Sure why not the user Peter Hucker runs multiple gpu's, though none this powerful, in a machine with risers, I think he puts them all on the same Project but with a few <exclude_gpu>lines in an app_config file you can even run each gpu on it's own Boinc Project, assuming there are 7 gpu Projects left. I guess with all the PrimeGrid sub-projects there ARE more than 7 gpu projects still running. I would guess the other problem would be getting enough gpu tasks from a Project to keep the gpu's running full blast.
I certainly think it would with the exception of getting a project to be able to provide the GPU tasks with enough work tasks to keep your 7 GPUs busy. Like Mikey says:
mikey wrote:
..... I would guess the other problem would be getting enough gpu tasks from a Project to keep the gpu's running full blast.
I think the best idea would be to have each individual RTX 4090 GPU running a different project and hope for the best.
This is a big buck system! These prices are an estimate based on Amazon and Newegg.
7 x $2,000 = $14,000 - RTX 4090 GPUs (at around MSRP)
1 x $219 = $219 - Samsung 980 PRO 2TB SSD
1 x $1,000 = $1,000 - ASUS Pro WS WRX80E-SAGE SE Motherboard
No I've had 4 gpu systems before when the motherboards supported that.
The only way to do 7 is on a server motherboard with either single slot cards with waterblocks with single slot I/O brackets or via PCIE cables in a mining chassis for air cooled cards.
Would have to figure out a place for the mining chassis and figure out the power circuits. Garage would be the easiest location for space and power. But would have to figure out internet connection out there.
And the main limitation is my pathetic internet connection. It is really insufficient for the five BOINC computers, internet surfing and content streaming I currently do.
No I've had 4 gpu systems before when the motherboards supported that.
The only way to do 7 is on a server motherboard with either single slot cards with waterblocks with single slot I/O brackets or via PCIE cables in a mining chassis for air cooled cards.
Would have to figure out a place for the mining chassis and figure out the power circuits. Garage would be the easiest location for space and power. But would have to figure out internet connection out there.
And the main limitation is my pathetic internet connection. It is really insufficient for the five BOINC computers, internet surfing and content streaming I currently do.
My mining chassis, which is now in a box, was less than 3 feet by 3 feet in size and could handle 7 gpu's so size isn't going to be too much of an issue, though none of the gpu slots can handle quad size gpu's
I am not even sure I would feel safe with that running at 100%, 24/7. I would (if we were building this system...) wait until the RTX 6000 Ada comes out. Might not be faster (would have more CUDA cores available), but runs at a lower power and Nvidia should have refined the power/adapter issue.
No I've had 4 gpu systems before when the motherboards supported that.
The only way to do 7 is on a server motherboard with either single slot cards with waterblocks with single slot I/O brackets or via PCIE cables in a mining chassis for air cooled cards.
Would have to figure out a place for the mining chassis and figure out the power circuits. Garage would be the easiest location for space and power. But would have to figure out internet connection out there.
And the main limitation is my pathetic internet connection. It is really insufficient for the five BOINC computers, internet surfing and content streaming I currently do.
Keith, I would think for the "pathetic internet connection" that you have, you could replace two of your multiple GPU machines with one of these BIG BUCK machines. And I'd be happy to take at least one of your machines off your hands.
Tom M wrote:Would this even
)
Sure why not the user Peter Hucker runs multiple gpu's, though none this powerful, in a machine with risers, I think he puts them all on the same Project but with a few <exclude_gpu>lines in an app_config file you can even run each gpu on it's own Boinc Project, assuming there are 7 gpu Projects left. I guess with all the PrimeGrid sub-projects there ARE more than 7 gpu projects still running. I would guess the other problem would be getting enough gpu tasks from a Project to keep the gpu's running full blast.
Tom M wrote: Would this even
)
I certainly think it would with the exception of getting a project to be able to provide the GPU tasks with enough work tasks to keep your 7 GPUs busy. Like Mikey says:
I think the best idea would be to have each individual RTX 4090 GPU running a different project and hope for the best.
This is a big buck system! These prices are an estimate based on Amazon and Newegg.
7 x $2,000 = $14,000 - RTX 4090 GPUs (at around MSRP)
1 x $219 = $219 - Samsung 980 PRO 2TB SSD
1 x $1,000 = $1,000 - ASUS Pro WS WRX80E-SAGE SE Motherboard
4 x $480 = $1,920 - EVGA SuperNOVA 1600 T2 80+ TITANIUM 1600W PSU
1 x $7,145 = $7,145 - AMD Threadripper PRO 5995 WX CPU
1 x $480 = $480 - G.Skill Trident Z Neo 128 GB (4x32GB) of DDR4-3200 memory CL16-18-18-38 1.35V
1 x $200 = $200 - Mining Rack for 8 GPUs (approximate cost)
Grand total = $ 24,964 plus tax (approximate $2,247 @ 9%) = $ 27,211
https://cdn.wccftech.com/wp-content/uploads/2022/11/NVIDIA-GeForce-RTX-4090-7-GPUs-Running-PugetSystems-Content-Creation-Benchmarks-_1-1456x819.png
BTW, who do you think within this BOINC project would be willing to put it all together and run it constantly?
Proud member of the Old Farts Association
GWGeorge007 wrote: BTW,
)
Keith of course!! LOL
mikey wrote: GWGeorge007
)
Howling!
Keith has been very careful to limit himself to 3 GPU systems. So getting him into 7 rtx 4090's really looks unlikely. :)
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
No I've had 4 gpu systems
)
No I've had 4 gpu systems before when the motherboards supported that.
The only way to do 7 is on a server motherboard with either single slot cards with waterblocks with single slot I/O brackets or via PCIE cables in a mining chassis for air cooled cards.
Would have to figure out a place for the mining chassis and figure out the power circuits. Garage would be the easiest location for space and power. But would have to figure out internet connection out there.
And the main limitation is my pathetic internet connection. It is really insufficient for the five BOINC computers, internet surfing and content streaming I currently do.
GWGeorge007
)
George,
I'm happy to assemble it. Hell, I'll even commit to running the power hungry beast, as long as somebody else donates the hardware.....
LOL
Keith Myers wrote:No I've
)
My mining chassis, which is now in a box, was less than 3 feet by 3 feet in size and could handle 7 gpu's so size isn't going to be too much of an issue, though none of the gpu slots can handle quad size gpu's
I am not even sure I would
)
I am not even sure I would feel safe with that running at 100%, 24/7. I would (if we were building this system...) wait until the RTX 6000 Ada comes out. Might not be faster (would have more CUDA cores available), but runs at a lower power and Nvidia should have refined the power/adapter issue.
Gandolph1 wrote: GWGeorge007
)
Greg, if I could find somebody else for you to get you the hardware, I'd be more than happy to. And thanks for volunteering to assemble and run it.
As for me... I couldn't even afford the electric bill. I would be happy to assemble it, and let you or somebody else run it though.
ROFLOL ;^)
Proud member of the Old Farts Association
Keith Myers wrote: No I've
)
Keith, I would think for the "pathetic internet connection" that you have, you could replace two of your multiple GPU machines with one of these BIG BUCK machines. And I'd be happy to take at least one of your machines off your hands.
:^)
Proud member of the Old Farts Association