VOGONS


First post, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie

Hello VOGONS members!

I have a question. It is well regarded that the stock heatsinks that are oh so wrongly thermal-epoxied to both of the Voodoo5 5500's GPU dies do a poor job at cooling those hot chips, and that rapidly chilling and then popping them off to replace them with third-party heatsinks does the GPU right, without being able to tell how much it helps, since the card itself has no way to convey the temperature of the GPU dies. But, has there been anyone who has conducted an almost Gamers Nexus-style experiment, using temperature probes or thermal couples to measure the temperature of the GPU die as best as possible before removing the stock heatsink, and then afterwards with a new aftermarket heatsink using thermal paste? It has been nearly two decades after the release of the V5 5500, and I'd be surprised if no one has thought of trying to do such a thing yet.

The reason why I'm asking this question is I might get a V5 5500 in the future, and I want to know if the risk of modding the card would be significantly worth it just to cool the card better. If some one has done such a test, and came back with a 5C difference, I would say that such a mod would be unnecessary.

Any help would be appreciated. Thanks!

Last edited by ultra_code on 2019-01-13, 21:39. Edited 1 time in total.

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 1 of 38, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

Well, my original V5-5500 received a pair of SS7 heatsinks shortly after I purchased it new.

The heatsinks barely even get warm when passive. I use a 80mm fan blowing on them to keep them nice and cool.

The stock coolers are junk.

Yamaha YMF modified setupds and drivers
Yamaha XG resource repository - updated November 27, 2018
Yamaha YMF7x4 Guide
AW744L II - YMF744 - AOpen Cobra Sound Card - Install SB-Link Header

Reply 2 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
cyclone3d wrote:

Well, my original V5-5500 received a pair of SS7 heatsinks shortly after I purchased it new.

The heatsinks barely even get warm when passive. I use a 80mm fan blowing on them to keep them nice and cool.

The stock coolers are junk.

A prime example of qualitative testimony. Nothing against you, but that doesn't tell me anything. It tells me it helps, but by how much? Numbers, numbers are what I'm looking for here.

EDIT: mixed up quantitative and qualitative like an idiot. 😜

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 3 of 38, by doaks80

User metadata
Rank Member
Rank
Member
the_ultra_code wrote:

and that rapidly chilling and then popping them off

How does one do that exactly? Liquid Nitrogen? Freezer? Do they need to be hot beforehand?

Personally I use 2 x 80mm fans in a PCI slot blowing directly onto the card, and heatsinks on the rear of the card with air-flow from side-panel fans to keep it cool. What I know is those rear heatsinks get amazingly hot, too hot to touch for more than a second, and I wonder where all that heat would be going otherwise...

k6-3+ 400 / s3 virge DX+voodoo1 / awe32(32mb)
via c3 866 / s3 savage4+voodoo2 sli / audigy1+awe64(8mb)
athlon xp 3200+ / voodoo5 5500 / diamond mx300
pentium4 3400 / geforce fx5950U / audigy2 ZS
core2duo E8500 / radeon HD5850 / x-fi titanium

Reply 4 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
doaks80 wrote:
the_ultra_code wrote:

and that rapidly chilling and then popping them off

How does one do that exactly? Liquid Nitrogen? Freezer? Do they need to be hot beforehand?

Either you stick the card in an anti-static bag and then into a freezer for two hours, and then attempt to remove the heatsinks, from what I read, or you do this: https://youtu.be/5hB4jOCHLHc.

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 5 of 38, by havli

User metadata
Rank Oldbie
Rank
Oldbie

When removing the original glued heatsinks, use extreme caution. I have one dead V5 5500 AGP where apparently the previous owner tried to remove heatsink from the primary VSA chip.... and killed in the process. See the photo here:

http://abload.de/img/pict1434uzsxw.jpg

He damaged some of the traces on the lower side and even one of the BGA balls below. Unfortunately this part of the chip is the section where memory bus is located. 😵

--------------
As for the quality of the original heatsinks - they are not so good, however I believe the main problem is the plastic package of the VSA-100 chip itself. It isn't very heat conductive, therefore no matter what cooler you put on top, the chip will be still hot inside. And not just inside - you can measure temperature at the back side of the PCB just below the chip.

I remember when using Primecooler PC-NB2 https://www.czc.cz/primecooler-pc-nb2/33658/produkt on V4 4500, the heatsink was almost cold but back side below the chip was still hot. I don't have any numbers at hand... but maybe if I can dig up some overkill heatsink that fits V4, I can compare it to regular V4 4500 (and measure the temperature with infra-thermometer).

