Minging ERGO with Vega 56's: Problems, in need of assistance

[marctheshark]

Hello Everyone,

I am in the process of building out my rig of 9 Vega 56’s and 1 470 atm. I can only run 6 GPU’s at a time currently.
Motherboard- H110 Pro BTC+ ASRock (P1.60 03/23/2018)

CPU - 4 × Intel® Core™ i5-7400 CPU @ 3.00GHz AES

Disk Model- USB SanDisk 3.2Gen1 15.4GB

My problem is multi fauceted:

Problem A: No matter what strap or OC I have or defaults, my rig will run for an hour or so then GPU 5 crashes.

Attempted Solution: Change out GPU5, its cables, riser, etc - Didnt work

Attempted Solution: Remove GPU5 all together and run only 0-4 GPU’s, this gives me another problem

Problem B: Now this problem is that my one of the GPU’s unknown atm can load up the miner and gets stuck, when I added the 6th GPU (#5) it never has that problem then eventually GPU will crash lol.

I did some testing where I would run all 6 with and w/o OC settings where I would only apply the settings to a single Gpu so I know all are stable with the current settings. any ideas?

I did notice though in the logs that right before this error (GPU 5 crashing) occurs the Watts for that GPU drops alot:

The VDD@ 925 for GPU5 as been my attempt to fix it, thinking it needs more voltage

AND occassionally I will get this error too:

[rkulov]

Watts drop when gpu crashes so i guess they need more juice. Your main problem isnt this. The vega cards are not suited fpr ergo mining. Its due to their design and for autolykos its not good. Its like they have half the memory they have in reality.

[marctheshark]

Yeah I have a sneaking suspicion that it could be the power(using 110V atm), cause never had an issue before when run less than 6 gpus. as for the driver or the stuck GPU I need to do some isolating on a benchmarking rig I need to build ha. Whats weird about the GPU5 issue is that it usually runs fine for about an hour and then crashes.

I have never heard of that about ERGO and the Vegas, do you have any source materials for your claim? Interesting in reading that, do you have Vegas, if so what do you mine?

[rkulov]

cant find the post, and i remember that i read it yesterday here in hiveOS forum. maybe i am mistaken about the GPU but i remember that some card performed bad due to its architecture. something about the size of its memory blocks being 128kb and Autolykos using 32kb blocks. so for a 32kb block to be mined you use your whole 128kb block of memory, for other cards the memory blocks are 64kb so you use 50% more blocks for the same volume of Vram.

If you get 2 times as much as when you mine ETH then its ok.

[nanoTWminer]

You’re wrong, what you’re referring to is RDNA1 and RDNA2.
VEGA is very good at autolykos, says Teamredminer.

[rkulov]

Actually you are right. They get 30% more. I guess i didnt understand correctly or its for some ther gpu.

[marctheshark]

I think what I am experiencing is a power issue since I am using 110V so my power supply is 900W. I dont have a metter but people seem to think we use more Wattage than hiveos reports so it all check out if that is the case. I wasnt able to replicate the issues on my test bench.

Yeah I thought that vega and autolykos worked well especially since a some community people have been posting about it with some solid/promising results imo.

[rkulov]

Yeah i investigated a bit more about the GPUs , 3070 also performs a bit better than most on Autolykos. The reported power consumption on HiveOS is only for the GPUs, so you need to add more power for CPU , RAM, HDD,Mobo which by my estimates are around 150+ all. and it depents also on the PSU and the coeficient of the power transformation. For example, HiveOS says my GPU gets 95W but my PSU has coeficient 80% and the real consumation of the GPU is actually 118W.