Hey everyone, small contributor to a few BOINC projects (SETI and Rosetta), but first time on the message boards. Just wanted to know if anyone could steer me in a direction to building BOINC-optimized rigs. I am a Desktop Technician at a University and we think it would be neat to build something to harness the power of roughly 40 GPU's we have sitting around. We upgraded an AutoCAD computer lab, but the GPU's they came with weren't great so we upgraded them all. Now we have a surplus of normal consumer grade (but brand new) GPU's.

We don't have a crazy budget, but if we could build a 5 GPU rig to test it out and scale up from there would be great. Maybe later I could convince them to buy us that new ASUS B250 Mining MOBO for 19-GPU power. haha

Thanks for any suggestions!

First decide which projects interest you. Then see which of those projects support gpu applications. Check to see if any require double-precision calculations in the gpu. That should narrow it down for you.

Also, a lot of the newer gpu applications require one full cpu core to support the gpu task, so you would need to insure you have enough cpu cores to support each gpu task.

Some gpu applications can support multiple concurrent gpu tasks per gpu card and some others only work well with a single task occupying and using 100% of the gpu's processing power.

Lots of choices and decisions.

If you're not sure which project would be best for those GPUs, you can use WUprop (http://wuprop.boinc-af.org/results/delai.py) to see what projects run on them and what the crunching performance is like.

---

My Detailed BOINC Stats

I'd check first to get permission.

---