GPU client beating the crap out of my graphics card

A forum just for SPCR's folding team... by request.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
haysdb
Patron of SPCR
Posts: 2425
Joined: Fri Aug 29, 2003 11:09 pm
Location: Earth

GPU client beating the crap out of my graphics card

Post by haysdb » Thu May 07, 2009 10:03 pm

I have the clocks on one of my GeForce GTS 250 graphics cards turned down as low as the EVGA Precision tool will allow, and it's still running 81 degrees and the fan at 83%.

I am using the just released drivers, the official Windows 7 drivers from NVidia. I'm seeing some strong PPD numbers, but don't want to deal with this kind of heat and noise 24x7.

Suggestions, other than rolling back to an earlier set of drivers, not running FAH, or turning my AC down to about 70?

garacs1
Posts: 31
Joined: Tue Jul 04, 2006 11:48 pm
Location: Italy

Re: GPU client beating the crap out of my graphics card

Post by garacs1 » Fri May 08, 2009 1:34 am

haysdb wrote:I have the clocks on one of my GeForce GTS 250 graphics cards turned down as low as the EVGA Precision tool will allow, and it's still running 81 degrees and the fan at 83%.

I am using the just released drivers, the official Windows 7 drivers from NVidia. I'm seeing some strong PPD numbers, but don't want to deal with this kind of heat and noise 24x7.

Suggestions, other than rolling back to an earlier set of drivers, not running FAH, or turning my AC down to about 70?
I know that some could say this is not an option, but have you considered going liquid cooling? Even the worst and cheapest waterblock in the market will make your card stay cooler, and as per the radiator fan, well, a silent 120mm fan is easy to find, even searching on SPCR's fan review database.

Plus, if you go the step further, you could easily cool the whole system down (CPU, mosfets, NB ecc) with a dedicated liquid cooling system. There are millions of solutions on the net.

My 2 cents.

Garacs1

PScal
Posts: 1
Joined: Wed Mar 04, 2009 5:55 am
Location: Little Rock, AR

Post by PScal » Fri May 08, 2009 6:00 am

I have fitted an aftermarket cooler on my GTS 250 and noticed a significant decrease in temperature and noise compared to the stock one. With the cooler I chose (Zalman VF 1000) my idle temps are 36-38c with the fan turned to low. While folding (windows 7, nvidia drivers, gpu2 client) the temp goes no higher than 59c (fan turned up to 100%). The fan is whinier than the stock cooler but it's much quieter and less hair dryer sounding.

The VF 1000 did not come with enough heat sinks to cover the voltage regulators so I also bought the Thermalright HR-03 GTX 285/260 Heatsink Pack.

I am getting roughly 5k ppd with my gts 250, but it depends on the size of the work unit. How many are you getting?

haysdb
Patron of SPCR
Posts: 2425
Joined: Fri Aug 29, 2003 11:09 pm
Location: Earth

Post by haysdb » Fri May 08, 2009 6:41 am

I have no interest in water cooling, but thanks for the suggestion.

I have an Arctic Cooling Twin Turbo that I could put on it. I bought it for an 8800GT but that card did not offer PWM fan control so I didn't bother and replaced that card and a passively cooled 9600GT with a pair of GTS 250s.

There isn't enough space below the second card for an aftermarket cooler, but something weird about all of this is that it's only the top card that I'm having an issue with. The second/lower card is running fine, at 66C at 60% fan. The top card, even with the clocks turned all the way down, is running about 10 degrees hotter with a higher fan speed. It makes me wonder if the thermal bond between the GPU and heatsink has degraded.

Points-wise, the numbers are all over the map depending on the WU. If I calculate points based on "All frames," FahMon reports 5484 PPD for one card and 5724 for the other.

I think it's time to contact EVGA support to see what they say. It's looking to me like a hardware failure of the cooling unit.

I think I will stop both GPU clients and swap the WU assignments, then check the temperatures again. That will tell me whether the specific Work Units are a factor.

garacs1
Posts: 31
Joined: Tue Jul 04, 2006 11:48 pm
Location: Italy

Post by garacs1 » Fri May 08, 2009 7:22 am

haysdb wrote:I have no interest in water cooling, but thanks for the suggestion.

I have an Arctic Cooling Twin Turbo that I could put on it. I bought it for an 8800GT but that card did not offer PWM fan control so I didn't bother and replaced that card and a passively cooled 9600GT with a pair of GTS 250s.

There isn't enough space below the second card for an aftermarket cooler, but something weird about all of this is that it's only the top card that I'm having an issue with. The second/lower card is running fine, at 66C at 60% fan. The top card, even with the clocks turned all the way down, is running about 10 degrees hotter with a higher fan speed. It makes me wonder if the thermal bond between the GPU and heatsink has degraded.

Points-wise, the numbers are all over the map depending on the WU. If I calculate points based on "All frames," FahMon reports 5484 PPD for one card and 5724 for the other.

I think it's time to contact EVGA support to see what they say. It's looking to me like a hardware failure of the cooling unit.

I think I will stop both GPU clients and swap the WU assignments, then check the temperatures again. That will tell me whether the specific Work Units are a factor.
Well before contacting EVGA support, consider that if the upper card runs 10°C hotter than the lower card, much probably it's because the lower card does not let enough freash air to be sucked by the upper card heatsink fan. Basically, while the lower card gets fresh air, the upper card gets air that is heated up by the lower card, which does get hot even in the back side. Consider eventually to put a fan to blow fresh air directly on the side of both cards, from the side panel. Of course, without having an idea of your case and components installed in it, it's hard to give a better idea.

Regards

Garacs1

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Fri May 08, 2009 7:48 am

Makes me wonder what would happen if you removed the stock cooler and used Artic Silver when you reseat it.

I'm sure they chose a cheaper cooler, but I wonder if bad installation isn't also affecting your temps.

AZBrandon
Friend of SPCR
Posts: 867
Joined: Sun Mar 21, 2004 5:47 pm
Location: Phoenix, AZ

Post by AZBrandon » Fri May 08, 2009 9:11 am

garacs1 wrote:Well before contacting EVGA support, consider that if the upper card runs 10°C hotter than the lower card, much probably it's because the lower card does not let enough freash air to be sucked by the upper card heatsink fan. Basically, while the lower card gets fresh air, the upper card gets air that is heated up by the lower card, which does get hot even in the back side. Consider eventually to put a fan to blow fresh air directly on the side of both cards, from the side panel. Of course, without having an idea of your case and components installed in it, it's hard to give a better idea.
Agreed. Back when I ran two cards, the temps were something like this with all solid slot covers in place:

Top card: 85*
Bottom card: 65*

I removed the slot covers (see picture) and the split dropped to only 10 degrees instead of 20 degrees. Like garacs1 said, heat rises, and heat comes off even the back of the PCB, which is something that you can't do much about besides opening up slot covers to try to get the case fans to draw more air in via that path. With the evga, I would hope you could get the temperature split even lower since they exhaust at least half the heat out the back, as opposed to mine which dump all the heat into the case.

If you want to be SURE it is a positional problem and not a physical card problem, that's a piece of cake to test: just reverse the two cards positions. If the top card is STILL the hotter card, then you know there's nothing wrong with the cards, and it's all about the position. Alternately, just remove the bottom card and see what happens to the temps on the top card.

Image

haysdb
Patron of SPCR
Posts: 2425
Joined: Fri Aug 29, 2003 11:09 pm
Location: Earth

Post by haysdb » Fri May 08, 2009 10:37 am

Good input. There is one slot separating the cards, but it does make sense that the lower card could be heating the upper one. OTOH, I didn't have this issue until recently. The upper card may be have run a wee bit warmer, but nothing like this.

I may try taking the case apart and re-seating the stock cooler.

First, I may try just removing the lower card. Actually, first I may just stop FAH on the lower card and see how much the temperature of the upper card is effected.

Thanks for the suggestions.


Edit: swapping the cards is a great suggestion.

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Fri May 08, 2009 8:02 pm

AZBrandon, forgive the newbish question, but what kind of cooler is that on your cards in the picture? Looks almost like a water block.

