Some users in the Pent were running with 1x risers on former mining machines ok.
Do you happen to know what the crunch time comparison was for tasks using a particular model GPU?
If one is desperate to get extra points for a competition, it's probably quite 'OK' if those extra points (that you otherwise wouldn't have) take twice as long to garner in the x1 slot than they would have in an x16 slot.
Some users in the Pent were running with 1x risers on former mining machines ok.
That made me curious, so I set up tests, the first card in the x16 slot, the second with riser in the x1 slot. The first card showed to run unaffected, so time as usual. The second card, if an older model, took something like a quarter more time than expected, or, if a current model, took about ten percent longer than expected. I don't currently have a PSU to support another card per x1 slot, also no core to support additional tasks, and the rig was a temporary construction, so no additional tests. So I'd conclude that anyone who has an unused GPU around and has fun setting up a little rig, could do so. The reduced performance of the second card translates into lower power efficiency while on the other hand you save the second machine.
mmonnin wrote:Some users in
)
Do you happen to know what the crunch time comparison was for tasks using a particular model GPU?
If one is desperate to get extra points for a competition, it's probably quite 'OK' if those extra points (that you otherwise wouldn't have) take twice as long to garner in the x1 slot than they would have in an x16 slot.
Cheers,
Gary.
mmonnin schrieb:Some users in
)
That made me curious, so I set up tests, the first card in the x16 slot, the second with riser in the x1 slot. The first card showed to run unaffected, so time as usual. The second card, if an older model, took something like a quarter more time than expected, or, if a current model, took about ten percent longer than expected. I don't currently have a PSU to support another card per x1 slot, also no core to support additional tasks, and the rig was a temporary construction, so no additional tests. So I'd conclude that anyone who has an unused GPU around and has fun setting up a little rig, could do so. The reduced performance of the second card translates into lower power efficiency while on the other hand you save the second machine.
Recent progress with my
)
Recent progress with my project:
I got the 6 PCIe risers, Geekbench shows no noticable performance overhea:
https://browser.geekbench.com/user/216324
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8721668
GPU004 = GT420 in MOBO's x16 slot
GPU007 = GeForce 605 in riser in MOBO's x1 slot
I chose SETI because the estimated runtimes were shorter, I'm going to run this for a while in "test" mode.
So far I have 2 valid tasks YAAY
I chose win10 since I'm expecting to use only proprietary drivers with these older cards I'm getting, I'll attach a link to my project later.
Amazing how much windows has progressed, I hope this driver handling extends to AMD/NVIDIA mixes as well ^^
EDIT:
My project is on hold as I got a job and don't have that much free time anymore :(