Author

Topic: RX580 rig is crashing with connected 2 and more GPUs (Read 179 times)

newbie
Activity: 15
Merit: 0
I connected 4th GPU and it has been mining without any problems for 1 hour. Thx Vann.
Params:
Quote
-cclock 0,0,0,0 -mclock 0,0,0,0 -powlim 0,0,-5,-5
I'm setting up another absolutelly the same rig with other 4 ASUS RX580 GPUs to compare stability.
newbie
Activity: 15
Merit: 0
VannI connected now 3 x RX580 where GPU1 and GPU2 is with Samsung and moded Bios and GPU3 is a new one with next params:
Quote
-cclock 0,0,0 -mclock 0,0,0 -powlim 0,0,-5
Seems to be working stable, at least 15 min without crashing. I'm monitoring it, hopefully it's a solution.

When I decrease cclock or mclock (as example to 1950) on GPU3 it gives incorrect shares or the rig simply dies. =(

crairezx20 but it is an issue even with 2 x RX580 (with Hynix memory).
I do have amp meter, this is what I have:
3 x GPUs (28Mh/s, 28Mh/s, 22 Mh/s): 575-621 W
4 x GPU (28Mh/s, 28Mh/s, 22 Mh/s, 22 Mh/s): 759-782 W

This is with the default setting, without undervolting or overclocking. But two first GPUs have moded Bios.
legendary
Activity: 1638
Merit: 1046
I think this is wattage issue do you have amp meter to monitor how much watts you get in your rig since you only have 1000w psu so your power source is low and your wattage from your GPU is high that is why sometimes some of your Graphics card down to 0 mh/s
Try to change your psu to 1600w or more..  because 1 rx580 is 350w so if you have 4 rx 580 in 1 rig  4(rx580)x350w=1400w
So you need to add more psu or buy almost 1600w PSU..
hero member
Activity: 1036
Merit: 606
Even with the stock Bios you still need to adjust the core and memory frequency. I use Windows v1703 with the blockchain drivers with Afterburner to set the core/memory and power limit. If you are using Claymore to set the power limit -powlim -5 decreses the TDP by 5% and -powlim 5 increases it by 5%. With RX 580's dual mining I set the power limit to -15% with a -150 mV core undervolt in Afterburner. Some cards like some of my RX 570's I need to increase the power limit to -5% for the cards to be stable.
newbie
Activity: 15
Merit: 0
Vann I guess there is nothing to do with mod Bios, I just want to run new cards with default Bios and keep them stable.
I did the Bios mod only on cards with Samsung memory and they works stable if I don't mix them with new cards. The biggest issue for me why the rig crashes when there are only new GPUs with default Bios.

I'm reading again your message about "increasing the the power limit to -5%". Should it be -powlim 5 pr -powlim -5?
newbie
Activity: 15
Merit: 0
Ok, then I guess we're done here.
I think you got what means "everything" in this conext. All what I know of.
hero member
Activity: 1036
Merit: 606
It also could be the Bios mod you did on the cards. I would try the Polaris Bios Editor v1.6.2 'one click timing patch' on the original Bios. PBE automatically detects the memory type and v1.6.2 applies the bundled performance straps to the 1750 MHz and up timings. The current v1.6.6 adjusts the 1500 MHz and up timings, but the 1750 MHz straps are more stable for most cards.

https://github.com/jaschaknack/PolarisBiosEditor/tree/9ec64066eecdb55ac86da7bc82181eaab2161d51

newbie
Activity: 15
Merit: 0
OpenCL hangs are typically from too much overclock or undevolting. Try incresing the the power limit to 5% or lowering the memory to 1950 MHz on the GPU that is crashing.
Thank you for quick response. Yeah, I tried to increase the power limit to +25%. No result. =(
newbie
Activity: 182
Merit: 0
Quote
I tried already everything

Ok, then I guess we're done here.
hero member
Activity: 1036
Merit: 606
OpenCL hangs are typically from too much overclock or undevolting. Try increasing the the power limit to -5% or lowering the memory to 1950 MHz on the GPU that is crashing.
newbie
Activity: 15
Merit: 0
Hello,

My RX580 rig is crashing when I connect 2 and more GPUs.

