GPU client showing low PPD

[APOLLO]

I have an 8800GT card that is processing very low compared to the other 3 cards in this system. The PPD is about 1500 lower than normal for a 450 pointer. Temperatures are reported very high for this card (107deg) however, nothing makes sense about the readings. First, the fan is working at over 2000 RPM and I removed it twice from the system to blow dust out of it. There are two 120mm fans blowing air on all the cards in this system for additional cooling. Ambient temperatures are pretty much the same for all the cards. Riva Tuner shows that performance 3D is set for this card. What could be the problem?

 

[sirmonkey1985]

card could be dying if its hitting 107C.. or the heatsink needs to be remounted.. but i have a feeling the cards throttling due to the high temps.. even though rivatuner shows it at 3D clocks try using gpu-z to see if its throttling..

 

[APOLLO]

card could be dying if its hitting 107C.. or the heatsink needs to be remounted.. but i have a feeling the cards throttling due to the high temps.. even though rivatuner shows it at 3D clocks try using gpu-z to see if its throttling..

OK, I figure the low PPD and high temps could very well be related. I dropped the shader clocks to default and the temps did not drop a single degree. Moreover, the fan speed reported by RT is nearly the same as the other cards. It is very weird. I'd think that the fan speed would drop or fluctuate if that was the problem, and that's what happened to another card last year until I replaced the HSF. Could a sudden increase in heat still happen if the HSF components and other temperature factors were normal?

 

[sirmonkey1985]

as far as i know the fan does not change on the card if you are using stock cooling.. the lowest it will go is 30% and it wont increase to 100% til you get near the thermal limit or it may not change at all.. i never used the stock cooler on my 8800GT.. day i got the card i replaced it with a duorb cooler.. i know my GTX 260 fan doesnt change at all it sits at 40% until my system BSOD's..

the last resort in all of this would be to bake the card since this is pretty much the symptoms of a card on its way out and is usually fixed by baking it.. but screw around with the other options before doing that..

 

[APOLLO]

as far as i know the fan does not change on the card if you are using stock cooling.. the lowest it will go is 30% and it wont increase to 100% til you get near the thermal limit..

I have all my fans at 100% no matter what season it is. Can the temperature increase that much even if the cooling components are operating normally and all other cards in the system are not affected? I have a reading of 2065 RPM for the fan. I don't believe the card is throttling at all. I have seen cards throttle before and the PPD fluctuated wildly. This card is just running at a steady PPD but it is far lower than normal. When I reduced the clocks to stock frequencies, the PPD did not lower, but remained the same. There was no decrease in reported temps either. I am thinking that the temp sensor is defective, but how can this affect performance if there doesn't appear to be any throttling?

 

[Hallis]

Maybe a false reading? Would never hurt to pull the heat sink off and put some fresh thermal compound on.

 

[APOLLO]

I installed GPU-Z and the temp reading is nearly identical to RT. I restarted the system and the GPU client seems to be operating at full speed despite the high temperatures being reported by two monitors. I really do believe these are false readings. If the card was indeed failing, I would see EUEs galore since it happened to me before on two occasions. I have not had an EUE in a long time on this card. There is something else that had affected its performance.

How do I tell if the card is throttling?

 

[Mr. Pedantic]

Open up the hardware monitor on Rivatuner. It should show on your clocks, but also look at the % load.

 

[lehmann]

just for the hell of it swap slots. update drivers.

 

[Nathan_P]

I have all my fans at 100% no matter what season it is. Can the temperature increase that much even if the cooling components are operating normally and all other cards in the system are not affected? I have a reading of 2065 RPM for the fan. I don't believe the card is throttling at all. I have seen cards throttle before and the PPD fluctuated wildly. This card is just running at a steady PPD but it is far lower than normal. When I reduced the clocks to stock frequencies, the PPD did not lower, but remained the same. There was no decrease in reported temps either. I am thinking that the temp sensor is defective, but how can this affect performance if there doesn't appear to be any throttling?

Your card has a problem somewhere, a 8800gt fan runs at 3000rpm at full tilt even if it has a non reference cooler. Its either clogged with dust or you have a problem in the driver set up somewhere. I recently had an issue with my GTS250 where the fan was stuck on 38% no matter what i set it too in the control panel. I damn near fried the card and ultimately had to pull it for a while. There may be a simplier cure such as removing the drivers, running driver sweeper and reinstalling but i had other issues and ended up nuking the windows install.
If its dust just pull the card, remove th shroud and remove all the dust using tweezers and compressed air.
Folding under full load with the dreaded 511 pointers last year on my 3 8800gt's in a single box i never saw temps above 88deg C on the middle card.
If you can, get some additional colling into the box, preferably pulling heat straight off the top of the cards.

HTH

 

[FrEaKy]

BAKE IT!

 

[APOLLO]

Your card has a problem somewhere, a 8800gt fan runs at 3000rpm at full tilt even if it has a non reference cooler. Its either clogged with dust or you have a problem in the driver set up somewhere.

The latter is very possible. It is definitely not clogged see below. All my single-slot cards have fans running just over 2000 RPM when new, not 3000 RMP. Are you referring to dual-slot cards perhaps?

I recently had an issue with my GTS250 where the fan was stuck on 38% no matter what i set it too in the control panel. I damn near fried the card and ultimately had to pull it for a while. There may be a simplier cure such as removing the drivers, running driver sweeper and reinstalling but i had other issues and ended up nuking the windows install.

There are many possibilities and I'm trying to investigate them all but it will take time. This system (Skulltrail) has a checkered history from very low CPU performance in folding clients to weird memory amount recognition, and not to mention issues rebooting, etc. It might be the motherboard itself that's at the center of my issues. I do have 4 GPUs in it after all..

If its dust just pull the card, remove th shroud and remove all the dust using tweezers and compressed air.

I removed the card twice in a week to blow dust from it. When I redid it yesterday there was hardly any dust that had accumulated inside. If there's a real heat issue then it's due to other factors that are yet to be determined.

Folding under full load with the dreaded 511 pointers last year on my 3 8800gt's in a single box i never saw temps above 88deg C on the middle card.
If you can, get some additional colling into the box, preferably pulling heat straight off the top of the cards.

The card has to be misreporting the temps. There are two 120mm fans blowing air on it and it is not throttling. Last year, I had another card that was identical make and model and it was throttling when it hit 99-100 degrees. This card doesn't throttle at a reported 107-108. It's not as hot as it is reporting.

The fact that I see no drop in temp reading when it is dropped to factory clocks is further evidence there's a problem with the readings. Also, as I posted above, the PPD reports as steady albeit low. When I experienced throttling in the other card last year, the PPD was fluctuating like a yo-yo. Anyway, I might reseat or install another HSF anyway, just to eliminate that possibility.

BAKE IT!

LOL, the number of cards to bake in my storage bin is piling up. So far, there's one dead card (no display), one card that displays but EUEs immediately and this card which is working, for the most part...

 

[Nathan_P]

The latter is very possible. It is definitely not clogged see below. All my single-slot cards have fans running just over 2000 RPM when new, not 3000 RMP. Are you referring to dual-slot cards perhaps?...

No, i've had both single and dual slot leadteks and they all read 3000Rpm at full bore

There are many possibilities and I'm trying to investigate them all but it will take time. This system (Skulltrail) has a checkered history from very low CPU performance in folding clients to weird memory amount recognition, and not to mention issues rebooting, etc. It might be the motherboard itself that's at the center of my issues. I do have 4 GPUs in it after all..?

It could be the M/B, as i've run 3 without issue and others have run 4 without problems, i do know from past posts that skulltrail never really lived up to expectations

I removed the card twice in a week to blow dust from it. When I redid it yesterday there was hardly any dust that had accumulated inside. If there's a real heat issue then it's due to other factors that are yet to be determined.

OK dust is not a factor

The card has to be misreporting the temps. There are two 120mm fans blowing air on it and it is not throttling. Last year, I had another card that was identical make and model and it was throttling when it hit 99-100 degrees. This card doesn't throttle at a reported 107-108. It's not as hot as it is reporting.

its possible that the sensor is misreporting - the only suggestion i have is to try several different monitoring programs to see if another one reports different temps

The fact that I see no drop in temp reading when it is dropped to factory clocks is further evidence there's a problem with the readings. Also, as I posted above, the PPD reports as steady albeit low. When I experienced throttling in the other card last year, the PPD was fluctuating like a yo-yo. Anyway, I might reseat or install another HSF anyway, just to eliminate that possibility.

LOL, the number of cards to bake in my storage bin is piling up. So far, there's one dead card (no display), one card that displays but EUEs immediately and this card which is working, for the most part...

