New version of my PPD Estimator

[[Ion]]

Sandy Bridge i7 support now included

[D_A]

Just a little thing, a mite if you will.

Perhaps mentioning somewhere that any numbers from this thing are only valid for Windows systems in general?

[Panther_Seraphi]

Scratch one of the ''it sucks vote'' was meant to be its great.

[[Ion]]

Just a little thing, a mite if you will.

Perhaps mentioning somewhere that any numbers from this thing are only valid for Windows systems in general?

Good point. Although, at least in the .exe version (most up-to-date), the only two options are x86 Windows and x64 Windows.

Is Linux PPD really that different anyways?

Scratch one of the ''it sucks vote'' was meant to be its great.

No worries

[shoota]

Good point. Although, at least in the .exe version (most up-to-date), the only two options are x86 Windows and x64 Windows.

Is Linux PPD really that different anyways?

Yeah in some projects it can be quite different.

[[Ion]]

Yeah in some projects it can be quite different.

I didn't really realize that, I've run Windows pretty much exclusively :o

[shoota]

I didn't really realize that, I've run Windows pretty much exclusively :o

then you're getting points I forget which projects they are but Linux will crunch more wu's but get awarded less points than a comparable windows system. kinda stupid if you ask me but i'm sure there's a reason

[D_A]

Linux 64bit crunches WCG work faster on all projects but most notably on HCC where it's up double the speed, give or take services and other apps etc. This means that each unit takes around half the time. Combine that with the benchmarks being not much different between Win64 and Lin64 and the Lin64 machine will always claim less for a given unit.
As for why the benchmarks are so close when the real world performance is so different ... well there's the fact that benchmarks are always a synthetic representation of a machine's performance and don't really show anything, then there's the conspiracy theory about how a Microsoft plant is diddling the code to inflate the benchmarks and make sure Windows appears competitive (I'm not convinced of this by the way, though MS did exactly this kind of thing with the World Standards Organisation not too long ago so ...) but overall I don't know. Whatever it is it's not a simple issue.

[[Ion]]

then you're getting points I forget which projects they are but Linux will crunch more wu's but get awarded less points than a comparable windows system. kinda stupid if you ask me but i'm sure there's a reason

I'm pretty sure it's faster on C4CW, I'm not sure what else.

Linux 64bit crunches WCG work faster on all projects but most notably on HCC where it's up double the speed, give or take services and other apps etc. This means that each unit takes around half the time. Combine that with the benchmarks being not much different between Win64 and Lin64 and the Lin64 machine will always claim less for a given unit.
As for why the benchmarks are so close when the real world performance is so different ... well there's the fact that benchmarks are always a synthetic representation of a machine's performance and don't really show anything, then there's the conspiracy theory about how a Microsoft plant is diddling the code to inflate the benchmarks and make sure Windows appears competitive (I'm not convinced of this by the way, though MS did exactly this kind of thing with the World Standards Organisation not too long ago so ...) but overall I don't know. Whatever it is it's not a simple issue.

That's astonishing!
So should dedicated WCG rigs run Linux x64 unless there's a compelling reason to do anything else? Even if PPD is about the same, it seems to me like it must be beneficial to the people at WCG to get twice as much work done, no?

[shoota]

But ppd between the two OSes isn't the same though right?

[[Ion]]

I think I pulled about the same PPD under Kubuntu 10.10 x64 as Win7 x64 on my i7 860 just running HCC. Kubuntu got far more WUs done, Win7 got far more points per WU.

[[Ion]]

Back from the dead with a new version! Check the OP for a link to download the newest edition, which supports Sandy & Ivy Bridge (including -E variants) CPUs and the AMD FX4/6/8 CPUs (first and second generations!)

[Rob_B]

Thanks.

For a moment I thought some thing was wrong. But when it asks for the number of cores, it really means cores and not threads.


BTW you need to update your signature.

[[Ion]]

Thanks.

For a moment I thought some thing was wrong. But when it asks for the number of cores, it really means cores and not threads.


BTW you need to update your signature.

Yeah. I'm still working on getting support for non-HT Intel chips too, but I don't have any i5s available. And to avoid having three different entries each for Sandy/Ivy/FX-Gen1/FX-Gen2 I decided to have just a single entry and make the user enter the number of cores. Now it also works for the Sandy & Ivy 8+ core Xeons too

I do--thanks!

[Otis11]

Thanks for the great tool.

[[Ion]]

OP updated with new version, information, and download link!