AZBrandon
Friend of SPCR
Posts: 867
Joined: Sun Mar 21, 2004 5:47 pm
Location: Phoenix, AZ

Post by AZBrandon » Sat May 09, 2009 8:42 am

KansaKilla wrote:AZBrandon, forgive the newbish question, but what kind of cooler is that on your cards in the picture? Looks almost like a water block.
It's a Zalman VF-2000 and looks copper, but since there's no VF-2000 CU listed on the Zalman website, I suspect it's just anodized aluminum.

Zalman VF-2000 info page

I run mine undervolted to 5v and it's enough to keep my 9800GTX+ within reasonable temperatures when folding. As mentioned, the only big disadvantage over the dual-slot coolers is that it just dumps the heat into the case, where it goes up into the CPU which is directly above it.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Fri May 29, 2009 5:10 am

AZBrandon wrote:
KansaKilla wrote:AZBrandon, forgive the newbish question, but what kind of cooler is that on your cards in the picture? Looks almost like a water block.
It's a Zalman VF-2000 and looks copper, but since there's no VF-2000 CU listed on the Zalman website, I suspect it's just anodized aluminum.

Zalman VF-2000 info page

I run mine undervolted to 5v and it's enough to keep my 9800GTX+ within reasonable temperatures when folding. As mentioned, the only big disadvantage over the dual-slot coolers is that it just dumps the heat into the case, where it goes up into the CPU which is directly above it.
Actually, those are VF830, which is OEM only. The fins are made of aluminium, if you compare the color of the copper heatpipes and the fins you'll find the latter to be too much yellow. It's available in different colors, look here.
They're pretty much a budget version of the retail VF950, link.

The VF2000 CPU/GPU heatsink is 45 mm high and takes up two slots, and have four heatpipes.

I'm looking for a VF2000 review, but I've only seen the one where it's compared to a stock Intel cooler and a VF830. I'd like to see a comparison with the VF1000.

haysdb
Patron of SPCR
Posts: 2425
Joined: Fri Aug 29, 2003 11:09 pm
Location: Earth

Post by haysdb » Fri May 29, 2009 7:02 am

Update: Confirming that the upper card is running hotter because of it's location. I swapped the cards and the temperatures remained constant, i.e. the upper card running several degrees warmer than the lower card.

For the summer I have little choice but to turns the clocks down. I've turned the clocks on the upper card as far down as the EVGA Precision tool will allow, and fiddle with the clocks on the lower card but they are generally about 10% higher.

With the fans manually set at 73%, the temperature reported for the upper card is 80, and 78 for the lower card. I have been trying to keep them at a maximum of 80 degrees, just because that seems like a reasonable round-number to shoot for.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Sat May 30, 2009 6:03 am

I found in the long term my temps kept rising. I took the cover off one of my 8800GSs and was shocked at how much dust collected in it. I cleaned it out and temps dropped back to where they had started. It took about a year.

Given that room temps have risen in the month since I purchased my GTX260, and the card has risen accordingly. Its now 69C, which is OK.

I had lowered my OCing during warm periods and even shut off pcs during major heat spells. I also agree anything below 80C should be OK, but found stock clocks at higher temps more reliable than OCd ones.

haysdb
Patron of SPCR
Posts: 2425
Joined: Fri Aug 29, 2003 11:09 pm
Location: Earth

Post by haysdb » Sat May 30, 2009 7:54 am

I fiddled around with overclocking in the beginning but I came to feel it wasn't worth the trouble, and the risk of trashing work.

Tobias
Posts: 530
Joined: Sun Aug 24, 2003 9:52 am

Post by Tobias » Sat May 30, 2009 10:22 am

Have you tried ducting? If you have two open slots between the cards you could create a cardboard duct creating an airpath between the heatsink and the outside of the case pushing the heat directly outside and create a duct for cool air to the fan.

I have a hard time believing that the lower card would heat up the upper card simply by proximity/radiation. It must be a question of the lower card restricting airflow to the upper card, or reducing the quality (temperature) of the air of the upper card.

haysdb
Patron of SPCR
Posts: 2425
Joined: Fri Aug 29, 2003 11:09 pm
Location: Earth

Post by haysdb » Sat May 30, 2009 11:06 am

Hot air rises. Makes sense to me. I'm just surprised at the degree at which the upper card seems to be heated by the lower card.

The positive-pressure case design should ensure that there is good airflow through and between the cards.

Post Reply