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

Folding on 8x slots

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

sporktar

Member
​
Joined
Jul 10, 2010
What kind of performance drop could I expect folding on 8x slots? I'm considering buying my coworkers server, which has a supermicro X8DTN+ . If I understand the product page correctly, the motherboard has a 16x slot, that gets split into two 8x slots when a 2nd 16x card is plugged into the related slot. My plan is to use the server as a linux environment for projects, and fileserver stuff. I'm also planning on moving one of my 970's to it for folding duties. When i do get around to upgrading my main rigs gpu, the 2nd 970 would potentially get moved to this server, bringing both lanes down to 8x.

Is this a feasible plan, or would it be better to just stick to running 1 card?
 
I don't think there is a loss going to x8 with a 970, I thought that was only seen on the 980ti's and titans.
 
My 980ti's fold as 8x and don't seem to see a dip in ppd as compared to 16x cards.

sorry, that's on Gen3, not Gen2.
 
Last edited:
sooooooo, it looks like it's me that's going to step into the breach on this one.
my file server hangs on the wall and looks to have the right gear to test this and it's just kind of hanging around, looking pretty most of the time.
I moved the 970 down one slot and it looks like it runs at x8 in this slot, agreed?
 

Attachments

  • gpu-z.PNG
    gpu-z.PNG
    123.9 KB · Views: 112
caddi daddi said:
sooooooo, it looks like it's me that's going to step into the breach on this one.
my file server hangs on the wall and looks to have the right gear to test this and it's just kind of hanging around, looking pretty most of the time.
I moved the 970 down one slot and it looks like it runs at x8 in this slot, agreed?
Click to expand...

Looks like it's running at 8x to me!
 
we'll just let this bake a few days and see what it does then we'll move it back up to the x16 slot for a few days.
I think i'm about #82 or so on the team so you can keep abreast there.
 
caddi daddi said:
we'll just let this bake a few days and see what it does then we'll move it back up to the x16 slot for a few days.
I think i'm about #82 or so on the team so you can keep abreast there.
Click to expand...
Got your stats page bookmarked, I'll be keeping an eye on it! Thanks for running some real tests on this. I should be picking my server up tonight after work, but all my gpus are still tied up in my watercooling loop.
 
Caddi, what about the different projects/cores that the GPU will be working on during this test period? Will that modify the PPD average, or is the goal to identify large(er) point gaps?
 
pci8.PNG
Well i feel pretty dumb now. I was doing some research to find why my m.2 speeds were so slow, and found out I'm already folding on 8x on both my cards. This would explain why I've always gotten lower speeds than most people get on their 970's. Looks like I'll be putting my 2nd 970 in my server sooner than I thought...
 
well, that's three days at x8, I'll pause it for a day and move it up to an x16 slot and run another 3 days at that.
luck of the draw on work units is just a fact of life, so in this short test we just have to use larger margins of error.
but a 970 is not going to fill the pcie lanes anyway, we're just proving it out.
the folding slot on this rig is configured for a 980 and not a 970 so it's not going to make max ppd anyway.
 
i thought with FAH they loaded the work into the GPU memory to run on the core. maybe im missing something about running FAH on the GPU works but i mean, yea the wu for a cpu core is on the HD. then everything gets loaded into memory and ran on the CPU so hd speed wouldnt make a difference, following the logic i see about PPD based on slot speed. someone more "in the know" on GPU folding prolly needs to fill me in...
 
hdd or ssd does not matter in how fast a work unit completes, it does effect total power draw and heat from the rig so here in the south an ssd is very slightly better.
cpu clock will give you nothing either so on my folding test platform I under clock and under volt the cpu to reduce the heat output a little.
system memory makes no difference either so I just slap in some cheap, stable, slow, 1333 ram, a single 2 gig stick is plenty.
so, for the NVidia stuff it's a matter of cutting my quad cores to dual cores, slowing the cores and clocking the cards.
gpu memory clocking has shown a slight advantage for me but not enough for me to give up any gpu core clock or worry much about.
 
so off we go for a few days at x16, all I have done is move the riser cable to an x16 slot.
 

Attachments

  • gpu-z.PNG
    gpu-z.PNG
    33.1 KB · Views: 81
good catch, but the board seems to have figured it out.
 

Attachments

  • gpu-z.PNG
    gpu-z.PNG
    22 KB · Views: 77
as we have seen before, x8, x16, with a 970 it doesn't matter one bit, but it never hurts to go back and prove it out.
one thing this does prove it that having the slot configured properly for the card is what really counts, the slot was configured for a 980.
 
Good to know! thanks for testing it out for us. Helped share some knowledge, and the team got an extra few hundred thousand points out of the ordeal!
 
You must log in or register to reply here.

Similar threads

torin3
Best motherboard for multi-GPUs
Replies
7
Views
721
Blaylock
Blaylock
K
NVME to PCIE adaptors
Replies
17
Views
833
EarthDog
EarthDog
don256us
SOLVED GPU "disabled" in client.
Replies
7
Views
2K
don256us
don256us
tacitus
Building a silent rack-mounted workstation with the Silverstone RM51
Replies
4
Views
2K
Zerileous
Zerileous
J
Gtx1080ti aorus extreme running at 8x pcie 3.0
Replies
5
Views
3K
jazztecho
J
Back