MOO Wrapper Happy Holidays!

[spingadus]

So, my last wu (Link) took 948.36 seconds and was awarded 3,686.40 credits.

3,686.40/948.36*86400=335848 credits/day

Previous results:

16,653.60/4,315.63*86400=333409 credits/day

So, the rate is the same, apparently I just got a shorter wu. This is with "Holiday" credits.

Considering that 'holiday' credits are about 2.5x normal and my gtx is dual core working on 1 task:

335848/2.5/2=67169.6 credits/day per core. This is basically what I had estimated before. So, actually nothing has changed.

will try running the benchmark as soon as I can get my client to download the ...stream.exe file. It's not currentl y in my moo!wrap directory for some reason.

[spingadus]

Be sure to give moo enough CPU. I usually give them 3 cores, so that even with moo running, the CPU is no where near 100%. Then I add back CPU threads one at a time until run times start to climb, then go back one notch.

It has 2 free threads atm and the cpu % fluctuates between 89-90%.

[zombie67]

I don't think you can look at any one unit to measure performance on moo. Mine are all over the place. You really need to measure day vs day.

[Al]

So, my last wu (Link) took 948.36 seconds and was awarded 3,686.40 credits.

3,686.40/948.36*86400=335848 credits/day

Previous results:

16,653.60/4,315.63*86400=333409 credits/day

So, the rate is the same, apparently I just got a shorter wu. This is with "Holiday" credits.

Considering that 'holiday' credits are about 2.5x normal and my gtx is dual core working on 1 task:

335848/2.5/2=67169.6 credits/day per core. This is basically what I had estimated before. So, actually nothing has changed.

will try running the benchmark as soon as I can get my client to download the ...stream.exe file. It's not currentl y in my moo!wrap directory for some reason.

Correct me if I'm wrong, but I think you should be looking for the cuda version, not stream...I think stream is for benchmarking the ati's only. I could be wrong on that though.

[spingadus]

Correct me if I'm wrong, but I think you should be looking for the cuda version, not stream...I think stream is for benchmarking the ati's only. I could be wrong on that though.

Yeah, I think you are right. It wouldn't download. I even turned on cpu processing in the project configuration and eventually tried a reset.

I ran the cuda bench and it said to use #1. So, I've set that.

[STE\/E]

Personally I think the Benchmark thing to choose a CPU Core is a lot of wasted time, also shutting down 1-2 or 3 Cores. I tested shutting down Cores (3 on my HT 12 Thread E980x's) and didn't see 1 bit of time difference or Credit difference for either my 6990 or 590's ... I also did a little looking around at guy'z Box's that say their using this Core or that Core & my Wu's run just as fast as theirs just using AUTO Select unless they have theirs Overclocked more than mine.

I also discovered a long time ago that setting the # of CPU's to use to <ncpus>0</ncpus> that the GPU's WU from any GPU Project will automatically use the amount of CPU that it needs to use without any input from me. I don't know how that works since I set it to use 0 CPU but it must be something the BOINC Client Automatically adjusts for when running GPU Wu's because at times depending on the GPU Wu's I'm running I have 2-3 or 4 CPU's running on a Quad Core or 4-5 or 6 running on a 6 Core Box.

I know that works for Client 6.12.34 & up but not sure about anything lower than that ...

[Crazybob]

I ran the bench on this dual system and the #2 core looked to be the best. Problem is that if you have more than 3 systems running Moo and you set each to a different location and set the core #, won't systems set to the same location run differently with the same core number selected?

[Al]

Is anyone else starting to see a consistent, albeit sight decrease in credits/wu? Mine are coming in at the low 13000's. Had been in the 16's, 15's and some 14's. Not complaining, just an observation.

[coronicus]

aye.. also seems like they stopped sending out work, just noticed

also, no colatz or MW... so what should i use for backup?

[Fire$torm]

Personally I think the Benchmark thing to choose a CPU Core is a lot of wasted time, also shutting down 1-2 or 3 Cores. I tested shutting down Cores (3 on my HT 12 Thread E980x's) and didn't see 1 bit of time difference or Credit difference for either my 6990 or 590's ... I also did a little looking around at guy'z Box's that say their using this Core or that Core & my Wu's run just as fast as theirs just using AUTO Select unless they have theirs Overclocked more than mine.

I also discovered a long time ago that setting the # of CPU's to use to <ncpus>0</ncpus> that the GPU's WU from any GPU Project will automatically use the amount of CPU that it needs to use without any input from me. I don't know how that works since I set it to use 0 CPU but it must be something the BOINC Client Automatically adjusts for when running GPU Wu's because at times depending on the GPU Wu's I'm running I have 2-3 or 4 CPU's running on a Quad Core or 4-5 or 6 running on a 6 Core Box.

I know that works for Client 6.12.34 & up but not sure about anything lower than that ...

I think some of the factors that influence Moo production include hardware make and system setup. I have 3 5830s, one is a HIS and other two are Sapphire units. And each computer is different, i7 920 (Socket 1366) running Win7 x64, i7 860 (Socket 1156) running WinXP Pro x64 Corp. and a Phenom II 810 Running WinXP Pro x64. When the HIS was the only one I had I "thought" I was getting good credits but I didn't know anything about the benchmarking or why to reserve CPU cores. After I started running the Sapphire cards I discovered they were way outperforming the HIS. Maxwell and Bryan turned me on to maxing out the HIS which increased that cards output somewhere in the 20~30 range. Sorry I forgot the exact amount.

Your mileage may vary....