Folding Forum

The biggest protein yet. Worth 2539 points. "Only" 38 % bigger than the next biggest protein.

The deadline of 5 days is pretty tight for the size. Only 2 % of the protein is completed per hour on my 2.4 ghz iMac.

50 hours for the total WU ... a little more than 2 days ...

That sounds correct to me

Yeah, I thought about it some more.

The time is about right.

I just got one of these today. I have a four core 2.66 GHz Mac Pro, and it's going to take over 50 hours here as well, so F@H isn't making good use of my cores.

And those 50 hours will only earn 2539 points compared with 2.5 x 1760 = 4400 points I've been getting per 50 hours on Project 2605

AlanH wrote:I just got one of these today. I have a four core 2.66 GHz Mac Pro, and it's going to take over 50 hours here as well, so F@H isn't making good use of my cores.

And those 50 hours will only earn 2539 points compared with 2.5 x 1760 = 4400 points I've been getting per 50 hours on Project 2605

Welcome to the forum AlanH.

F@H is always making good use of your computer, regardless of your PPD!!! At first glance, the points on 3060 seem a tad low. But just because the points on 2539 are higher than the benchmark doesn't make the points on 3060 too low. Before guessing any further, how does your system compare to the SMP benchmark system?

How do you decide the credit value of SMP work units?

Hi 7im

Thanks for the welcome, although I was actually around on the previous board. This is just my first post since you moved here.

My comment about the utilisation of my system related to the fact that my completion time with 4 cores looks to be similar to that reported for a 2 core iMac with comparable CPU speed. The extra 2 cores don't seem to contribute.

My Mac Pro has 2 x 2.66 GHz dual core Xeon (Woodcrest) with 4 GBytes of FBDIMM and 1.33GHz bus. So my CPUs are 14% faster than your reference system, and I have 20% less memory. Looking at memory usage in the Activity Monitor, I doubt if my smaller memory would have much effect, if any, since F@H only uses 30 MBytes per process. My Mac spends most of its day waiting for me to do anything at all, and then waiting for my next mouse click/key press. It's not a heavily used system currently.

My system has been folding p2605 units for a little while now, and was consistently turning them in every 20 hours. They were worth 1760 points, so your reference system must have folded them in 24 hours, making my system 20% faster than the reference.

The p3060 unit I completed yesterday took 48 hours. Its points value of 2539 indicates that your reference system completed it in 2539/1760x24 hours = 34.6 hours, so my system is now 18% slower than the reference system. I'm now folding a second one. so my average contribution to our team's efforts looks like it will drop by over 35% now that I seem to be targeted for p3060 instead of p2605.

AlanH wrote:My Mac Pro has 2 x 2.66 GHz dual core Xeon (Woodcrest) with 4 GBytes of FBDIMM and 1.33GHz bus. So my CPUs are 14% faster than your reference system, and I have 20% less memory. Looking at memory usage in the Activity Monitor, I doubt if my smaller memory would have much effect, if any, since F@H only uses 30 MBytes per process.

The size of the memory isn't going to matter. The cache will, however. How much cache does your CPU have?

The speed of some WUs is the same when you compare a 4MB cache with smaller cache, but other WUs see a rather large speed difference. (Similarly, a quad should have 2x4MB.)

I have 4 MBytes cache per CPU, same as the reference system. My Mac Pro is a stock dual 2.66 GHz Woodcrest system. I don't even know whether cache size is an option from Intel. It's certainly not a build option from Apple. Everything except the items I mentioned is going to be the same as the reference system. Oh! And I'm running Leopard.

I should correct my previous statement - my first p3060 unit took 48 hours, as stated, but this is 28% slower than the reference system, not 18% as I wrote.

My second p3060 unit has finished today, and actually went a bit faster - 45 hours. That's still 23% slower than the reference system.

Maybe these p3060 units are more variable in their workload than previous ones, and the benchmark run on the reference system happened to use a fast one?

I now have a p2605 unit again. It aborted at 17% and restarted, but I'll still probably get a better rate of production on this unit than I have on my p3060 units.