I just upgraded my SLI 570's to a single GTX980, till I can afford a second one. the problem I'm having in that I'm now running 3 WU on my GPU but it is not producing the RAC I think it should taking 20 minutes sometimes to do one WU. so how do I find the bottleneck? I was "told" that my 980 was working about 85% of max. if I drop it to only 2 WU it would only be working at 50% or so of max. is that right? HELP!!!!

---

Comparing your times to a host with 4 GTX 980s your times are rather similar.

I think the current bottleneck for the newest generations of NVIDIA hardware is the application. In that architecture changes to increase the GFLOPS of the hardware are not as easy to to take advantage of codewise. Which is one of the reasons that running multiple tasks is required to fully utilize the GPU.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Couple of things,

Depending what what kind of MB you get, the time to complete 1 work unit can be anywhere from 6 minutes to 18 minute or more.

When you start to do more than 1 work unit per card that time increases with each new consecutive work unit added on.

What you want to see is the total time to complete divide by the number of instances(how many work units per card)

That number should be lower than running 1 work units by itself.

Next,

Have you played with the mbcuga.cfg yet? You should change the setting in that. Use Notepad to open it and edit the last 3 lines.

processpriority = high
pfblockspersm = 16
pfperiodsperlaunch = 400

make sure to remove the ;;;

Post if you have some more questions


Zalster

My GTX970 takes up to 20 mins doing a WU running 3 at a time, shouldn't the GTX980 be somewhat faster?

Here's my WUs

http://setiathome.berkeley.edu/results.php?hostid=5284345&offset=0&show_names=0&state=4&appid=

A bit faster, but as Zalster has said, the bottleneck just now is that the current apps don't utilise the GTX980's capabilities properly, and it very much depends on the hardware feeding the GPU for the actual performance. I'm currently watching my "daily driver", which has recently gained its send GTX980 catch up with my "big cruncher", which has a pair of GTX780s. Th GTX980s are only running two-per, while the on the GTX780s I'm ru