The Chips should be at their hottest in the center. You can test your thermometer with a cup of boiling hot water. Move it across the surface of the water and see where the laser is when you read 98-100 Celcius and if you even read 100C. Then use some Ice Cubes, the melt water coming off of it should be very close to 0C. Nvm you addressed it at 8:00 :D
This seems to be the case there. When he is measuring the bottom of the chip he is actually measuring the PCB under the chip and when he is measuring the top he is measuring the center of the chip. I wonder if the thermometer came with a manual which would specify the correct measuring distance where the laser and the aperture would converge?
It's not just that these have an offset, but the measurement is a cone area that gets bigger the further away tou hold it, idealy you put the aperture area right onto the thing to measure if possible.
What did he say that thing was? $11 or something? Those "Shenzhen Special" tools are usually piles of hot garbage. It would be a small miracle if the thing is accurate to within +/- 10% on the temp readout.
This is a normal construction detail of these IR thermometers. They have no way to track how far the object you're measuring is! Also the optics of the IR unit is pretty good, so you're not limited to measuring at a single given distance. For this reason they simply choose to aim the laser parallel to the pickup, and you have to account for the offset between the laser pointer and the pickup in the device manually. As to how accurate they are? Well in spite of being from Shenzhen, they're usually about the same as you'd expect, and you don't expect an IR thermometer to ever give faithful readings, because even in the best case, these are black body radiation measurement devices, and you don't have black bodies in real life. You use it to zero in on issues and get a ballpark figure, and then you should tape a K-type thermocouple where you need it. Those are precise by nature.
Was never a problem in period. I ran a P2-350 and later a P3-700, Rage 128, Voodoo 2 in SLI, NIC, SCSI, Sound and modem cards filling all the slots. No overheating issues with only an 80mm intake fan and that massive SCSI cable blocking the airflow!
If you look at the front of the thermometer you'll see that the laser sits above the IR sensor. The laser is just a rough guide of where you're aiming. If you look at the side of the thermometer there's a sticker showing how large area the sensor will "see" at certain distances. The temperature reading will be an average of the area the meter sees.
I have a generic version of the same IR thermometer and due to it having a laser pointer and the sound it makes when you release the trigger and it holds the last recorded temperature for display, my cat comes running every time I use it because she wants to chase the laser that she knows is always around when she hears that sound. Sometimes she comes running just because she hears the click of the trigger pull. Its a legit 100% working cat call, without fail she comes running when she hears that "Beep Beep". lol
Just a quick note on the temperature readings.. The thermometer you purchased has a fixed emissivity setting of 0.95 which will give you an inaccurate reading on some materials, particularly when pointing it at metal parts with a much lower emissivity rating (EG metallic pins). The extra few degrees you get by pointing the thermometer off the chip is probably that.
No, it's not that, although it does fail on metal objects, it's the offset between the pointer and the pickup - they are aimed in parallel, but on the device they are not coincident, the laser pointer is at the top and the pickup is a couple cm below it. This is why when he was aiming the laser on the pins, it picked up exactly the middle of the chip! You're supposed to simply take that into account.
The reason why your laser temperature meter reads off from the laser dot it because of a Parallax error caused because you were measuring the chip from a very short distance. the error will lesson as the distance grows. even expensive units can suffer the same. Old point-and-shoot camera viewfinders suffered the same phenomenon
The chip measures hotter at the top, because the sensor of the thermometer is underneath the laser. So basically that's when the sensor is aligned with the center of the chip.
Yep, had this exact same card back in the day and I remember it ran pretty hot. Really pissed though, I lent it to a friend after I upgraded, and they said it didn't work bet never ended up giving it back so I could test it myself. So now I no longer have a piece of gaming history! :(
Amazing ancient technology. There are no BGA connections, no FLIPCHIP silicon assembly, ordinary poison tin PB and 22 year old equipment still working: D While buying 22 years ago as a top card, such a VOODOO 2 cost a fraction of what today's top graphics cards will live 1/10 of the time as such a classic 3Dfx. My VOODOO 1 has factory heat sinks and is quite cool. As for the pyrometer. It is this laser pointer that is simply not aligned with the sensor axis and rather shows the measurement location to the eye. You have to shoot this meter like cowboys from your hips: D: D
I was brave. I have put a voodoo2 to my I7-4770 pc with windows 10 :D A dont know the temperatures but it worked!!! Driver was a bit tricky. BTW I could archive 50 FPS in quake 3 (640x480)
lol you are a brave one xD some LGA 1150 motherboards still have like 3 PCI slots (5V keyed i assume), so plugging a V2 in there is tempting. but yeah would be interesting to know the temps though, not sure i'd want to risk one since they are hard to find here.
I'm thinking about doing something very simular on my rigs. But rather then putting the fans like you did sens I have pretty much all the slots filled with cards network audio raid controllers and so on. My idea is to mount two 90mm fans vertical blowing from the side in between the cards. Forcing the hot air lit forward towards the hard-drive bays. That way all the cards do get cooling and making a bracket is probably not going to be to hard. I have two rigs one is going to be overkill fast. Windows xp and Core2quad and the other is going to be overkill sempron 3000+ speed. Så both need cooling. For me voodoo are a soft spot I wish they would have kept voodoo around and Nvidia would have continued the series :) 3dfx was awesome and still is :)
+PhilsComputerLab Another fun video! Interesting too. Back in the day I was having heat problems with that very card. Games kept stuttering and crashing during extended gaming sessions. At first I just took the sides of the case off. Ended up attaching 3 low profile heatsinks to each of the 3 chips and got a "Hamster-Wheel" type blower next to it to pull air out of the case. After that, no problems. I like the fan mounting kit you showed in the video. Very clever bit of kit!
Just like a laser sight for a gun, the best the laser can do is run parallel to the heat sensor. It will never be accurate, but it will always be equally inaccurate. The alternative would be a tilted mount that's only accurate at a given distance and wildly inaccurate at any other distance, or an auto ranging mechanism that actually adjusts the angle of the laser to hit the target at a given distance.
I wouldn't call your sensor inaccurate as it was probably made for testing things at much greater distances, The 2 Cm distance is just the location of the sensor and the laser pointer.
One idea would also be to use some thermal adhesive (Basically thermally conductive two-part epoxy - Arctic Silver and Arctic Alumina come in adhesive versions for example) and use that to affix some small heatsinks to the chips. This might not necessarily be ideal, though - it's a permanent modification to the card. But it would certainly provide a marked increase in thermal dissipation. You could also use some chipset heatsinks with integrated fans as well if you found ones with decent or replaceable fans. The not so good ones make a clearly audible high-pitched whine. Those little 40mm fans sure can manage to be pretty loud.
Does a pretty good job going from no air moving near the gpu too a directed airflow across the gpu. on top of dedicating a fan or two too move air across a sli voodoo 2 setup id also mount on a bunch of slim heatsinks.
You need to place some Aluminum radiators glued on the chips with Arctic Alumina thermal compound which is meant to stick forever the radiators on such chips. They are too hot.
I coupled two of these cards in SLI mode in a PIII 450 Mhz setup. To keep them a little cooler, I cut a hole in the case side panel, and installed a low profile 120mm fan to the inside of the panel, pushing fresh air directly into the gap between the two cards. Did the same in the middle of the top of the case, to suck the warm air out. So far no, problems. But I am not sure about the temps. Cards are still OG, with no heat sinks installed. Games run fine in 1024x768 rez.
Back then the Harddrive and PSU was the noisemaker. :) Heat rises upwards. Law of physics. Maybe if you placed the card vertical, the heat is spread more evenly on the chip. But I believe the heat will dissipate into the air better on this way you placed the card. The other (normal) way, it will just heat up the PCB more. Actually I never thought of that. It makes perfectly sense. Same with a CPU cooler. Will always be most hot on the top. if placed in a Tower case. With a fan.. not so noticeable, since the heat is being pushed / pulled out of the heatsink. So Passive cooled chips especially with heartsinks, will be better cooled when you place it with the chip/heatsink upwards. But dust buildup is another story. Seems like there is a plus and a minus to that solution. A fan is always a good idea - so we can move the heat away.The price we pay is more noise/ dust. Thank for a good informative video - again. :)
I have that exact model of thermometer and I can tell you, the infrared appeture measures heat about 2-3cm below the laser point. Try it with a hot knife oriented horizontally and test to see where it starts picking it up
Had two of these puppies back in the day, it was my first build. Paired them with a Celeron, cos i didn't know any better. Used to play Carmageddon with software rendering, then switched to Glide, can't remember if it actually used SLI though. The difference was amazing.
Not used my Voodoo 2 card in a while "creative". But do have a voodoo 3 "creative" on a PIII 733mhz. And never actual did consider temps on the card. It's at around 50 to 55 when playing using a non laser tool to measure temp. Tho it's in a modern case with good air flow and a 120mm fan on the door pointing at it. My Voodoo 1"diamond" with a 166mmx is also in a modern case with good airflow to. Though did not test that atm. But I'm fairly sure it would run above 60 with Unreal. But I mostly only play i76 and NFS2SE on it.
Heatsinks are higly recommended. Ensures the life of your card, don't emit noise and is cheap, so it's a no brainer. There are tons of them in different sizes on Ebay.
I have heat sinks. Just not a fan setup under it like on the video. My Voodoo 3 has it by default. And the voodoo 1 card got an after marked variant put on in 1999.
I added heatsinks to my Voodoo 2 SLI setup. I bought the heatsinks from Ebay and it was basically a perfect fit on the 3dfx chips. The heat sinks are barely warm with the finger test when playing Unreal (92 MHz GPU).
I bought some black aluminum heat sinks off of ebay from China, and some thermal epoxy off of newegg. I glued them on my Voodoo2s for my SLI setup, and a fan at the front of the case, blows toward the back of the case to help cool them. There are also some VGA slot coolers which can be mounted perpendicularly to the cards, so they blow downward over cards.
It wouldn't be that difficult to design and print a shroud to clip on to the Voodoo2 and mount a small 60mm fan. I might consider it since I have four Voodoo2 boards.
GGigabiteM that would be something I would buy!.. I do not know anyone or own a 3d printer unfortunately.. I don't like the idea of sticking on heatsinks, just for the fact the cards i have are in new condition.
Nice video. I've always wondered how hot these chips with no cooling get. Have ever used those little stick on heat sinks you can get? I've used them a few times on sound blaster cards like the Audigy 4 which get quite hot.
Nice vid. I've also done that in the past and I was also surprised with the results. Actually my Geforce 4 Ti is running like this. I would say in some cases you don't need a extra heatsink, a direct airflow like this will do a nice job.
A note about the accuracy of the thermometer, it is probably pretty accurate at about a meter, digital thermometers like that typically aren't meant to be used that close to the item being measured. Still a nice tool to have to get a good idea how hot things are. I remember back when I had a Voodoo 2, Voodoo 3, and Voodoo 5 I used to make my own GPU coolers with a mounting bracket and fans to basically do what that device does... I actually miss those days :)
So I've used this thing in another project, and someone else mentioned it here, the results can vary depending on what material you are measuring. For example a aluminium heatsink throws it off totally. Has something to do with emission. So yea, these are the things you only learn after the fact :D The distance thing is purely so that the laser pointer is accurate. I now just turn it off and search for the hot spot by hand. You can go straight up to what you measure, it's actually better.
i had voodoo rush, voodo3d, voodoo2, and later was given 3 and 4th gen cards, honestly, the matrox m3d/apocalypses 3d using powervr chipsets where better for titles that had support for them(quite a few actually had powersgl exectuables), and it had better ogl/d3d support with last gen drivers then early sdfx offered, you could/can also stack as many of the powervr cards like the matrox m3d in a system and they will work in "sli" no dongles or any of that, i tested up to 4 and saw VERY good scaling in every sgl and many ogl/d3d titles back in the day, i would LOVE to see one of the all in one cards that had the powervr chip used in the matrox m3d on it(like a voodoo rush but power vr) tested like this, but also tested with more cards with the same chipset added to the system... i know with 2 of those cards i could do 1024x768 maxxed out in ultim@race and a slew of other titles, unreal was AMAZING with those cards(blew the quality 3dfx could offer away...no blatant 16bit dithering for example... and omg.. the kyro2....i would have loved to see powervr tech get the next upgrade to have hardward TnL and how much it would have made ati and nvidia look over priced...in a proper system the kyro2 sure as hell did...
Should get some of those cheap aluminum RAM and chipset coolers from AliExpress and stick them on this card without the fan and see how much cooler they run.
A K-Type thermocouple would probably be better (and just as cheap) for measuring the temperature of a setup like this, just secure it in place on the gpu with a blob of thermal putty.
I'm curious if your thermometer is just bad. I did the exact same test with a higher ended model that uses 2 lasers to help give you the correct distance for a proper measurement which should be around 14 inches or so. On a Tualatin P3 1.4ghz i often measured only around 55-60c on my voodoo 2s. At over 80 you should not have been able to hold your finger there as long as you did. Regardless, the cards need some form of active cooling period, even voodoo 3 with its minor fixes to this bug and heatsink need it as they still get far too hot imo. Not having airflow of some sort over your voodoo card is asking for an early grave :(
I don't see how an open case with no airflow will cool better than a closed case with _proper_ airflow. At any rate, this also highlights what is, in my opinion a design flaw in *all* graphics cards to date, which, in most PC cases, will be mounted with the chips and heatsink on the underside, which doesn't help with cooling.
Just take a look into the BTX design. It promoted a different design that basically had a single straight airflow through the case. But it was loud and Pentium 4's let the market, so everybody kept going with ATX. But there are some cases that allow different mounts for the motherboard. upside-down, rotated by 90°, laying flat, etc. Proper airflow depends on getting cold air to the parts and getting rid of the warm air. How a system handles that is not all too important. People say hit air rises up. That is generally true, but once the air is moved actively, it doesn't effect anything. People say a fan at the side destroys the airflow. Not wrong, but depending on the parts that are blown onto it can actually help with temperatures. An open setup has no actively "channeled" air movement, but alot of air around. Can help, can make things worse. Depends on the setup. And about the orientation of expension cards in systems, go back to 1991 and tell Intel that the orientation of PCI is bad.
In many cases (pun intended) the internal air temperature will be roughly 10 to 15C hotter than the rooms ambient air temps, even the best cases will still be a few degrees Celsius warmer, due to the difficulty in getting the warm air out. Which does make sense when you think that almost all cases will have just 1 exhaust, and even that is not guaranteed, and even if the case does have an intake fan, it will most likely be sandwiched up against the HDDs, blocking off most of the air flow from it. Gamers Nexus demonstrates this in their case reviews, where poor airflow can cause thermal throttling with even the best CPU and GPU coolers.
The ATX form factor is severely flawed. Intel had absolutely no forward planning when designing that. Which is really retarded because ATX was designed to overcome the limitations of AT and give more headroom for the future but in the end did not.
The older AT design did put the chips on the top surface of the expansion cards in a tower case, but in the ATX form factor with PCI slots this was reversed. Keep in mind that at the time ATX was introduced, the vast majority of PCs were desktop models where the expansion cards were vertically oriented, so it didn't really matter which side of the PCB the chips were on. The main consideration in regards to cooling with the ATX form factor was to get the CPU as close to directly under the power supply (in a desktop case) as possible to pull air over it and out of the case as efficiently as possible. Mid tower and full tower cases were intended for enthusiast/server setups and would have been expected to have more robust cooling installed.
I ended up putting copper heatsinks on both my voodoo 2s and 3d printed an 80mm fan mount that attaches to either card while in SLI and blows cool air over the heatsinks. It let me overclock my Voodoo 2s by quite a bit to better accompany the PIII 800 Mhz.
Seems interesting to me that it got so much hotter with high resolution and a faster CPU. I guess if they knew they were going to be used in such fast machines they likely would have included some small heat sinks of the chips. It is still, not that hot in the grand scheme of things, especially considering the size of the lithography. Not sure how useful the video is actually but it still was interesting to me. :)
Got my voodoo 2 today for my timemachine project to fill 2 roles. First role is that it fills the gap with my Nvidia 6800GT to be able to play dos titles and older titles the NVidia 6800GT is to new for. The second role is for the summer months, i decided that since my NVidia 6800GT died in the warm weather its probably not a good idea to have that one in the system in the summer and i have a S3 Trio64 graphics card. Which would mean the summer version of this build would be perfect for the 3.11 / 98 era. And the winter build would be perfect for the 98 / XP era. With both versions handling dos very well. The CPU though is a 2,6GHz CPU so its going to run quite hot. I assume that puts me into the territory where i have to cool it. But whats the best approach here? Just passive heatsinks and tape, or active cooling? My PCI slots are very busy so i don't have much space.
All IR thermometers work like that. They can't put the laser inside the IR sensor, it will always shoot high and parallel to the center of the sensor. This will make it seem more "off" the closer you are to the target. It should have also explained in the manual how far away to hold the device from the thing being measured based on the size of the thing being measured. Smaller things need to be closer, etc. It's basically a camera with no zoom. It was hard to tell exactly how far away the chip was from the sensor, but it was probably good... maybe a scratch closer would have been better.
hah that's the same termometer i have in my hands right now. about the temperature being lower on the bottom, it's because the thermometer measures the temperature just BELOW the laser dot. so when you go to the bottom of the chip, you're actually measuring the pcb temperature. look at the front of the thermometer, it measures with that same offset. i know, it's a bit annoying. it should have another laser below the measuring lens so you could frame what you want between the lasers. well whatever it still works fine.
This video inspired me to have a go at it myself! 😄 But now I must ask.... Does an SLi setup run cooler by default? For if not, I am getting surprisingly different results from this very same test... Just performed this test on my V2 SLi (2x12MB) P3-733 Rig... Unreal running at 800x600 in HQ texture mode. Later I've ran it on 1024x768 as well, without any change worth mentioning apart from lower fps... I get an average of about 62 degrees celcius with my fans turned off... And when I turn my (totally differently installed) fans back on, I get an average of 40 degrees celcius... This, while my room without aircon is a very humid 30 degrees in a currently very tropical Netherlands... a little warmer than Phil's A/c'd 27... Hence my initial question... 😆
Just stuck old cpu heatsink and fans to my old voodoo 2's in sli with memory heatsink as well . Where my back up cards after I changed to geforce 256 which I had to keep getting replaced as they died for a hobby
All I did for mine was screw my 2200RPM 120mm radiator fan onto the side of the case, leaving the panel off. Both Voodoo 2 and the Diamond Viper V550 no longer have thermal issues after that.
One slot cooler I would definitely recommend that's a lot like that is the vantech Spectrum dual 70 mm slot cooler it has a whole lot of air flow with an adjustable rate of speed it ranges from about 15 to $20 depending on where you get it but I would definitely recommend it back when Pentium 4 with hyper-threading Dell gx270 I got an x 800xl to compliment the video card before I got the cooler sad about 50 to 60 Celsius I was able to drop it about 20% it hangs around 40 Celsius maybe a little bit lower
sweet! I've known the cards to run hot, but didn't think they were getting near 80c. I use a copper heatsink option on my voodoo 2. Might have spent more than $11.00, but it works well. I can't help but wonder which is the better option. the Iceberq 4 cooler, with a copper Cooler Master ram sink kit, or the fans. I didn't use the fan option based on the location of my card, and not having a free slot next to it. Hey Phil, maybe you could answer an age old question about the voodoo cards. Does it matter which PCI slot it goes in?
I'd try sticking a fan on the far side of the card or cards so it sort of blows air like a reference cooler on nvidias graphics cards. I----- like that above is your card on its own seen from the side & you place a fan on the right side of that making it kind of look like I----[] with [] being the fan ofc.
Those old mid to late 1990's 3D accelerator cards always lacked active cooling. What a mistake. Then again perhaps they did it on purpose so your components wouldn't last long so you had to make a hardware upgrade. I've put fans on my Voodoo 1 cards before as well. The Voodoo 3 cards were insanely hot ..even with their heat sinks (and mine came with a big one too). Yeah I would say that the Voodoo 3 cards were the worst of the 3Dfx cards when it came to heat
Up until the late 90s, there was still a stigma against using fans in desktop PCs. This was partly because most fans of the day used really noisy ball bearings and poor blade design that resulted in humming and whirring that echoed around in the case to amplify the sound. It was also partly to blame on design philosophy of the time, fans were just considered "not cool" to have. It's one of the reasons why towards the late 90s that CPU heatsinks grew to monstrous proportions, to try and stave off the need for a fan. I can remember some Slot 1 PII-PIII heatsinks that were so large that you couldn't hold them in one hand. I still have a PIII-500 with a monster Compaq heatsink riveted to it.
Many of the graphics chips of that time period were designed to run at up to 90°C without any problems or reduced lifespan. The bigger problem with the hardware of that era was with thermal stress from the chips heating and cooling repeatedly causing mechanical failure, as opposed to overheating (unless there was no airflow at all).
Mechanical stress failures related to heat was not as common of a problem back then because most large ICs were in DIP or PQFP packages. Those that were in BGA form factor had a similarly low failure rate because leaded solder was used, which was ductile enough to cope with the mechanical stress without breaking.
we don't know what is the reference operating temperature provided by chip spec. It may very well be that these temperatures are approriate and cooling it down doesn't bring any benefit. Also notice in a closed pc case there is a natural air flow, which means measuing temperature without the case is .... not really indicative.
It wouldn't bring a performance benefit, as these temperatures are indeed appropriate, but it will bring a longevity benefit, as the speed of transistor/FET degradation due to migration approximately doubles with every 10°C increase.
In, say, your i7-2600 rig's PCI slot... using unsigned drivers for W10... (think i've seen voodoo drivers for newer versions of NT) or running winxp... would there be a serious risk of the card being damaged? would the PCI bus be a significant bottleneck at any point?
Thanks for the information, i've got a Voodoo 2 coming in the mail in the next couple of days and i was worrying about the possibility of overheating, i was thinking of ordering a bunch of 14x14x6mm aluminium heatsinks for the chips, are those ideal for a card like this?
I'd go for 35x35x10 or 40x40x10mm heatsinks. More surface area in a single heatsink dissipates heat better. instead of you putting on like 3 or 4 per chip you put on a single large one.
Hmmm.... I think I need to get some permanent attached heatsinks on my Voodoo2 cards. Then get a fan solution like that. I have noticed this heat problem on a P3-1000. And that is why I only use my cards on K6-3 400.
