Overclock.net › Forums › Overclockers Care › Overclock.net Folding@Home Team › GPU not fully utilized folding, (again)
New Posts  All Forums:Forum Nav:

GPU not fully utilized folding, (again) - Page 3

post #21 of 27
Any clue on this topic? I just managed to face same issue, one of the nvidia perform average 78% core load only.

I bought 2 nvidia cards, 1050ti and 1070. I have tested both separately in my i7 7700 @ 3.60GHz, with good estimated PPD: 1050ti = 150k, 1070 = 600k without any OC yet, both cards > 95% core load.

Then I manage to put both cards into my mining rig (old computer: gigabyte EP43-UD3L with Core2Quad Q9550 @ 2.83GHz) with PCIe riser. Running Windows 10 as well, 1050ti = 120k (slightly lower but still acceptable) at > 95% core load, but the problem is 1070 = 400k with only ~78% core load. But the weird issue is I have full 100% core load for both cards by using ccminer or other miners.

I have tried some of the suggested solutions such as swapping PCIe slot position, power option to High Performance, but still no improvement frown.gif
post #22 of 27
Quote:
Originally Posted by tonywonlost View Post

Any clue on this topic? I just managed to face same issue, one of the nvidia perform average 78% core load only.

I bought 2 nvidia cards, 1050ti and 1070. I have tested both separately in my i7 7700 @ 3.60GHz, with good estimated PPD: 1050ti = 150k, 1070 = 600k without any OC yet, both cards > 95% core load.

Then I manage to put both cards into my mining rig (old computer: gigabyte EP43-UD3L with Core2Quad Q9550 @ 2.83GHz) with PCIe riser. Running Windows 10 as well, 1050ti = 120k (slightly lower but still acceptable) at > 95% core load, but the problem is 1070 = 400k with only ~78% core load. But the weird issue is I have full 100% core load for both cards by using ccminer or other miners.

I have tried some of the suggested solutions such as swapping PCIe slot position, power option to High Performance, but still no improvement frown.gif

 

I don't know about mining, but for F@H on Nvidia cards, a CPU thread runs at 100% for each GPU. On newer CPUs the CPU thread is mostly waiting for the GPU to complete its tasks, so the folding performance depends solely on the GPU. For an older, lower-clocked CPU I suspect that the CPU thread ends up being the bottleneck if the GPU is fast enough. IOW your Q9550 is able to keep up with the 1050ti, but it is bottlenecking the 1070.

Neurosis
(7 items)
 
Monochrome
(28 items)
 
Patchouli
(11 items)
 
CPUMotherboardGraphicsGraphics
i7 5820K Gigabyte GA-X99 Gaming G1 EVGA GTX-980ti EVGA GTX-980ti 
RAMHard DriveOS
Corsair DDR4 32GB Samsung 950 Pro NVME Windows 10 Pro x64 
CPUMotherboardGraphicsGraphics
i7 2600K @4.5GHz Gigabyte P67A-UD7 B3 EVGA GTX 780 Ti SC 3GB EVGA GTX 780 Ti SC 3GB 
RAMHard DriveOptical DriveCooling
2x4GB Super Talent DDR3-1600 CL9 Western Digital Caviar Black WD1001FALS LiteON iHBS212 EK Supreme HF Copper/Acetal 
CoolingCoolingCoolingCooling
4x Noiseblocker NB-BlackSilentPro PLPS 120mm PW... Bitspower Water Tank Z-Multi 150 Aquacomputer Flow Rate Sensor "High Flow USB" G1/4 Bitspower black compression fittings 
CoolingCoolingCoolingCooling
Tygon silver tubing 0.5" ID 0.75" OD 2x Watercool HEATKILLER® GPU-X³ GTX TITAN/780/7... Alphacool HF D5 acetal pump top Aquacomputer D5 pump (USB/Aquabus) 
CoolingCoolingOSOS
EK CoolStream XTC 420 6x Noiseblocker NB-BlackSilentPro PK-PS 140mm P... Windows 7 x64 Pro Arch Linux x64 
MonitorKeyboardPowerCase
Hewlett Packard LP2475W Ducky Shine II Seasonic X750 Lian Li PC-A77A Silver 
MouseAudioAudio
Logitech MX Revolution RME Fireface 800 Dynaudio BM6A 
CPUMotherboardGraphicsGraphics
i7 2600K Gigabyte GA-Z77X-UD5H Intel HD3000 EVGA GTX980 
RAMHard DriveHard DriveCooling
Corsair Vengeance Western Digital Caviar Black WD1600JD Western Digital WD20EARS (2), WD20EARX (1), WD2... Thermalright HR-02 
OSPowerCase
Ubuntu Server 12.04 LTS x64 Seasonic X760 Lian Li PC-A77F 
  hide details  
Reply
Neurosis
(7 items)
 
Monochrome
(28 items)
 
Patchouli
(11 items)
 
