Folding Forum

Hi, maybe it does not belong here, because it seems project related, not to specific WU

I am running beta 6.x clients on my Macs since december now, mostly had WUs from projects
26xx, giving 1760 points.
On my MacMini, CoreDuo 1.66 these ran at 33 to 35 pph, today I got my first WU of p3052
(run 9, clone 24, gen 94) which credits 1440 points, but pph is only about 27 to 28.
That means, it will run about as long as the 1760 point units. Is this miscalulated credit
or is it because the CoreDuo is not really uptodate? Does this project benefit more from
Cure2Duo machines?

Not a real problem, but my ppd goes down...
Anybody experiences with this project.
The logs do not state any errors or such, it is just slow compared to p26xx WUs

Lutz

Ok, nobody any idea?

I finished this WU now with a pph of 27.7.
While running I switched from 6.02 to 6.10 beta, no change in speed.
I noticed the for tasks of FahCore_a1.exe to run at about 50% for one
instance and 30 to 35 for the other three only, not totally to the limit.
Now I have a 2605 WU again, where CPU usage is at 50,50,45 and 40%,
about 33 pph. So, seems to be a issue with the project 3052 perhaps,
and the limitations of the old CoreDuo CPU. Will have to wait until
my MacPro at work with two Xeons will get a 3052 WU, maybe they
benefit from a large cache and a better RAM bus.

EL

I saw a similar report elsewhere.

It's likely that the lower PPD is due to hardware differences between your machine and the benchmark machine (such as a smaller cache) and if that's true, there's nothing that can be done about it. At this point, nobody really knows.

Yeah I was averaging ~3600 PPD with the 1760pt work units (on a Q6600 machine), but am now only getting ~2300PPD with these 1440pt work units.

Can something be adjusted her? Thats a HUGE disparity.

As Bruce just posted, differences in hardware from the Benchmark machine can make significant differences in PPD for the SMP clients. The biggest noted difference is the Benchmark having a 4 MB cache on the processor. And on some projects, that memory bandwidth makes a big difference in PPD if your processor only has 2MB or less.

The only adjustment you could make would be to purchase hardware similar to the benchmark computer.

7im wrote:The only adjustment you could make would be to purchase hardware similar to the benchmark computer.

Ok, i assumed something like this. not a real problem, I just noted the considerable speed loss on this project (or WU).
Never saw this on my MacPro at work with 2 Dualcore Xeons at 2.6 GHz.

In fact, I've planned to build some Q6600 machine in the near future anyway

Well if we got a bunch of time posted per different machine/work unit a fahinfo, we might
be able to see something.
Though I have not at this time, figured out what if any good the info you get there is.

If I look at the avg of say 3062. There are six submitted records. X2 AM2, Phenom X4, C2D allen, C2D Con, C2D Con M, C2Q kent.
PPD/GHZ 197.60 - 173.58 - 237.34 - 281.29 - 535.36 - 268.10

So what good is this info? You don't know the Ghz, running two clients on quads, or
one. Just guess?

Per se: the Phenom 173.58 x 2.0 Ghz = 347.16 x 4 cores = 1388.64 PPD
And my C2D 6320 281.29 x 2.8 Ghz = 787.612 x 2 cores = 1575.224 PPD

If this is correct, the only thing I see is "don't get Phenom" as yet.