HW museum.cz - my collection of PC hardware

Reply 6 of 38, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++
the_ultra_code wrote:
cyclone3d wrote:

Well, my original V5-5500 received a pair of SS7 heatsinks shortly after I purchased it new.

The heatsinks barely even get warm when passive. I use a 80mm fan blowing on them to keep them nice and cool.

The stock coolers are junk.

A prime example of qualitative testimony. Nothing against you, but that doesn't tell me anything. It tells me it helps, but by how much? Numbers, numbers are what I'm looking for here.

EDIT: mixed up quantitative and qualitative like an idiot. 😜

Well.. I could do some actual tests as I have a spare 5500 that has the original coolers on it. What benchmarks you want me to test with?

1. Original.. no extra cooling
2. Original - 120mm fan also blowing down across the front and back of the card
3. Modded - no extra cooling
4 Modded - 120mm fan blowing down across the front and back of the card

I might even have an IR temp gauge laying around somewhere.

Yamaha YMF modified setupds and drivers
Yamaha XG resource repository - updated November 27, 2018
Yamaha YMF7x4 Guide
AW744L II - YMF744 - AOpen Cobra Sound Card - Install SB-Link Header

Reply 7 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
havli wrote:

When removing the original glued heatsinks, use extreme caution. I have one dead V5 5500 AGP where apparently the previous owner tried to remove heatsink from the primary VSA chip.... and killed in the process. See the photo here:

http://abload.de/img/pict1434uzsxw.jpg

He damaged some of the traces on the lower side and even one of the BGA balls below. Unfortunately this part of the chip is the section where memory bus is located. 😵

Good warning. Thankfully, the methods I stated shouldn't cause that kind of damage.

havli wrote:

As for the quality of the original heatsinks - they are not so good, however I believe the main problem is the plastic package of the VSA-100 chip itself. It isn't very heat conductive, therefore no matter what cooler you put on top, the chip will be still hot inside. And not just inside - you can measure temperature at the back side of the PCB just below the chip.

A great point.

havli wrote:

I remember when using Primecooler PC-NB2 https://www.czc.cz/primecooler-pc-nb2/33658/produkt on V4 4500, the heatsink was almost cold but back side below the chip was still hot. I don't have any numbers at hand... but maybe if I can dig up some overkill heatsink that fits V4, I can compare it to regular V4 4500 (and measure the temperature with infra-thermometer).

If you could do that, that would be a great start in trying to answer my question (especially since the V4 4500 and V5 5500 use the same GPU chips, no?). Would very much appreciate it.

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 8 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
cyclone3d wrote:
Well.. I could do some actual tests as I have a spare 5500 that has the original coolers on it. What benchmarks you want me to t […]
Show full quote
the_ultra_code wrote:
cyclone3d wrote:

Well, my original V5-5500 received a pair of SS7 heatsinks shortly after I purchased it new.

The heatsinks barely even get warm when passive. I use a 80mm fan blowing on them to keep them nice and cool.

The stock coolers are junk.

A prime example of qualitative testimony. Nothing against you, but that doesn't tell me anything. It tells me it helps, but by how much? Numbers, numbers are what I'm looking for here.

EDIT: mixed up quantitative and qualitative like an idiot. 😜

Well.. I could do some actual tests as I have a spare 5500 that has the original coolers on it. What benchmarks you want me to test with?

1. Original.. no extra cooling
2. Original - 120mm fan also blowing down across the front and back of the card
3. Modded - no extra cooling
4 Modded - 120mm fan blowing down across the front and back of the card

I might even have an IR temp gauge laying around somewhere.

All of it! That would be very informative, and if you could do such things, I would very much appreciate it. I would assume you would measure the temperature of the back of the card too, no?

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 10 of 38, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++
the_ultra_code wrote:

Any updates?

I kinda have a full time job and a family.. Haven't had time to mess with it yet.

Yamaha YMF modified setupds and drivers
Yamaha XG resource repository - updated November 27, 2018
Yamaha YMF7x4 Guide
AW744L II - YMF744 - AOpen Cobra Sound Card - Install SB-Link Header

Reply 11 of 38, by Unknown_K

User metadata
Rank Oldbie
Rank
Oldbie

Stock fans worked fine for me (cards of that era ran hot). A newer well ventilated case would be better then risking ruining a chip freezing the OEM heatsinks off a card that is going up in value yearly. I kind of like the way the OEM heatsinks look anyway.

Collector of old computers, hardware, and software

Reply 12 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
cyclone3d wrote:
the_ultra_code wrote:

Any updates?

