Fah Smp Afiinity Changer - It Works!

SpoogeMonkey

2[H]4U
Joined
Feb 20, 2005
Messages
2,934
I've been seeing some discussion about the Affinity Changer program and decided to give it a try today. Being the skeptic that I am after all the bogus "ram doubler" programs from the mid 90's I wasn't expecting much. I must say that I'm very impressed with this little gadget.

My E6420 running at 3.2ghz was crunching a 2653 wu in about 13:30 per frame. After installing the affinity changer service, times dropped to 11:45.....a good 1:45 faster and 12.5% speed increase for free!!! I highly recommend everyone with an SMP setup to give this gem a try; with any luck we can increase our ppd by an easy 20,000 by nothing nothing.

Check the link for all the details.
 
Yeah, it does indeed work. I use it on all of my windows machines. I wonder if notfred does this on his CD? If not, adding something similar to it for Linux would be a major performance boost due to the 4 instances not having to fight over the cores. Hmm...
 
Anyone know if this also helps with VMs? Looking at the documentation doesn't look like
the program is designed for VM. But what about manually locking affinity to cores, anyone have any
1st hand knowledge, does this improve times? Trying it out now on a
quad core running 2 Vms, keeps fingers crossed....
 
Anyone know if this also helps with VMs? Looking at the documentation doesn't look like
the program is designed for VM. But what about manually locking affinity to cores, anyone have any
1st hand knowledge, does this improve times? Trying it out now on a
quad core running 2 Vms, keeps fingers crossed....

Manually locking does indeed help -- it's exactly what the program does. The only difference between the program and doing it manually is that the program rechecks affinity every 10 minutes to make sure another program/the OS hasn't changed the affinity.
 
This works with Dual Cores? :confused: Guess I will have to try it out when I get home then.
 
Just when I thought for sure everyone was using this, up pops another thread about it. Back when I started using it, I found the increase to be 9.4% So yes, it's well worth 7.3 seconds to download it and install it.

It's main purpose is getting the most out of a dual core.

 
is this a program? or hardware?

It’s a tiny well written program. Dl and install. No need to reboot or restart your clients. Give it about 10 to 12 min to kick in.

No joy on VMs, those you must set by hand however once set they stay set unless you reboot or shut down the VMs.

It runs as a service and requires no care or love;)


 
It's a program that runs as a service and assigns affinity to the 4 fahcores that run. It assigns the cores to cores (funny huh?) based on memory use. Highest and Lowest goes to core 0 and the two in the middle go to core 1. Bam, nice gain in PPD for doing almost nothing. Links to download are in the OPs link.

 
If I ran this affinity changer with a quad, can I run two SMPs at the same time without using a VM?

Or am I missing something?

For quad core processors, first SMP client is assigned to 1st and 3rd core, 2nd SMP client assigned to 2nd and 4th core. Assigning affinities on QUAD if you run one SMP client does not make sense.
 
Killer[MoB];1032151419 said:
Just when I thought for sure everyone was using this, up pops another thread about it.
Well, not everyone can use it. I'm either running AMD machines or dual quads. Xeons don't seem to benefit at all. I tried with any number of clients and it doesn't make a difference. So, it's either single C2D or single C2Q processors that will see a benefit, nothing else TMK. Despite the fact a lot of people are running them, that's still a very narrow range of processors.

 
It hasn't made a difference with my E6700 machine at work. Maybe Dell's are immune to the benefits? :p


 
If I ran this affinity changer with a quad, can I run two SMPs at the same time without using a VM?

Or am I missing something?

You could always run dual SMP clients without a VM. Only one problem. When you shut down one client one or both will crash. If it’s a folding only machine then the pain is minimal but if you use it every day……..not recommended.;)

 
Does this make a difference on a vista x86 machine as well? I have noticed that vista is a little slower at folding than xp was :(
 
I'm giving a try right now with my own work computer running XP with E6300. I'll let everyone know if it improve ;)

 
It hasn't made a difference with my E6700 machine at work. Maybe Dell's are immune to the benefits? :p



I'm running it on 2 E6400 Dell's and it works great! Dropped it from 21 minutes per frame to 18 minutes per frame :eek:
 
Does this make a difference on a vista x86 machine as well? I have noticed that vista is a little slower at folding than xp was :(

I have it running on a Dell E1501 with a T7500 processor running Windows Vista 32-bit and it works great!
 
After 2 hours, the PPD has raised by 100-150 approximately and I shaved 2-3 minutes off per frame. It does work indeed and I will install this on all the Windows SMP boxes.

 
After 2 hours, the PPD has raised by 100-150 approximately and I shaved 2-3 minutes off per frame. It does work indeed and I will install this on all the Windows SMP boxes

Hmm, I think it's working for me, too. I looked through my log file and over night my frame times are excellent. I think I'm not "seeing" a difference during the day because I'm actually using my work computer and SMP is being pushed to the background.



 
Killer[MoB];1032151419 said:
Just when I thought for sure everyone was using this, up pops another thread about it. Back when I started using it, I found the increase to be 9.4% So yes, it's well worth 7.3 seconds to download it and install it.

It's main purpose is getting the most out of a dual core.


I guessed that the main benefit would be for quad cores...thats why I never paid much attention to it before. Who woulda thunk it worked so well for a simple dual core :D

This is another little item that needs to be edited into the SMP faq.......the more you know.....


