Help needed! 4GB GPUs can't be mined on?

[deecept]

Hi,

New user on the forum, thanks for having me =)

Having a weird issue with one of my rigs - see image #1 below
PhoenixMiner/Claymore it seems to show the same issue, but Claymore has more information, see image #2, tokenminer even more (see error text in bottom of post)
Seems to do with the miner not being able to save DAG to GPU memory? Maybe not purged after previous session?
10x 4GB RX570s, the machine has 8GB of RAM.
Running  Win10 with latest driver.
Tested with Claymore, Phoenix, tokenminer, ethminer - Clearly not a miner issue
Testet reinstall driver (with DDU), reboots and increase of swap file to 32GB+.

Anyone seen this before?

Img #1
https://i.imgur.com/H2mbS8K.png

Img #2
https://i.imgur.com/TZVHPKe.png

Including error text for search
In PhoenixMiner: unable to use opencl device radeon rx 570 series for ethash mining
In Claymore: unable to use opencl device radeon rx 570 series for ethash mining / OpenCL error -61 - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.
In tokenminer: OpenCL device Ellesmere  has insufficient GPU memory.536870912 bytes of memory found < 1073739904 bytes
 of memory required
In tokenminer: No GPU device with sufficient memory was found. You may need to change the --opencl-platform argument to 1. (this changes nothing BTW)

[1714961159]

1714961159

[1714961159]

1714961159

[1714961159]

1714961159

[butka]

Anyone seen this before?

Img #1
https://i.imgur.com/H2mbS8K.png

Img #2
https://i.imgur.com/TZVHPKe.png

Including error text for search
In PhoenixMiner: unable to use opencl device radeon rx 570 series for ethash mining
In Claymore: unable to use opencl device radeon rx 570 series for ethash mining / OpenCL error -61 - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.
In tokenminer: OpenCL device Ellesmere  has insufficient GPU memory.536870912 bytes of memory found < 1073739904 bytes
 of memory required
In tokenminer: No GPU device with sufficient memory was found. You may need to change the --opencl-platform argument to 1. (this changes nothing BTW)

I'm not familiar with Radeon mining, but from the error message, I think it is pretty obvious that there's a memory problem. Your graphic cards seem to have enough RAM (4Gb should be enough). Another thing you should check is virtual memory. Have you checked your virtual memory (Page Size), maybe there lies the problem?

[DevelopmentBank]

The error message is pretty descriptive as it and all you have to do is as it says, "check your drivers."

If you really already tried DDU and reinstallation of latest drivers, I would recommend using the Aug. 23 AMD Blockchain Drivers. They should still be available for download online as they are pretty popular for their stability and use in mining.

Also, did you update your windows 10 OS recently? That latest update screwed up everything. Google ethermining and win10 update errors.

[deecept]

Thanks for replying! Really do appreciate the time taken, and help offered =) Response below

I'm not familiar with Radeon mining, but from the error message, I think it is pretty obvious that there's a memory problem. Your graphic cards seem to have enough RAM (4Gb should be enough). Another thing you should check is virtual memory. Have you checked your virtual memory (Page Size), maybe there lies the problem?

As stated in my post, I have tried upping from 24000-32000 MB paging file to 32000 - 48000 MB. No difference.
Yes, I think there is some issue with maybe the motherboard, as the same error is for all cards in the rig, and isn't really anything pointing to this being a software issue.

The error message is pretty descriptive as it and all you have to do is as it says, "check your drivers."

If you really already tried DDU and reinstallation of latest drivers, I would recommend using the Aug. 23 AMD Blockchain Drivers. They should still be available for download online as they are pretty popular for their stability and use in mining.

Also, did you update your windows 10 OS recently? That latest update screwed up everything. Google ethermining and win10 update errors.

I really have tried DDU, and reinstalled drivers. Both new and old. Trying the blockchain driver now, haven't yet because 1. Newest worked fine before this issue, and 2. I have 10 cards in the rig, and the blockchain driver only supports up to 8 GPUs.
Win10 is not updated (disabled updates), and is sitting on version 1709.. this same setup, and versions works across all rigs I own.

