• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Processor scaling

E

elleana

2[H]4U
Joined
Apr 15, 2006
Messages
2,590
(Apologies for the numerous questions - I am trying to get a feel of how to optimise the various parts I have in my boxen to maximise PPD)

How well does PPD scale as mhz goes up? Like if I overclock by say, 10%, will I see a corresponding 10% increase in my PPD? On a somehat related note, how does cache affect PPD? I know double cache won't mean double PPD, but say, between a E7200 and E8400? What about between a quad and dual core running at the same speed. If the quad runs two smps compared to one on the dual core, will the PPD also double or is there some overhead?

And of course - Intel vs AMD? Is there anything in the architecture of either side of the fence that F@H is optimised for? Would a 3 core AMD proc make sense for 2 cores running a SMP client and the 3rd core for 2 GPU clients?
 
elleana said:
(Apologies for the numerous questions - I am trying to get a feel of how to optimise the various parts I have in my boxen to maximise PPD)

How well does PPD scale as mhz goes up? Like if I overclock by say, 10%, will I see a corresponding 10% increase in my PPD? On a somehat related note, how does cache affect PPD? I know double cache won't mean double PPD, but say, between a E7200 and E8400? What about between a quad and dual core running at the same speed. If the quad runs two smps compared to one on the dual core, will the PPD also double or is there some overhead?

And of course - Intel vs AMD? Is there anything in the architecture of either side of the fence that F@H is optimised for? Would a 3 core AMD proc make sense for 2 cores running a SMP client and the 3rd core for 2 GPU clients?
Click to expand...

OC'ing the cpu on smp folding gives one a nice boost. But I don't think a 10% OC nets 10% more folding. Maybe 7%.

As for cache. More is better. I have a E6400 and E6600. When smp folding the E6600 whoops the E6400 mhz to mhz. Where as the E6600 produces right at half of what the Q6600 does when smp folding mhz to mhz. That would be the cache in action.

Intel just out folds AMD hands down. I run my Q6600 right now smp folding on 3 cores and one feeding a gpu in XP. If I was on Vista I would fold all cores smp and feed the gpu with a part of a core. Best setup would be two gpus on two cores with VMware running on two cores.

There are no set hard rules when it comes to getting the most out of your boxen. What works for me could be a failure for you. what works for others sometimes won't work well for me. You just have to build and play.
 
I just keep tinkering with the OC, the what core use what, and so on. Once I get a combo that works..... I walk away, And hope a power outage doesn't come.

I've found that it much easier to have CPU boxen and GPU boxen. I get a headach every time I try and run both. One always seems to steal from the other. (probably just me)

 
There is almost a straight line for PpDpGhz within the same family of CPU's.
So a 10% faster CPU will produce ~10% more points.

Cache. The bigger the better.
Some protien dont need that much cache to fold in and will roughly have the same PpDpGhz across a wide range of CPU's.
Others need a lot of cache and the CPU's with a small cache will take a hit in terms of PpDpGhz.

Back in the days of Tinkers, AMD's folded at twice the speed of Intel.
But at the moment the Core2 CPU's rule.

You can look up the average PpDpGhz figures of the various CPU crunching the various Protiens over at FahInfo.

I'm useing Q6600's & nVidia cards on my mixed client boxen to fold on under XP.
I've found that ............
You can run upto 4 GPU clients on one CPU core without taking a hit PpD wise.
You cannot run the GPU and SMP clients on the same core without taking a big hit PpD wise.
Its easier to set affinities running the client inside a VM, you just set the VM, than it is running a naked SMP client.

Luck ............. :D
 
For CPU folding, the best you're going to do is probably a Q6600. It's going to get the most points for the power, it's cheap and it's a damn good overclocker. I have a 50% overclock on two of my Q6600's and they probably get close to 50% performance increase. Most of the "loss" is normally from inefficiency due to the work units or the cores itself.

For the best points, I run dual Linux VMs and each one with its own client. This usually gets the best PPD as it helps limit the inefficiency inherent with the multicore client. Also, clients reporting the system as a dual core machine have a tendency to get the work units with the best PPD.

Something to keep in mind is that the A1 core is rather inefficient when running more than 2 cores. The A2 core is supposed to take care of this but I haven't tested it yet as I still run all my clients in VMs which means they only run as dual cores. However, the A2 core does give much better PPD even while running on 2 cores and does max out both cores much closer to 100%.

Cache within the same architecture does seem to have an affect up to a point. As stated, an E6400 (which I used to have) doesn't fold as fast as an E6420 and the only difference is the amount of cache. Any more than 2 meg cache per core doesn't make a difference.

There is no contest between Intel and AMD for folding. The Core2 architecture rapes AMD at this point in time. Neither the Athlon X2s or the Phenoms have a chance. AMD has a lower IPC and doesn't clock as high in most cases which puts Intel on top.

 
You must log in or register to reply here.
Back
Top