 
I agree this should be added in the SMP optimizations toolbox.

Sepaking of guides, where's schenksmill and the updated guide ???

 
You could always run dual SMP clients without a VM. Only one problem. When you shut down one client one or both will crash. If it’s a folding only machine then the pain is minimal but if you use it every day……..not recommended.;)
My workaround that problem is disconnecting or disabling the Internet or LAN connection. That will stop the clients but not crash them. Then you can proceed to shut them down safely. I have not found any other method to run multiple SMP clients without running a risk of one or more crashing when shutting them down, but this method has worked well.

After 2 hours, the PPD has raised by 100-150 approximately and I shaved 2-3 minutes off per frame. It does work indeed and I will install this on all the Windows SMP boxes.
Can you let us know if you install the AC on dual Xeon machines (providing there are some)? TIA.

 
My workaround that problem is disconnecting or disabling the Internet or LAN connection. That will stop the clients but not crash them. Then you can proceed to shut them down safely. I have not found any other method to run multiple SMP clients without running a risk of one or more crashing when shutting them down, but this method has worked well.

Good idea there.

I made up a bat file I run as a scheduled task every few hours that backs up my two folding directories. Just a little safety net incase I have a problem. So far so good though. :cool:



 
My workaround that problem is disconnecting or disabling the Internet or LAN connection. That will stop the clients but not crash them. Then you can proceed to shut them down safely. I have not found any other method to run multiple SMP clients without running a risk of one or more crashing when shutting them down, but this method has worked well.

Can you let us know if you install the AC on dual Xeon machines (providing there are some)? TIA.


Who is telling I got dual Xeon machines ?? I don't have any at work.

 
Who is telling I got dual Xeon machines ?? I don't have any at work.
Nobody actually. But a few who are running work machines do, and since I didn't know the specs of your work machines I thought it could be a distinct possibility. I was interested to know other experiences with the AC and dual quads.

I don't know where you work or what you do exactly. I'm in eastern Canada (Quebec) and most of the members here seem to be US citizens.
 
I work for the Ministere of Justice of Quebec. All the PC are all for basic office duties so no Xeons. I don't have access to the servers since it's another team who manage them. I'm a senior IT Technician and I'm in charge of managing the renewal of the computers fleet so that explain how I can fold on many pre-production boxes while it await assignment.

 
I agree this should be added in the SMP optimizations toolbox.

Sepaking of guides, where's schenksmill and the updated guide ???

busy busy....spring break turned into very not spring break, hence my lack of posting lately....will get to it soon hopefully
 
That's ok :) Just add the Affinity Changer tidbit in the guide.

 
As you can see by my sig I'm not into the world record of OC's I keep my lowly machines as cool as possible for running F@H 24/7.:cool:

Being kind of late to the Affinity deal I installed it on the only 24/7 windose machine I have folding ATM a E6600, OC'ed 2.8 GHz and it went from about 16 mins a frame to about 14 mins a frame. (when I say about, depending on the wind direction :rolleyes: , I sometimes get better and sometimes worse frame times)

What impressed me was I get about 12 min a frame in LinuxSMP, but I was only getting about 16 min a frame in winders. Now, with this infinity program I narrowed the frame times to about 2x mins difference (about 14 mins a frame) which is copacetic to the max to me :p

Man, I wish the Q9450's would hurry up and hit the wilds, I almost bought a Q6600 G0 on Ebay last night.(just one nervous finger click away) I was rationalizing 9x multi compared to a 8x multi, but I furget the low vcore of the 45 nm chip. :p (I'm tryin' not to own too much of the local power company ;))
 
Tried to get affinity changer to do anything, but it does absolutlely nothing.
My personal vote is that it's completely worthless.
 
Tried to get affinity changer to do anything, but it does absolutlely nothing.
My personal vote is that it's completely worthless.

You're just old and if it doesn't work it's because it's your fault. :p

I have no experience with this since I can't actually use it and my interjection is mostly worthless and can be ignored if you choose to do so.:)

 
Tried to get affinity changer to do anything, but it does absolutlely nothing.
My personal vote is that it's completely worthless.

You have to install it first. Just downloading it does nothing. ;)



 
Killer[MoB];1032156291 said:
You have to install it first. Just downloading it does nothing. ;)




It has no effect. Prove otherwise.
 
It has no effect. Prove otherwise.

I have 3 systems I am running it on. Two of them are dual core intel chips. On my E6600 it is 9.4% faster after installing affinity changer. On my E4500, I don't know the exact times but it is faster after installing affinity changer.

The third system is a quad. As pointed out before, it does nothing on a quad running one instance of WinSMP. I have 2 instances of WinSMP running using affinity changer. I'm getting frame times of 11:12 to 11:32. So depending on how much use the machine gets, PPD can be as high as 4394 to 4524 PPD. Running just one instance, I was getting frame times of around 8:10 which is 3103 PPD for the machine.

Is that proof. It's enough for me. Plus there are posts going back several weeks on this forum and back to last year on other forums to back up that it does work.
Plus, before the little program was made, you could manually set affinity with the same results as long as you set the right fahcores to the right cpu cores. That's just a pain though. AC does the work for you.
 
I gained about a minute per frame on this poor 2140. I wonder if this helps the small cache cpu's or the larger ones more?
 
Back
Top