COUNTERINTUITIVE FOLDING RESULTS

[Wheels_78]

I've set up my 560 TI, and my Athlon x3 455 (unlocked and overclocked to 3.8Ghz) both to fold as per the tutorials here on HC. The only option I did not copy exactly from the tutorials was setting the CPU client to run as a service so that it would begin at start-up.

I'm using HFM.Net to monitor, again, as per the HC tutorial and the results I'm seeing just don't make sense to me.

The 560 Ti is kicking my CPUs ass all over the place, as in 560 TI = 14,945 PPD and the CPU = 3500 PPD. Maybe this is as it should be? I don't understand the demands of folding so maybe my intuition is backwards here but I'm thinking the unbalance should be the other way around.

The CPU goes to 100% on all four cores as soon as I start the client so I know it's working. What's going on here, any ideas?

[lowfat]

CPU folding is exponential due to a bonus system. Unless you have a Core i7 or faster you really don't get all that many points with them. So your points sound pretty bang on the money in terms of where they should be.

[Wheels_78]

Ok, glad to hear that. Thanks for the reply.

[_dangtx_]

3500 ppd is quite decent for an amd quad(current gen) , so dont worry about it

[geokilla]

You better be using Beavers_Gone_Bananas xD

[Wheels_78]

sure am ;)

[FiXT]

Great to have you onboard Wheels :D PPD is look great :)

It definitely can throw a guy for a loop though when you see your GPU out performing your CPU.

[frontier204]

Hmm my Phenom 9950 OC to 2.7 GHz on Linux is giving me 6300 PPD. Note that's on a rig that doesn't even have a GUI installed on its OS.
I know your GPU client is eating away somewhere between 500 and 2000 PPD depending on WU, but shouldn't you still have >4000 PPD?

I wonder if it's the Phenom's 2 MB L3 cache or the Linux that's making most of the difference? (because you have a 40% clock speed advantage + one AMD generation over my rig)

[Wheels_78]

It got a little better Frontier. I managed to clock up another 100mhz ( not sure how much difference that made) and over the longer haul it now seems to be doing 4500 - 5000 most of the time

[10e]

It also depends on the work units. Some naturally run a little bit faster than others, and even a small change in time per frame (or percentage) can make a noticeable difference.

I don't think my 3.6Ghz 955BE was doing more than 5,500 PPD, but that was with only GTX200 or G92 clients running, and they use less of a CPU core per GPU client than the GTX4xx and higher clients do.

Nice PPD from two clients though.