Topic: Any word on amd vega hash rates? - page 47. (Read 202722 times)

You have 2x8pin really? Because I have 6+2 and a 6 pin on which I use 6 to 8 pin adaptor

So a y-splitter is no good, but the 8 pin to 2x-8pin that comes with my EVGA is okay?

I hope so, cause that's what I've been using.

Using Y-spliiter for rx Vega is very dangerous. It can't handle the power spikes fluctuations, however you can use dual cerfified cable like the one EVGA has with 1 end 6+2 pin and the other end with 6pin and use a 6 to 8 pin adaptor

Please beat this dead horse for me.. 
I bought a Y-Splitter, 8-Pin pcie to 2x 8-pin pcie. 

https://www.amazon.de/adaptare-35118-Grafikkarte-Stromkabel-8-pin-PCIe-Stecker-Stecker/dp/B01N5JJ7DD
If that link gets taken down, search B01N5JJ7DD on amazon.de

No, I'm not advertising, I just want to be clear about my cables, because the whole . 

Is it the PSU that will be causing problems or the connections on the cable that I've got?  (or am I fine?)

I imagine that the problem would lie with the connection at the PSU end (where there is up to 250W going through those small wires), am I wrong?  What's wrong with my thinking that a more quality Y-Splitter would be able to handle the wattage? 

PS I've been mining about a week with two of these cables for 2 cards, and the other 2 cards using 2 separate cables each which came with the PSU (2x 8-pin to 8-pin).  Am I playing with fire?

Yes unless you want to risk damage the cards, or burn the whole place down. It's up to you.

Guys, I just got my 6 Vega 56 cards... I wanted to power them with an EVGA 1600 G2 but I just don't have enough VGA power connectors! Do I really have to connect 2x 8-pin per GPU? Is that needed for mining?

I do it for my Linux Ethereum rig. It undervolts the HBM2 and I can reduce the core speed to 991MHz to undervolt the core.
Unfortunately I cannot get more than 38MH/s. 2 Vega 64 is 76MH/s and 400Watt at wall using rocm kernel and ethminer.

https://www.youtube.com/watch?v=WozmsD7oKrM

Thoughts?

Or maybe a better question - odds it will get us off blockchain drivers?

I have modified the script slightly to work with 1 of my Vega card. You can modified it accordingly to fit your rig.

Download link below:
https://drive.google.com/open?id=0B2XFJ__I0jzLWlBZMTVCMWZDZ3YwRDZiRkh6cldKWnc1Wjlz

Some changes I made:

If you are executing the run_HashMonitor.cmd, make sure you change the file name in it accordingly to reflect CastXMR_HashMonitor_v2.6.ps1 (Or any name that you saved as).
Notable change I have made in this ps script includes:
Line 168 - 172 : I am using the script to call Cast-XMR directly, and Line 169 is the usual config including my wallet address, mining pool etc (the usual you put in a .bat) & also --remoteaccess in this case. Line 170 - 172 is to assign the localhost & port accordingly (7777)
Line 187: The main address for local/remote monitoring
Line 191 - 201: I didn't use $hdiff variable, but make a new variable for min Vega Hashrate before restart $global:vegahash @ line 201 (multiply by 1,000 since the reported value in the web is in actual hash rate)
Line 392 - 397 : Initial Vega hashrate for comparison (Not really needed since we have a reset trigger down below, but it's the same as the next)
Line 476 - 479 : Vega hashrate used for monitoring in order for restart.
Line 483 - 484 : Nothing major, just changing monitoring from 1 minute to every 30 secs
7: Line 486 : Most important part, condition to restart (Here to check my Vega hashrate against my preset target -> ~1.75kH/s @ Line 201)
This should set you up in the right direction to start with;
Do take note that I am running only 1 Vega to check against, hence I can use the total_hash_rate for comparison; if you are running multiple Vegas rig, you might need to divide it by number of GPU instances before you make the comparison. Alternatively is to loop through the "devices" in the hashrate report JSON to check if any of them drop, but I find total_hash_rate to be easier. You can also use total_hash_rate_avg in the report, as current total_hash_rate might be too sensitive for occasional dip.
And remember to change the CMDLine in Line 169, or else you would be mining to my wallet instead.
Let me know if you have any further question. I will try my best to help set up.

