Quote from: pbfarmer on March 27, 2019, 09:44:31 PM

Quote from: kerney666 on March 27, 2019, 09:34:50 PM

Quote from: pbfarmer on March 27, 2019, 09:10:42 PM

Quote from: kerney666 on March 27, 2019, 08:31:59 PM

Quote from: dafuzz_rsa on March 27, 2019, 08:08:52 PM

Quote from: kerney666 on March 27, 2019, 07:59:03 PM

Quote from: dafuzz_rsa on March 27, 2019, 07:50:34 PM

snip

Yeah there's def a variance here. Ref V56s flashed to V64 bios and true Vega 64s all seem to be straightforward to get hashing > 19 kh/s. I have a few things to test, but I need to get a few Vega 56s going that aren't flashed with the V64 bios.

Are you on win10 or linux? And, I assume you've tried configs like L24+24 and 28+28?

Yea i tried them, best for me seems to be L28+26
on Linux

Hmm. I really need to get an amdgpu-pro environment going to be able to test better on linux as well. The memory subsystems on win vs linux are much more similar now than they were in e.g. the Blockchain driver days, but I wonder if the miner really behaves the same way on win vs linux for this algo at peak pressure.

Hold my beer...

Haha, yeah I was hoping you'd join the party! Very interesting to see if you can spot any clear differences between parameters such as OS/environments/drivers/gpu types/etc.

Haven't really done much in the past w/ CN light variants, as I got the impression they were a bit more power hungry... Is that the case w/ this (e.g. compared to cnv4)? Just so I can get a sense of where to start my voltage settings.

Not trivial to measure, there are a few different trade-offs in play, but I'd actually say they're about on par. CN-turtle uses the CNv8 main loop which (in our case) draws less power than the CN/r main loop, but CN-turtle does for sure put more load in general on the gpu than the 2MB pad variants.