ETH Mining Q- 9x 290/380 = ONLY 100mh/s?! Pls help, pics and details inside

[danosphere]

I'm hoping someone can help me, I bought a fully assembled ETH miner with 6x AMD 290/380 4GB gfx cards. The rig was happily mining at a steady 120mh/s for a year and then things started acting up. First thing that happened was one of the cards suddenly started reported ~2.x mh/s rather than the normal 20mh/s, knocking the power down to a total ~100mh/s. I wasn't too worried about this, so i just let it keep hashing away. It was when a second card failed, this time reporting 0 mh/s that I decided it was time to fix things.

So I picked up 3 matching gently used 290/380 4GB cards for a decent price and swapped out the 2 cards that were producing <20mh/s with new cards. This did nothing to solve the problem. I asked around on slack and some people suggested I swap out the cheap "ribbon risers" with "proper risers" so I did that, also did nothing to solve the problem.

So now I'm here seeing if some pros can help me out. The weird twist to this situation is that with all 6 cards in the rig, including the one that is under-hashing, I can get a total ~104mh/s. I don't know if these operate in pairs or something but if I disconnect the 1 card that is under-hashing, the entire rig drops down to 80mh/s with a DIFFERENT card now under-hashing- I'm completely baffled by this; put back the "slow" card and its back to an almost-full-power ~104mh/s.

So I now have 9 of these higher-end gfx cards but for the life of me can't figure out why I can't get that last 20mh/s of power back. Is it possible the PCI port labeled #1 in the pics is just like... dead or something? Is there anyway to test that? Any help or guidance at all would be appreciated, TIA!

Top view of rig, you can see the "under-hashing" card as the fans don't need to be running often at 2mh/s!


Bottom view of the rig, shows the USBs connected to the "better risers". Red indicates connection to the "under-hashing" card.


Mining console, using Claymore dual miner in ETH-only mode. Changing the mining software has ZERO impact on the hashing results. Note GPU0!?


The "extra gear" I've gone thru/blown $$ on trying to get that 20mh/s back!


Thanks again in advance for any help, really stumped here and tired of wasting time+$$$ on things that don't work!  

[1714054897]

1714054897

[cryptominer420]

Load msi afterburner and turn the mem down 1 card at a time and see which card hashes slower, when you find the one that is hashing at 2MH turn the fan on it to 100% that should identify the failing card as the order in claymore generally isn't the slot order (this is a OCL bug).

[64dimensions]

In my experience it's usually a bad riser. This is the first thing to check.


You didn't mention your OS, but if linux, run aticonfig --list-adapters and see which cards are working or if Windows like the poster suggested run MSI after burner.


Other possibilities:

Check MB defaults for the PCIe slots.

Check to see if your HD is getting full.

Flaky DRAM chip which would effect your virtual buffer.

Overheating card: therefore the card is throttling back. This happens because of dust buildup or a fan is going bad, or the thermal compound is going bad.


Remote chance:

If you are windows and don't have auto update disabled, check to see if the driver has changed.

[danosphere]

Thanks everyone for the replies, did some more diagnostics and here are teh results:

Load msi afterburner and turn the mem down 1 card at a time and see which card hashes slower, when you find the one that is hashing at 2MH turn the fan on it to 100% that should identify the failing card as the order in claymore generally isn't the slot order (this is a OCL bug).

Assuming that GPUn in claymore's mh/s stats match GPUn in claymore heat/fan speed stats then using the above test method I deduced GPU0 to surprisingly actually be 0 as labeled in the OP pics.  GPU4 is 2 as labeled in the OP pics. This is weird because GPU4 is also reporting a hashrate of 20 mh/s but somehow doing it without a fan? Something is odd but those are the results

Using that info, and the fact GPU0 was reporting 0 mh/s I cycled through 3 different cards none which yielded higher than 2mh/s. 

In my experience it's usually a bad riser. This is the first thing to check.


You didn't mention your OS, but if linux, run aticonfig --list-adapters and see which cards are working or if Windows like the poster suggested run MSI after burner.


Other possibilities:

Check MB defaults for the PCIe slots.

Check to see if your HD is getting full.

Flaky DRAM chip which would effect your virtual buffer.

Overheating card: therefore the card is throttling back. This happens because of dust buildup or a fan is going bad, or the thermal compound is going bad.


Remote chance:

If you are windows and don't have auto update disabled, check to see if the driver has changed.

These PCI slots and cards were all working fine before, so its not like MB defaults changed turning off a slot, also claymore recognizes the card it just doesn't make the hashes it should and it is limited to the same PCI0 slot it seems. Could the actual MB slot just be bad somehow? Not the riser, the MB connection itself?

Its a windows 10 OS running pretty bare bones, I have auto updates turned off. Even if it was a driver thing that would impact all cards I think, not just 1. I thoroughly air-cleaned the thing so its not dust imo.

---

I'm going to try swapping #5 and #0 and see if I still end up with 0 mh/s on GPU0 or if it moves to GPU5. Also seeing the card from slot 0 in slot 5 and producing hash would add more evidence to the investigation. Thanks for any additional help or insights!

[BenCodie]

Have you try with only one card that giving you problem?
Remove all the cards except those giving you problem and check what they gives.
1- Change the power connectors to these cards
2- Might be drivers issue Re-install Drivers

[danosphere]

Have you try with only one card that giving you problem?
Remove all the cards except those giving you problem and check what they gives.
1- Change the power connectors to these cards
2- Might be drivers issue Re-install Drivers

Shockingly this was the solution. Re-installed the recommended 15.12 AMD driver and it fixed the issue; all the cards came to life running a full 20 mh/s+ on every card. Thank you very much to all who replied. Closing this thread for future reference.