[Eisenkopf89]

Did you check your risers? I had the same Problem on my RX550 Rig that the Cards fail in mem. Also i would push up more virtual mem to 60GB+

[markiz73]

Try to buy a USB drive for 8 GB and install the Hive OS on it. The software is free for 3 farms. If in Linux there are failures then it is necessary to search for errors in the computer components

[sxemini]

Mod bios on gpu? Which Mainboard? Which cpu? config files?
DDU in safe mode? we need more infos?
Which claymore version/phönixminer
You try it with 1 gpu and then add the next one and then next and so on.
I think it is a driver issue and not hard to solve.

Check your win 10 version. it is very hard to deactivate windows updates completely. the most descriptions are not sufficient.

@markiz73
sorry stupid idea. there are so many things that he can test before installing a new os XD


sorry for my bad english

[FFI2013]

Try to buy a USB drive for 8 GB and install the Hive OS on it. The software is free for 3 farms. If in Linux there are failures then it is necessary to search for errors in the computer components

I've been running Hiveos for months and it's rock solid on all my rig plus the dev is always pushing updates and keeps the miners updated I recommend hiveos to anyone looking for a stable mining os https://bitcointalk.org/index.php?topic=2350786.0

[markiz73]

@markiz73
sorry stupid idea. there are so many things that he can test before installing a new os XD

sorry for my bad english

I have many colleagues-miners using several systems on farms. This is very convenient in many cases. Windows they load from the hard drive and Linux from the USB drive. But let everyone make his choice.

[nshyamkrishna]

Hi,

Edit your miner starting batch file ( eg. start.bat) and edit the line..

setx GPU_SINGLE_ALLOC_PERCENT

it should be 100 .. not 1 ..

setx GPU_SINGLE_ALLOC_PERCENT 100


good luck.

[swogerino]

Also don't forget to increase the virtual memory as that is the first recommendation from Claymore miner himself. If I were you I would completely reinstall Windows 10 and first thing I would do after that is:

services.msc

scroll down to Windows Update, right click and STOP it, then DISABLE it completely.

I just did this yesterday and my friend computer is running super stable with latest Claymore miner and latest MSI Afterburner together with 18.3.4 ATI driver.

[yhoga45]

I have 4GB cards and it's still mining with ethash algorithm, have you increase the virtual memory size? You can try with that and try with Blockchain Driver August, a lot of people used this driver a long time ago and now I using Adrenaline Driver 18.3.3 and working same like Blockchain Driver but you must change it to compute mode

[crairezx20]

It looks like your GPU, not getting enough power, make sure you fix the cable to your GPU's the power from the riser and GPU is different so make sure to check the 8 or 6 pin power cable to your GPU or replace the risers, usb cable, power cable and make sure to fix these cables properly then start to check again if it works then if not your OS system files are corrupted you must format it again with fresh OS install.

[Mohondoa]

I think that happen because your virtual memory not enough. You can set 16gb minimum and 32gb maximum.
But if you have to do that, i can suggest you need to clean reinstall gpu driver, delete folder amd after DDU.
Good luck.

[leonix007]

best way to troubleshoot

You try it with 1 gpu and then add the next one and then next and so on.

you have 10GPUs installed, possibilities on mem shortage for 4GB RAM, by doing the suggested above, you can prove this

[nshyamkrishna]

best way to troubleshoot

You try it with 1 gpu and then add the next one and then next and so on.

you have 10GPUs installed, possibilities on mem shortage for 4GB RAM, by doing the suggested above, you can prove this

Look at the image 1 he posted guys..

he set the  GPU_SINGLE_ALLOC_PERCENT  value to 1%

thats the actual problem.. If not pls correct me..

[Raja_MBZ]

best way to troubleshoot

You try it with 1 gpu and then add the next one and then next and so on.

you have 10GPUs installed, possibilities on mem shortage for 4GB RAM, by doing the suggested above, you can prove this

Look at the image 1 he posted guys..

he set the  GPU_SINGLE_ALLOC_PERCENT  value to 1%

thats the actual problem.. If not pls correct me..

Exactly.

setx GPU_SINGLE_ALLOC_PERCENT 100

This is the edition required. Right now, it's only taking a maximum 1% of each GPU's performance, which won't mine any coin.