I kinda have a full time job and a family.. Haven't had time to mess with it yet.

🤣 I forgot not everyone is single and unemployed like me. 🤣

It's okay - it's okay. Take your time. Whenever is fine. No rush. Just curious.

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 13 of 38, by havli

User metadata
Rank Oldbie
Rank
Oldbie
the_ultra_code wrote:

If you could do that, that would be a great start in trying to answer my question (especially since the V4 4500 and V5 5500 use the same GPU chips, no?). Would very much appreciate it.

Yes, both are using the same chip and temperature differences on the V4 should apply to the V5 also.
Perhaps today I'll have some time to try it. I found a Zalman VF900, with some luck, it should be possible to mount it on the V4.

HW museum.cz - my collection of PC hardware

Reply 14 of 38, by EdmondDantes

User metadata
Rank Member
Rank
Member
doaks80 wrote:

How does one do that exactly? Liquid Nitrogen? Freezer? Do they need to be hot beforehand?

You show the card some Rule 34 then make it take an ice shower. The heatsink pops right off. 😉

(More seriously, on a V3 I managed it putting the card in a freezer overnight and then prying at the heatsink using a really thin blade from... whatever you call those knives with the removable blades. Box-cutting knives?)

Reply 15 of 38, by lvader

User metadata
Rank Member
Rank
Member

I'm not a fan (pardon the pun) of replacing the cooler on these cards, sure they run a bit hot I've not had any problems with the stock coolers. It would put me off buying one (seeing a replacement cooler)

Reply 16 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
lvader wrote:

I'm not a fan (pardon the pun) of replacing the cooler on these cards, sure they run a bit hot I've not had any problems with the stock coolers. It would put me off buying one (seeing a replacement cooler)

That's why I want hard numbers, to get an accurate idea on how hot the GPU chips get with the stock heatsinks, and therefore see if any one should be concerned with stock cooling performance.

Builds
ttgwnt-6.png
kcxlg9-6.png

Reply 17 of 38, by havli

User metadata
Rank Oldbie
Rank
Oldbie

Ok, here are my numbers measured on V4 4500 AGP:

1. 3dfx / STB V4 4500 AGP with original heatsink and fan = 65 °C
2. 3dfx / STB V4 4500 AGP with original heatsink and Sunon Maglev fan = 60 °C
3. 3dfx / STB V4 4500 AGP with Zalman VF900 = 50 °C
4. PowerColor EvilKing IV with original heatsink and fan = 67 °C
5. PowerColor EvilKing IV with original heatsink and fan and 80mm fan blowing directly at the card = 69 °C

Here is link to the Sunon fan - it is one of the few 40x40x6mm fans available.... which means it can be used as a direct replacement of the original AAvid including the screws.
https://eu.mouser.com/ProductDetail/Sunon/MF4 … we3wNSdxSk5Fw==

The Zalman VF900 was originally mounted on Radeon X800 GTO and it is some kind of OEM version. However it should be identical to regular VF900. I used Arctic MX4 thermal paste. And fan was connected directly to MB fan header - which means 12V, full power.

Temparature was measured using infra-thermometer at the back side of the PCB, directly under the VSA chip. Before the measuring, I started UT99 training map and let it run for ~15 minutes at 1280x1024 2xAA. All tests were performed on Athlon 600, MSI K7 Pro board (AMD 751 chipset), 256 MB SDRAM, win XP and using open benchtable with no extra cooling (except case 5).

Just out of curiosity I added PowerColor Voodoo4 also and it seems to perform slightly worse than reference 3dfx V4. Also it is interesting adding extra fan next to the card actually makes the cooling worse. Perhaps this is due to having optimal cooling conditions as it is on the open benchtable and when running the V4 in regular case it could be the opposite.

So as a conclusion - bigger heatsink definitely makes a difference, however the risk of irreparable damage is still here. Perhaps using only a better performing fan could be a good compromise. I am using pair of similar Sunons (I don't remember the specs, bought the 10 years ago 🤣 ) on my V5 AGP and the work reasonably good. http://hw-museum.cz/vga/62/3dfx-voodoo5-5500-agp

I guess on Voodoo5 temperatures might be few degrees higher as there are two chips crowded in small area. Perhaps tomorrow I could measure also V5 PCI with AAvids and V5 AGP with Sunons. But I won't disassemble my main V5 system, so it might not be completely accurate - V5 AGP in AXP 3000+ system and V5 PCI in this Athlon 600.

