Claymore's Dual Ethereum AMD+NVIDIA GPU Miner v15.0 (Windows/Linux)

[thevictimofuktyranny]

Got bad news on RX Vega 56 - managed to do a test with the Beta Blockchain Drivers (solely released to fix the Dag file memory issue for the RX 580/480 and 570/470 GPUs).

RX Vega 56

-24% Powertune
HMB2 925mhz
Default Clock ranging between 1136mhz to 1260mhz.
PSU Efficiency is 90%
Power Draw 152watts

Miner produced a Claymore result of 35.4mhs for 153watts.

Pool shares 40 minutes never above the rate for 36mhs.
Pool shares next 40 minutes around the rate for 25.5mhs.

Average pool shares found averaged around 31mhs for Ethereum - the same as the main official drivers. The Beta Blockchain Drivers are for the RX 580s and RX 570s, when used on RX Vega GPUs they produce fake hashrate figures, which are incorrect by 4mhs.

Beta Blockchain drivers do reduce powerdraw by 13watts per GPU, compared to the official drivers for RX Vega 56. This is why people are using them!

Nothing like a definitive 40min no name pool test to prove things.  

RX Vega has 2 stacks of 4GB HBM2 memory modules - how would it have Dag size memory problem?

You can stick the entire Ethereum Dag file on one module or the other module.  

I not going to waste extra money on electricity for 1 week of testing, when most people have figured out those hashrate screenshots are fakes for Ethereum mining.

HBM is stacked memory.  I don't think its one 4GB Module its like four 1GB modules stacked ontop of each other.

You do realise the R9 290 4GB was not affected by the Dag File size issue this year!

[miner1337]

Vega 64 here.  

All the numbers below are dual mining eth+dcr.  

I also feel like the blockchain driver hashrate in Claymore is not real.  Two weeks of mining on ethermine with blockchain drivers and stuck at 24hr avg of 32 MH but reported is always 37 MH.  Perhaps that is expected?  32 MH is what I get with non-blockchain drivers and at least over 48 hours I got 24 avg of 32MH, perhaps that was luck?  

I run at -22% power and otherwise leave clock speed and voltage alone.  Not by choice.  Any changes to voltage wrecks hashrate.

On blockchain drivers.  Anytime I adjust any voltage or clock speed on either core or memory the cards hashrate drops ~-3MHs and gets stuck there.  Memory locks to 800Mhz on anything higher than 945Mhz.   I have not tried non-blockchain drivers.  Given I hear folks are hitting 40+ with non-blockchain drivers in recent news I might have to give it a try.

[P00P135]

Got bad news on RX Vega 56 - managed to do a test with the Beta Blockchain Drivers (solely released to fix the Dag file memory issue for the RX 580/480 and 570/470 GPUs).

RX Vega 56

-24% Powertune
HMB2 925mhz
Default Clock ranging between 1136mhz to 1260mhz.
PSU Efficiency is 90%
Power Draw 152watts

Miner produced a Claymore result of 35.4mhs for 153watts.

Pool shares 40 minutes never above the rate for 36mhs.
Pool shares next 40 minutes around the rate for 25.5mhs.

Average pool shares found averaged around 31mhs for Ethereum - the same as the main official drivers. The Beta Blockchain Drivers are for the RX 580s and RX 570s, when used on RX Vega GPUs they produce fake hashrate figures, which are incorrect by 4mhs.

Beta Blockchain drivers do reduce powerdraw by 13watts per GPU, compared to the official drivers for RX Vega 56. This is why people are using them!

Nothing like a definitive 40min no name pool test to prove things.  

RX Vega has 2 stacks of 4GB HBM2 memory modules - how would it have Dag size memory problem?

You can stick the entire Ethereum Dag file on one module or the other module.  

I not going to waste extra money on electricity for 1 week of testing, when most people have figured out those hashrate screenshots are fakes for Ethereum mining.

HBM is stacked memory.  I don't think its one 4GB Module its like four 1GB modules stacked ontop of each other.

You do realise the R9 290 4GB was not affected by the Dag File size issue this year!

Im aware.  You do know that r9 290 is totally different memory and 512bit bus, right?  You keep stating stupid facts that have correlation to your arguments lol.  Maybe you should start with why there was a DAG issue with some AMD cards and work your way backwards.

