Overclock.net banner
81 - 86 of 86 Posts

·
Premium Member
Joined
·
6,081 Posts
@mmonnin @Jpmboy what GPUs are you guys running on Einstein? I am just curious to see how my single Fury X, which is now exclusively crunching Einstein, stacks up against some different GPUs. :D
A single RX 580 at around 500k RAC. I just turned back the Power since it's started to reach the 80s in NoVa.
 

·
Premium Member
Joined
·
4,955 Posts
Making a run to 1B on Einstein.

Is anyone running more than 2 concurrent FGRP GPU apps? Running 3 is more efficient on my Radeon VII, but the driver will randomly crash. The driver crash happens when a task errors out, and it leaves the driver in an unrecoverable state, even if the GPU that crashed is reset. 🤔 Only way to get any of the GPUs back up and crunching is to reset the machine.

This happens while running in Linux using the ROCm stack for OpenCL. Since I am doing a bunch of other testing with ROCm, I haven't tried to uninstall ROCm and crunch with the OpenCL bits from the amdgpu-pro driver.

Just curious if anyone is able to crunch 3x or more tasks, running the AMDGPU-Pro driver in Linux or in Windows.
 

·
Premium Member
Joined
·
4,955 Posts
I ran just a single task in Windows on my R7 and the driver randomly crashed when running [email protected] MW and Collatz have been fine. Sometimes the driver recovered, others a BSOD.
Bumped it down to two tasks per GPU and I still get random crashes on a random GPU. With two tasks I can recover from the crash by resetting the GPU that crashed without having to reboot the whole system. A bit of an improvement but still annoying.
 

·
Premium Member
Joined
·
4,955 Posts
I'm going to put a few Vega 64s on, more or less, full time Einstein duty @mmonnin is creeping back up on my #1 spot


Efficiency is not too bad. Running 2x tasks, the average runtime per task is about 415 seconds, and that is with the Vega 64s power limited to 115W.
 
81 - 86 of 86 Posts
Top