Topic: [Cast XMR] high speed XMR/CryptoNight miner for RX Vega GPUs (2 KHash/s) - page 67. (Read 206388 times)

I will try it, thank you

I shared my modified version of JJ Hash Monitor for XMR-Stak awhile back, modified it to work with Cast-XMR.

https://drive.google.com/drive/folders/1TnicWxpB1TtGQCe_cuXND8ha-XsgQsU1

You can go through the changelog and update it to latest Cast-XMR accordingly.

This script works great

https://github.com/fiddlefingers/Fiddlefingers-cast_xmr-hashrate-monitor-and-restart-tool

Did someone find a solution on sporadic issue Vega 64 of FE recongnized as gfx900 instead gfx901?

Same problem here. The duration is not always the same.


Im also interested, because of the problem above.
May glph3k can include it like claymore.

Is there any free scripts available to restart CastXMR when it hangs or speed drops to 0?

When I mine with cast it hangs up after running for a couple of hours. The longest I’ve managed to get it to run is like 8 hours. Anyone else experiencing this problem? Typing S doesn’t do anything I have to exit the application and restart it.

Currently using 2 Vega 56s
Temps stay around 63

Gents,

I was able to "partially" resolved issue with combination of 1 x Vega FE + 4 x Vega 64.

1. Go "safe mode" DDU any drivers available.
2. Reboot
3. Install Blockchain drivers (only AMD driver, nothing else "sorry wattman and other features". Important here was, evrything run via RDP (remote session), if I plug in monitor to any GPU's installation crashes with BSOD.
4. Disable in registry Crosslink.
5. Reboot
6. Configure OverdriveNT (my version is 0.2.5). Vega 64 and FE have same clocks for core and memory and mV
7. NEVER disable/enable Vega FE on this "bad blockchain" drivers, if something is stuck... reboot rig.
8. Configure enable/disable for Vega 64, worked without any issues.
9. Run miner via batch file: (a. Disable/enable only Vega64, b. Overdrive, c. Miner)

Everything should be done, only remotely... No monitors plug in to any GPU, got artifacts and crash when execute miner.

I have 1 caveat: Vega FE hash rate is 1400-1500, because I can't touch HBCC (no wattman), and another thing. Because I have Vega FE if I will install AMD Radeon Settings for some reason I don't have Wattman there at all... Only Pro Settings for all cards.

3.

Vega is the only card that has a new kind of memory called HBM.
All reference design has Samsung HBM chips in it.
It doesn't matter if it is Sapphire,  Asus or 56 and 64.
All reference card was made by AMD itself and it has Samsung in it.

Most of the non reference cards have hynix HBM chips.

Samsung good
Hynix bad

How bad I don't really know. Apparently Hynix memory will do around 900hz . Samsung could do 1200 but usually stays at 1100.

Avoid XFX non reference!!!

That's all

About the problem mentioned.  Is the card in question  an GFX900 instead GFX901?
My guess is, it's not the card. It's something else. If you change the card position,  do it keep the lower hash?
I do have an problematic Sapphire 64, same card as the other 5 in the rig. It is the only that is identified as GFX900 and the only that gave me errors.

Lol, just saw u do have 2 GFX900.  That's your problem. This card is more tricky.  Mine is doing same hash but with 2% error.
No one knows the difference from gfx 900  to gfx 901!
There are 56, 64 and FE as 900 but the majority in the market is 901.

 

what should be the memory type to attain the mentioned hash rate and what memory types to avoid , can u tell me.

I was in blockchain drivers too in my rig of 10 rx580 since january!!

I change to last one amd 18.2.3.

And i recommend a lot!! Real stable and the hashrate is the same, no errors at all.

In vega i dont test. Im waiting for my to come from amazon  to test

But give it a try and test the compute mode in new drivers.

Old blockchain (have compute mode by default).

Try to find the Vega Thread in this site forum too and as...

 

There is no compute option in wattman, also because 1 of the card is Vega FE wattman even not showed up.

Memory type? gpu-z should do it. Hwinfo might.

Is there a way to find this out by the way?...

Have you enable the compute option in wattman?

I think it's not due to wattman...I already tried uninstalling wattman and use overdriveN tool but still the hashes go down...now I have steady hashes with wattman installed..

https://i.imgur.com/gallery/xIH2r

I think you forgot something. Check this guides, here is all what you need

type vega mining guides into google(they refer to the original post below)

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

The reddit guide says to use wattman. I dont recommend it. U only need the driver but no amd software.

Need help...

New Rig, new fresh Windows 10 Pro 1703
CPU - Ryzen 5 1600/8GB RAM

4 GPU's: 1 Vega Frontier, 3 Vega64 (ASUS+XFX).

Blockchain drivers.

Here is weird results:

C:\Mining\cast_xmr-vega>cast_xmr-vega -S cryptonight.usa.nicehash.com:3355 -u xxx.amd -G 0,1,2,3 --fastjobswitch --forcecompute
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  Cast XMR 0.8.5 CryptoNight mining optimized for AMD Radeon Vega GPUs

  Coded by [glph3k] for more info visit http://www.gandalph3000.com
  Donating 1.5% mining power to developer. Thanks.

  For current stats press 's'
  For seamless exit press 'q'

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

[15:12:58] Initializing GPU, loading kernel ...
Fast job switching mode enabled.
Detected OpenCL Platform: OpenCL 2.0 AMD-APP (2442.12)
[Suspicious link removed]pute Driver detected.
GPU0: gfx900 | 64 Compute Units | Memory (MB): 8176
GPU1: gfx901 | 64 Compute Units | Memory (MB): 16368
GPU2: gfx900 | 64 Compute Units | Memory (MB): 8176
GPU3: gfx901 | 64 Compute Units | Memory (MB): 8176
[15:13:00] Connecting to cryptonight.usa.nicehash.com:3355 ...
[15:13:00] Connected to pool.
[15:13:01] New job received.
[15:13:01] Difficulty changed. Now: 200007.
[15:13:14] GPU1 | 51°C | Fan 1024 RPM | 287.1 H/s
[15:13:16] GPU3 | 38°C | Fan 0 RPM | 239.1 H/s
[15:13:17] GPU2 | 57°C | Fan 1312 RPM | 212.6 H/s
[15:13:18] GPU0 | 43°C | Fan 4666 RPM | 209.8 H/s
[15:13:25] GPU1 | 52°C | Fan 1015 RPM | 359.0 H/s
[15:13:27] GPU3 | 38°C | Fan 0 RPM | 312.0 H/s
[15:13:35] New job received. Avg Job Time: 34.0 sec
[15:13:35] GPU0 | 45°C | Fan 4668 RPM | 158.9 H/s
[15:13:35] GPU0: 28 ms needed to switch to new job
[15:13:35] GPU3 | 39°C | Fan 0 RPM | 357.4 H/s
[15:13:35] GPU3: 36 ms needed to switch to new job
[15:13:35] GPU1: 49 ms needed to switch to new job
[15:13:35] GPU1 | 51°C | Fan 1016 RPM | 325.4 H/s
[15:13:35] GPU2: 135 ms needed to switch to new job
[15:13:35] GPU2 | 56°C | Fan 1324 RPM | 155.9 H/s
[15:13:50] New job received. Avg Job Time: 25.0 sec
[15:13:50] GPU1: 67 ms needed to switch to new job
[15:13:50] GPU1 | 51°C | Fan 1010 RPM | 126.7 H/s
[15:13:50] GPU0: 151 ms needed to switch to new job
[15:13:50] GPU0 | 46°C | Fan 4680 RPM | 113.7 H/s
[15:13:50] GPU2: 172 ms needed to switch to new job
[15:13:50] GPU2 | 55°C | Fan 1325 RPM | 114.3 H/s
[15:13:50] GPU3: 200 ms needed to switch to new job
[15:13:50] GPU3 | 38°C | Fan 0 RPM | 113.4 H/s
[15:13:54] New job received. Avg Job Time: 18.0 sec
[15:13:54] GPU0: 53 ms needed to switch to new job
[15:13:54] GPU2: 72 ms needed to switch to new job
[15:13:54] GPU3: 138 ms needed to switch to new job
[15:13:54] GPU1: 205 ms needed to switch to new job
[15:13:54] GPU1 | 52°C | Fan 1015 RPM | 472.3 H/s
[15:14:10] GPU1 | 51°C | Fan 1020 RPM | 189.4 H/s
[15:14:11] GPU2 | 56°C | Fan 1316 RPM | 212.8 H/s
[15:14:12] GPU3 | 38°C | Fan 0 RPM | 211.8 H/s
[15:14:12] GPU0 | 45°C | Fan 4681 RPM | 209.8 H/s
[15:14:26] GPU1 | 51°C | Fan 1014 RPM | 244.1 H/s
[15:14:29] GPU2 | 55°C | Fan 1324 RPM | 198.9 H/s
[15:14:29] GPU3 | 38°C | Fan 0 RPM | 199.4 H/s
[15:14:30] GPU0 | 45°C | Fan 4682 RPM | 198.3 H/s
[15:14:45] GPU1 | 51°C | Fan 1019 RPM | 210.8 H/s
[15:14:48] GPU2 | 55°C | Fan 1320 RPM | 194.6 H/s
[15:14:48] GPU3 | 38°C | Fan 0 RPM | 193.4 H/s
[15:14:49] GPU0 | 45°C | Fan 4671 RPM | 189.7 H/s


.... ETC

Can someone have any idea, why hash soooo looooow...

Thanks.

After some driver reinstalls and test, it seems it comes from card 2. Cast XMR works fine when I exclude card 2 from mining... (-G 0,1,3,4)...
And it does artefact when I use only card 2...

I hope this card is not died
Which software could I use to test it completely ?

I will try to flash a new bios and to change its position in the rig.

Thanks

edit : Looks like it's solved ! Despite I had set 0 to crossfire parameters in regedit, it seems to be still activated  (thanks GPU-Z) !
I installed AMD soft to desactivate it and it's OK !

i bet your first card is the coolest and those in the middle are 10°c hotter
your observation is called airflow i assume
switch position of cards and check temps again if you want to go sure