Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Nvidia Tesla P100 and V100 not working #205

Open
AugustJMote opened this issue Feb 14, 2018 · 7 comments
Open

Nvidia Tesla P100 and V100 not working #205

AugustJMote opened this issue Feb 14, 2018 · 7 comments

Comments

@AugustJMote
Copy link

I compiled ccminer with CUDA 9.1 and gcc 5.4.

It generates a simple "Out of memory error" and closes.
On the Tpuvrot version they both attempt to mine, but generate constant "nonce does not validate" errors.

I tried editing the make file, but I might not have done so correctly.

Any guidance would be appreciated, thank you

@KlausT
Copy link
Owner

KlausT commented Feb 14, 2018

The default intensity is probably too high. Try setting it manually.
What algo are you trying to mine?

@AugustJMote
Copy link
Author

Neoscrypt
I've tried intensities between 15 and 28, but I can try more shortly.

@KlausT
Copy link
Owner

KlausT commented Feb 14, 2018

Did it say something about cuda_neoscrypt.cu ?
Maybe it's not your cards but your system that's out of memory.
In this case changing the intensity wouldn't help.

@AugustJMote
Copy link
Author

It seems it was an issue with intensity after all. It seems odd considering how high the Tesla k80 liked intensity. After adjusting the intensity, the p100 could do 1.4 MH/s and the v100 could only seem to do 0.6 MH/s, which seems odd considering the Titan V can do 2.6, but it might be user error again. Thank you for your help.

@KlausT
Copy link
Owner

KlausT commented Feb 16, 2018

If those cards have the same high memory latency than the 1080Ti, they might be faster when we make a little code change.
After this line:

unsigned int cc = props.major * 10 + props.minor;

I could add this line:
if(cc==70 || cc==60) use_tpruvot=true;

After this change the intensity wouuld have to be adjusted again, probably.

@TheRealMrQ
Copy link

Will the equihash algo be affected in the same way, as to make a similar code change, for 1080ti/v/p100 cards?

@AugustJMote
Copy link
Author

Unless I misunderstand, the code change appears to just use tpuvrot's implementation of the neoscrypt miner. It wouldn't affect any other algo. The code didn't change much for the p100, but boosted the v100 to Titan V levels. If anyone has access to either of these devices, I'd probably recommend Ethereum mining since their HBM2 seems to benefit Eth massively while their low core clocks hurt everything else.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants