Author

Topic: Any word on amd vega hash rates? - page 127. (Read 202731 times)

legendary
Activity: 1510
Merit: 1003
September 23, 2017, 03:05:01 AM
Hi All. Looking to hopefully resolve an issue I have using Claymore's 9.7 miner.
Whenever I boot up the software, I receive a, "Error: AMD HSA Code Object loading failed".
I have browsed the Wiki, this page, and other sites and cannot find a reason behind the error nor a solution.
Here's to problem solving!
Claymore is using direct code encapsulation in his miners so you need to wait for next version of his xmr miner that will support Vega.
full member
Activity: 252
Merit: 100
September 23, 2017, 02:50:12 AM
Hi All. Looking to hopefully resolve an issue I have using Claymore's 9.7 miner.
Whenever I boot up the software, I receive a, "Error: AMD HSA Code Object loading failed".
I have browsed the Wiki, this page, and other sites and cannot find a reason behind the error nor a solution.
Here's to problem solving!
legendary
Activity: 1510
Merit: 1003
September 23, 2017, 02:33:59 AM
Whats the best hash rate for 580 per watt nowadays.    I was quite impressed to get 32mh on ETH with under 130w and I should be able to improve on that as I didnt do HBCC or use block drivers.    Apparently the memory voltage is still not right but for me dropping it to 900 which I think is the lowest meant that stable clocks raised up quite a bit.  So thats core and HBM and ironically less voltage helps the card do higher clocks which gave me about a 25% boost I guess.

Still havent really tested properly or got round to trying XMR.   I probably should as I have a 50w 8 core broadwell xeon cpu here as well that could probably mine it I was thinking.   For starters, should I use nicehash or is that the wrong track, I heard its easy for noobs :p
Strange, I think I can understand the words in the text above but can't catch the meaning ... maybe cause English is not native for me ...
STT
legendary
Activity: 4088
Merit: 1452
September 23, 2017, 02:22:41 AM
Whats the best hash rate for 580 per watt nowadays.    I was quite impressed to get 32mh on ETH with under 130w and I should be able to improve on that as I didnt do HBCC or use block drivers.    Apparently the memory voltage is still not right but for me dropping it to 900 which I think is the lowest meant that stable clocks raised up quite a bit.  So thats core and HBM and ironically less voltage helps the card do higher clocks which gave me about a 25% boost I guess.

Still havent really tested properly or got round to trying XMR.   I probably should as I have a 50w 8 core broadwell xeon cpu here as well that could probably mine it I was thinking.   For starters, should I use nicehash or is that the wrong track, I heard its easy for noobs :p
legendary
Activity: 1764
Merit: 1002
September 23, 2017, 02:09:24 AM
Are Vegas worth it compared to 480/580? I mean price to hashrate.
with xmr performance right now, vega its better
sr. member
Activity: 244
Merit: 250
September 23, 2017, 12:20:23 AM
Are Vegas worth it compared to 480/580? I mean price to hashrate.
newbie
Activity: 36
Merit: 0
September 23, 2017, 12:13:54 AM
Alright actual question here.  Might be worth another thread.

I restarted my rig.  4/6 Vega 64's are hashing at 1900+ still.  2 are not.  No matter what I do with the HBCC slider for those cards, enable, disable, I cannot get the hash rate back.  The card is otherwise clocked correctly, I get the same ETH hash rate as I would expect.  

Any tips?  Full clean re-install of AMD drivers?  What I wouldn't do for some non-flashy software that would let you control these settings.
I think adding extra 8gb of RAM will make things easier.
Also I think in fact your cards aren't running 1000mhz gpu clock whatever monitoring software says. It is much higher. You wouldn't get 1900+h/s with only 1000mhz gpu clock.
And actual gpu voltage also is not 950mV

As for clock speeds and voltages.  I agree, mostly.  I mean Wattman and HWMonitor both claim the clock speed is 1000mhz when I am hashing at 1900H.  Although I appreciate both could be wrong.  Perhaps that is why I feel my power draw is to high.  1450W with the settings I previously linked, 300W idle, means roughly 191W per card, definitely to high.

So any tips for checking the actual speeds?


As for the HBCC issue.  I have acquired 8GB more RAM to test this later today.  It turns out on the 2 problematic cards I cannot actual adjust HBCC.  1 card it is disabled and I cannot enable it, period, always stays disabled.  The other card it is enabled but the slider is at minimum, any adjustments to the slider causes it to reset back to minimum.  I am thinking if RAM does not fix it then I will DDU and start from scratch?




Yes, if you tried different drivers it is better to do full uninstall before installing new one. I use amd cleanup utility but not sure if it is a right choice ))

I tried AMD cleanup and DDU.  Went through a fresh install of the blockchain drivers both times.  Same two cards refuse to accept adjustments to their HBCC settings.

Here is what happens.  Anytime I change HBCC for a card the screen flickers, Wattman restarts.  Except for those two cards.  When I change their HBCC setting Wattman just closes.  Given these two problematic cards was at one point working with HBCC I figured it was some driver issue.  I am definitely at a loss now. 

