Author

Topic: [Cast XMR] high speed XMR/CryptoNight miner for RX Vega GPUs (2 KHash/s) - page 149. (Read 206388 times)

newbie
Activity: 88
Merit: 0
New version, good. I will test ASAP.
With 0.4, i have 1834 with MSI Vega 64 (and recommended setting).
Very stable.

Vega64 is now doing 2150h/s xmr with xmr-stak-amd.
1408gpu/1150mem@875mV, approx 150watt
How did you set 1150 for mem?
Check overclock.net vega bios thread Wink
Thanks. Found it  Smiley
legendary
Activity: 1510
Merit: 1003
New version, good. I will test ASAP.
With 0.4, i have 1834 with MSI Vega 64 (and recommended setting).
Very stable.

Vega64 is now doing 2150h/s xmr with xmr-stak-amd.
1408gpu/1150mem@875mV, approx 150watt
How did you set 1150 for mem?
Check overclock.net vega bios thread Wink
newbie
Activity: 88
Merit: 0
New version, good. I will test ASAP.
With 0.4, i have 1834 with MSI Vega 64 (and recommended setting).
Very stable.

Vega64 is now doing 2150h/s xmr with xmr-stak-amd.
1408gpu/1150mem@875mV, approx 150watt
How did you set 1150 for mem?
legendary
Activity: 1510
Merit: 1003
New version, good. I will test ASAP.
With 0.4, i have 1834 with MSI Vega 64 (and recommended setting).
Very stable.

Vega64 is now doing 21502050h/s xmr with xmr-stak-amd.
1408gpu/1150mem@875mV, approx 150watt
full member
Activity: 162
Merit: 100
New version, good. I will test ASAP.
With 0.4, i have 1834 with MSI Vega 64 (and recommended setting).
Very stable.
hero member
Activity: 935
Merit: 1001
I don't always drink...
Just as an experiment I thought I'd see how well 4x Vega 56 would work in a win 10 system with 8gb RAM and a 65GB dedicated swap drive for the virtual memory.  Turns out that the HBM2 slider in Radeon Settings will only slide over to about 12 GB and that the system won't run with 4 cards for too long.  It is stable for 3 cards, however.
member
Activity: 71
Merit: 10
Thanks for a new powerful Vega miner! :-)

But for me it doesn't work with my Vega 56.
First of all, i have to explain that my system has as first GPU a Nvidia GTX 970 and second the RX Vega 56.
If i start the miner without the -G parameter, i get the error message:

"compute Driver NOT detected."

and my GTX is listed ("GPU0: GeForce GTX 970 | 13 Compute Units")
If i start with -G 1 (or -G 2, -G 3 and so on), everytime I get the following errer message:

Selected OpenCL device index 1 doesn't exist.

Isn't that scenario supported by this mining software? Anyone have a similar issue?

Thanks.


Use --opencl argument, see this post. https://bitcointalksearch.org/topic/m.23295977
newbie
Activity: 11
Merit: 0
Allocate more Virtual memory
YEAH!! That's it 8 GPU work like a work like a charm!!!!!! Thanks rednoW
but some GPU still has a lower hash rate (~1100-1300) even i toggle and apply the all the HBCC

It is a known problem for > 4 amd gpus with blockchain drivers.
You can try latest http://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-ReLive-Edition-Beta-for-Windows-10-Fall-Creators-Update-Release-Notes.aspx and try to enable both Compute mode and HBCC in Readeon Settings, maybe they fixed it ... but I'm not sure.
If you are going to try then don't forget to competely clean old drivers before install with DDU or amd cleanup utility cause new drivers use different power managment and can be uncompatible with softpowerplay mod used for old blockchain drivers.

Thanks for your reply, i will try it!
hmm..And now I face the other issus what the console is work but no more refresh(no share send to the pool anymore) while running...

Update 1:
Brief : it would crash when one of the GPU is down and continued more than 1 mins.
One GPU is not working (0 H/s) constantly in about 1mins before this issue comes up.
The GPU Clock of that GPU is frozen at 1530mhz.

Update 2:
Brief : the console frozen without any issue.
Now 2 rigs are running cast_xmr.
miner2 currently working fine(at the top)
miner1(at the bottom) the console no freshing and when i press 'q' the console refresh all previous unshow status then quite the app.
https://photos.app.goo.gl/akNyF4JvOCeX9uBq2

Update 3:
Brief : The app will crash very soon when one the of the GPU's temp value shows 511.
crash if i press 'q' or close the app or restart the comp..
The driver not working after crash and require to re-install driver.

