04-18-2017, 06:00 AM
Hashcat Cluster
|
04-20-2017, 08:20 AM
The reason is what I've described with the results of my tests if AMDGPU-Pro. The power-saving is simply too aggressive and there's no way (yet) to turn it off.
04-23-2017, 09:48 AM
It has to be aggressive, otherwise the card would start a fire.
Update :
Hi, No fire yet but we now have 34 cards running at our warehouse. It's running fine but we had to fit a custom made factory screen door to ventilate the 50' long warehouse. Currently waiting for stock to go up to 45 cards. This would brings us to 7.6Mh/s of WPA2 for reference. Currently when I flash the hashcat profile I get 5.4Mh/s WPA2. That's rather good. Highest temperature delta is 50C from ambiant when using hashcat and 40C when mining. Lowest delta is 30C from ambiant when using hashcat 27C when mining. By the way, cards are fluctuating quite a lot on linux with amdgpu-pro on some algos. I reckon the problem will be fixed by AMD at some point. For now staying on 16.40 ( most stable and optimal for mining ) Also, no such fluctuation when mining ethereum or zcash. Thank you'all for your input. Cheers, David.
06-13-2017, 04:59 AM
I'm afraid you entirely ignored the point, rjtoubib. Of course the cards are fluctuating quite a lot while running Hashcat. That's because of PowerTune. And of course you haven't experienced a fire yet, that's also because of PowerTune.
The card itself is electrically capable of drawing way more power than the cabling can actually support, making it a fire hazard. The firmware, however, throttles the GPU back when it detects its drawing too much power, preventing aforementioned fire. But this throttling means a reduction in performance. The more power-hungry an algorithm is, the harder the card will throttle. This is what is meant when atom says "The power-saving is simply too aggressive" and also when I say "It has to be aggressive, otherwise the card would start a fire." |
« Next Oldest | Next Newest »
|