Topic: Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux) - page 584. (Read 6590757 times)

worked thanks, why does it have to be higher thats crazy.

try 32,000 pagefile.

Im using 8 GTX 1070 GPUS and have set page file to 16000mb im getting this error: can someone help please


UDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.
CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.
Setting DAG epoch #143 for GPU0
GPU 0 failed
GPU 0, CUDA error 11 - cannot write buffer for DAG
GPU 7 failed
Setting DAG epoch #143 for GPU7
GPU 7, CUDA error 11 - cannot write buffer for DAG

You are wrong. powered risers use between 40W - 50 W each, which is NOT reported in GPU-Z. This is easy to verify in a dual PSU setup like I have where all the risers are connected to the same PSU as the motherboard and the secondary PSU only powers the tops of the cards.

Watttool doesn't support the 5xx series does it?

The risers themselves do NOT use anywhere close to even 40W.  The power load for risers is negligible even with six of them.  It's the GPUs themselves that are using the power.

I don't think the power sensors in GPUs are all that accurate anyway plus they fluctuate so wildly.

I have switched all my RX BIOSes to Anorak's powersave versions.  I lose about 3-5% hash rate but it saves almost that much in power and reduces heat output and wear on the cards.

When I see 6 GPUs consuming over 1000W and then I look at an Antminer that can produce 4x more revenue on 1200W I wonder why I'm even bothering with GPUs...

I did not think the raisers would consume that much power. It makes sense now. Thank you.

The power usage shown in GPU-Z is for the card VGA power connectors. You also need to add 40-50W for each riser. 1000-1100 W is about right for a six card RX 580 rig when dual mining.

Can anyone comment on the power consumption for the RX580 cards. I have two rigs each has 6 RX580 and both are dual mining ETH+LBC (Eth: 176 Mh/s + LBC: 510 Mh/s). Each rig consumes about 1020 Watts reading from the wall-meter while GPU-Z reads less than 130 Watts for each card. I think they should all be around 6*130 + 70(system) ~ 850 Watts total. My setting (-cclock 1150 + -mclock 2100 -cvddc 870 -mvddc 870)

It can fluctuate because of high tempreture or too low power limit. As I've read the vega 56 consumes much more power with vega 64 bios.

Try gpu on around 1200 Mhz.

set fan using watttool, work for me.

-tt not working properly. I have cards that produce very different amounts of heat, and turn their fans seemingly without any logic. I even have a RX460 that has 32°C and 90% fan. Anything I could do about it? Setting fan speed doesn't work either, not from claymore bat file nor from wattman.

It makes no difference where you put it in the command line. As noted in the Claymore README, Nvidia cards don't support undevolting. You can adust the power limit % on Nvidia cards using Afterburner in Windows, which reduces the power consumption target when set at a negative value.

Where do you put the cvddc settings, at the beginning of the file or the end? And also can you undervolt Nvidia cards too?

@rednoW:

Thanx a lot, should've looked for that particular thread.

Cheers

2 brnsted

Move to vega thread - this problem is solved.
You should get 44mhs eth with 150-170watt
https://bitcointalksearch.org/topic/m.22158253

Hey guys, there seem to be a lot of people here and elsewhere that managed to get the Vega 56 to a smooth 39MH/s- ish performance at very efficient power draw numbers.
If reports are true, some of which must be running at 150W draw, which I find impressive.

I am currently running into huge problems, especially with AMD Wattman which feels like not correctly applying my settings done (especially on voltage).

I did this:

- flashed Vega 64 bios on one of my my Vega 56.
- Windows: Clean install
- Installation of latest blockchain drivers + AMD Wattman

Running Claymore Miner, just on a stock Vega 56 and a Vega 56 w/ V64 BIOS.

I start miner, get 37 MH/s. However 300W+ from the wall for one card.

- Decrease voltage: No change in Power draw
- Set P6 / P7 states to max core clock: 1000 (first, gets applied)
- Set Memory to 1100MHz clock, resp 950MHz for the V56 (gets applied)
- Decrease power limit to less than -10% -> Memory jumps back to earlier state, which however I am not able to adjust
- Even then, GPU clock starts ticking off at about 1500 again.

So it seems to me that some of the settings are being overridden.
Max I could get was 40MH/s but at almost 400W, nowhere close to reports of 39 MH/s at 150W

Anyone else experiencing fluctuating core / mem clocks or non applied voltage settings?

Thanks and cheers,
brnsted

You may need to underclock.

post your build specs