I think it has something to do with Crossfire.  Everytime I reinstall the drivers Crossfire enables.  Disabling it is a pain in the butt, almost always crashes the machine.  Sometimes it changes which card it is crossfired after it crashes.  Finally it will get disabled after a few tries and I can carry on mining.  It is always one of these two cards that the main gpu tries to crossfire with.
hero member
Activity: 615
Merit: 500
September 22, 2017, 11:22:10 PM
I have a Vega64 for sale in the US. Please PM me if interested.

https://bitcointalksearch.org/topic/m.22019288

I am willing to use escrow via Phil.
legendary
Activity: 4256
Merit: 8551
'The right to privacy matters'
September 22, 2017, 06:22:01 PM
Everyone seems mining monero right now with vega...
any other algo that performs well with vega now?

Monero

Is better then other coins.
member
Activity: 125
Merit: 10
September 22, 2017, 06:02:51 PM
Everyone seems mining monero right now with vega...
any other algo that performs well with vega now?
newbie
Activity: 49
Merit: 0
September 22, 2017, 04:56:45 PM

I use Claymore to quickly enable these settings on all cards in case Wattman crashes and restores all to default (screw AMD Wattman crappy interface).


But you still use xmr-stak-amd to mine right?

Mind to share the claymore settings you're using (no need to show values). I just want to stop stressing with wattman UI
full member
Activity: 252
Merit: 100
September 22, 2017, 03:15:05 PM
what pool you use?Huh
and use claymore or xmR-STAK-AMD

who is the best for vega 64???
legendary
Activity: 1510
Merit: 1003
September 22, 2017, 02:34:28 PM
As for clock speeds and voltages.  I agree, mostly.  I mean Wattman and HWMonitor both claim the clock speed is 1000mhz when I am hashing at 1900H.  Although I appreciate both could be wrong.  Perhaps that is why I feel my power draw is to high.  1450W with the settings I previously linked, 300W idle, means roughly 191W per card, definitely to high.

So any tips for checking the actual speeds?

Did you compare wattage and speed running default (except for the mem speed, make it also 1025mhz to equalize with second setup) and your "downvolt" setup? Did you see wattage decrease applying core clock 1000 and core voltage 950?
full member
Activity: 252
Merit: 100
September 22, 2017, 02:18:00 PM
After a week of testing here are my results.

Blockchain drivers.
6xVega 64
memclock 1025
mem voltage 950
coreclock 1000
core voltage 950
power limit seems to have almost no effect with these settings.  Between 0 and -10 I saw no change in power consumption.
HBCC enabled, max for all cards.
Fan speed is ~3200RPM.  The cards in the middle of the rig run hotter and therefore the fan hits ~3400RPM.  The temps sit stables ~66C.

I use Claymore to quickly enable these settings on all cards in case Wattman crashes and restores all to default (screw AMD Wattman crappy interface).

I only have 8GB of System Memory but 64GB of Virtual Memory so I can still use HBCC apparently.  When I am mining with 2016+1600 intensity on 6 cards I use up ~56GB of virtual memory.

I am getting 1950H per card, total 11700H.  At my Kill-O-Watt meter I read 1450W.  Idle is 300W.  Clearly I can do better power wise but I am not sure how.

Every now and then one card drops down and I have to disable/enable HBCC on every card, this has only happened once in the past week.

The pool I am using reflects these numbers, after a week I am seeing on average slightly higher than 11700.

Hopefully this helps others.  All for sharing information.  The virtual memory trick makes me think we might be able to bump up intensity even further.


hi
how i set core voltage and mem voltage??

i use msi afterburn and not see this settings
member
Activity: 202
Merit: 10
Eloncoin.org - Mars, here we come!
September 22, 2017, 12:29:49 PM
Can potentially save me a lot of time later.

When using AMD Cleanup or DDU, should I:
- remove all but 1 GPU
- Disable 4G Encoding
- boot to safe mode. 
- run cleanup
- Then reinstall as normal

Or can I just do it all from a normal windows boot without rebooting.  I imagine I'll get a black screen if I run DDU and reboot with 4G Encoding enabled. 

I know, I know, I should try it.  12 hours from now I will. 

Generally speaking I do not remove any of cards. Just leave them there and follow the recommended operation on the DDU. Rebooting is absolutely needed. I do not know 4G setting though. Usually I will leave a card on primary pci-e slot to make sure it will boot up.
newbie
Activity: 36
Merit: 0
September 22, 2017, 12:16:55 PM
Can potentially save me a lot of time later.

When using AMD Cleanup or DDU, should I:
- remove all but 1 GPU
- Disable 4G Encoding
- boot to safe mode. 
- run cleanup
- Then reinstall as normal

Or can I just do it all from a normal windows boot without rebooting.  I imagine I'll get a black screen if I run DDU and reboot with 4G Encoding enabled. 

I know, I know, I should try it.  12 hours from now I will. 
newbie
Activity: 60
Merit: 0
September 22, 2017, 12:05:01 PM
Alright actual question here.  Might be worth another thread.

