Very unstable hashrates when using risers (i mean really unstable)

spypredatorspypredator Member Posts: 6
Hi guys,

First of all, I would to thank you all for this wonderful community where a lot of interesting information can be found. :)

I don't want to seem desperate, but I kind of am. Despite all the information on the site, I’m stuck and I really need your help on my first rig.

After battling a week to run my rig (I had a defective motherboard AND a defective RAM from Amazon) I finally got it working.

My configuration:
  • 4 off ASUS - ROG-STRIX-RX570-O4G-GAMING,
  • MSI Z270 A Pro,
  • RAM 4GB Corsair - CMK4GX4M1A2400C14,
  • EVGA - SuperNOVA 750 G2,
  • Windows 10 / Claymore Dual Miner.
  • Latest AMD driver from this month
Until today, my rig was running with 2 RX 570 mounted on the motherboard because I had no risers. Everything was running perfectly fine with both GPU overclocked at 27mh/s.

Today, I received 4 USB risers that I ordered from Amazon. I plugged all the 4 GPUs on the usb risers and configured the bios so that the GPUs would be detected (at first I had a code 12 error, but I solved it thanks to the instructions here : 6 GPUs on windows 10 problem, New 6-GPUs Mining Rig issues).

Here comes the problem:

When running Claymore Dual Miner, all the 4 GPUs start with their maximum hashrates but then rapidly drop to 4-2 hashrates. Sometimes, one of them jumps to its maximum hashrates (27-26) but then drops again. overall I have a very unstable hashrates with an average hash rates around 10 mh/s (after 1Hour).

At first I thought it was a power supply problem. So I tried with only one GPU and the problem was still there.

Then I realized that the computer was very slow, so I checked the CPU and memory usage. There I saw that when Claymore Dualminer was running, the CPU usage was at 100%, which it shouldn’t be (see below screenshot)


(sorry for the omelette du fromage interface)

Do you have any idea of what is wrong in my case?

Thank you all in advance for your help.

Comments

  • AsterAster Member Posts: 2
    Did you check your risers ? It seems they are faulty.
  • FurryBuxyFurryBuxy Member Posts: 42
    remove all risers and try one riser at a time and see if that works out
  • spypredatorspypredator Member Posts: 6

    currently my cards are plugged this way:
    1GPU on the first PCI express 16x port
    1GPU on the first PCI 1x port
    1GPU on the second PCI 1x port
    the last GPU on the second PCI express 16x port

    with this configuration, I have 3 cards running in Claymore Dual Miner at 27 Mh/s and one card, which every time start at 27, but then drops until 12 Mh/s.

    This is all the things I’ve done :

    Starting from bios:
    • I set all the pci port to gen1
    • I modified the pci bus latency timer from 32 to 96
    • I’ve enabled the Above 4G Decoding options
    • I’ve disabled all not needed functionalities on the mother board like sound card, thnunder bolt…
    Regarding the software part:
    • I have the latest AMD driver
    • All the cards are patched
    I tested one by one each risers. When plugged in, all of them make this process appear in the task manager: "system interrupts".


    With some configuration (depending which card is on which PCI port), the system interrupt task sometimes uses up to 70% of my cpu!

    In Hwinfo, I can see that the cards generate a lot of memory errors (exept one, I guess the one which is directly plugged to the mother board). Also, the gpu usage is at 100%.


    Is it possible that the problem are faulty risers? Has anybody had the same issue?
  • spypredatorspypredator Member Posts: 6
    Found the problem myself:

    The instability was due to high memory overclocking on two of my cards.
    These two cards were generating a lot of errors. I couldn't see this until I enabled the GPU memory errors.

    Once I overclocked to an optimized frequency, the error diminished a lot (from 1K/s to 1/min) and this resulted in reducing the CPU usage of the process called "system interrupts". Before, this CPU usage was around 70% but it used to come with another process which both combined would cap the CPU.

    When my CPU usage reached 100% then claymore would kind of stop working, and all the hash rates of the four cards would drop around 1-2MH/s.

    When the most faulty card was put on risers, then it was worse. That's why, when swapping the position of my graphic cards, I found that the best configuration was to put one of the cards on the PCI-express 16x port. This specific card turn out to be the most faulty.

    Now, with the correct clocking, all the cards are stable.

    Hope that this will help someone someday.

    PS: All this time lost because I used the overclocked bios of one card, which was fine for that card, on all other cards, without checking them one by one. Lesson learned.
or to comment.