New Linux SMP WU's

shadco · Post by **shadco** » Tue Apr 17, 2007 12:53 pm

I just started catching these.

I've got one on an Opty SMP rig and one on a C2D.

First impressions are they are gonna yield lower ppd, I could be wrong though.

Code: Select all

2608	 171.64.65.56 	P2608	126010	 4.00	 4.00	 1385.00	100	GRO-SMP	Description	kasson
2609	171.64.65.56 	P2609	126010	4.00	4.00	1523.00	100	GRO-SMP	Description	kasson

Post by **DMB2000uk** » Tue Apr 17, 2007 12:57 pm

I've had some 587point WU's before now, is this what they are?

Dan

shadco · Post by **shadco** » Tue Apr 17, 2007 1:02 pm

DMB2000uk wrote:I've had some 587point WU's before now, is this what they are?

Dan

Nope

1385 pts and 1523

Post by **DMB2000uk** » Tue Apr 17, 2007 1:07 pm

I though that was what those numbers were after I'd posted

Dan

shadco · Post by **shadco** » Tue Apr 17, 2007 1:16 pm

Nice ppd drop with these

Code: Select all

[18:33:29] Project: 2609 (Run 0, Clone 30, Gen 0)
[18:33:29] 
[18:33:29] Entering M.D.
[18:33:36] Protein: Protein
[18:33:36] Writing local files
[18:33:37] t)
[18:33:37] ra SSE boost OK.
[18:48:38] t triggered.
[18:53:02] Writing local files
[18:53:02] Completed 5000 out of 500000 steps  (1 percent)
[19:08:03] Timered checkpoint triggered.
[19:12:45] Writing local files
[19:12:45] Completed 10000 out of 500000 steps  (2 percent)

shadco · Post by **shadco** » Wed Apr 18, 2007 5:26 am

Time to completion on a 2608 looks like 33% longer on a 2608 vs 2605 680 minutes vs 1030 minutes.

2.588 pts per minutes for 2605

1.344 pts per minute for 2608

These are based on an E6400 running at 3.2Ghz.

I'll do calcs on my opty's later but the drop off there looked to be far less

Post by **Apoptosis** » Wed Apr 18, 2007 7:36 am

interesting... thanks for sharing the stats!

shadco · Post by **shadco** » Wed Apr 18, 2007 8:08 am

Apoptosis wrote:interesting... thanks for sharing the stats!

OK I was off on my calcs (not enough coffee)

Here's the deal

E6400 @ 3.2Ghz

2605=1133 mins 1.553 pts per min
2608=1717 mins .8066 pts per min

Opteron 170 @ 2.8Ghz

2605=1900 mins .926 pts per min
2608=2300 mins .602 pts per min

That's enough measurebation for now.

Moral of the story as Dan said mo cache is better

shadco · Post by **shadco** » Sat Apr 21, 2007 6:32 am

With the addition of the last Opteron I shold have leveled off at about 11000 ppd average based on the old points allocation scheme.

Now it looks like I'll level off at about 7200 with the new points allocation scheme.

armstpe · Post by **armstpe** » Sat Apr 21, 2007 8:51 am

YEE GAD'S

Just got a 2609 on my X2 [email protected]. I dont' know if it will finish in time. Deadline is 4 days and by current time per frame will take 3.8 days to complete.

On my two Opteron 170's there is no problem completing on time. These things are very cashe sensitive.

On my Opty 170 with bad memory running at 2.5Ghz (same as the 3800) it will take 2.7 days.

Paul

armstpe · Post by **armstpe** » Sat Apr 21, 2007 9:20 am

Just some wandering thoughts

It has come out in various threads on the folding forum that the prefered deadlines and the final deadlines for the SMP workunits had been changed so that they were the same.

As far as I can see that only applied to these two work units. I wonder if this is telling us something?

Paul

armstpe · Post by **armstpe** » Wed Apr 25, 2007 6:26 am

PHEW

My X2 3800 completed it's work unit with 8% of time remaining. Now it's running a 2608.

Paul