PCI-e bandwidth experiment

[hiigaran]

Hey guys, I'm trying to gather some data on GPU projects and need a little help from anyone who can spare a few minutes.

For anyone running a BOINC project that is using the GPU, could you run it exclusively, then note the PCI-e bus/bandwidth usage via MSI Afterburner or any similar software? Please post the amount of bandwidth used, along with the project, your GPU, and both the size and version of the PCI-e slot it is connected to.

The goal, obviously, is to try and get a general idea of how much bandwidth is used during the running of a work unit and determine which projects use the least amount.

[PoNyePiC]

I've never done GPU-based BOINC projects, having only done CPU-only Rosetta.  If you're still looking for an answer, let me know what projects you want to test and I will try it.  For the record, it'll be a PCIe 3.0 at 16x speed

[hiigaran]

Basically any GPU project you can run. Though from what I've been seeing from similar responses on other forums, things don't look too promising.

[RoBorg]

I'm on holiday at the moment but I'll give it a go when I'm back next week.

[10esseepony]

I can tell you that Folding@Home takes a serious performance hit at 1x, whereas Milkyway does not. (PCIe 2.0, GTX1070 and R9-280X) It's been a long time ago that I did the test, so no data, sorry.

[Rainboom_Dash]

I can tell you that Folding@Home takes a serious performance hit at 1x, whereas Milkyway does not. (PCIe 2.0, GTX1070 and R9-280X) It's been a long time ago that I did the test, so no data, sorry.

It's the BUS usage you look at, right?

It's currently saying it's at about 37% with folding@home with PCIe 3.0 16x

Don't know anything about BOINC... and I haven't actually tested 4x or anything