6x RX570 Sapphire Nitro+ 4gb / GPU0 Issues

[Valentine_13]

Hi,

I started mining 2 weeks ago.  Still Looking to find the sweetest spot.
Atm I dual mine at around 30.5mhs-480mhs per GPU.

The RIG is pretty much stable, restart or crash time to time.

6x RX 570 Sapphire Nitro+ 4GB (2 Elpida, 4 Hynix)
Win10 + Claymore 10 + 08.08 drivers
Modded bios
OC 1209/2100 => MSI Afterburner Beta 16
Undervolting directly in the miner 950mv

My ask is more about my GPU0.
It mines about 1-1.5mhs below average. Also some fluctuations do exist.
I understand it would come from the dummy HDMI plug I attached to it.
I tried to boot on the MB GPU but then I can’t undervolt my cards neither in the miner nor in MSI afterburner.
The RIG won’t start if no monitor is plugged, that’s the reason I’ve a dummy plug, also for teamviewer (RIG is in the cellar).

Any other idea you could bring forward?
Thanks
Val.

[1715699934]

1715699934

[1715699934]

1715699934

[hagbase]

No your rig is not stable if it crashes or restarts.

How many GPU Mem Erros in Hwinfo? 
Try 1100/2000 and latest blockchain drivers.

[P00P135]

Disable ULPS in afterburner and make sure the GPU's aren't in crossfire.

[opies32765]

Sounds like you have overclocked too much. Either increase the voltage on the cards or underclock the GPU0 for example you could do

2050,2100,2100,2100,2100,2100

[Valentine_13]

No your rig is not stable if it crashes or restarts.

How many GPU Mem Erros in Hwinfo? 
Try 1100/2000 and latest blockchain drivers.

I meant, hashrate was stable, at least for 5 GPU's.
Had a lot of errors. Millions on GPU0-1.

I installed the september drivers, hashrate was absolutely not stable and undervolting not working. Put back the 08.08 one.
I lowered OC to get less errors. It's now between 0 and 15 errors after 8 hours.

Disable ULPS in afterburner and make sure the GPU's aren't in crossfire.

That was already disabled.

Sounds like you have overclocked too much. Either increase the voltage on the cards or underclock the GPU0 for example you could do

2050,2100,2100,2100,2100,2100

I increased the voltage and underclocked a bit all my GPU's.
Losing around 5-6 mhs total with 0 errors. But my GPU0 is still very volatile, between 22 and 29 mhs...

:/

Just you to know, it's not the card, as i tried tu put 3 different card as GPU0.
Always the same problem.

[hagbase]

Your mainboard/PSU?

I'm using these cards too and hashrate only drops for 1 sec if i turn on my display and only on the gpu with display.   But i'm running @1100/2000.
 

[Valentine_13]

AsRock h81 pro btc r2.0
Kolink 1200w

At wall, the rig drains around 1'000w.

[hagbase]

Dual/single Mining?
GPU Power Chip Draw in HWInfo?

Turn off iGPU if your cpu got one.
Set PCIe to Gen1 or 2.

[Valentine_13]

- Dual
- Pwer draw average 130w on GPU0, 115w on the others

- iGPU, will check tonight when got home
- Gen1/2, same tonight

[hagbase]

Switch also to single mining and check if your gpu ist still unstable.
If it's stable in single mode than dual mining draws too much power.

[Valentine_13]

Single mining at OC 1900 or 1800, the GPU0 is unstable.
I don't get it.

[hagbase]

Powered Risers i guess? Check connection. Also don't use extra 2x 4pin power connetion on Asrock h81 if you use powered risers.
Your dummy plug is in gpu0 or 1(pcie 16x slot)? My monitor is on gpu1(16xslot).

[coinfoundry]

I'd try to figure out the problem methodically by removing all cards except one. Validate stability. Add another. Rinse repeat until you find the culprit.

[Valentine_13]

Powered Risers i guess? Check connection.

Yes, already did.
I also switched 3x cards ... same problem each time, so obviously could be the config or the slot on the MB.

Also don't use extra 2x 4pin power connetion on Asrock h81 if you use powered risers.

Ah no? I've to check that.

Your dummy plug is in gpu0 or 1(pcie 16x slot)? My monitor is on gpu1(16xslot).

My dummy plug is GPU ....... 1 .... when 2 or more card are plugged !
Gosh, I feel terrible, sorry. During my whole process I thought 16x slot was always GPU0 ...

[JuanHungLo]

I had similar issues.  Mine turned out to be a defective riser:

https://www.amazon.com/gp/product/B01GU94QSQ/ref=oh_aui_detailpage_o03_s00?ie=UTF8&psc=1

I replaced it with another of the same, problem disappeared.

[Valentine_13]

Thanks guys.
Just an update.

- GPU0 is the GPU on the 16x slot
- Monitor is now on a iGPU
- Took off the 4x molex I had on the MB next to slot GPU1
- All bios reflashed with the anorak 1500 mem
- OC 1200/2000 getting around 28.75 mhs per card, if I push to high (2150), it gives a lot of mem errors

I'm still not convinced about the bios.

Cheers !