Topic: Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux) - page 605. (Read 6590565 times)

Hello.

With GTX 1070 I got strange errors after some time (very random) :

08:02:50:826   d98   parse packet: 247
08:02:50:826   d98   ETH: job is the same
08:02:50:826   d98   new buf size: 0
08:02:55:248   14ec   GPU 4, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:248   14ec   GPU 4, GpuMiner kx failed 1
08:02:55:248   14ec   Set global fail flag, failed GPU4
08:02:55:248   14ec   GPU 4 failed
08:02:55:248   1b40   GPU 4, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:248   1b40   GPU 4, GpuMiner kx failed 1
08:02:55:248   1b40   Set global fail flag, failed GPU4
08:02:55:248   1b40   GPU 4 failed
08:02:55:264   d70   GPU 7, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:264   d70   GPU 7, GpuMiner kx failed 1
08:02:55:264   d70   Set global fail flag, failed GPU7
08:02:55:264   dbc   GPU 6, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:264   dbc   GPU 6, GpuMiner kx failed 1
08:02:55:264   dbc   Set global fail flag, failed GPU6
08:02:55:264   d70   GPU 7 failed
08:02:55:264   dbc   GPU 6 failed
08:02:55:264   17c4   GPU 6, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:264   17c4   GPU 6, GpuMiner kx failed 1
08:02:55:264   11a4   GPU 7, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:264   17c4   Set global fail flag, failed GPU6
08:02:55:264   17c4   GPU 6 failed
08:02:55:264   11a4   GPU 7, GpuMiner kx failed 1
08:02:55:264   11a4   Set global fail flag, failed GPU7
08:02:55:264   11a4   GPU 7 failed
08:02:55:295   1b20   GPU 2, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:295   1b20   GPU 2, GpuMiner kx failed 1
08:02:55:295   1b20   Set global fail flag, failed GPU2
08:02:55:295   1b20   GPU 2 failed
08:02:55:295   1b24   GPU 2, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:295   1b24   GPU 2, GpuMiner kx failed 1
08:02:55:295   1b24   Set global fail flag, failed GPU2
08:02:55:295   1b24   GPU 2 failed
08:02:55:311   1124   GPU 3, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:311   1124   GPU 3, GpuMiner kx failed 1
08:02:55:311   1124   Set global fail flag, failed GPU3
08:02:55:311   1124   GPU 3 failed
08:02:55:311   1b28   GPU 3, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:311   1b28   GPU 3, GpuMiner kx failed 1
08:02:55:311   1b28   Set global fail flag, failed GPU3
08:02:55:311   1b28   GPU 3 failed
08:02:55:311   17e4   GPU 0, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:311   17e4   GPU 0, GpuMiner kx failed 1
08:02:55:326   17e4   Set global fail flag, failed GPU0
08:02:55:326   17e4   GPU 0 failed
08:02:55:326   1b38   GPU 0, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:326   1b38   GPU 0, GpuMiner kx failed 1
08:02:55:326   1b38   Set global fail flag, failed GPU0
08:02:55:326   1b38   GPU 0 failed
08:02:55:326   12bc   GPU 5, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:326   12bc   GPU 5, GpuMiner kx failed 1
08:02:55:326   12bc   Set global fail flag, failed GPU5
08:02:55:326   12bc   GPU 5 failed
08:02:55:326   17b0   GPU 5, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:326   17b0   GPU 5, GpuMiner kx failed 1
08:02:55:326   17b0   Set global fail flag, failed GPU5
08:02:55:326   17b0   GPU 5 failed
08:02:55:342   17e0   GPU 1, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:342   17e0   GPU 1, GpuMiner kx failed 1
08:02:55:342   17e0   Set global fail flag, failed GPU1
08:02:55:342   17e0   GPU 1 failed
08:02:55:342   ab8   GPU 1, GpuMiner cu_k01 failed 4, unspecified launch failure
08:02:55:342   ab8   GPU 1, GpuMiner kx failed 1
08:02:55:342   ab8   Set global fail flag, failed GPU1
08:02:55:342   ab8   GPU 1 failed


Strange is, that all GPU fails at once. I'm using Nvidia 384.94 drivers.

Any suggestions?

Hey, I connected a rx 480 to a PCI slot. Not pcie. It smoked! Now my card is not recognized. I've tried in different pc's. Is there a fuse or capacitor that might have blown on the cards board?

Well my findings are the opposite, vega 56 with moded bios can do 42-43Mhs. I can't measure the power consumption. Also the 64bios can let you clock the memory on the 56 to 1100Mhz.
Dual-mining  works fine with all drivers.

Setting lower voltages in wattman seems to lower the power consumtion, but I relly can't test much without a damned watt-o-meter.

I'm running the aug. 11 Radeon Pro public beta for Vega FE.

I was running blockchain drivers but they require too much cvddc and the hbm2 gets over 80C and starts to throttle.

An earlier reading from the wall suggested 511w but that was with my previous power supply which was in the process of dying and a much higher cvddc setting.  Setting of 1050, where with the aug 11 vega fe beta drivers I can run cvddc at 950.

Curious if anyone has actually tested vs Genoil and the claims of much higher hash rates

Yes I am 

Replacing thermal paste did wonders for me... Reduced the average temperature by ~5-8 degrees.

Haven't checked the power consumption change yet, as I've lost my power meter. Hoping it'd have dropped as well.

Are you doing this on Blockchain drivers?

Me. And I have to completly shutdown the computer (PSU switch) for that to stop happening.
Also, I've set the default rom, just for testing, without any OC. I still get invalid shares on ethereum and a warning to check my overclock. (eu1.ethermine.org)


In case anyone is having the issue with rx500 series gpu's that can't reduce voltage below +/-0.9v, that is caused by the VRAM voltage. So if you have your vram at 2000mhz @ 1v, your core won't go any lower than 1v (under load, it drops a little). Because the default vram voltage on my sapphire rx580 4gb pulse is 950mv, the lowest possible voltage is 950mv. So, I did this:

-Download OverdriveNTool from guru3d foruns.
-Disable ALL P-States on memory and core except the higher state. (to disable click on P0, P1, P2,...)
-Set your gpu clock frequency and desired voltage on the last P state (P7). My example: 1250MHZ @ 900mv
-Set your ram clock frequency and the same voltage as your gpu on last P state (P2). My example: 2050mhz @ 900mv
-Save and give a name to that profile. My example: "2050"
-Now put overdriveNtool on the same folder as your "start.bat".
-Edit your .bat file and remove all overclock options from the claymore miner and add this before starting the miner:
That will reset the cards, apply the profile named "2050" to gpu0+1+2+3+4 and wait 5 seconds before starting claymore.
Enjoy your lower power consuption. Make sure that you tweak the dcri value if you change your gpu clock. Higher clock = higher dcri

please, browse throw my recent posts (you can easily find it in my profile), I'm tired to repeat.

cool. so when you get chip@160W you see VRM power@70W (don't really get the "error" part) what is the ETH hash rate that you are getting? Also how does it look when you do dual mine?

You can try to do DDU(s), + safe mode, remove cards physically, reset mobo, uninstall AB and similar, do some proper "card completely removed and added back" kind of stuff. Also mind that voltages are not over 1v. Then try 1080MHz:1900MHz @ 875mv

The more total power is the more is the diff. But if it show 160watt it is no more then 70 watt error.
When it shows 110 watt there is only 30 watt error.
It is my own measurements on Vega.

I would guess (extrapolating on the polaris power consumption) that memory part while running eth would require more than 60W. I'd guess 80+W is more likely

currently Vega won't even boot with modded bios - there is some kind of protection against bios mod

So it is not only me???

I had mentioned it a few days ago and Claymore proposed this:


I didn't get the chance to set original ROM but I did remove overclock. Not only it didn't help but it got worse! In fact, while I was trying different OC settings I noticed that the lower I was setting my clocks, the more unstable my rig became.

My rig usually freezes after a GPU hangs or after a share gets rejected. The clocks I am using and the ROM I am using have been stable for months now with previous drivers and miner version

One more thing I have noticed is that after a rig restart (following a freeze) the miner starts, the clocks and voltages are being set normally but I get~ 3.5% less performance than usual. Once I restart the miner it gets back at 100%. Does anyone else have the same issue?

Issues with Vega 56:

Blockchain drivers lockup if attempting to mine with more than 1 card.
Blockchain drivers crash dual mining.

Using 17.9.2:
Can be bios modded with Vega 64 bios, however this makes mining slower slightly, and changing mem clock has no effect.
Slightly slower than blockchain drivers, but works better, and can dual mine.
Lowering power limit or dual mining without raising ~+20 makes mem constantly throttle to 700MHz.
Setting voltage doesn't really seem to affect power usage at all.
Clockspeed seems to have 0 impact on power usage, very little impact dual mining.
Setting clockspeed doesn't set at a fixed speed, it somewhat affects it, but it has a mind of its own.
Mem can be set ~950MHz, possibly up to ~980MHz.
CM 9.8 slightly faster ETH it seems.
Best profit seems dual mining ~34.5MHz ETH, 140MHz LBRY, +20% power, 1300 core, 9xx voltage.


Anyone have anything to add or suggestions?

Has anybody bothered to mod their Vega BIOS instead of messing with the software drivers and miner config? Memory straps, GPU underclock, VRAM overclock and Voltage offset lowered?

Wolf has posted enough info to go at it with the hex editor.

Yes, it's working. But freezes happens. Not so stable, even if mem errors in HWInfo are zeroes.

It is 17.8.2 or 17.9.1 drivers. blockchain drivers don't allow to show gpu total power