Troubleshooting horrible performance

[Railgun]

So, I've recently restarted my SETI efforts and have joined OC's team to help move things along. I've installed this on two boxes, however, the second installation is giving me well below sup-par performance. I'm basing this on other clients I've found running the same CPU and some similar GPUs.

For my main box, it's an i7-970 with 2xGTX460s and an 8800GT. I may or may not be getting the most out of it, but it seems to be doing well enough.

The second box is an i7-920 with an 8800GT and that's where my concern is. As you can see (http://setiathome.berkeley.edu/hosts_user.php?userid=9481809), the second box is barely doing anything...respectively. Second to that, it seems to be making very few connections back to the server to upload any info.

I installed this client and its respetive optimizations as I did on my main box, but I'm thinking something is amiss. Obviously the 8800 isn't going to perform as well as the 460s, however, I'm looking at this from a CPU perspective and again, comparing to others'. If I can borrow one of you for the moment...taking a look at this (http://setiathome.berkeley.edu/hosts_user.php?userid=98489), with no GPUs, I'm getting killed (not that I'm competing per se, but I'd expect better numbers ouf of my boxes than this).

I'm not doing anything out of the ordinary, nor am I using any scheduler. Based on what I've read, I suppose I can give that a shot, but in the interim, I'm open to some suggestions. Other threads I've found don't really touch on this kind of thing and some of the older threads are...well...well out of date.

[Tyrinon]

Looking at the 920 times, they would be in-line for a stock running system. I do see there is a bunch of pending work to be compared with other results. Those can take a few days to clear up and points to be assigned. For work units, Berkeley sends out four wu's and when three are returned and match up, then credits are assigned. You may just want to give it a few days to get things going. :)

[SameOldFatman]

I agree with the Evil one, you have more pending than validated, the numbers will go up

[Railgun]

Fair enough. I'm not the most patient person ever that's for sure.

That being said, in terms of squeezing as much out of this, and in light of other threads, is there a one stop shop with recommend settings? I've seen a lot of discontinuity regarding this. eg threading vs not. Using all cores vs saving one for handling GPU work. Multiple GPU WUs, etc.

Second to that, I'm wondering if there aren't other things at work here. Pretty much right away my main box produced pretty large numbers whereas my server...not so much. Again, perhaps I'm getting hung up on the other numbers I'm seeing on others' boxes.

Lastly, I think my main box has quite a few pending as well.

But I digress...and shall wait.

Thanks in advance.

[QuietIce]

A couple of things come to mind, here. First, it usually takes a couple of weeks for the RAC to stabilize. Variations in return times for validation will eventually even out but there can be big variations there.

Second, the first rig is doing most of it's work with the GPUs. Few people understand at first how far GPUs out-strip CPUs in Crunching.

Here are some Validated work units I grabbed to compare your two rigs. The times seem fairly consistent to me. If anything the second rig seems to slightly out-perform the first, which might be expected if you haven't idled a core to feed the GPUs. ;)

11,695.26 10,444.08 96.94
11,196.10 10,685.82 105.41

8,431.06 7,848.36 114.82
8,754.69 8,569.84 123.97

2,421.64 2,359.52 28.91
2,696.52 2,677.84 32.57


You might want to check out the SETI FAQs in my sig. They explain more about the validation process and how work units are credited ... :)

[Railgun]

Thanks...it was just a little impatience that was getting to me, that's call. :P