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

AMD Vega64 CN-GPU low auto intensity. #2517

Closed
Ryo-RU opened this issue Aug 23, 2019 · 1 comment
Closed

AMD Vega64 CN-GPU low auto intensity. #2517

Ryo-RU opened this issue Aug 23, 2019 · 1 comment

Comments

@Ryo-RU
Copy link
Contributor

Ryo-RU commented Aug 23, 2019

TLDR: When started V56 xmr-stak generates proper intensity in config. But V64 - not.
Algo: CN-GPU
Tested on 2 different devices on HiveOS and W10
[Imgur link]

Details:
This seem to appear on 2.10.7 (2.10.5 generates 1536 auto-intensity)
Auto intensity: 1024 (the rest is fine: worksize 16, memchunk 2, 1 thread)
Manually found that intensity 2016 is the best spot there.

(2023 / (16*64) * 64*16 = 1024 for Vega 64
(2023/ (16*56)) * 56*16 = 1792 for Vega 56

Bcs of 1,9 turning into 1 here = 64*16 gives 1024 as a result.

Both have same GPU codename: gfx900

4024/2 = threads = 2012
(2012 / 16)) * 16 = 2000
Ending with at least 2000 intensity in one thread with ws16 will be optimal with extra reserve space to prevent possible crashes.

Thanks to Villos1 (HiveOS team) and newtoetn from discord for feedback

psychocrypt added a commit to psychocrypt/xmr-stak that referenced this issue Aug 24, 2019
solve fireice-uk#2517

In the case we can utilize more than 75% of the compute units we will
not enforce a multiple of the compute units.
@psychocrypt
Copy link
Collaborator

fixed in #2518

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

No branches or pull requests

2 participants