HW museum.cz - my collection of PC hardware

Reply 18 of 38, by The Serpent Rider

User metadata
Rank l33t
Rank
l33t

I started UT99 training map and let it run for ~15 minutes at 1280x1024 2xAA. All tests were performed on Athlon 600, MSI K7 Pro board (AMD 751 chipset), 256 MB SDRAM

Pro tips:
UT99 - bad benchmark to heat up GPU, due to being extremely CPU heavy game.
2x AA and high resolution - limitations of the ROP performance and memory bandwidht are not optimal for full GPU load.

Conclusion: needs better CPU and GPU heavy game at low resolution without AA.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 19 of 38, by ultra_code

User metadata
Rank Oldbie
Rank
Oldbie
havli wrote:
Ok, here are my numbers measured on V4 4500 AGP: […]
Show full quote

Ok, here are my numbers measured on V4 4500 AGP:

1. 3dfx / STB V4 4500 AGP with original heatsink and fan = 65 °C
2. 3dfx / STB V4 4500 AGP with original heatsink and Sunon Maglev fan = 60 °C
3. 3dfx / STB V4 4500 AGP with Zalman VF900 = 50 °C
4. PowerColor EvilKing IV with original heatsink and fan = 67 °C
5. PowerColor EvilKing IV with original heatsink and fan and 80mm fan blowing directly at the card = 69 °C

Here is link to the Sunon fan - it is one of the few 40x40x6mm fans available.... which means it can be used as a direct replacement of the original AAvid including the screws.
https://eu.mouser.com/ProductDetail/Sunon/MF4 … we3wNSdxSk5Fw==

The Zalman VF900 was originally mounted on Radeon X800 GTO and it is some kind of OEM version. However it should be identical to regular VF900. I used Arctic MX4 thermal paste. And fan was connected directly to MB fan header - which means 12V, full power.

Temparature was measured using infra-thermometer at the back side of the PCB, directly under the VSA chip. Before the measuring, I started UT99 training map and let it run for ~15 minutes at 1280x1024 2xAA. All tests were performed on Athlon 600, MSI K7 Pro board (AMD 751 chipset), 256 MB SDRAM, win XP and using open benchtable with no extra cooling (except case 5).

Just out of curiosity I added PowerColor Voodoo4 also and it seems to perform slightly worse than reference 3dfx V4. Also it is interesting adding extra fan next to the card actually makes the cooling worse. Perhaps this is due to having optimal cooling conditions as it is on the open benchtable and when running the V4 in regular case it could be the opposite.

So as a conclusion - bigger heatsink definitely makes a difference, however the risk of irreparable damage is still here. Perhaps using only a better performing fan could be a good compromise. I am using pair of similar Sunons (I don't remember the specs, bought the 10 years ago 🤣 ) on my V5 AGP and the work reasonably good. http://hw-museum.cz/vga/62/3dfx-voodoo5-5500-agp

I guess on Voodoo5 temperatures might be few degrees higher as there are two chips crowded in small area. Perhaps tomorrow I could measure also V5 PCI with AAvids and V5 AGP with Sunons. But I won't disassemble my main V5 system, so it might not be completely accurate - V5 AGP in AXP 3000+ system and V5 PCI in this Athlon 600.

YAY! Awesome work! Finally, some hard numbers. 😀

I'm actually kinda relieved that the numbers aren't that insane. Worried the Voodoo was going to get into the 70s or something like that, but the 60s is perfectly fine. Plus that fan replacement is truly a great idea. Obviously, though, that aftermarket cooler did do the best, but a 5C difference from the new-fan replacement is, I think, not worth the risk.

But, more testing is always appreciated, as more data is better data. Keep up the good work!

The Serpent Rider wrote:
Pro tips: UT99 - bad benchmark to heat up GPU, due to being extremely CPU heavy game. 2x AA and high resolution - limitations of […]
Show full quote

I started UT99 training map and let it run for ~15 minutes at 1280x1024 2xAA. All tests were performed on Athlon 600, MSI K7 Pro board (AMD 751 chipset), 256 MB SDRAM

Pro tips:
UT99 - bad benchmark to heat up GPU, due to being extremely CPU heavy game.
2x AA and high resolution - limitations of the ROP performance and memory bandwidht are not optimal for full GPU load.

Conclusion: needs better CPU and GPU heavy game at low resolution without AA.

Good to know, although I don't see why testing at lower resolutions and quality settings would put more load on the GPU. If anything, that's reducing the load on the GPU and thus not pushing the card to its limits.

Builds
ttgwnt-6.png
kcxlg9-6.png