SRBMiner Cryptonight AMD GPU Miner V1.9.3 - native algo switching

[willycut]

Hi, with version 1.5.4 i have 10.8 kh/s but with latest version 1.5.5  i got 10.68 and some job timeouts, same settings.
Also get more pool rejects.

[1679596676]

1679596676

[willycut]

@doktor83.  Is there a virtual memory requirement or free ssd/hdd space requirement when compiling the kernels and copying over the device?

Virtual mem is same as for every other Cryptonight miner(the more, better ) , kernel compile happens on GPU, after that it's written to the hdd, miner directory, and it takes up approx 2.1-2.3mb / binary.
Or what was your question ?

Cause this is weird behavior I observed today.  Same I/W/T settings 59/8/1 on 4gb RX 570 which was working fine.  Restarting the miner it gave the error related to below error message,
https://i.imgur.com/TqAdUMm.png

sometimes drivers or windows behaves strange, i bet if you delete the .srb files and reboot it will all be good.

Nope deleting the .srb file didn't solve it.  Rebooting did.  it seems more like a driver/miner didn't clean up memory properly?

probably. these kind of 'errors' happen a lot after driver crash/recovery.

Invalid buffer size usually means you're trying to use an intesity that is too high for the card. Happens often on CN-Heavy coins. That's probably the case since you're using 4GB cards.

Is that so?  Check my post again.  I mentioned it was working and didn't work later during relaunch.
Too lazy to take another screenshot but here are 6 x 4gb card doing the same settings on CNH.  Anyways,  I got the RX570's running at 59 intensity too.
https://i.imgur.com/lavRfGo.jpg

I mine with 59.5 intensity with rx's 570 4gb, what i found is we have to increase virtual memory with 13 cards i have set 76 gb and all is fine.

[lebuawu2]

new driver 18.5.1 increase hashrate CN-V7 about 80 h/s for RX Vega. CN-Heavy directly with intensity 56 will get more than 1400 h/s but still a bit behind compare with intensity 60 and go back to intensity 56.

[heavyarms1912]

@doktor83.  Is there a virtual memory requirement or free ssd/hdd space requirement when compiling the kernels and copying over the device?

Virtual mem is same as for every other Cryptonight miner(the more, better ) , kernel compile happens on GPU, after that it's written to the hdd, miner directory, and it takes up approx 2.1-2.3mb / binary.
Or what was your question ?

Cause this is weird behavior I observed today.  Same I/W/T settings 59/8/1 on 4gb RX 570 which was working fine.  Restarting the miner it gave the error related to below error message,

sometimes drivers or windows behaves strange, i bet if you delete the .srb files and reboot it will all be good.

Nope deleting the .srb file didn't solve it.  Rebooting did.  it seems more like a driver/miner didn't clean up memory properly?

probably. these kind of 'errors' happen a lot after driver crash/recovery.

Invalid buffer size usually means you're trying to use an intesity that is too high for the card. Happens often on CN-Heavy coins. That's probably the case since you're using 4GB cards.

Is that so?  Check my post again.  I mentioned it was working and didn't work later during relaunch.
Too lazy to take another screenshot but here are 6 x 4gb card doing the same settings on CNH.  Anyways,  I got the RX570's running at 59 intensity too.
https://i.imgur.com/lavRfGo.jpg

I mine with 59.5 intensity with rx's 570 4gb, what i found is we have to increase virtual memory with 13 cards i have set 76 gb and all is fine.

Yup I know it works on 59.5 as well.  Just that 59 is good enough for me. 
It wasn't virtual memory.  I have already figured out the reason and mentioned in my previous posts.  Virtual memory is set fine and not even used.  I have set it at 40GB which works fine (8 x 4GB + 1 x 8GB Nvidia).

[doktor83]

Hi, with version 1.5.4 i have 10.8 kh/s but with latest version 1.5.5  i got 10.68 and some job timeouts, same settings.
Also get more pool rejects.

nothing that can impact mining speed has been changed from 1.5.4, also a log would be nice to see too.
Which pool, and which share reject reason ?

[rusman2]

Is that so?  Check my post again.  I mentioned it was working and didn't work later during relaunch.
Too lazy to take another screenshot but here are 6 x 4gb card doing the same settings on CNH.  Anyways,  I got the RX570's running at 59 intensity too.
https://i.imgur.com/lavRfGo.jpg

I replied in the reddit thread, but will repeat it here in case others won't see that thread.
It's the difference between 570 4GB and 5804GB. I believe that the increased cores, TMUs and CUs allow the 580 to use higher intensity. There also might be more of the 4GB memory able to be utilized for processing on the 580.

Card Type                                 RX570         vs           RX580
Shader Processing Units (cores)   2048      vs               2304
Texture Mapping Units                 128      vs                 144
Compute Units                             32          vs               36

https://www.techpowerup.com/gpudb/2938/radeon-rx-580
https://www.techpowerup.com/gpudb/2939/radeon-rx-570

There also might be more of the 4GB memory able to be utilized for processing on the 580. Without running an OpenCL command, you can look at GPU-Z Advanced tab > OpenCL drop down. Scroll down and it lists "Max Memory Allocation". All of my cards each show 3264MB (or 3422552064 bytes if you calculate it out). Can you look at your 580 4GB cards and see what this shows for them?

[hesido]

Really nice update, will try it as soon as possible (after a crash or before a pool switch )

Doktor, how fast does the vmem have to be? Should it better be on the SSD, or can I use a cheap hard drive just to support the hashing process... Does the miner have to access the vmem so often and so frequent that the only viable option is an ssd? I'd be really glad to know, because I don't want to buy a second ssd just to accomodate the vmem.

[Vers]

with 1.5.5 sometimes there are errors (Invalid job id). 1.5.4 works without errors.
http://i105.fastpic.ru/big/2018/0524/5a/4d767ea5c53537545597a8bc1ce01c5a.jpg
http://i104.fastpic.ru/big/2018/0524/08/a153fbcf6f27e67228c6561f7c16fb08.jpg

[alibabacool]

Hi, Doktor!
In new version (1.5.5) there is some bug:
Sometimes I get errors (invalid job id)
https://drive.google.com/file/d/1to6uThqMp075p1E5Zn0WEe784GIW00vy/view
https://drive.google.com/file/d/1AGkC5GfAXUAzn6tV4K4Occ7fN0Hx-Com/view
I get this on diffrent pools and algos.

[madmossy]

Had exactly the same issue, pool rejected shares (invalid job id) on 3 different rigs.

Reverted back to 1.5.4 and all is good again.

[UAGet]

Nope deleting the .srb file didn't solve it.  Rebooting did.  it seems more like a driver/miner didn't clean up memory properly?

Try this driver. I tested a lot of drivers (adrenaline, blockchain, etc.) - these drivers are the best for my graphics cards. Maybe it will help you.

[UAGet]

Hi, Doktor!
In new version (1.5.5) there is some bug:
Sometimes I get errors (invalid job id)
https://drive.google.com/file/d/1to6uThqMp075p1E5Zn0WEe784GIW00vy/view
https://drive.google.com/file/d/1AGkC5GfAXUAzn6tV4K4Occ7fN0Hx-Com/view
I get this on diffrent pools and algos.

I also confirm this error

[Sgsg666]

Hi, Doktor!
In new version (1.5.5) there is some bug:
Sometimes I get errors (invalid job id)
https://drive.google.com/file/d/1to6uThqMp075p1E5Zn0WEe784GIW00vy/view
https://drive.google.com/file/d/1AGkC5GfAXUAzn6tV4K4Occ7fN0Hx-Com/view
I get this on diffrent pools and algos.

Usually, "errors" happen when your GPUs aren't underclocked/overclocked on stable settings. very high difficulty, high ping to pool. different cards for each rig could also affect(cant tell yet). Try setting static difficulty.

[abg00]

Hi, Doktor!
In new version (1.5.5) there is some bug:
Sometimes I get errors (invalid job id)
https://drive.google.com/file/d/1to6uThqMp075p1E5Zn0WEe784GIW00vy/view
https://drive.google.com/file/d/1AGkC5GfAXUAzn6tV4K4Occ7fN0Hx-Com/view
I get this on diffrent pools and algos.

I also confirm this error

same error in normalv7 mode
ok in heavy mode

[UAGet]

very high difficulty, high ping to pool.

Difficulty: 1.95 G
Algo:cryptonight-heavy

[doktor83]

I guess i know what's causing this, will fix it in next ver.

You guys get a job timeout before this, am i right?

Are you setting it manually in pools config?

If not, can you try to put "job_timeout" : 900 , that is 15 minutes, and test to see if you still get timeouts and invalid job rejects?

If no timeouts i will know this was the problem
thanks

example:

pools.config

{
"pools" :
[
   {"pool" : "xmr-eu1.nanopool.org:14444", "wallet" : "4A5hJyu2FvuM2azexYssHW2odrNCNWVqLLmzCowrA57xGJLNufXfzVgcMpAy3YWpzZSAPALhVH4Ed7x o6RZYyw2bUtbm12g.donation", "password" : "x", "job_timeout" : 900}
]
}

[willycut]

I guess i know what's causing this, will fix it in next ver.

You guys get a job timeout before this, am i right?

Are you setting it manually in pools config?

If not, can you try to put "job_timeout" : 900 , that is 15 minutes, and test to see if you still get timeouts and invalid job rejects?

If no timeouts i will know this was the problem
thanks

example:

pools.config

{
"pools" :
[
   {"pool" : "xmr-eu1.nanopool.org:14444", "wallet" : "4A5hJyu2FvuM2azexYssHW2odrNCNWVqLLmzCowrA57xGJLNufXfzVgcMpAy3YWpzZSAPALhVH4Ed7x o6RZYyw2bUtbm12g.donation", "password" : "x", "job_timeout" : 900}

]
}

I thought the same and increased it to 420...

[doktor83]

I thought the same and increased it to 420...

try something bigger, def is 300. Like 600 or 900

[nordmann666]

are you in touch with developer of minerguarddog?

i have problems with monitoring the hashrate cause "Warning: Minimum Hashrate threshold in invalid. Hashrate threshold check is disabled"

[doktor83]

are you in touch with developer of minerguarddog?

nope,you should contact him.