Are you 100% sure that the card is undervolted? Almost every time I quit xmr-stak-amd the driver glitches and resets one of my cards to the default voltages and also ramps the fan down to 400rpm. That is on both of my rigs (1 card on each rig resets, I think it is always the same one). I just restart, run mythic's driver reset bat, run hellae's registry file, restart, run mythic again, reset voltages to manual and fans to 3000 and then it is fixed.
I still do get some GPU COMPUTE ERRORS but it is more like 10 per week, mostly on one of the two rigs. Each rig is running 4 vega 64s.

Thanks for your fast answer mate!!!

Now I'm running cast-xmr, but I found a bit difficult to understand how to modify the final script for using another miner instead of xmr-stak-amd.

Has anyone the script modified? If the answer is right, could anyone upload it?

Yes, vega miniguide works for me.

DDU, install bc august vga driver only, no extras, softpowerplay reg mod disable crossfire on registry keys, overdriventoll and be happy !

Detail, i use gatelessgate-0.1.3-pre6b-win64 miner for 2 rx580 and 4 vega56:

  "algorithm": "cryptonight",
  "rawintensity": "892,892,1920,1920,1920,1920",
  "worksize": "8",
  "gpu-threads": "2",

Hello mates,

I also have experienced the problem of reduced hashrate without apparently reason, I have followed all the steps found on this guide:

https://www.reddit.com/r/MoneroMining/comments/74hjqn/monero_and_vega_the_definitive_guide/

And my 2 vegas can achieve 1920 h/s aprox but a few hours later (like 3-4 hr) the performance reduces to 1620 h/s and I have to move hbcc slider a bit and restart miner again. Also I receive too many errors mining.


So, my question is: maybe the solution is doing all the steps shown here??

http://vega.miningguides.com


What do you think about it?

Thanks in advance

Has anybody tried the latest driver 17.11.4?

Arround 800W~820W for all 6 cards on software of my nobreake and more ~200W for the system (cpu, mobo, ram, risers and SSD 120GB).

Flash the v64 Bios increase voltages e clocks = More hot, more power usage and a bit more hs 1950~1980, so its not good for me. With v64 bios i can set mem at 1100@950mv and gpu core 1300@905mv

Sapphire's Rx580 Nitro+ 4GB Elpida/Hynix memshift 1500 and setings below:

I'd also love to know how to get 900+ H/s with a 580!  That's fantastic.

are you plugged into a switched on monitor which may be using your cpu video (not vega). When I tried unplugging the hdmi cable I it dropped to 1800 from 2100 straight away, when replugging in the hdmi I had to restart xmr-stak (daemon mode) to get back to 2100.

The stock bios for v56 seems to do pretty well in that regards, did you monitor their power usage and see if it's somewhat lesser than v56 flashed with 64 bios?

Aside, I am actually more interested with your rx580, did you do memory strap on that to achieve such high hashrate?

Thanks.

My 4 Vega 56 (2 XFX, 2 Power Color) runs fine at 1830~1890 with bc vga driver only on Win 10 x64, no amd settings or extras installed, softpowerplay reg mod and OverdriveNtool, no bios mod or V64 bios, no monitor, no hdmi dummy, just connect to remote desktop to manage the rig whit more 2 x rx580 Nitro+ (6 cards total) .

Ovntool settings:
GPU_P7=1407;950
Mem_P3=925;900

Attention Detail :

If hashrate drops or every system restart just go to the device manger disable all Vega cards and enable, the HBBC turns on.

I'm having the same issue. 1400 core, 1100 mem, 0.925v in gpuz. It starts at 2050h/s then drops to 1700. The core hot spot is 65 degrees, while HBM is 75 degrees,

It's the newegg $700 Vega FE, but right now it underperforms a Vega 56.

2016/2016 was also the best settings for me. 4 Threads instantly crash with any intensity, 3 threads of 500 intensity each gave about 1600h/s.