Hi guys, I recently built a rig using 8x Sapphire Pulse RX 6600 XT, but for some reason one GPU is not overclocking. I’m able to get 32 MH/s on all of them except this one. I can confirm that both the riser and the GPU are working. I’m running two 660 W platinum PSU, which seems to be the issue, but I can’t figure out why. I should be good on power. As soon as I unplug any of them, it starts working. Just like if 7 cards is the limit.
you should be able to lower the core to 900 without any loss of hashrate. which miner are you using?
have you looked in the miner to see if one of th4e values (core clock/memory clock) didn’t take on that card? have you tried any other miners?
Hello, thank you for the reply. I did a few modifications and I still get a black sheep at lower hashrate. I lowered the core freq to 900 and set the core voltage to 650. I also changed from gminer to teamblackminer. I’m getting higher hashrate overall, but still have that one card at 20.
Did you verify the correct core and memory settings were setting on the problem card inside the miner itself? Or does it have a much lower core clock than the others?
I’m not sure how to verify. My miner is showing all zeros for all my GPUs. Not sure why… The only place I can see it is in the dashboard. I assume it doesn’t overclock since it’s lower than the others.
It’s always that one card, but if I unplug any other one, it starts going up just like the other ones. It makes me think it’s related to my power… but I should have plenty with 2 PSUs.
Oh that’s what I did, using Teamblackminer it shows all zeros for all the cards. Perhaps TBM cannot read the values, but hiveos can set them? Anyway, it definitely does something and like I said, if I unplug one of them, no matter which one, it’s all fine. I feel like going above 7 triggers something. Could it be BIOS related? I was a little concern about going dual PSUs and my motherboard doesn’t seem to have a property to have it balanced. After trying it I think it’s automatically balanced but maybe not?
Just tried forcing PCIe gen1, gen2 and gen3, still the same result. It really does seem like a power issue. Any advise on that? Maybe if I’m able to lower the power consumption it’s gonna work?
thats why the consumpption is reported higher then, you should be able to erase them. for accurate info you should use a wall meter though, amd cards generally under report wattage, for example my 6600s in software say 39-41w, and at the wall draw ~61w each
Hi msylvestre, what CPU and Motherboard are you using? Are you using PCIE Splitter cards and/or riser cards?
I think your card with the low hashrate may be suffering contention on the PCIE lanes coming from the chipset (as opposed to PCIE lanes directly from the CPU)
@Batfink I have a Ryzen 5600x with an Asus ROG B550-F (no wifi). I also use both PCIe splitters and risers. If thats the case and theres contention… any idea on how I can fix that?
@keaton_hiveon I ordered a wall meter. I’ll let you know when I get it.
Hmmm…Ryzen 5600x has 20 PCIE lanes (16x GPU + 4x storage/NVME, etc) + whatever the B550 chipset can provide (10 I think if I remember correctly), so you should have more than enough lanes. btw, the Intel 2.5GBps NIC will take a PCIE lane (from the B550 chipset). This still leaves you with plenty of lanes.
I did download the manual for that motherboard (I am looking at B550 mobos anyway)…
Just be aware that there are some slots on that mobo that share bandwidth (this might be your issue)
if I was going to use the PCIE slots/M2 slots I would probably use them in the following order (fastest 1st);
PCIEX16_1 slot (use with a 4-port splitter)
M.2_1 slot (use with a M.2 to PCIE 4x adapter + 4-port splitter)
The two slots above take PCIE lanes directly from the CPU so will not suffer the latency of the PCH (Platform Controller Hub)
PCIEX16_2 slot (use with a 4-port splitter)
[don’t use PCIEX1_1, PCIEX1_2, PCIEX1_3 slots if using PCIEX16_2 - they share bandwidth]
M.2_2 slot (use with a M.2 to PCIE 4x adapter + 4-port splitter)
The last two slots above provide PCIE lanes via the PCH (the B550 Chipset) so they will be slightly slower than PCIE lanes coming directly from the CPU.
I reckon the B550 chipset should easily support 16x GPUs with the right CPU and mobo. I have not tried this yet but I am tempted with the cost of B550 mobos pretty low at the moment
Nice choice of CPU and mobo btw!
Just for your reference I use these M2 adapters and 4-port splitters. Not had a single bit of trouble from either of them. Plug-and-play (which suprised me, I was expecting a battle to get them to work);
Hey, thanks a lot for that. I think you might be onto something. I have two 4 port splitters and I had them in PCIex_slot1 and PCIex_slot2. Reading at your comment, they’re definitely sharing. I didn’t actually know it would fit in PCIe 16. However, I changed them to both 16 lanes and I’m still having the same problem. I also tried PCIe16_slot1 and PCIex_slot3 and I get the same issue as well. Am I using the wrong ones?
Note that I do have an nvme drive in the first m.2 slots. Would that be causing an issue?
here, right now I have them in PCIEX16_1 and PCIEX16_2. Before changing anything I was using PCIEX1_1 and PCIEX1_2.