[manotroll]

when my internet falls and back the boards do not return to render as they were rendering before.

do you have any command to restart the miner in case of a drop in yield or to rebound the os?

[maxnaughty]

The MEGA files are infected with viruses / trojans, but the GOOGLE ones are fine....

[redmonski]

@claymore, sir please add support for pascalcoin solomining. 

[tharp]

I don't know if it is specific to me, but I also have issues installing drivers in win10 for vega.  I get one driver installation, after that something always goes wrong and I have to reinstall windows.

Awesome it's not just me. I have the same issue's that I get 1 installed fine anything after that it's a 50/50 chance that the driver installs correctly for the other 9 Vega gpu's. I haven't tried another driver yet besides the Aug23 blockchain driver. Have you? Trying to compare notes to see what else to test to try and get them hashing. I've spent WAY to much time on get more that 1 Vega GPU working...

I'm running 2x Vega FE.  The public drivers I've run are Vega FE release drivers, Vega FE beta(public) and the aug. 23 block chain drivers.  

I got both cards up to 42.4Mh/s last night.  Windows updated and restarted the computer and I can't get them past 36mh/s now.  Very frustrating.

(edit) speed for 8 RX vega is 319/MHs and 10AMPS

Well FUCK this but her ya go found the fix to this... Get Windows installed and install team viewer from there and install drivers from there. Only install 8 GPU's more than that it will CRASH on you when you do the driver install. Don't be afraid to reboot the system while the driver install isn't going on a you can't connect via team viewer. Also apply claymore clocks and mem clocks first then use overdriveN about 1min after to fix the clock speeds.

https://imgur.com/a/CkRan

regards,
tharp

[Metroid]

@Claymore, anyway you could add a reset overclock settings command at the beginning of the program? So we don't need to wait the card to crash till new settings are applied or maybe a keyboard command in which will reset the overclock settings to default and will exit the program?

[rednoW]

Vega 64 here.  

All the numbers below are dual mining eth+dcr.  

I also feel like the blockchain driver hashrate in Claymore is not real.  Two weeks of mining on ethermine with blockchain drivers and stuck at 24hr avg of 32 MH but reported is always 37 MH.  Perhaps that is expected?  32 MH is what I get with non-blockchain drivers and at least over 48 hours I got 24 avg of 32MH, perhaps that was luck?  

I run at -22% power and otherwise leave clock speed and voltage alone.  Not by choice.  Any changes to voltage wrecks hashrate.

On blockchain drivers.  Anytime I adjust any voltage or clock speed on either core or memory the cards hashrate drops ~-3MHs and gets stuck there.  Memory locks to 800Mhz on anything higher than 945Mhz.   I have not tried non-blockchain drivers.  Given I hear folks are hitting 40+ with non-blockchain drivers in recent news I might have to give it a try.

blockchain drivers have buggy clocks/voltage control and monitoring but there IS a way to make a serious downvolt without loosing the speed. As I already posted in vega thread - 44.2 mhs @ 160watt from the wall (1408gpu/1100hbm2@0.9v).
There IS a big problem to make clocks low with this drivers so I still need to keep them high (1408mhz) and this is bad for dual-mining.
To be stable at 1408gpu/1100hbm2 in dual mode I need 0.975v core voltage and this leads to high 300watt consumption (43eth+240lbc)

new 17.8.2 and 17.9.1 drivers have clocks/voltage issues fixed but speed is not so good and you need some hack to keep memory oc (1100 instead of 800) when running low gpu clocks.
Max I was able to get is 43 mhs @140-150 watt (1000gpu/1100hbm2@0.85v) and 41mhs+210lbry @230watt (1300gpu/1100hbm2@0.9v)

[adaseb]

The MEGA files are infected with viruses / trojans, but the GOOGLE ones are fine....

This is why I kept repeating over and over and over again to Claymore.

POST THE HASHES OR THE SIGNATURES OF YOUR EXECUTABLES and he never listens.

His MEGA account might of been compromised.

[Orielres]

HBM is stacked memory.  I don't think its one 4GB Module its like four 1GB modules stacked ontop of each other.

You do realise the R9 290 4GB was not affected by the Dag File size issue this year!

Does it mean there is no need to upgrade the drive for 290/390 owers?