Yea in a closed system I can see temps going to 90c or higher, and that is a worry. I don't think anything fancy is needed, just something that cools a little.
Could you do one of these for the s3 Savsge4? Also, I had one other question; I'm new to building "retro nouveau" PCs. If I have the choice of using either PCI or AGP for a graphics card, which is the better slot to use if I'm installing Windows 98 SE on the machine?
Powersupply ? you mean pc too test it in rather im guessing. Is it a agp or pci one btw ? if its a pci card it will work in just about any pc with a pci slot even something like my main pc with a I5 2500 non k 8 gb ram & a r9 280.
70C isn't terribly hot. My mobile i7 runs at 103C (!!) before it really starts to throttle down much. Even then its only losing 2-300mhz off of max turbo with 4 cores at load. And I have had it running at full load for long periods of time with no issues. My mobile quadro runs at 80-90C easily. I don't know what the design limits for the Voodoo 2 are, but I would imagine that 70C is likely in the safe limit. If it were that great of a concern, they would have put a heat sink on them. I think that you are definitely right that inside a case with poor airflow they would likely climb a bit higher as well. Most GPUs I have used can easily hit 90C without too many issues. Obviously if you want to overclock you would likely want to look at cooling for more headroom. Edit: I found a reference that Voodoo3 ran comfortably overclocked to 180mhz in the 80-90C range.
I come back 10 month later with a new product, that maybe it can be interesting: robot-electronics.co.uk/usbtpa64.html This pcb camera have a AMG8833 thermopile array, it have only 64 pixel, but it's cool!
Are there any PCI riser cards that measure current or that would allow you to measure current? That would correlate directly to power usage and heat dissipation with the card.
I'd imagine that an engineer who develops PCI cards has such a beast, but I've never seen one. It would actually be quite difficult to make such a board because unlike PCIe where all of the power rails are in the front of the slot, PCI has power rails sprinkled all over the slot. All of the common power rails from an ATX power supply are available on the PCI bus except -5v, so you'd have to monitor all of those lines individually and add the result up. Compounding that difficulty, extending the PCI bus with a riser card can lead to bus instability from connection issues and from circuitry added to the riser to measure power draw.
I think designing such a riser should be pretty easy - you would need some decoupling on the voltage rails, and it would get a bit bulky, but you can route all the data lines straight through while compensating for the distortion. There's only 3 power rails you're interested in, 3V3, 5V and 12V, -12V would only be of interest to maybe something like Ethernet and RS232 cards, so we can skip that. You can easily merge all the pins corresponding to a given voltage onto one point on the input and fan it out on the output - the multiple pins are only there to guarantee good connection and for current rating. Should i do that, would anybody buy one?
-12v is commonly used on sound cards, so it is somewhat important. I'm sure someone would buy it, but I wouldn't plan on making a huge run of cards. You'd probably sell a few dozen initially and have orders sprinkled about down the line. An AGP/PCIe riser for the same purpose would probably be more popular, as hardly anyone uses PCI these days, let alone slots being available on modern motherboards.
The negative voltage rail is needed for the sound chip and sometimes the amplification circuitry. There's a video somewhere on YT demonstrating several sound cards with the negative voltage rails taped off, which causes varying degrees of malfunction.
I had a non-reference Voodoo2 years ago and it came with green anodized aluminum heatsinks factory fitted. they were pretty small but im sure they made a difference. it was a Gainward Dragon 3000
Useful for Overclocking
I got exactly this card yesterday. It generally works fine, but it does lock up in 3DMark 2000 on an Athlon XP@1500MHz. It gets crazy hot during the bench.
the laser is off target as you hold it very close , look at the sticker on the side , it explains that
I found some 28x28x10mm heatsinks with 3M 8810 thermal tape on Amazon that are the perfect size for the Voodoo ICs. Just sharing with the community in case anyone else would like to add some passive cooling to their retro voodoo cards. Hope this helps.
I can understand it running cooler with the slower processor, there is less data to crunch with the slower processor than with the faster ones, so you do not get to utilize the card to its full potential. This was actually a great way of showing how a processor bottle neck can affect a video card, without having to see benchmark results :)
Like some other people mentioned, it doesn't measure exactly at the laser dot, but the laser is a sort of an indication where you are aiming. The measurement is just below the laser dot. I got a IR thermometer as well, and it usualy helps to move it around a bit around the area you want to measure until you hit the highest temperature. The closer to the object, the more accurate as it measures in a cone shape (closerby is a smaller cone, is a more accurate reading). My old 3S805C 1MB onboard chip of my 486 would hit 70ºC even when just 'idling' on the desktop (chips didn't have power saving modes back then, they always ran full speed all the time, no matter the load).
Yea I stopped using the laser. But the other issue that is doesn't measure all materials the same. For example a silver aluminium heatsink, it totally throws off the device.
I got a bit more expensive IR thermometer, and it can even do shiney surfaces (silver/alluminium heatsinks for example). Only glass throws it off. I never use the laser on the deveice, it drains battery too fast, potentionaly dangerous (for your or someone else's eyes), and the stuff I measure is always closeby so I can aim without a laser :)
I have a voodoo 2 with an athlon 1,4 ghz,,,I think I should add a fan :P Thanks for the video :)
Heatsinks are a better first step and don't emitt noise. There are tons of them in different sizes on Ebay.
Get a thermal probe from banggood, you can tape the probes onto the components to measure the temperature more easily. Kapton tape will hold under temperature, and they have a dual chsnnel model, too.
This was my first gpu! Years old by the time I got my hands on one but man I miss those days.
I feel like this is the perfect opportunity to make use of your batch of heat spreaders and thermal pads on those chips
Phil, this kind of contactless thermometers are practical for some applications where area to measure is large and material emissivity is well known and adjusted for. While the latter *might* not be a problem, the distance that you measured from and the fact that laser is not really collimated with the actual sensor will seriously impair the measurement.
Yes I've mentioned this in the video. I won't use the laser the next time, people get too caught up over it. It's a simple case of searching for the "hottest spot".
"So look, its not totally useless.." best product endorsement ever.
In all seriousness though, $11, some quirks are to be expected.
Your thermometers infrared aperature and where the laser points where certainly not aligned correctly. The Chips should be at their hottest in the center. You can test your thermometer with a cup of boiling hot water. Move it across the surface of the water and see where the laser is when you read 98-100 Celcius and if you even read 100C. Then use some Ice Cubes, the melt water coming off of it should be very close to 0C.
Nvm you addressed it at 8:00 :D
This seems to be the case there. When he is measuring the bottom of the chip he is actually measuring the PCB under the chip and when he is measuring the top he is measuring the center of the chip.
I wonder if the thermometer came with a manual which would specify the correct measuring distance where the laser and the aperture would converge?
On the side of the thermometer is a little chart explaining the measuring distance
It's not just that these have an offset, but the measurement is a cone area that gets bigger the further away tou hold it, idealy you put the aperture area right onto the thing to measure if possible.
What did he say that thing was? $11 or something? Those "Shenzhen Special" tools are usually piles of hot garbage. It would be a small miracle if the thing is accurate to within +/- 10% on the temp readout.
This is a normal construction detail of these IR thermometers. They have no way to track how far the object you're measuring is! Also the optics of the IR unit is pretty good, so you're not limited to measuring at a single given distance. For this reason they simply choose to aim the laser parallel to the pickup, and you have to account for the offset between the laser pointer and the pickup in the device manually.
As to how accurate they are? Well in spite of being from Shenzhen, they're usually about the same as you'd expect, and you don't expect an IR thermometer to ever give faithful readings, because even in the best case, these are black body radiation measurement devices, and you don't have black bodies in real life. You use it to zero in on issues and get a ballpark figure, and then you should tape a K-type thermocouple where you need it. Those are precise by nature.
Was never a problem in period. I ran a P2-350 and later a P3-700, Rage 128, Voodoo 2 in SLI, NIC, SCSI, Sound and modem cards filling all the slots. No overheating issues with only an 80mm intake fan and that massive SCSI cable blocking the airflow!
If you look at the front of the thermometer you'll see that the laser sits above the IR sensor. The laser is just a rough guide of where you're aiming. If you look at the side of the thermometer there's a sticker showing how large area the sensor will "see" at certain distances. The temperature reading will be an average of the area the meter sees.
That fan bracket is handy, haha
It would've been interesting to see the temperatures of other components, for example the memory on the V2.
I have a generic version of the same IR thermometer and due to it having a laser pointer and the sound it makes when you release the trigger and it holds the last recorded temperature for display, my cat comes running every time I use it because she wants to chase the laser that she knows is always around when she hears that sound. Sometimes she comes running just because she hears the click of the trigger pull.
Its a legit 100% working cat call, without fail she comes running when she hears that "Beep Beep". lol
Just a quick note on the temperature readings.. The thermometer you purchased has a fixed emissivity setting of 0.95 which will give you an inaccurate reading on some materials, particularly when pointing it at metal parts with a much lower emissivity rating (EG metallic pins). The extra few degrees you get by pointing the thermometer off the chip is probably that.
No, it's not that, although it does fail on metal objects, it's the offset between the pointer and the pickup - they are aimed in parallel, but on the device they are not coincident, the laser pointer is at the top and the pickup is a couple cm below it. This is why when he was aiming the laser on the pins, it picked up exactly the middle of the chip! You're supposed to simply take that into account.
The reason why your laser temperature meter reads off from the laser dot it because of a Parallax error caused because you were measuring the chip from a very short distance. the error will lesson as the distance grows. even expensive units can suffer the same. Old point-and-shoot camera viewfinders suffered the same phenomenon
The chip measures hotter at the top, because the sensor of the thermometer is underneath the laser. So basically that's when the sensor is aligned with the center of the chip.
Yep, had this exact same card back in the day and I remember it ran pretty hot. Really pissed though, I lent it to a friend after I upgraded, and they said it didn't work bet never ended up giving it back so I could test it myself. So now I no longer have a piece of gaming history! :(
:(
i still have two of them diamond monster 3D II 8mb
Amazing ancient technology. There are no BGA connections, no FLIPCHIP silicon assembly, ordinary poison tin PB and 22 year old equipment still working: D
While buying 22 years ago as a top card, such a VOODOO 2 cost a fraction of what today's top graphics cards will live 1/10 of the time as such a classic 3Dfx.
My VOODOO 1 has factory heat sinks and is quite cool.
As for the pyrometer. It is this laser pointer that is simply not aligned with the sensor axis and rather shows the measurement location to the eye.
You have to shoot this meter like cowboys from your hips: D: D
I was brave. I have put a voodoo2 to my I7-4770 pc with windows 10 :D
A dont know the temperatures but it worked!!! Driver was a bit tricky.
BTW I could archive 50 FPS in quake 3 (640x480)
lol you are a brave one xD some LGA 1150 motherboards still have like 3 PCI slots (5V keyed i assume), so plugging a V2 in there is tempting. but yeah would be interesting to know the temps though, not sure i'd want to risk one since they are hard to find here.
I'm thinking about doing something very simular on my rigs. But rather then putting the fans like you did sens I have pretty much all the slots filled with cards network audio raid controllers and so on.
My idea is to mount two 90mm fans vertical blowing from the side in between the cards. Forcing the hot air lit forward towards the hard-drive bays. That way all the cards do get cooling and making a bracket is probably not going to be to hard.
I have two rigs one is going to be overkill fast. Windows xp and Core2quad and the other is going to be overkill sempron 3000+ speed. Så both need cooling.
For me voodoo are a soft spot I wish they would have kept voodoo around and Nvidia would have continued the series :) 3dfx was awesome and still is :)
+PhilsComputerLab
Another fun video! Interesting too. Back in the day I was having heat problems with that very card. Games kept stuttering and crashing during extended gaming sessions. At first I just took the sides of the case off. Ended up attaching 3 low profile heatsinks to each of the 3 chips and got a "Hamster-Wheel" type blower next to it to pull air out of the case. After that, no problems. I like the fan mounting kit you showed in the video. Very clever bit of kit!
Yea and it keeps the card untouched without having to glue.
I've considered adding heatsinks to my voodoo II cards but not done it yet, just made sure case has plenty of airflow
Just like a laser sight for a gun, the best the laser can do is run parallel to the heat sensor. It will never be accurate, but it will always be equally inaccurate.
The alternative would be a tilted mount that's only accurate at a given distance and wildly inaccurate at any other distance, or an auto ranging mechanism that actually adjusts the angle of the laser to hit the target at a given distance.
I wouldn't call your sensor inaccurate as it was probably made for testing things at much greater distances, The 2 Cm distance is just the location of the sensor and the laser pointer.
One idea would also be to use some thermal adhesive (Basically thermally conductive two-part epoxy - Arctic Silver and Arctic Alumina come in adhesive versions for example) and use that to affix some small heatsinks to the chips. This might not necessarily be ideal, though - it's a permanent modification to the card. But it would certainly provide a marked increase in thermal dissipation. You could also use some chipset heatsinks with integrated fans as well if you found ones with decent or replaceable fans. The not so good ones make a clearly audible high-pitched whine. Those little 40mm fans sure can manage to be pretty loud.
I was a little surprised how effective the fan cooler was, since it is literally just two 80mm fans bolted together
Does a pretty good job going from no air moving near the gpu too a directed airflow across the gpu.
on top of dedicating a fan or two too move air across a sli voodoo 2 setup id also mount on a bunch of slim heatsinks.
Those day add a heat-sink and PSU fan to voodoo 2 is a must
You need to place some Aluminum radiators glued on the chips with Arctic Alumina thermal compound which is meant to stick forever the radiators on such chips. They are too hot.
Put some small aluminum heatsinks on it like some raspberry pi builds. Should improve thermals massively.
I coupled two of these cards in SLI mode in a PIII 450 Mhz setup. To keep them a little cooler, I cut a hole in the case side panel, and installed a low profile 120mm fan to the inside of the panel, pushing fresh air directly into the gap between the two cards. Did the same in the middle of the top of the case, to suck the warm air out. So far no, problems. But I am not sure about the temps. Cards are still OG, with no heat sinks installed. Games run fine in 1024x768 rez.
Back then the Harddrive and PSU was the noisemaker. :)
Heat rises upwards. Law of physics.
Maybe if you placed the card vertical, the heat is spread more evenly on the chip. But I believe the heat will dissipate into the air better on this way you placed the card. The other (normal) way, it will just heat up the PCB more.
Actually I never thought of that. It makes perfectly sense.
Same with a CPU cooler. Will always be most hot on the top. if placed in a Tower case. With a fan.. not so noticeable, since the heat is being pushed / pulled out of the heatsink.
So Passive cooled chips especially with heartsinks, will be better cooled when you place it with the chip/heatsink upwards. But dust buildup is another story. Seems like there is a plus and a minus to that solution.
A fan is always a good idea - so we can move the heat away.The price we pay is more noise/ dust.
Thank for a good informative video - again. :)
I have that exact model of thermometer and I can tell you, the infrared appeture measures heat about 2-3cm below the laser point. Try it with a hot knife oriented horizontally and test to see where it starts picking it up
Do watch the rest of the video :)
is the diagram on the side a guide to figuring out where the 2 beams cross?
Had two of these puppies back in the day, it was my first build. Paired them with a Celeron, cos i didn't know any better. Used to play Carmageddon with software rendering, then switched to Glide, can't remember if it actually used SLI though. The difference was amazing.
One small tip, you can buy large quantities of raspberry pi heatsinks on aliexpress and other sites. You can put them on the chips!
all so can get some Low profile heat sinks and glue them out with thermal adhesive
I prefer to keep the cards in pristine condition.
Not used my Voodoo 2 card in a while "creative". But do have a voodoo 3 "creative" on a PIII 733mhz. And never actual did consider temps on the card. It's at around 50 to 55 when playing using a non laser tool to measure temp. Tho it's in a modern case with good air flow and a 120mm fan on the door pointing at it. My Voodoo 1"diamond" with a 166mmx is also in a modern case with good airflow to. Though did not test that atm. But I'm fairly sure it would run above 60 with Unreal. But I mostly only play i76 and NFS2SE on it.
Hey I can't seem to find a proper Interstate76 copy. Do you have the original cd?
Yeah. Still got my original copy from back in the day. No box tho.
Heatsinks are higly recommended. Ensures the life of your card, don't emit noise and is cheap, so it's a no brainer. There are tons of them in different sizes on Ebay.
I have heat sinks. Just not a fan setup under it like on the video. My Voodoo 3 has it by default. And the voodoo 1 card got an after marked variant put on in 1999.
I added heatsinks to my Voodoo 2 SLI setup. I bought the heatsinks from Ebay and it was basically a perfect fit on the 3dfx chips. The heat sinks are barely warm with the finger test when playing Unreal (92 MHz GPU).
Cool! Question is how much a P1-class system bottlenecks a Voodoo2.
Now come up with a cooling solution for SLI
A fan to the side of the case that blows onto both cards.
suppose that might work.. but if you have a period correct case (like i do) wont work that well without hacking up the case.
I bought some black aluminum heat sinks off of ebay from China, and some thermal epoxy off of newegg. I glued them on my Voodoo2s for my SLI setup, and a fan at the front of the case, blows toward the back of the case to help cool them. There are also some VGA slot coolers which can be mounted perpendicularly to the cards, so they blow downward over cards.
It wouldn't be that difficult to design and print a shroud to clip on to the Voodoo2 and mount a small 60mm fan. I might consider it since I have four Voodoo2 boards.
GGigabiteM that would be something I would buy!.. I do not know anyone or own a 3d printer unfortunately.. I don't like the idea of sticking on heatsinks, just for the fact the cards i have are in new condition.
Nice video. I've always wondered how hot these chips with no cooling get.
Have ever used those little stick on heat sinks you can get?
I've used them a few times on sound blaster cards like the Audigy 4 which get quite hot.
Nice vid. I've also done that in the past and I was also surprised with the results. Actually my Geforce 4 Ti is running like this. I would say in some cases you don't need a extra heatsink, a direct airflow like this will do a nice job.
When I saw your chips were running in the 30s I though to myself "Dang boy! How cold is your A/C?"
Then I released you're reading in Celsius.
"oh.."
A note about the accuracy of the thermometer, it is probably pretty accurate at about a meter, digital thermometers like that typically aren't meant to be used that close to the item being measured. Still a nice tool to have to get a good idea how hot things are. I remember back when I had a Voodoo 2, Voodoo 3, and Voodoo 5 I used to make my own GPU coolers with a mounting bracket and fans to basically do what that device does... I actually miss those days :)
So I've used this thing in another project, and someone else mentioned it here, the results can vary depending on what material you are measuring. For example a aluminium heatsink throws it off totally. Has something to do with emission. So yea, these are the things you only learn after the fact :D
The distance thing is purely so that the laser pointer is accurate. I now just turn it off and search for the hot spot by hand. You can go straight up to what you measure, it's actually better.
i had voodoo rush, voodo3d, voodoo2, and later was given 3 and 4th gen cards, honestly, the matrox m3d/apocalypses 3d using powervr chipsets where better for titles that had support for them(quite a few actually had powersgl exectuables), and it had better ogl/d3d support with last gen drivers then early sdfx offered, you could/can also stack as many of the powervr cards like the matrox m3d in a system and they will work in "sli" no dongles or any of that, i tested up to 4 and saw VERY good scaling in every sgl and many ogl/d3d titles back in the day, i would LOVE to see one of the all in one cards that had the powervr chip used in the matrox m3d on it(like a voodoo rush but power vr) tested like this, but also tested with more cards with the same chipset added to the system... i know with 2 of those cards i could do 1024x768 maxxed out in ultim@race and a slew of other titles, unreal was AMAZING with those cards(blew the quality 3dfx could offer away...no blatant 16bit dithering for example... and omg.. the kyro2....i would have loved to see powervr tech get the next upgrade to have hardward TnL and how much it would have made ati and nvidia look over priced...in a proper system the kyro2 sure as hell did...
Just a heads up. All IR thermometers no matter how cheap or expensive are not accurate that close. Read the label on the side. Good video anyway.
Should get some of those cheap aluminum RAM and chipset coolers from AliExpress and stick them on this card without the fan and see how much cooler they run.
Phil heres a tip, make a 3d printed stand for your temperature gun to constantly monitor the gpu temp while trying games.
So many memories with the Radeon 7000
So the video card gets hotter the harder you load it? Who would've guessed :P
Really interresting video, I would'nt mind seeing more of these.
HAha, there are some that say the V2 doesn't get hotter, so that was a good reason for me to check it out :D
I do understand why you did it. And now we have hard evidence that it's true :)
Pentium 3 500mhz, my first cpu ^^
Same =D Dell Optiplex GX1
A K-Type thermocouple would probably be better (and just as cheap) for measuring the temperature of a setup like this, just secure it in place on the gpu with a blob of thermal putty.
I'm curious if your thermometer is just bad. I did the exact same test with a higher ended model that uses 2 lasers to help give you the correct distance for a proper measurement which should be around 14 inches or so. On a Tualatin P3 1.4ghz i often measured only around 55-60c on my voodoo 2s. At over 80 you should not have been able to hold your finger there as long as you did.
Regardless, the cards need some form of active cooling period, even voodoo 3 with its minor fixes to this bug and heatsink need it as they still get far too hot imo. Not having airflow of some sort over your voodoo card is asking for an early grave :(
Typically you have to hold those temperature probes 30cm from the target for the laser dot to be accurate
Yea it's not, but on large surfaces it's not a big deal. You have to get close because the chip is quite small.
Great information, 80C+ in open air is a little surprising. Did you happen to look at how warm the memory chips got?
Good point, I didn't check the memory chips...
I don't see how an open case with no airflow will cool better than a closed case with _proper_ airflow.
At any rate, this also highlights what is, in my opinion a design flaw in *all* graphics cards to date, which, in most PC cases, will be mounted with the chips and heatsink on the underside, which doesn't help with cooling.
Its kind of bizarre, when you look at ISA cards and the components are on top.
Just take a look into the BTX design.
It promoted a different design that basically had a single straight airflow through the case.
But it was loud and Pentium 4's let the market, so everybody kept going with ATX.
But there are some cases that allow different mounts for the motherboard. upside-down, rotated by 90°, laying flat, etc.
Proper airflow depends on getting cold air to the parts and getting rid of the warm air. How a system handles that is not all too important.
People say hit air rises up. That is generally true, but once the air is moved actively, it doesn't effect anything.
People say a fan at the side destroys the airflow. Not wrong, but depending on the parts that are blown onto it can actually help with temperatures.
An open setup has no actively "channeled" air movement, but alot of air around. Can help, can make things worse. Depends on the setup.
And about the orientation of expension cards in systems, go back to 1991 and tell Intel that the orientation of PCI is bad.
In many cases (pun intended) the internal air temperature will be roughly 10 to 15C hotter than the rooms ambient air temps, even the best cases will still be a few degrees Celsius warmer, due to the difficulty in getting the warm air out.
Which does make sense when you think that almost all cases will have just 1 exhaust, and even that is not guaranteed, and even if the case does have an intake fan, it will most likely be sandwiched up against the HDDs, blocking off most of the air flow from it.
Gamers Nexus demonstrates this in their case reviews, where poor airflow can cause thermal throttling with even the best CPU and GPU coolers.
The ATX form factor is severely flawed. Intel had absolutely no forward planning when designing that. Which is really retarded because ATX was designed to overcome the limitations of AT and give more headroom for the future but in the end did not.
The older AT design did put the chips on the top surface of the expansion cards in a tower case, but in the ATX form factor with PCI slots this was reversed. Keep in mind that at the time ATX was introduced, the vast majority of PCs were desktop models where the expansion cards were vertically oriented, so it didn't really matter which side of the PCB the chips were on. The main consideration in regards to cooling with the ATX form factor was to get the CPU as close to directly under the power supply (in a desktop case) as possible to pull air over it and out of the case as efficiently as possible. Mid tower and full tower cases were intended for enthusiast/server setups and would have been expected to have more robust cooling installed.
I ended up putting copper heatsinks on both my voodoo 2s and 3d printed an 80mm fan mount that attaches to either card while in SLI and blows cool air over the heatsinks. It let me overclock my Voodoo 2s by quite a bit to better accompany the PIII 800 Mhz.
Gabe Newell
So, when will Half-life 3 be released?
c'mon, you can tell us. :)
Seems interesting to me that it got so much hotter with high resolution and a faster CPU. I guess if they knew they were going to be used in such fast machines they likely would have included some small heat sinks of the chips. It is still, not that hot in the grand scheme of things, especially considering the size of the lithography. Not sure how useful the video is actually but it still was interesting to me. :)
Got my voodoo 2 today for my timemachine project to fill 2 roles.
First role is that it fills the gap with my Nvidia 6800GT to be able to play dos titles and older titles the NVidia 6800GT is to new for.
The second role is for the summer months, i decided that since my NVidia 6800GT died in the warm weather its probably not a good idea to have that one in the system in the summer and i have a S3 Trio64 graphics card. Which would mean the summer version of this build would be perfect for the 3.11 / 98 era. And the winter build would be perfect for the 98 / XP era. With both versions handling dos very well.
The CPU though is a 2,6GHz CPU so its going to run quite hot. I assume that puts me into the territory where i have to cool it. But whats the best approach here? Just passive heatsinks and tape, or active cooling? My PCI slots are very busy so i don't have much space.
My LGA1151 mainboard has a PCI slot! Let's see, how hot it gets with an i7-7700k!
All IR thermometers work like that. They can't put the laser inside the IR sensor, it will always shoot high and parallel to the center of the sensor. This will make it seem more "off" the closer you are to the target. It should have also explained in the manual how far away to hold the device from the thing being measured based on the size of the thing being measured. Smaller things need to be closer, etc. It's basically a camera with no zoom. It was hard to tell exactly how far away the chip was from the sensor, but it was probably good... maybe a scratch closer would have been better.
Yea I got a ton of comments :D I ended up just going with a standard thermal probe. Got one with 2 and another with 4 channels :D
hah that's the same termometer i have in my hands right now.
about the temperature being lower on the bottom, it's because the thermometer measures the temperature just BELOW the laser dot. so when you go to the bottom of the chip, you're actually measuring the pcb temperature. look at the front of the thermometer, it measures with that same offset. i know, it's a bit annoying. it should have another laser below the measuring lens so you could frame what you want between the lasers. well whatever it still works fine.
Thanks Phil, I like that fan setup it`s a good solution for a single Voodoo2 setup. do you think it would do much for a SLI setup?
That should, work, but there won't be many PCI slots left for like a sound card.
This video inspired me to have a go at it myself! 😄
But now I must ask.... Does an SLi setup run cooler by default? For if not, I am getting surprisingly different results from this very same test...
Just performed this test on my V2 SLi (2x12MB) P3-733 Rig...
Unreal running at 800x600 in HQ texture mode. Later I've ran it on 1024x768 as well, without any change worth mentioning apart from lower fps...
I get an average of about 62 degrees celcius with my fans turned off... And when I turn my (totally differently installed) fans back on, I get an average of 40 degrees celcius...
This, while my room without aircon is a very humid 30 degrees in a currently very tropical Netherlands... a little warmer than Phil's A/c'd 27...
Hence my initial question... 😆
Well done. Nice test.
Unlike what he said if you actually have a fan or 2 in your case you'll actually see cooler temperatures as there will be air flow in the case.
Thank you for the test. But I am Missing a table or diagram to visualise the data.
That's because you're pretty stupid tbh... or a troll. Either way stop being silly Conrad.
question: isn't this because of the higher bus speed ? When changing the agp divider the card might not heat up as much? Have you tried that?
Just stuck old cpu heatsink and fans to my old voodoo 2's in sli with memory heatsink as well . Where my back up cards after I changed to geforce 256 which I had to keep getting replaced as they died for a hobby
i glued heatsinks on mine back then :D
I would find it interesting to see the temperatures of the chips if you stick just heatsinks to it, but not useing a fan. Just passive heatsinks.
The cards are now worth a lot, I rather keep them pristine.
All I did for mine was screw my 2200RPM 120mm radiator fan onto the side of the case, leaving the panel off. Both Voodoo 2 and the Diamond Viper V550 no longer have thermal issues after that.
Seeing that those brackets with fans didn't exist yet in the days of the Voodoo2, maybe you should test that on a modern passively cooled gfx card?
I have the canopus pure 3d2, this voodoo2 version has a small fan on the main chip. Afterwards i installed surplus heatsinks on the other chips.
One slot cooler I would definitely recommend that's a lot like that is the vantech Spectrum dual 70 mm slot cooler it has a whole lot of air flow with an adjustable rate of speed it ranges from about 15 to $20 depending on where you get it but I would definitely recommend it back when Pentium 4 with hyper-threading Dell gx270 I got an x 800xl to compliment the video card before I got the cooler sad about 50 to 60 Celsius I was able to drop it about 20% it hangs around 40 Celsius maybe a little bit lower
sweet! I've known the cards to run hot, but didn't think they were getting near 80c. I use a copper heatsink option on my voodoo 2. Might have spent more than $11.00, but it works well. I can't help but wonder which is the better option. the Iceberq 4 cooler, with a copper Cooler Master ram sink kit, or the fans. I didn't use the fan option based on the location of my card, and not having a free slot next to it.
Hey Phil, maybe you could answer an age old question about the voodoo cards. Does it matter which PCI slot it goes in?
Hmm I did have cases when moving the PCI slot resolved the card not working, but apart from that, it shouldn't matter much.
I'd try sticking a fan on the far side of the card or cards so it sort of blows air like a reference cooler on nvidias graphics cards.
I-----
like that above is your card on its own seen from the side & you place a fan on the right side of that making it kind of look like I----[] with [] being the fan ofc.
AB-BH6 can run 1Ghz P3 on v1.02? Isn't Coppermines over 700MHz officially only runs on v1.1/v1.2?