Reseating is worth a shot.
Dumb question but have you checked the nvidia control panel just to make sure that SLI and Physix are actually disabled

 

[APOLLO]

No, i've had both single and dual slot leadteks and they all read 3000Rpm at full bore

The only single-slotters I ever acquired where XFX and BFtech. Both manufacturers seems to have their fans running 2100-2200 RPM at full speed on their single-slot 8800GT cards according to RT. I would love if I can get the fan speed up to 3000 PPD. Stability and OC would be higher.

 

[Nathan_P]

The only single-slotters I ever acquired where XFX and BFtech. Both manufacturers seems to have their fans running 2100-2200 RPM at full speed on their single-slot 8800GT cards according to RT. I would love if I can get the fan speed up to 3000 PPD. Stability and OC would be higher.

Hmmm i wasn't aware of a difference, i thought that all single slots cards used the reference cooler but i must be mistaken.

Just had a thought, when i was having heat issues with my 250 i used the cpu usage slider (systray client) to lower the CPU usage to 70-80% - it reduced the heat enough that my card could run safely at 5.5k PPD. It may be worth playing around with that to see if you get an improvement

 

[APOLLO]

Hmmm i wasn't aware of a difference, i thought that all single slots cards used the reference cooler but i must be mistaken.

The reference cooler is different with a much more flimsy fan, IIRC. Unless there's an error with the RT monitoring, I have always seen those RPM figures and I ran more than a half dozen of these cards from both manufacturers. These two were by far the most commonly available in my area and the EVGA version but I never purchased the EVGA singles. I do have dual-slot 8800GTs from EVGA but those are entirely different cards with totally different stock clocks, etc.

Just had a thought, when i was having heat issues with my 250 i used the cpu usage slider (systray client) to lower the CPU usage to 70-80% - it reduced the heat enough that my card could run safely at 5.5k PPD. It may be worth playing around with that to see if you get an improvement

Unfortunately, I use the console client exclusively. The other 3 cards are working perfectly in this system, so it must be something else possibly hardware related to the card itself.

 

[APOLLO]

Just had a thought, when i was having heat issues with my 250 i used the cpu usage slider (systray client) to lower the CPU usage to 70-80% - it reduced the heat enough that my card could run safely at 5.5k PPD. It may be worth playing around with that to see if you get an improvement

You may be on to something with CPU usage. I stopped one of my other r clients for several minutes and the GPU client in question jumped to nearly normal performance levels. This almost disproves the high temperature/throttling possibility. It is very possible that 4 clients assigned to one core might be excessive depending on the GPU WU. I will need to do more exhaustive testing to ascertain all the dynamics involved here. Too many variables to sort out as of yet. There's still no good explanation for the temp reading though, even if there turns out to be something like a conflict.

 

[Nathan_P]

You may be on to something with CPU usage. I stopped one of my other r clients for several minutes and the GPU client in question jumped to nearly normal performance levels. This almost disproves the high temperature/throttling possibility. It is very possible that 4 clients assigned to one core might be excessive depending on the GPU WU. I will need to do more exhaustive testing to ascertain all the dynamics involved here. Too many variables to sort out as of yet. There's still no good explanation for the temp reading though, even if there turns out to be something like a conflict.

What cpu are you using to driver these 8800gt's, it sounds like something somewhere is bottlenecking the card to some degree. Another thought what M/B are you running?

 

[APOLLO]

What cpu are you using to driver these 8800gt's, it sounds like something somewhere is bottlenecking the card to some degree. Another thought what M/B are you running?

Current processors are dual E5110 Woodcrests @2.5GHz and motherboard is Intel D5400XS.

 

[Nathan_P]

Current processors are dual E5110 Woodcrests @2.5GHz and motherboard is Intel D5400XS.

Well cpu power shouldn't be a problem.
If the card runs are near normal spped if another card is removed and allows the card i question to run at nearly full speed then i would suggest an issue with the motherboard.
Perhaps a bios update may solve the problem but i think that the board is struggling to cope with all 4 PCIe slots populated.

 

[APOLLO]

Well cpu power shouldn't be a problem.
If the card runs are near normal spped if another card is removed and allows the card i question to run at nearly full speed then i would suggest an issue with the motherboard.
Perhaps a bios update may solve the problem but i think that the board is struggling to cope with all 4 PCIe slots populated.

I'm leaning towards that possibility although it won't explain the temp readings, but that could be a different issue such as a malfunctioning sensor. As of the last 12 hours, the card is working normally at full PPD despite the apparent temps. What has changed is the type of WUs that I received in that time.