New VMPlayer supports 4 cores!

Vaulter98c

[H]ard|DCer of the Month - October 2009
Joined
May 21, 2008
Messages
5,817
Browsing around the interwebs because I'm tired of looking for hardware right now, and I seen this..

LINK

Can someone here try this? I'm at work and my Quad is sitting on a desk...

Few things to remember...Everything is the same as the old VMsetup method found HERE except...

1) set the config to " numvcpus = "4" " instead of 4
2) when you do the config in your browser, leave it at 4 not 2
3)???
4)Profit?

I would also suggest you put atleast 1 Gig, maybe 1.5-2 for the memory. Early reports show slighlty higher numbers, more responsive computers, and lower heat, and increased stability. Ive read that all you have to do is install it over your old client, but youll need to delete your current VM (just the work) and restart it with 4 cores to keep from having a corrupt WU. I'm seeing Q66's at 3.2 Ghz getting like 4500 PPD, clients finishing in like 10 hours

Any takers here? If this works, I suggest we update the guide so people can get an easier set up

Obviously, this only applies for 4+ core users, dual core people should not apply, lower scores have been reported
 
I'll try this and report back in an hour, I have a quad core 9100e.

EDIT: Wait... and do what with it?
 
I'm sorry, I have never heard of doing this. I always thought the CPU client would use all 4 cores, by default.

EDIT: Wait? You're posting from work?
 
yup. from work lol

Anyways, here's the concept.

THe standard client sucks. Can't use a CPU that well and is only 1 core. So, there is the SMP client. Well, the Windows one is pretty bad too. Hard to set up, not that stable, no where near efficent (sp), can't even use what it has efectivly. Linux client can do that, but most can't be bothered to run Linux, and it hinders GPU folders anyways, so we use VM's. Notfred is the name of the VM we use because it is pre packed and all we have to do is a bit of config then it's done.

This was good, but, the free version of the VM software (VMPlayer) could only support 2 cores, so if you had a 4 or 8 core rig, you had to run 2,3, or even 4 VM's. While this would work, and put out nice numbers, it killed system resources and the system itself was pretty laggy. Some people didnt even have enough RAM to run multiple VM's

Now, you can use the free client and run 1 4 core VM rather than 2 2 core VM's, this cuts down on RAM needed and seems to lower heat/increase system responsiveness (sp). All you need to do is follow that guide I linked to, but be cautious of where I noted changes and youll be fine

If you already have a VMPlayer setup running, then you need to install over the old client and dlete your work, then once you edit the files and reload, youll start fresh with a 4 Core VM

Hope that helps. Ive got a similar write up I'm working on in my F@H paper, so this was fresh in my head :)
 
ugh... work......folding. .. ....hard. .. . .[H]. . . .. .work . . . . . lazy . . . . ..aviod...too...late...

I'd downloading the stuff, but SWMBO is using the quad core right now... to surf yahoo news...

Does a dual core Atom count (just for kicks)? (it has HT...)

good luck on your writeup!
 
Last edited:
nope, you can only run standard client on the Atom (Ive tried lol, im on mine rigt now)
 
nope, you can only run standard client on the Atom (Ive tried lol, im on mine rigt now)
That brings up a good point. How many people are running the standard single-core client? If anyone has free cores idling, it's time to take a look at the old-fashioned, long-in-the-tooth console client of yore. Yes, the good old standby can grant several hundred additional PPD per core running the new double gromacs WUs. My S-940 Opteron @2.8GHz is getting nearly 400PPD on these, with just a single core processing. Intel hardware will get quite a bit higher. So, if SMP is not possible for whatever reasons, I think these are a good supplement to the main meal. :cool:
 
so far so good. all 4 cores at 100%. you can actually edit all the VM settings thru a GUI in the player now.
 
OK, good to know, I'll change the VM section in my write up, we should let Cap know he needs to update his guide lol
 
So, if SMP is not possible for whatever reasons, I think these are a good supplement to the main meal. :cool:
Excellent point, let me go over to my parents house across the street right now and reinstall the single core client I used to run on it. I guess I will be back in business sooner with some points than I thought. :rolleyes:
 
That brings up a good point. How many people are running the standard single-core client? If anyone has free cores idling, it's time to take a look at the old-fashioned, long-in-the-tooth console client of yore. Yes, the good old standby can grant several hundred additional PPD per core running the new double gromacs WUs. My S-940 Opteron @2.8GHz is getting nearly 400PPD on these, with just a single core processing. Intel hardware will get quite a bit higher. So, if SMP is not possible for whatever reasons, I think these are a good supplement to the main meal. :cool:

So then averaged out let's say we can get 200PPD for each single core client. If every active folder (all 1014 of us) can add a single client, that's over 200K and we just countered fahman.

See what I just did there? :p
 
So then averaged out let's say we can get 200PPD for each single core client. If every active folder (all 1014 of us) can add a single client, that's over 200K and we just countered fahman.

See what I just did there? :p

-_-
 
My S-940 Opteron @2.8GHz is getting nearly 400PPD on these, with just a single core processing. Intel hardware will get quite a bit higher.
Not really. I ran two uniprocessor clients on a 1.8GHz C2D for a few days and I was getting about 200PPD per client. But every little bit still counts.
 
Doesn't the GPU client still need a bit of the CPU?

EDIT: nevermind, I just took a look at my task manager.

I forgot to check before posting...
 
Not really. I ran two uniprocessor clients on a 1.8GHz C2D for a few days and I was getting about 200PPD per client. But every little bit still counts.
That's strange, it should be higher unless it was because of the much lower clock speed compared to my Opterons. Where these the Double Gromacs or Amber WUs? The typical run-of-the-mill WUs have terrible PPD, but the latter mentioned two special types of WUs give good PPD for the standard client.
 
I just set up the client with default settings and the -advmethods flag. It was running whatever it got.
 
That's strange, it should be higher unless it was because of the much lower clock speed compared to my Opterons. Where these the Double Gromacs or Amber WUs? The typical run-of-the-mill WUs have terrible PPD, but the latter mentioned two special types of WUs give good PPD for the standard client.
Depending on core usage on the other cores, my i7 at 3.6GHz gets about 550ppd on big WUs. With normal or small WUs, it drops to about 220. So...yeah.
 
Strange behavoir...I switched over to a single vm across 4 cores last night and the vm screen confirm 4 cores found and 1 SMP across 4 cores. However, only 2 cores are showing 100% load the other 2 are bouncing around from 0-30%...I also have 3 GPU's running. Also Fahmon reports 6500ppd like this which seems way out of line. What's going on you think?
 
do you use WIN-AFC? If so, remember to change the setting in there, its still set up for 2 cores only
 
In the affinityinput.txt file, you should have a line which you put in, saying

*\vmware-vmx.exe := PAIR0+PAIR1 [assign=2, priority=idle]

If you have 8 cores, then you should also have PAIR3+PAIR4 there as well. The line basically means that vmware-vmx.exe has access to all the CPUs set under the PAIRs you've put in. The assign=n bit means that each instance gets n number of cores. So to get it to use all 4 cores, you want to change it to 4. Something like this:

*\vmware-vmx.exe := PAIR0+PAIR1+PAIR2+PAIR3 [assign=4,priority=Idle]

for my i7.
 
if you have a quad, it should look like this when youre done

*\vmware-vmx.exe := PAIR0+PAIR1 [assign=4, priority=idle]

for anyone that may have a Quad core only and is trying this
 
Thanks Vaulter and Pedantic - it works just like it should now. I am seeing 200-300ppd more on 1 VM than I was on 2 combined and there is definitely more free RAM. This was a great techie find - good job.
 
glad it helps, glad we could help

is your system a bit more responsive in normal tasks aswell? I'm Thinking of re-doing Capreppy's guide and then sending it back to him, or atleast letting him know whats up, and I want to know how its working for you
 
Thanks for the info Vaulter98c. I will try to update the guide in a couple of weeks when I am home again. I will definitely be giving this a try. I am running workstation on one of the machines and it works, but it is a lot of overhead for a dedicated machine.
 
Running Solid for a day and a half now on 4 cores. Showing just over 4K PPD with SMP Gromacs CVS Core and P2671 (R14, C98, G110) work unit. This is on my stock Q9950.
 
Sounds good. How is system responsiveness? I just want to clarify this for when Cap redoes the guide, because that is probly one of the largest gripes (and RAM usage) against multiple VM's
 
System is very responsive. No noticeable lag under normal usage.

I am also running the GTS250 on this same system, and getting over 6K PPD without running WinAFC. Just changed the Core priority within the settings of the GPU client.
 
is your system a bit more responsive in normal tasks aswell? I'm Thinking of re-doing Capreppy's guide and then sending it back to him, or atleast letting him know whats up, and I want to know how its working for you
Far more responsive. Since I'm only using about 70% of the RAM I was before, and I'm not running out, it's much better. No more random lags and stutters. And I'm getting about 25% more ppd now too.
 
Good news all around I see. Thanks for trying this out guys!!
 
OK, I loaded up the new version (2.5.3) and on the notfreds screen it says it only finds two CPUs

:confused:

What stupid thing did I forget?
 
did you edit the config (the one you open in wordpad per the guide) ?
 
Ohh I messed up a few things :eek:

Lets see how this works......
 
Anybody try this with a i7? I've got 2 VMs running using 4 cores each (2 logical/2hyperthreaded) using that little edit trick where you manually change the number of cores from 2 to 4 that was posted a couple of months ago. i7 is at 3.2 Ghz and I get around 3400 per VM right now. I'd be interested in seeing how it performed running 2 VMs set up like above. I'd be the first to give it a try, but I'm at work until mid November.


 
Anybody try this with a i7? I've got 2 VMs running using 4 cores each (2 logical/2hyperthreaded) using that little edit trick where you manually change the number of cores from 2 to 4 that was posted a couple of months ago. i7 is at 3.2 Ghz and I get around 3400 per VM right now. I'd be interested in seeing how it performed running 2 VMs set up like above. I'd be the first to give it a try, but I'm at work until mid November.


If you're already running 4-core VMs, this method would basically be doing the exact same thing. You wouldn't see a difference.
 
Anybody try this with a i7? I've got 2 VMs running using 4 cores each (2 logical/2hyperthreaded) using that little edit trick where you manually change the number of cores from 2 to 4 that was posted a couple of months ago. i7 is at 3.2 Ghz and I get around 3400 per VM right now. I'd be interested in seeing how it performed running 2 VMs set up like above. I'd be the first to give it a try, but I'm at work until mid November.
At 3.6GHz I get about 5000ppd for the VM not running with the GPU client, and 3800ppd for the one running with. But this shouldn't give you any benefits if you're already using 4 cores per VM.

OK, I loaded up the new version (2.5.3) and on the notfreds screen it says it only finds two CPUs



What stupid thing did I forget?
I thought it was version 3.0?
 
I thought it was version 3.0?
It is supposed to be :eek:

All fixed now

officerbarbrady.gif
 
Back
Top