Specs:
ASUS H270 PLUS
Intel Celeron 3930
4GB DDR
1000W Chieftec APS-1000CB
2 x ASUS DUAL-RX580-O8G (Samsung)

Driver: Win10-64Bit-Crimson-ReLive-Beta-Blockchain-Workloads-Aug23 (The same issue with: win10-64bit-radeon-software-adrenalin-edition-17.12.2-dec19)
Miner: Claymore 10.2 or 9.8

Story
Initially I had 2 x RX580 (Samsung) and was mining SOLO ETH. From the box it gave 25Mh/s, I moded BIOS and it gives 28MH/s, without overclocking and etc. So, it was working really stable for 3-4 weeks.

A few days ago I bought 6 more ASUS DUAL-RX580-O8G (Hynix). I added to the rig 2 more to have 4 GPU in total in one rig. Everything was fine for first 5-6 hours and then it started crashing randomly. As example it can work 5 min and then one of the cards randomly gives 0 Mh/s and watchdog is restarting the miner.

I tried already everything, different set of GPUs, with moded BIOS, with only 2 GPUs, with different rizes, with GPUs connected into mobo, different cables from PSU, reinstalled Windows, different miners and etc.
The result is the same.

But...:
- when I connect only one GPU, independenlty wich one, it's super stable
- when I connect two firstly bought GPUs (Samsung), which were working stable from the beginning, it's super stable


Quote
Log:
11:08:51:224   1bb8   ETH: 12/31/17-11:08:51 - New job from eu1.ethermine.org:4444
11:08:51:231   1bb8   target: 0x0000000112e0be82 (diff: 4000MH), epoch 160(2.25GB)
11:08:51:237   1bb8   ETH - Total Speed: 50.393 Mh/s, Total Shares: 3, Rejected: 0, Time: 00:04
11:08:51:242   1bb8   ETH: GPU0 27.922 Mh/s, GPU1 22.471 Mh/s
11:08:59:024   1bb8   got 248 bytes
11:08:59:037   1bb8   buf: {"id":0,"jsonrpc":"2.0","result":["0xb00bace67f1c64c81166ad3209ddd6fe196ab937aaa61fe78b2132972ad976ca","0x10fc9a2e5b65ea3f990c50787c055ec381519a57803a159640ee6ceac8ea4f5d","0x0112e0be826d694b2e62d01511f12a6061fbaec8bc02357593e70e52ba","0x49af9d"]}
11:08:59:052   1bb8   parse packet: 247
11:08:59:066   1bb8   ETH: job changed
11:08:59:075   1bb8   new buf size: 0
11:08:59:083   1bb8   ETH: 12/31/17-11:08:59 - New job from eu1.ethermine.org:4444
11:08:59:091   1bb8   target: 0x0000000112e0be82 (diff: 4000MH), epoch 160(2.25GB)
11:08:59:123   1bb8   ETH - Total Speed: 27.923 Mh/s, Total Shares: 3, Rejected: 0, Time: 00:04
11:08:59:129   1bb8   ETH: GPU0 27.923 Mh/s, GPU1 0.000 Mh/s
11:08:59:886   1bb8   ETH: checking pool connection...
11:08:59:891   1bb8   send: {"worker": "", "jsonrpc": "2.0", "params": [], "id": 3, "method": "eth_getWork"}
11:08:59:962   1bb8   got 248 bytes
11:08:59:969   1bb8   buf: {"id":3,"jsonrpc":"2.0","result":["0xb00bace67f1c64c81166ad3209ddd6fe196ab937aaa61fe78b2132972ad976ca","0x10fc9a2e5b65ea3f990c50787c055ec381519a57803a159640ee6ceac8ea4f5d","0x0112e0be826d694b2e62d01511f12a6061fbaec8bc02357593e70e52ba","0x49af9d"]}
11:08:59:985   1bb8   parse packet: 247

...

11:10:05:555   1b58   WATCHDOG: GPU 1 hangs in OpenCL call, exit
11:10:05:561   1b58   watchdog - thread 3 (gpu1), hb time 76391
11:10:05:571   1b58   WATCHDOG: GPU 1 hangs in OpenCL call, exit
11:10:05:885   1b5c   GPU0 t=67C fan=67%, GPU1 t=52C fan=44%
11:10:06:696   1b58   Restarting OK, exit...



Thank you and Happy New Year!
Jump to: