Bitcoin Forum

So I'm trying to get a decent miner to work on my rig of 6xVega 56 (flashed to 64).

Tried Avermore 1.4.1: doesn't start after loading the kernel. Screen stays blank.
Tried Zawawa's GGS but it's not optimised and does no more than 10 MH/s per card.

Any more suggestions?

You could try https://github.com/KL0nLutiy/sgminer-kl/releases/tag/kl-1.0.5fix

turn down the memory a ton like back to 900

vegas are beasts at x16r , on par with a 1080ti

You just have to lower memory alot, almost to stock

i also run a lower power limit, hash rate is 15mhs-23

Thanks I will.
I suspect it will fail too though. All SGminer forks have been hanging so far.

I've lowered it to 900 already. Will try even lower (and lower power limit too). Which miner are you using?

Vegas are not on par with 1080tis. They are atleast 30% lower hashrate than 1080tis especially since there is no consistently updated amd miner for x16r.. x16r is not memory intensive, AMD wont be on par with nvidia.

Tbf Avermore is not bad, It just needs a few updates. As it stands a RX 570 is about 15% slower than a gtx1060 and a 580 about 10% where they should be about on par (given a few more optimisations in avermore I think theyre would be)

30 percent slower? maybe 10 percent , I get 22 mhs on my vegas a low of 15, max i have seen on 1080 ti is 17 to 28

Averages are a lot lower than that 22 mhs. Lowest i see on my 1080tis is 18, and have seen numbers over 35 and this is at 140watts. The averaged number for vega 64 was around 17-18mh, while the average for a 1080ti is over 25mh.

The first time you run Avermore it has to build the kernals. This can take like 5min or more so just wait out the black screen.

what about mining Raven with RX580's, can anyone confirm and publish some statistics if they are more profitable mining Raven with AMD's

compared to mining ETH, which has become really only barely profitable...

There is a initialization problem with the SGminer versions and Vegas. It doesn't matter how long I wait, the screen will stay blank.
One thing that will make it work, is to get GGS to initialise the cards, then closing it, and starting SGminer (whichever fork). That works every time.

However, I cannot get close to the speeds Marvell2 is claiming he's achieving. I'm averaging 12.5MH/s per card (whether I use Avermore, MadKernel or KL0nLutiy1's version). Not sure what I'm doing wrong... :(

EDIT: what intensity/threads/clocks etc are you using? Some SGminer versions have a --benchmark function, some don't. When the miner fails, I'm never sure it's because of the parameters, the clocks, or the initialisation... :(

EDIT2: It seems like using Overdriventool to set the clocks and voltages doesn't play ball with SGminer. I need to start GGS, apply clocks within it close it and don't touch anything else for SGminer to initialise the GPUs and start hashing... wtf.

Try -w 64 -g 2 -X 320 (worksize is 64, 2 threads, Xintensity 320) and you could use powerplay table like for monero.

Thanks. This gives me very inconsistent hashrates... Starting around 13.5, going down to 7... and totally different figures between GPUs... NB: I'm using Avermore 1.4.1 with --benchmark flag. Clocks at 1428/900 MHz @ 915/905 mV respectively. Driver version 18.6.1.

EDIT: it would seem that low worksize tends to increase this hashrate inconsistency. But regardless, whenever I find a setting that gives me close to 14MH/s, hashrate will drop a minute or so later.  It's just weird. I run my fans at high speed so core temps are around 70C. I don't understand whether there is throttling or whatever is going on.

UPDATE:

4 days later, still struggling. Whilst the RX 570 rigs are hashing fairly stably at just over 7MH/s per card on average, the Vega rig is still a handfull.

Regardless of the fact that initializing the GPUs is almost a complete hit-and-miss (SGminer won't initialize them at all - need to start and close GGS or SRBminer to allow SGminer to start hashing later on), I've been trying to get higher hashrates by using --thread-concurrency. Whilst it looks promising, the miner doesn't manage to start all GPUs, or they go SICK/DEAD very quickly.

Something is preventing that rig from working properly, I just can't put my finger on it. Swap file is set to 70GB. Not sure what else I can try... :(

EDIT: right, so I managed to get all 6 Vegas to hash. But it seems hashrate is really capped around 13.5MH/s in benchmark mode. Strangely, when one or two cards were going SICK/DEAD, at least one or two of the others were hitting speeds of 17+... Is there a bandwidth issue of some sort?

A while back I also tried mining x16r on my Vega rig and I had the same issue where one of the cards went SICK/DEAD after a while.
I tried downclocking and upping the voltage but never got it to work propperly so I gave up on this.

Well, I managed to get the Vega rig to hash in a stable fashion using -X 256, -w 256, --thread-concurrency 20480 (could be any number there, it doesn't seem to make any difference), -g 2, --gpu-platform 0, -d 0,1,2,3,4,5. Clocks at 1428/900 MHz @ 915 mV respectively. PL@-20. Driver version 18.6.1.

...but still only approx 13.5MH/s avg hashrate per card. Nowhere near the 17-23 that were claimed by other users. And I have no idea why.