CPUMotherboardGraphicsGraphics
i7 5820K Gigabyte GA-X99 Gaming G1 EVGA GTX-980ti EVGA GTX-980ti 
RAMHard DriveOS
Corsair DDR4 32GB Samsung 950 Pro NVME Windows 10 Pro x64 
CPUMotherboardGraphicsGraphics
i7 2600K @4.5GHz Gigabyte P67A-UD7 B3 EVGA GTX 780 Ti SC 3GB EVGA GTX 780 Ti SC 3GB 
RAMHard DriveOptical DriveCooling
2x4GB Super Talent DDR3-1600 CL9 Western Digital Caviar Black WD1001FALS LiteON iHBS212 EK Supreme HF Copper/Acetal 
CoolingCoolingCoolingCooling
4x Noiseblocker NB-BlackSilentPro PLPS 120mm PW... Bitspower Water Tank Z-Multi 150 Aquacomputer Flow Rate Sensor "High Flow USB" G1/4 Bitspower black compression fittings 
CoolingCoolingCoolingCooling
Tygon silver tubing 0.5" ID 0.75" OD 2x Watercool HEATKILLER® GPU-X³ GTX TITAN/780/7... Alphacool HF D5 acetal pump top Aquacomputer D5 pump (USB/Aquabus) 
CoolingCoolingOSOS
EK CoolStream XTC 420 6x Noiseblocker NB-BlackSilentPro PK-PS 140mm P... Windows 7 x64 Pro Arch Linux x64 
MonitorKeyboardPowerCase
Hewlett Packard LP2475W Ducky Shine II Seasonic X750 Lian Li PC-A77A Silver 
MouseAudioAudio
Logitech MX Revolution RME Fireface 800 Dynaudio BM6A 
CPUMotherboardGraphicsGraphics
i7 2600K Gigabyte GA-Z77X-UD5H Intel HD3000 EVGA GTX980 
RAMHard DriveHard DriveCooling
Corsair Vengeance Western Digital Caviar Black WD1600JD Western Digital WD20EARS (2), WD20EARX (1), WD2... Thermalright HR-02 
OSPowerCase
Ubuntu Server 12.04 LTS x64 Seasonic X760 Lian Li PC-A77F 
  hide details  
Reply
post #23 of 27
Quote:
Originally Posted by valvehead View Post

I don't know about mining, but for F@H on Nvidia cards, a CPU thread runs at 100% for each GPU. On newer CPUs the CPU thread is mostly waiting for the GPU to complete its tasks, so the folding performance depends solely on the GPU. For an older, lower-clocked CPU I suspect that the CPU thread ends up being the bottleneck if the GPU is fast enough. IOW your Q9550 is able to keep up with the 1050ti, but it is bottlenecking the 1070.

Thanks, I was suspecting CPU issue too, if this is the problem, then maybe let me OC the CPU abit and see if any improvement. I will update my status later.

PS: I also tried to install Ubuntu and run F@H on it, same problem but just slightly better...
post #24 of 27
Quote:
Originally Posted by tonywonlost View Post

Thanks, I was suspecting CPU issue too, if this is the problem, then maybe let me OC the CPU abit and see if any improvement. I will update my status later.

PS: I also tried to install Ubuntu and run F@H on it, same problem but just slightly better...

That motherboard only has a 2.0 16x slot so you'll lose a bit of performance from that. Plus probably some more by using a riser. I see the motherboard only has one 16x slot so the other card is def going to lose some from the 1x slot. Some WUs never loaded up my 1070 that well, esp Core18 tasks. Core21 was better.

https://www.newegg.com/Product/Product.aspx?Item=N82E16813128380
post #25 of 27
Quote:
Originally Posted by mmonnin View Post

That motherboard only has a 2.0 16x slot so you'll lose a bit of performance from that. Plus probably some more by using a riser. I see the motherboard only has one 16x slot so the other card is def going to lose some from the 1x slot. Some WUs never loaded up my 1070 that well, esp Core18 tasks. Core21 was better.

https://www.newegg.com/Product/Product.aspx?Item=N82E16813128380

I don't think this is the reason because the GPU core load is not fully utilized during folding, but it perform 100% on mining ETH or ZEC. Anyway, I might be free tomorrow to overclock my CPU clock and we will see if any improvement.
post #26 of 27
Quote:
Originally Posted by tonywonlost View Post

I don't think this is the reason because the GPU core load is not fully utilized during folding, but it perform 100% on mining ETH or ZEC. Anyway, I might be free tomorrow to overclock my CPU clock and we will see if any improvement.

Mining uses very little bandwidth and thats why it can be run with a 1x slot.
post #27 of 27
Quote:
Originally Posted by mmonnin View Post

Mining uses very little bandwidth and thats why it can be run with a 1x slot.

Just to update my status here, I have changed some configuration on my mining rig.

Both 1070 and 1050ti are connected through 1x slots with risers.
1st, I overclocked my CPU to 3.23MHz. And I can see significantly that 1050ti reached max at 100% (previously only 99% max), flactuating 80% to 100%, average >90%. Then I can't notice changes at 1070, maybe there is only little improvement. 2 core of CPU loaded 100% most of the time.

2nd, then I tried to put 1070 to x16 slot. 1070 reached max at 100%, flactuating 80% to 100%, most of the time at >90%. 2 core of CPU also loaded 100% most of the time.

So, conclusion is both CPU core speed and slot speed are important to folding coin (A pretty high requirement needed for mining). Thanks valvehead and mmonnin.

biggrin.gif
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Overclock.net Folding@Home Team
Overclock.net › Forums › Overclockers Care › Overclock.net Folding@Home Team › GPU not fully utilized folding, (again)