Topic: Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux) - page 1118. (Read 6590718 times)

Power off / Power on... Not what you wanted to hear but... It's a solution..

I get the same problem occasionally, Try looking at the problematic card, Down-clocking it etc.

Today got an error
OpenCL on GPU one crashed. Watchdog exited but the miner is unable to restart. Hands whole system.
Any solution?

I have a few questions regarding the Linux version:
- Is coloured text in the terminal possible? If so, how?
- How would I run the miner at system startup as user, inside screen ?
- For Claymore, can you build with a newer libcurl version so I don't have to install the libcurl-compat package ?
- Is temp/fan monitoring/change possible with AMDGPU driver ? If yes, does it require X to be running ?

Maybe Claymore or someone else can clear this up for me.

thank you. very good info.

One question. So if i want to maksimaze Mhash, i will go from stock MC (2000) to 2200 Mhz and i will lover GPU clock to 1250 or less (from stock 1346). how fair i can go with voltega if stock is 1150 mV?

i saw, that if i have 2200 MC and if i drop GPU clock to somewhere around 1050Mhz, i lost around 5-10 Mhash overall. but cannot confirm,that is less power hungry.
do you cheak at GPU-Z avg power? temporary i have around 150W (4 GPU),one gpu is showing 198w avg,which is a little to high. 1200 GPUC and 2200MC.

Use Wattman in Radeon settings.  For that card you can drop from stock (~1150mv) to ~1050mv at full 1342MHz, or drop clock to a lower OC and use 960-1020mv.  Depends on ASIC quality, but out of 10 Nitro cards these are what I use.  Usually at 1260MHz you can drop below 1000mv unless ASIC quality is really low.  Going below 1000mv you also need to drop memory voltage.  My worst card ASIC quality in the 0.4 percentile runs 1050mv @ 1306 Mhz.  980mv usually safe @ 1260Mhz.  I have ran semi stable down to ~940mv and 960 is often long term stable. Use GPU-Z to see actual voltage.

Hello

How i make that my bat closed and open after a gpu fall ?

Interesting. I would think the opposite or assume you can use some tricks to achieve the same without the increased power draw (like explicitly flushing or deliberately "starving" that queue).

I don't currently run a secondary coin because high memory clocks with tight timings are only stable at low temperatures (which is also beneficial for power efficiency on it's own) and it's hard to keep the reference 480s cool even at ~100% fan speed if they are busy with two coins. So, single-ETH seems to be better overall (for me, for now. may be winter will change that).

On my RX 480 Nitro+, increasing -dcri to about 40 for SIA will also increase the hashrate for ETH. MemClk = 2200 and core  = 1050 mV.

Moreover, at some (not large) -dcri values in dual mode you can get a bit more ETH hashrate compared to ETH-only mode (because of shorter queue to memory access), though I see it on 390 cards only.
Increase -dcri and ETH hashrate will fall.

may i ask what is the result with hash? dual ETH +SIA and power consuption?

thanks

Is it possible  that dual SIA mining NOT change hashrate  compared to ETH only?
Or there is something wrong with cards/drivers etc?
Gigabyte 380x.. about 20mhs a card with and without SIA..

when will you issue the next update (not pushing ... simply asking )

Thanks alot

I'm dual mining ETC + SIA

i saw that, but it is only for mining ETH. Not dual mining EHT +SIA.

I will still looking, maybe i miss something.

Read my previous post, I've just edited it 

i forgot to say that at 1250 GPU and 2200 MC system is stable...just need to undervolt it, but i do not now how, and how much mV per card?

Just one suggestion:

If your system is unstable, leave your memory setting at their default values.

Reduce your GPU speed and your voltages, both for GPU and Memory.

You might loose 5% ETC speed but you'll save 20-30% power

That's the trick!

No need for MSI Afterburner or Trixx or even WattTool. AMD's Wattman is more than enough to achieve this goal

Take a look here:

https://bitcointalksearch.org/topic/best-efficiency-settings-for-radeon-rx-480-1543701

Good luck!

Just one suggestion:

If your system is unstable, leave your memory setting at their default values.

Reduce your GPU speed and your voltages, both for GPU and Memory.

You'll might loose 5-10% speed but you'll save 20-30% power

That's the trick! Got it?

hello

i have one rig with 5 rx 480 nitro OC (not reference card)-saphire.

temporary i am mining ETH + SIA.

I have a set up 1250 GPU clock (defolt is 1346) and 2200 memory clock (defoult is 2000).
I have power at the wall 1,2 kw.
I did not change votaltage.
I have cca 128-135 Mhash for ETH and cca 2000 Mhash for SIA. (dcri 30).
Temp GU is aournd 68-72 C (fan from 80-95%).
Fot test i went with 1250 GPU and 2250 MC and i have 138MH+2500 sia (dcri around 35). but the system did not handle it and after few minutes restart windows 10.

can i do some change in voltage to have the same hash,but less power consuption? I foound several set up for rx 480 ref. card., but for that one not so much ,specially for dual mining. I am ssing TRIXX 3.0 for setin up GPU.

I am ussing two PSU: 650 (2 cards) and 750w (3 cards plus MB). At specification for GPU it say that one card need 225 w.

Also, when i raise DCRI to 32, system get unstable and few times GPU4 crash. Can this be a probem with PSU? The heat for to 74/75, but the system do not handle it. When i raise memory clock to 2250, ETH hash goes to 138, but is is not stable, which is strange, becouse i saw few test, where at 2250 was just ok.

I suspect that my PSU 750w (corsair RM), cannot handle 3 gpus at maks load.

Also, when in mine dual, hash is not stable. is that normal? i saw that i need to raise power limit to 20% higher. i did riase power limit tfor 10 %,but tht system start acting strange and again, it was not stable  and crash-restart( i am using risers with molex).

thank you in advance for any advice.