About the HBCC setting:
I noticed that only 4 cards can change and apply a new value.
My 2 rig (8 vega56 & 6 vega56) can only get  4 cards per each which has a optimized hashrate(19xx)
legendary
Activity: 1510
Merit: 1003
Allocate more Virtual memory
YEAH!! That's it 8 GPU work like a work like a charm!!!!!! Thanks rednoW
but some GPU still has a lower hash rate (~1100-1300) even i toggle and apply the all the HBCC

It is a known problem for > 4 amd gpus with blockchain drivers.
You can try latest http://support.amd.com/en-us/kb-articles/Pages/Radeon-Software-Crimson-ReLive-Edition-Beta-for-Windows-10-Fall-Creators-Update-Release-Notes.aspx and try to enable both Compute mode and HBCC in Readeon Settings, maybe they fixed it ... but I'm not sure.
If you are going to try then don't forget to competely clean old drivers before install with DDU or amd cleanup utility cause new drivers use different power managment and can be uncompatible with softpowerplay mod used for old blockchain drivers.
full member
Activity: 387
Merit: 102
2800-3200 RPM on VGA FAN its about 40-50%? Its make more noise?
newbie
Activity: 11
Merit: 0
hmm..always gets lower hash on specific GPU (raiser seems ok, and i have toggle and apply the HBCC Memory Segment )

OS : win 10 pro
Driver [Suspicious link removed]pute Driver Version 17.30.1029-170808a2-317304E-CrimsonReLive
GPU : 6 x MSI vega56
Overclock : -4%,945,-10%

https://lh3.googleusercontent.com/mN2AqnJ-hlJq0lds_N02_3nkaXP2SSSPDfRKyUxZSZTXWVrh5mbzidchaR_Kbdyi1wFYY7KyKg_PgYolnz4T7rhdMfkwyPel5zJfXU6Jg-Ur5qv9TyKYYvZw6l4loHL58VSdAabetHk

One has to toggle the HBCC Mememory Segment setting for each GPU (really annoying), those giving below 1800 Hash/s (GPU1, GPU3, GPU5) seems not have been toggled.

Thanks for reply, in facts i have toggled all of them.. and finally i solved the problem by replacing GPU1,3,5's raiser..

The next coming problem
when i assign more than 6 GPU and i got a message : CL_MEM_OBJECT_ALLOCATION_FAILURE when calling clEnqueueNDRangeKernel for kernel 0
hmmm.... i will use the claymore (8 GPU is working fine and stable but only 13xxh/s) until this problem is fixed.. Waiting for your solution!

Allocate more Virtual memory


YEAH!! That's it 8 GPU work like a work like a charm!!!!!! Thanks rednoW

but some GPU still has a lower hash rate (~1100-1300) even i toggle and apply the all the HBCC
newbie
Activity: 2
Merit: 0
Thanks for a new powerful Vega miner! :-)

But for me it doesn't work with my Vega 56.
First of all, i have to explain that my system has as first GPU a Nvidia GTX 970 and second the RX Vega 56.
If i start the miner without the -G parameter, i get the error message:

"compute Driver NOT detected."

and my GTX is listed ("GPU0: GeForce GTX 970 | 13 Compute Units")
If i start with -G 1 (or -G 2, -G 3 and so on), everytime I get the following errer message:

Selected OpenCL device index 1 doesn't exist.

Isn't that scenario supported by this mining software? Anyone have a similar issue?

Thanks.
full member
Activity: 252
Merit: 100
whats your real HashRate on Vega64 now with new version?Huh
 over 2khs?Huh?
because my is 1950

thanks a lot
legendary
Activity: 1510
Merit: 1003
hmm..always gets lower hash on specific GPU (raiser seems ok, and i have toggle and apply the HBCC Memory Segment )

OS : win 10 pro
Driver [Suspicious link removed]pute Driver Version 17.30.1029-170808a2-317304E-CrimsonReLive
GPU : 6 x MSI vega56
Overclock : -4%,945,-10%



One has to toggle the HBCC Mememory Segment setting for each GPU (really annoying), those giving below 1800 Hash/s (GPU1, GPU3, GPU5) seems not have been toggled.

