This new core...

[OSUguy98]

I got my first (to my knowledge, I've been out of town the last couple days) protein crunching away with the new core... but I noticed my PC was sounding a little odd... I have a bunch of thermal controlled fans and they were speeding up and slowing down much more frequently than usual... Normally, I could hear them slow down a bit in the few seconds that it takes for fah to finish a frame and get the next... but this was happening every few seconds...

So I took a look at the CPU performance graph in XP Pro... (I have a 2.4C clocked at 3.0) and fully expected to see the core cranking away with 49 or 50% utilization of the CPU (since I only have one instance running on it right now).... But, what I saw surprized me... it would jump to 50%... then fall back to 0%... then 12%... jumping around quite a bit... I took a look at Abit's Hardware Dr. to get a decent idea of the temperatures... Usually, the CPU is running in the 53C-56C range... but it was running around 43-45C... and I guess that's about the threshold for my fans to switch between speeds

The core seems to be putting out similar frame times to the 78 Core, even if it's not fully using the processor (of "half" of it with an HT P4)... Anyone else notice anything different with the new core?

 

[LPerry]

Currently this is normal for that version of the core. That may change.

 

[OSUguy98]

As long as it's normal

seems odd that it doesn't use the proc. quite as much, but does the same amount of work... but hey, as long as it's folding...

 

[HoosierDad]

Imagine how fast the WUs would complete if they used 100% of the CPU all of the time!!!

 

[BillR]

I'm sure they will soon get to the "core" of the problem

Sorry, couldn't resist

Fold on

BillR

 

[OSUguy98]

and I thought I was the only one who used horrible puns... (sorry Bill, I had to)

I thought that the two cores (78 and 79) were identical save the fact of the 79 core using doubles instead of floats...

maybe they tweaked the code a bit to make it crunch a bit faster? faster folding is always good

 

[MN Scout]

The processor graphs look better when they are straight lines.

Fold on.

-MN Scout

3.6Ghz folding for the [H]orde

 

[OSUguy98]

I'll agree to that.... flatline at 100%..... Now if I could just afford a Superdome or 10 to fold on..............

 

[gnewbury]

How about this e-mail I got at work:

Looking for FREE compute time on fast multi-processor, giga memory systems? If you have a project for NEXT Fiscal Year that fits into one of the following catagory: (you don't need to meet all, just real close to one of them.)

- Computational requirements - 15,000 GF-hrs over one year
- Peak performance - 10 GF system speed
- Memory - 5GB
- I/O requirements - 100GB on-line storage or 200 GB archival storage


If you have a project with any of these requirements or any questions about High Performance Computing Modernization Programs <snip>

DA%N think of the WU, heck it would even chew through Tinkers!!

 

[LPerry]

This weird use of the CPU is being looked into. This is new code and very experimental so there are bound to be some issues that show up. I have commented and posted about this issue to Stanford. I'm sure they will eventually improve this. It wastes a lot of potential CPU cycles, about 60-75% in my estimation, and needs a solution.

 

[OSUguy98]

Yup.... if it can finish a unit in about the same amount of time with only 1/3 of the CPU cycles................... then by all means crank that up to 100% and we'll finish them in 1/3 of the time