I restarted my rig.  4/6 Vega 64's are hashing at 1900+ still.  2 are not.  No matter what I do with the HBCC slider for those cards, enable, disable, I cannot get the hash rate back.  The card is otherwise clocked correctly, I get the same ETH hash rate as I would expect.  

Any tips?  Full clean re-install of AMD drivers?  What I wouldn't do for some non-flashy software that would let you control these settings.
I think adding extra 8gb of RAM will make things easier.
Also I think in fact your cards aren't running 1000mhz gpu clock whatever monitoring software says. It is much higher. You wouldn't get 1900+h/s with only 1000mhz gpu clock.
And actual gpu voltage also is not 950mV

As for clock speeds and voltages.  I agree, mostly.  I mean Wattman and HWMonitor both claim the clock speed is 1000mhz when I am hashing at 1900H.  Although I appreciate both could be wrong.  Perhaps that is why I feel my power draw is to high.  1450W with the settings I previously linked, 300W idle, means roughly 191W per card, definitely to high.

So any tips for checking the actual speeds?


As for the HBCC issue.  I have acquired 8GB more RAM to test this later today.  It turns out on the 2 problematic cards I cannot actual adjust HBCC.  1 card it is disabled and I cannot enable it, period, always stays disabled.  The other card it is enabled but the slider is at minimum, any adjustments to the slider causes it to reset back to minimum.  I am thinking if RAM does not fix it then I will DDU and start from scratch?




Yes, if you tried different drivers it is better to do full uninstall before installing new one. I use amd cleanup utility but not sure if it is a right choice ))

I think DDU is the best.
legendary
Activity: 1510
Merit: 1003
September 22, 2017, 11:56:44 AM
Alright actual question here.  Might be worth another thread.

I restarted my rig.  4/6 Vega 64's are hashing at 1900+ still.  2 are not.  No matter what I do with the HBCC slider for those cards, enable, disable, I cannot get the hash rate back.  The card is otherwise clocked correctly, I get the same ETH hash rate as I would expect.  

Any tips?  Full clean re-install of AMD drivers?  What I wouldn't do for some non-flashy software that would let you control these settings.
I think adding extra 8gb of RAM will make things easier.
Also I think in fact your cards aren't running 1000mhz gpu clock whatever monitoring software says. It is much higher. You wouldn't get 1900+h/s with only 1000mhz gpu clock.
And actual gpu voltage also is not 950mV

As for clock speeds and voltages.  I agree, mostly.  I mean Wattman and HWMonitor both claim the clock speed is 1000mhz when I am hashing at 1900H.  Although I appreciate both could be wrong.  Perhaps that is why I feel my power draw is to high.  1450W with the settings I previously linked, 300W idle, means roughly 191W per card, definitely to high.

So any tips for checking the actual speeds?


As for the HBCC issue.  I have acquired 8GB more RAM to test this later today.  It turns out on the 2 problematic cards I cannot actual adjust HBCC.  1 card it is disabled and I cannot enable it, period, always stays disabled.  The other card it is enabled but the slider is at minimum, any adjustments to the slider causes it to reset back to minimum.  I am thinking if RAM does not fix it then I will DDU and start from scratch?




Yes, if you tried different drivers it is better to do full uninstall before installing new one. I use amd cleanup utility but not sure if it is a right choice ))
newbie
Activity: 36
Merit: 0
September 22, 2017, 11:02:10 AM
Alright actual question here.  Might be worth another thread.

I restarted my rig.  4/6 Vega 64's are hashing at 1900+ still.  2 are not.  No matter what I do with the HBCC slider for those cards, enable, disable, I cannot get the hash rate back.  The card is otherwise clocked correctly, I get the same ETH hash rate as I would expect.  

Any tips?  Full clean re-install of AMD drivers?  What I wouldn't do for some non-flashy software that would let you control these settings.
I think adding extra 8gb of RAM will make things easier.
Also I think in fact your cards aren't running 1000mhz gpu clock whatever monitoring software says. It is much higher. You wouldn't get 1900+h/s with only 1000mhz gpu clock.
And actual gpu voltage also is not 950mV

As for clock speeds and voltages.  I agree, mostly.  I mean Wattman and HWMonitor both claim the clock speed is 1000mhz when I am hashing at 1900H.  Although I appreciate both could be wrong.  Perhaps that is why I feel my power draw is to high.  1450W with the settings I previously linked, 300W idle, means roughly 191W per card, definitely to high.

So any tips for checking the actual speeds?


As for the HBCC issue.  I have acquired 8GB more RAM to test this later today.  It turns out on the 2 problematic cards I cannot actual adjust HBCC.  1 card it is disabled and I cannot enable it, period, always stays disabled.  The other card it is enabled but the slider is at minimum, any adjustments to the slider causes it to reset back to minimum.  I am thinking if RAM does not fix it then I will DDU and start from scratch?


full member
Activity: 252
Merit: 100
September 22, 2017, 09:07:05 AM
What is a best for Vega 64 minning??
which coin - monero?
and who soft - claymore is best for vega 64?
Jump to: