• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

GPU MINING RIGs moving to FoldingHome , but PPD is super low , please help.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
TebaOPPA

TebaOPPA

New Member
Joined
Feb 11, 2023
Guys ! Hope all is well,,,

I currently have around 10 Rigs with 8GPU each that I am trying to dedicate them to Folding Home, however they are only getting around 6 million PPD , when my 3090TI single card PC is getting 10m PPD ,,, is there a way to set up these Rigs so each slot can be utilized more efficiently ?

All these Mining Rig GPUs are either 3080 or 3090s so I should be getting some pretty high PPDs but all these rigs are maxing out at 6M PPD ish( all slots are configured correctly so they are all running for sure )

** Currently 520 on the rank ;)

Thank you all in advance !
 
I'd imagine there's not enough PCIe lanes to go around/give you enough bandwidth for the cards. What motherboards are they on and with what CPUs? What slots are they in on each motherboard?
 
Most of these rigs are on ASRock H110 PRO BTC+ , and the CPU is on i3 usually not too high powered lol since they were meant for Mining,
( with the full intention of going back to Folding Home once ETH was done) which is what I am doing right now : )
Post magically merged:

Screen Shot 2023-02-11 at 8.19.52 PM.jpg
 
Linux, one real cpu core per gpu and pcie x8 gen 3, or better.

The FAH client uses one thread per gpu, but best to have a real core for each GPU. Some GPU projects will use more than one CPU cores, if available.

USB risers will restrict the data flow, 10-20-30% or more, depending on the project.

Windows and Linux will get about the same ppd on large atom count projects, but linux will do better on the smaller projects.

Big GPUs have best utilization/ppd on large atom count projects, lower utilization/ppd on smaller projects.
 
First, you should be folding for Team32 (us!!).. :p


Yep. You're short cores on the CPU and bandwidth for the PCIe slots to let these things run wild.

Surely we can get you up and running at your best, though, it's not going to be with that hardware.
 
LGA1151 i3 only has 2c/4t. Screen shot shows a CPU wu using one of the real cpu cores. Need to disable the cpu slot, which is using one of the 2 real cores.

The board specs show 12x PCI Express 2.0 x1 Slots.

Could try upgrading cpu to 6c/12t or 8c/16t to see how much that helps, but pcie x1 gen 2 is going to really hurt ppd, even on linux.
 
In the past I did observe folding to use a lot of PCIe bandwidth. Run GPU-Z and you can check it in real time for each GPU. Probably find any connected to x1 are maxed out. If you only have x1 lanes I'm not sure there is much to do to help the situation short of replacing the mobo.
 
I learn something every day.
I was unaware of motherboards like the ASRock H110 PRO BTC+ that have 1 PCIe 3.0 x16, 12 PCIe 2.0 x1 (never looked in to bitcoin mining).
I can only deduce that to run seven GPUs on that one MB, you have to be using risers.

​

HayesK has a great amount of experience and knowledge of Folding as evidenced by his signature. Seems as though he has identified two issues (lack of CPU cores and lack of PCIe bandwidth) preventing your GPUs from achieving their full potential.

My 3080 tends to average from 6 - 7 million PPD. It makes sense that a fully optimized 3090 would do significantly more.
I hope you figure out how to get all those GPUs Folding, and hope you join Team 32. As you can see we have some active members who know some "stuff".
 
Wow, see if you can get an i7 6/12 or 8/16 core CPU, that will help with the Gen 3.0 PCIe slot, don't know about the other slots.

12x1miner.jpg
 
Wow ya, those x1 slots are badly hampering your PPD. That CUDA stuff really makes use of lots of PCIe bandwidth (as mentioned already). It's why most of my systems have but 1 card per board, though running 2 in x8 might be doable as well as I think was mentioned in another thread.
 
In addition to the two above described issues (insufficient CPU cores and bandwidth), I might have identified an additional issue . . . . power.
If I read about PCI Express x1 slots correctly, all those slots share a maximum of 10 watts. So if you populate five slots, each slot only delivers two watts to the GPU in the slot. That might reduce performance in a 3090 that can consume over 300 watts (TGP is 320W).
A full-sized x16 graphics card may draw up to 5.5 A at +12 V (66 W).
 
Performance doesn't really brown out with a lack of power. Either gpus work (psu provides enough stable power) or they don't. I'd imagine it just uses more power through the pcie power connections... but not sure.

I'm floored they work at all as I thought they needed x4 slot (it was an SLI requirement).

BugFreak said:
Switch that team to 32 and I'll send you a 6600k for free. :)
Click to expand...
I'll send you a motherboard or two....
 
Last edited:
System ram also a consideration. Minimum would be 2GB per GPU (8 GPU x 2GB=16GB), although my suggestion would be 4GB per GPU (8x4GB=32GB).
 
TebaOPPA said:
Guys ! Hope all is well,,,

I currently have around 10 Rigs with 8GPU each that I am trying to dedicate them to Folding Home, however they are only getting around 6 million PPD , when my 3090TI single card PC is getting 10m PPD ,,, is there a way to set up these Rigs so each slot can be utilized more efficiently ?

All these Mining Rig GPUs are either 3080 or 3090s so I should be getting some pretty high PPDs but all these rigs are maxing out at 6M PPD ish( all slots are configured correctly so they are all running for sure )

** Currently 520 on the rank ;)

Thank you all in advance !
Click to expand...

Seems like with 80 GPUs you will need 80/3 = 27 motherboards that can populate three (3) GPUs per MB; and the cases, the power supplies, minimal memory per system, and at least one storage drive for the OS and minimal apps, and so on. That is a major capital investment; then their is the energy cost at say an average of 250 W per GPU x 80 = 20,000W!

I would like to know what would happen if you paused all the GPUs but one, and the also pause the CPU to see what PPD you get!
 
You must log in or register to reply here.
Back