GTX 1050 TI with only 55k PPD?

[AnimalFriend]

Hello everybody,

the PPD of my (fairly old) desktop seem to be much lower than expected (according to this forum) for an nVidia GTX 1050 TI card, since I only get 55k PPD from the GPU, while I have seen numerous posts in this forum claiming something around 200k.

I have:

• old intel i5-760 @ 2,8 GHz
• 8 GB memory
• new MSI 1050 TI w/ 4GB

I don't think there is a bottleneck on the PCIe 2.0 bus, as bus load is reported at 22%.

Windows 10 task manager reports 99% CUDA usage when an OpenMM 2.2 core is running, which also fully utilizes 1 core then.

I thought maybe it's because the CPU is too slow (since the GPU is practiacally only a vector processor in this scenario as I understand it), but then the CUDA utilization should not be maxed out IMHO.

I am running the 7.6.9 FaH client.

Any ideas? (besides "Duh man, get a decent computer!" )

Best regards
Animal

[foldy]

Have you set a passkey yet to get the bonus points?
https://foldingathome.org/support/faq/points/passkey/

[TPL]

Cuda is running on CPU...

[AnimalFriend]

Have you set a passkey yet to get the bonus points?
https://foldingathome.org/support/faq/points/passkey/

OMG - that was both embarassingly simple and stupidly obvious.

So with that changed, the estimated PPD jumped to 190k immediately.

Thx a lot man!
(At least my old box is still working as expected then )

[HaloJones]

The bonus will not apply immediately. You need to have successfully completed ten units to then qualify but that won't take long.

[PantherX]

Welcome to the F@H Forum AnimalFriend,

While I am glad that the issue was a simple fix by getting the passkey, please note that the usage of PCIe lane along can be slightly deceptive. Here's the therotical maximum speed for PCIe lanes:
PCIe 2 x16: 8 GB/s
PCIe 3 x16: 15.75 GB/s
PCIe 4 x16: 31.5 GB/s

So lets say that the PCIe lane is showing 22% usage. That means:
PCIe 2: 6.24 GB/s
PCIe 3: 12.285 GB/s
PCIe 4: 24.57 GB/s

As you can see above, the data transmission is significantly different based on what generation it is even though the reporting tools only show 22% usage.

When it comes to folding performance, check to see if the GPU Usage is >95% for most of the time. If it drops down to 0% and the CPU Usage spikes up, then it is normal as it is writing a checkpoint and performing sanity checks. However, if your GPU Usage is <95% then it could mean that your GPU is being starved due to bottlenecks or the WU is a small one. The small WU issue is mostly for high-end GPUs so may not be a factor in low to medium GPUs.