[mamoEB]

HBM is stacked memory.  I don't think its one 4GB Module its like four 1GB modules stacked ontop of each other.

You do realise the R9 290 4GB was not affected by the Dag File size issue this year!

Does it mean there is no need to upgrade the drive for 290/390 owers?

The 4GB limit will be hit around mid-September 2019

[vasilykozlov]

hey Guys. what is more profitable now? eth+lbc, or eth+sc?

[DaminouTav]

when my internet falls and back the boards do not return to render as they were rendering before.

do you have any command to restart the miner in case of a drop in yield or to rebound the os?

I have Teamviewer installed on my RIG to do tasks like this ^^
But it's strange, when my internet falls Claymore restart mining by itself ..

[thevictimofuktyranny]

Vega 64 here.  

All the numbers below are dual mining eth+dcr.  

I also feel like the blockchain driver hashrate in Claymore is not real.  Two weeks of mining on ethermine with blockchain drivers and stuck at 24hr avg of 32 MH but reported is always 37 MH.  Perhaps that is expected?  32 MH is what I get with non-blockchain drivers and at least over 48 hours I got 24 avg of 32MH, perhaps that was luck?  

I run at -22% power and otherwise leave clock speed and voltage alone.  Not by choice.  Any changes to voltage wrecks hashrate.

On blockchain drivers.  Anytime I adjust any voltage or clock speed on either core or memory the cards hashrate drops ~-3MHs and gets stuck there.  Memory locks to 800Mhz on anything higher than 945Mhz.   I have not tried non-blockchain drivers.  Given I hear folks are hitting 40+ with non-blockchain drivers in recent news I might have to give it a try.

blockchain drivers have buggy clocks/voltage control and monitoring but there IS a way to make a serious downvolt without loosing the speed. As I already posted in vega thread - 44.2 mhs @ 160watt from the wall (1408gpu/1100hbm2@0.9v).
There IS a big problem to make clocks low with this drivers so I still need to keep them high (1408mhz) and this is bad for dual-mining.
To be stable at 1408gpu/1100hbm2 in dual mode I need 0.975v core voltage and this leads to high 300watt consumption (43eth+240lbc)

new 17.8.2 and 17.9.1 drivers have clocks/voltage issues fixed but speed is not so good and you need some hack to keep memory oc (1100 instead of 800) when running low gpu clocks.
Max I was able to get is 43 mhs @140-150 watt (1000gpu/1100hbm2@0.85v) and 41mhs+210lbry @230watt (1300gpu/1100hbm2@0.9v)

That would be on UBIQ then?

And that is RX Vega 64 and not RX Vega 56?

[rednoW]

That would be on UBIQ then?

And that is RX Vega 64 and not RX Vega 56?

It is vega 64 liquid. I tested only eth with dagger-hashimoto.

[thevictimofuktyranny]

That would be on UBIQ then?

And that is RX Vega 64 and not RX Vega 56?

It is vega 64 liquid. I tested only eth with dagger-hashimoto.

And, are you confident you will be able to run it at that for 2 years 24/7 or was it just something you just did for benchmarking?

[rednoW]

That would be on UBIQ then?

And that is RX Vega 64 and not RX Vega 56?

It is vega 64 liquid. I tested only eth with dagger-hashimoto.

And, are you confident you will be able to run it at that for 2 years 24/7 or was it just something you just did for benchmarking?

I can't tell you cause now I'm running xmr 1600h/s @130 watt, 2 days stable ...

[thevictimofuktyranny]

That would be on UBIQ then?

And that is RX Vega 64 and not RX Vega 56?

It is vega 64 liquid. I tested only eth with dagger-hashimoto.

And, are you confident you will be able to run it at that for 2 years 24/7 or was it just something you just did for benchmarking?

I can't tell you cause now I'm running xmr 1600h/s @130 watt, 2 days stable ...

Can't fault that decision with those UV and OC settings, but the pools servers and implementations are terrible for Monero mining pools!

Ethereum mining pools are vastly superior - you can get within 1% of mining calculators estimates.

[rednoW]

Here it is independent post https://bitcointalk.org/index.php?topic=2002025.msg21691298#msg21691298 and some posts next confirming that 43-44 mhs eth @ 160watt is current Vega "not bad" result ))