Those old mid to late 1990's 3D accelerator cards always lacked active cooling.
What a mistake.
Then again perhaps they did it on purpose so your components wouldn't last long so
you had to make a hardware upgrade.
I've put fans on my Voodoo 1 cards before as well.
The Voodoo 3 cards were insanely hot ..even with their heat sinks (and mine came with a big one too).
Yeah I would say that the Voodoo 3 cards were the worst of the 3Dfx cards when it came to heat
Turning on V-Sync helps also, it limits the FPS.
Up until the late 90s, there was still a stigma against using fans in desktop PCs. This was partly because most fans of the day used really noisy ball bearings and poor blade design that resulted in humming and whirring that echoed around in the case to amplify the sound. It was also partly to blame on design philosophy of the time, fans were just considered "not cool" to have.
It's one of the reasons why towards the late 90s that CPU heatsinks grew to monstrous proportions, to try and stave off the need for a fan. I can remember some Slot 1 PII-PIII heatsinks that were so large that you couldn't hold them in one hand. I still have a PIII-500 with a monster Compaq heatsink riveted to it.
Many of the graphics chips of that time period were designed to run at up to 90°C without any problems or reduced lifespan. The bigger problem with the hardware of that era was with thermal stress from the chips heating and cooling repeatedly causing mechanical failure, as opposed to overheating (unless there was no airflow at all).
Mechanical stress failures related to heat was not as common of a problem back then because most large ICs were in DIP or PQFP packages. Those that were in BGA form factor had a similarly low failure rate because leaded solder was used, which was ductile enough to cope with the mechanical stress without breaking.
we don't know what is the reference operating temperature provided by chip spec. It may very well be that these temperatures are approriate and cooling it down doesn't bring any benefit.
Also notice in a closed pc case there is a natural air flow, which means measuing temperature without the case is .... not really indicative.
It wouldn't bring a performance benefit, as these temperatures are indeed appropriate, but it will bring a longevity benefit, as the speed of transistor/FET degradation due to migration approximately doubles with every 10°C increase.
In, say, your i7-2600 rig's PCI slot...
using unsigned drivers for W10... (think i've seen voodoo drivers for newer versions of NT)
or running winxp...
would there be a serious risk of the card being damaged? would the PCI bus be a significant bottleneck at any point?
Windows 98 is what I highly recommend, and a Pentium III is perfectly fine to saturate the V2.
Thanks for the information, i've got a Voodoo 2 coming in the mail in the next couple of days and i was worrying about the possibility of overheating, i was thinking of ordering a bunch of 14x14x6mm aluminium heatsinks for the chips, are those ideal for a card like this?
I'd go for 35x35x10 or 40x40x10mm heatsinks.
More surface area in a single heatsink dissipates heat better. instead of you putting on like 3 or 4 per chip you put on a single large one.
That size covers a quarter of one chip. 27x27mm is pretty close to exact, I've used 28x28x6mm.
Hmmm.... I think I need to get some permanent attached heatsinks on my Voodoo2 cards. Then get a fan solution like that. I have noticed this heat problem on a P3-1000. And that is why I only use my cards on K6-3 400.
Yea in a closed system I can see temps going to 90c or higher, and that is a worry. I don't think anything fancy is needed, just something that cools a little.
Some people watch a leather ball been kicked over the grass, but I rather watch a Vodoo 2 heating up!
Hey Phil, do you think adding heat sinks with thermal pads would have improved the temps
For sure!
sadly we can only 3d print plastic.. imagine if we could 3d print sheet metal or aluminium :D
Thermal camera could help you find hottest spot
Could you do one of these for the s3 Savsge4? Also, I had one other question; I'm new to building "retro nouveau" PCs. If I have the choice of using either PCI or AGP for a graphics card, which is the better slot to use if I'm installing Windows 98 SE on the machine?
I put one of these on a Core2 Duo. I had to heat sink all 3 processors and the memory chips as well.
I remember overclock my Voodoo2 to 98mhz (Bottom chip heat-sink/ fan mod)
The oldest voodoo I own is a voodoo 3 1000, don’t have a power supply lying around to test it though
Powersupply ? you mean pc too test it in rather im guessing.
Is it a agp or pci one btw ?
if its a pci card it will work in just about any pc with a pci slot even something like my main pc with a I5 2500 non k 8 gb ram & a r9 280.
70C isn't terribly hot. My mobile i7 runs at 103C (!!) before it really starts to throttle down much. Even then its only losing 2-300mhz off of max turbo with 4 cores at load. And I have had it running at full load for long periods of time with no issues. My mobile quadro runs at 80-90C easily. I don't know what the design limits for the Voodoo 2 are, but I would imagine that 70C is likely in the safe limit. If it were that great of a concern, they would have put a heat sink on them. I think that you are definitely right that inside a case with poor airflow they would likely climb a bit higher as well. Most GPUs I have used can easily hit 90C without too many issues. Obviously if you want to overclock you would likely want to look at cooling for more headroom.
Edit: I found a reference that Voodoo3 ran comfortably overclocked to 180mhz in the 80-90C range.
It is still Spring in Australia! LOL
It would be cool make a similar video with a thermal camera (like a flir)!
I had a quick look, they are not cheap.
Unfortunately yes, normaly they are expensive.
I come back 10 month later with a new product, that maybe it can be interesting: robot-electronics.co.uk/usbtpa64.html This pcb camera have a AMG8833 thermopile array, it have only 64 pixel, but it's cool!
I'd slap some stick on heatsinks on there, but I guess that would cover up the cool logos :P
I want to keep these cards pristine.
You could get logos laser etched onto the aluminium heatsinks, just saying...
What is this ATX test fixture you're using?
Are there any PCI riser cards that measure current or that would allow you to measure current? That would correlate directly to power usage and heat dissipation with the card.
I'd imagine that an engineer who develops PCI cards has such a beast, but I've never seen one.
It would actually be quite difficult to make such a board because unlike PCIe where all of the power rails are in the front of the slot, PCI has power rails sprinkled all over the slot. All of the common power rails from an ATX power supply are available on the PCI bus except -5v, so you'd have to monitor all of those lines individually and add the result up.
Compounding that difficulty, extending the PCI bus with a riser card can lead to bus instability from connection issues and from circuitry added to the riser to measure power draw.
I think designing such a riser should be pretty easy - you would need some decoupling on the voltage rails, and it would get a bit bulky, but you can route all the data lines straight through while compensating for the distortion. There's only 3 power rails you're interested in, 3V3, 5V and 12V, -12V would only be of interest to maybe something like Ethernet and RS232 cards, so we can skip that. You can easily merge all the pins corresponding to a given voltage onto one point on the input and fan it out on the output - the multiple pins are only there to guarantee good connection and for current rating. Should i do that, would anybody buy one?
-12v is commonly used on sound cards, so it is somewhat important.
I'm sure someone would buy it, but I wouldn't plan on making a huge run of cards. You'd probably sell a few dozen initially and have orders sprinkled about down the line.
An AGP/PCIe riser for the same purpose would probably be more popular, as hardly anyone uses PCI these days, let alone slots being available on modern motherboards.
-12V, on a PCI soundcard? I've only had SB Live/Audigy and haven't seen -12V utilised there! Nor on most late-era ISA soundcards that i looked at.
The negative voltage rail is needed for the sound chip and sometimes the amplification circuitry. There's a video somewhere on YT demonstrating several sound cards with the negative voltage rails taped off, which causes varying degrees of malfunction.
My Gainward Voodoo 2 has heat sinks :)
How about sticking heatsinks on the chips?
Yea a lot of people do that, but I prefer to leave the cards untouched.