Thanks for reply, in facts i have toggled all of them.. and finally i solved the problem by replacing GPU1,3,5's raiser..

The next coming problem
when i assign more than 6 GPU and i got a message : CL_MEM_OBJECT_ALLOCATION_FAILURE when calling clEnqueueNDRangeKernel for kernel 0
hmmm.... i will use the claymore (8 GPU is working fine and stable but only 13xxh/s) until this problem is fixed.. Waiting for your solution!

Allocate more Virtual memory
newbie
Activity: 11
Merit: 0
hmm..always gets lower hash on specific GPU (raiser seems ok, and i have toggle and apply the HBCC Memory Segment )

OS : win 10 pro
Driver [Suspicious link removed]pute Driver Version 17.30.1029-170808a2-317304E-CrimsonReLive
GPU : 6 x MSI vega56
Overclock : -4%,945,-10%

https://lh3.googleusercontent.com/mN2AqnJ-hlJq0lds_N02_3nkaXP2SSSPDfRKyUxZSZTXWVrh5mbzidchaR_Kbdyi1wFYY7KyKg_PgYolnz4T7rhdMfkwyPel5zJfXU6Jg-Ur5qv9TyKYYvZw6l4loHL58VSdAabetHk

One has to toggle the HBCC Mememory Segment setting for each GPU (really annoying), those giving below 1800 Hash/s (GPU1, GPU3, GPU5) seems not have been toggled.

Thanks for reply, in facts i have toggled all of them.. and finally i solved the problem by replacing GPU1,3,5's raiser..

The next coming problem
when i assign more than 6 GPU and i got a message : CL_MEM_OBJECT_ALLOCATION_FAILURE when calling clEnqueueNDRangeKernel for kernel 0
hmmm.... i will use the claymore (8 GPU is working fine and stable but only 13xxh/s) until this problem is fixed.. Waiting for your solution!

Updated:
It will also popup a message box and show the following message : The ram of the comp is not enough. Please close the application.
newbie
Activity: 13
Merit: 0
thanks a lot work good
newbie
Activity: 4
Merit: 0
2 codencoin
Your card's HBM memory is overheated and starts to throttle
Apply downvolt patch and make fan minimum 2800rpm

Thanks for the suggestion... turns out it was some kind of windows power setting that got me. I didn't have it set to high power. Found another thread that someone had the same kind of problem and changed their monitor to sleep "never" and it was fixed. I Adjusted all the power settings and it was fixed. Getting almost 1900 full time now.
member
Activity: 75
Merit: 10
With the AMD Blockchain driver available on Linux, any updates for Cast XMR in that regard?
member
Activity: 91
Merit: 29
[13:27:47] Shares: 681 Accepted, 19 Rejected | Total Hash Rate: 3959.2 H/s | Avg Search Time: 255.2 sec

Still finding that I am getting lots of Rejected shares.

It seems to be the same problems as was described earlier in the thread where the shares are expired and still completed and submitted anyway.

Try lowering your memclock.

That is a high rejection rate and I had the same issue when I ran my memclock too high. Doing 920 for memclock now and seeing about 1.5% rejected.
Code:
[16:04:39] Shares: 759 Accepted, 11 Rejected | Total Hash Rate: 5710.5 H/s | Avg Search Time: 35.2 sec

All there cards doing around 1900h/s.


The hashes are checked by CPU before sending to pool, as long as you don't see no "GPU error - nonce incorrect" after "Found Nonce" the GPU is working fine. When writing "submitting..." it is technical valid and sent to pool, but the pool is already working on a new job...
member
Activity: 71
Merit: 10
[13:27:47] Shares: 681 Accepted, 19 Rejected | Total Hash Rate: 3959.2 H/s | Avg Search Time: 255.2 sec

Still finding that I am getting lots of Rejected shares.

It seems to be the same problems as was described earlier in the thread where the shares are expired and still completed and submitted anyway.

Try lowering your memclock.

That is a high rejection rate and I had the same issue when I ran my memclock too high. Doing 920 for memclock now and seeing about 1.5% rejected.
Code:
[16:04:39] Shares: 759 Accepted, 11 Rejected | Total Hash Rate: 5710.5 H/s | Avg Search Time: 35.2 sec

All there cards doing around 1900h/s.

Jump to: