Need advice to maximize my ppd

swoody

Limp Gawd
Joined
Dec 27, 2008
Messages
352
My system is not utterly impressive, but I want to be able to make as much of a contribution as possible when I'm not using my computer. I have an AMD 64 X2 4600+ @ 2.85GHz (as far as I can overclock for now), an MTI NX8600GT with 1GB of memory, all on a 64bit Vista install.

Now how should I set it up, to maximize my ppd? An SMP Client on each core, and the GPU2? A single SMP and the GPU2? Just the GPU2? Just the SMP? What settings should I select when I install the clients? A higher priority? 'Big' file sizes? Anything else??

I know this is a GREAT forum with a LOT of very intelligent and clever people, so I'm looking for some good ideas :) Also, if you could let me know an approximate ppd for the setup you suggest, that would be awesome! :D Thanks in advance, and I'm looking forward to these responses.
 
If your just starting out I might recomend two standard clients and the GPU client. Set the GPU to "slighty higher" so the GPU gets fed what it needs. As far as PPD. I don't know, I don't have any hardware running like yours, maybe 2-3k ppd.

If you plan an upgrade for gaming/folding a stronger vid card wil be your best bet.

Let the clients run 24/7, Just pause the GPU when you load up a game.
 
Thanks for the quick reply Kendrak! Now by standard clients you mean the regular ones, or the SMP clients?
 
If your just starting out I might recomend two standard clients and the GPU client.

Out of curiosity, if I wasn't just starting out, would you recommend something different?


Anyone else have any ideas??
 
Anyone else have any ideas??

A single SMP client might work okay if you make sure to set the GPU client's priority higher (so the SMP client doesn't take up all of the CPU power). I'm not sure what kind of PPD that would give you though.
 
The SMP client that will use both your cores is best when running on linux. (windows one has/had issues.) So this requires setting up a VM and getting the SMP up and running it inside. If you have the knowhow, go for it. However I am not totally sure that the AMD chips will run VMs (someone else know?) so the whole issue might be moot.
Some of the WUs for the single clients give good ppd. And the single client is easy to set up, and does not need to be started stopped to free up resources for programs.
VMs can get sticky if you don't have plenty of RAM.

I will say this is just me and others will have their own opinions. There just isn't A answer.
 
Well I happen to be dual-booting this machine with Vista/Ubuntu, but I had problems with the 64bit version of Ubuntu, so I installed the 32bit instead. Also, I do have 4GB of RAM. Ubuntu only recognizes 3.2GB of RAM, but I may be able to get all 4 recognized with PAE.

Is there a way to run the GPU client in Wine, or would it be better off to run the SMP in a VM?

Now I was wondering running the two Regular clients, should I tell each Regular client that I have 4094 B of RAM, or should I tell each one I have 2047 B so they never clash with each other over memory??
 
Well I happen to be dual-booting this machine with Vista/Ubuntu, but I had problems with the 64bit version of Ubuntu, so I installed the 32bit instead. Is there a way to run the GPU client in Wine, or would it be better off to run the SMP in a VM? Also, I do have 4GB of RAM.

Now I was wondering running the two Regular clients, should I tell each Regular client that I have 4094 B of RAM, or should I tell each one I have 2047 B??

Its possible, but its not trivial to run GPU2 in WINE. Theres a howto here, but I can't find it now.
 
Well I happen to be dual-booting this machine with Vista/Ubuntu, but I had problems with the 64bit version of Ubuntu, so I installed the 32bit instead. Also, I do have 4GB of RAM. Ubuntu only recognizes 3.2GB of RAM, but I may be able to get all 4 recognized with PAE.

Is there a way to run the GPU client in Wine, or would it be better off to run the SMP in a VM?

Now I was wondering running the two Regular clients, should I tell each Regular client that I have 4094 B of RAM, or should I tell each one I have 2047 B so they never clash with each other over memory??

yes you can run the GPU in wine and there a guide somewhere.

The biggest problem I forsee is deadlines expireing on the clients installed on the OS that is not used as often or for a while.

I whatever is your main OS that you use install the clients on it and leave the other OS alone. If it is linux, get SMP up and see if you can get the Gpu up in Wine. I'm not a Linux expert (or even novice) so if you go that drection I'll let someone pick up from here.
 
I don't think I'll be able to run the linux SMP client, as on the download page it says: "(64-bit required for SMP)" I could still run 2 regular linux clients, but again, with the fact that Ubuntu only picks up 3.2GB of RAM, and that it's 32bit, would that make it more beneficial to run a VM in Windows, and let my work-horse GPU client run natively?
 
I whatever is your main OS that you use install the clients on it and leave the other OS alone.

Well luckily this machine is my gaming rig, and I don't have the time to game. So it will basically be a folding machine unless my fiancee hops on it to use the internet. So whether I leave it running 24/7 in Windows or Ubuntu makes no difference to me :) My laptop is the rig I use the most (surfing the web laying in your bed is much more comfy than an old office chair :D)
 
would that make it more beneficial to run a VM in Windows, and let my work-horse GPU client run natively?

Well, it depends on how well your system handles VMs. If it can run a VM without any major performance impact, then it would probably be faster to do that and run an SMP client. On the other hand, if your PC isn't very good at running VMs, then it would be best to just run a pair of uniprocessor clients along with the GPU.
 
What OS do you have running most of the time? The clients need time to do their thing.
 
Well, it depends on how well your system handles VMs. If it can run a VM without any major performance impact, then it would probably be faster to do that and run an SMP client. On the other hand, if your PC isn't very good at running VMs, then it would be best to just run a pair of uniprocessor clients along with the GPU.

Ok, noob question: how can I tell if the VM isn't running well? Do I have to keep an eye on my system resources? Will I just be able to tell that it's not running efficiently? Some other tricks?

What OS do you have running most of the time? The clients need time to do their thing.

Well right now I have it running Windows 24/7 to run the 3 folding clients, but obviously if I set the clients up in Ubuntu, then I'd be leaving that running 24/7. Like I had mentioned, I really don't use this computer, so it will be - pretty much - a dedicated folding rig :)
 
Ok, noob question: how can I tell if the VM isn't running well? Do I have to keep an eye on my system resources? Will I just be able to tell that it's not running efficiently? Some other tricks?

If it doesn't run well, it'll lag heavily. It should be pretty noticeable :).
 
If you got windows up, try playing with the VM.

I don't run any, but I've given up a fair bit of ppd for a GPU only solution.

I however will have to do something soon to keep my top 10 PPD status. It's getting vicious up there :eek:
 
Thanks for the help guys :) I'll try setting up the vm, and playing around with it to get it working right. I'll update you guys here on my progress. In the meantime, if you guys have any other tips or advice, I'm all ears :)
 
Ok, I don't know what happened, but the two regular Windows clients I'm running are acting strange. I have their affinities set - one client per core, and their priority set through the Task Manager to "Normal" but whichever client I set to Core0 gets ~85% of the CPU and the one set to Core1 gets ~15%, rather than being at 50/50 like they used to.

Also for some reason something is using ~70% of my second core (CPU1), and won't go away. I've checked in Task Manager, and it's not showing anything using any CPU, and I checked the Services, and disabled the folding services so they don't auto start, but after a reboot it's still there and not showing anything in Task Manager. Any ideas what this could be??
 
If it not a virus/mal-ware then it could well be Vista running some background upkeep.
Look in your "Scheduled Tasks" that nothing set to start/run.

You dont need to set affinity for the two CPU clients.
It only makes the odd second per Work-Unit difference.

Luck ............ :D
 
Ok, I don't know what happened, but the two regular Windows clients I'm running are acting strange. I have their affinities set - one client per core, and their priority set through the Task Manager to "Normal" but whichever client I set to Core0 gets ~85% of the CPU and the one set to Core1 gets ~15%, rather than being at 50/50 like they used to.

Also for some reason something is using ~70% of my second core (CPU1), and won't go away. I've checked in Task Manager, and it's not showing anything using any CPU, and I checked the Services, and disabled the folding services so they don't auto start, but after a reboot it's still there and not showing anything in Task Manager. Any ideas what this could be??

For the affinities, you'll have to use something like Bill2's Process Manager to keep them locked. Each time a work unit is finished, the core is restarted and the affinity is reset. However, like Tigerbiten said, it really doesn't matter much with the uniprocessor client.

As for the process using 70% of your core, it's probably the system idle process, but if you want to check for sure, go into Task Manager and check the box at the bottom that says "Show processes from all users", which will allow you to see every running process, including those controlled by the system.
 
As for the process using 70% of your core, it's probably the system idle process, but if you want to check for sure, go into Task Manager and check the box at the bottom that says "Show processes from all users"

It has been set to 'show processes from all users' and it's still not showing anything in Task Manager. It does show the 'Idle process' in the TM, but the System Resources tab, as well as the CPU monitor on my Sidebar is still showing that something is using 70% of my second core (cpu1). **Update: now it's using 80-90% :(??

If it not a virus/mal-ware then it could well be Vista running some background upkeep.
Look in your "Scheduled Tasks" that nothing set to start/run.[/URL]

My computer has been running since last night, so I'm not sure what upkeep would take this long :(
 
Ok, so I've been letting the clients run 24/7 for some time now, but I'm not getting the ppd that I thought I would. I have 2 standard Windows clients running (on AMD X64 4600+ oc'd to 2.88GHz), the GPU2 client (8600GT), and a standard client running on my laptop as well (1.7 Celeron M). My total ppd for all the clients is only coming to about ~1300-1400. Is this accurate, or is something not right here??:confused:
 
Ok, so I've been letting the clients run 24/7 for some time now, but I'm not getting the ppd that I thought I would. I have 2 standard Windows clients running (on AMD X64 4600+ oc'd to 2.88GHz), the GPU2 client (8600GT), and a standard client running on my laptop as well (1.7 Celeron M). My total ppd for all the clients is only coming to about ~1300-1400. Is this accurate, or is something not right here??:confused:

I know the 8600GT has 32 shaders. Most of the NV cards that are purchased for folding have 96 (8800GS) or higher (for instance a new gtx260 has 216) It's the shaders that do the work. The less the card has. the less the ppd.

Your probably about right.
 
1500 ppd seems pretty good. I get about the same with a 4 core smp and a gpu client running. It seems to vary a lot. The smp client is barely any better than multiple single clients and in the end the gpu client blows everything away so keep that one running!
 
OP, if I were you, I'd try 1 Windows SMP and 1 GPU client (with affinity set slightly higher). Windows SMP has been really stable for me for a while now and it's pretty much a set it and forget it.
 
1500 ppd seems pretty good. I get about the same with a 4 core smp and a gpu client running. It seems to vary a lot. The smp client is barely any better than multiple single clients and in the end the gpu client blows everything away so keep that one running!

With the rig in your sig? If so, that's not right at all. You should be getting three times that with the 8800GT alone.
 
If this ppd is accurate, then what would be THE best ppd GPU on the market that I can fit into a single pci-e slot? I have plenty of pci slots open, but only one pci-e.
 
Well, if you have room for a dual-slot card (one actual slot, but takes up the space of the adjacent slot as well), then the 9800GX2 would be the best card at the moment since you can run one GPU client for each GPU. The next best would be the GTX280. If you're willing to wait a while though and budget isn't an issue, you could wait until the GTX295 comes out, which is basically two GTX260s in one card (albeit with as many shaders as two GTX280s, so even better).
 
The GTX295 is actually supposed to be available everywhere on Friday after it's shown/announced @ CES
 
Any 8800 should be getting better than 3500PPD regardless of the WU mix. Hell, that Phenom should be able to do better than 1500PPD on its own.
 
Yeah... that would be nice. Realworld here.

Try running the gpu client by itself and see what you get ppd wise. It's possible that the smp client is starving the gpu....
 
Any 8800 should be getting better than 3500PPD regardless of the WU mix. Hell, that Phenom should be able to do better than 1500PPD on its own.

Well I'm not selecting the work units like some people do. I just let it run and do it's thing.
 
Well I'm not selecting the work units like some people do. I just let it run and do it's thing.
Even so, you should still be putting out a lot more than you are currently. Try tkam's suggestion and see if your output increases, since it is very possible that your SMP client isn't giving enough CPU time to the GPU client. If that is the case, you'll have to raise the priority of the GPU client.
He said he has a 8600GT not a 8800GT BIG diff.
We're talking about velusip now, who does have an 8800GT :).
 
Well, if you have room for a dual-slot card (one actual slot, but takes up the space of the adjacent slot as well), then the 9800GX2 would be the best card at the moment since you can run one GPU client for each GPU. The next best would be the GTX280. If you're willing to wait a while though and budget isn't an issue, you could wait until the GTX295 comes out, which is basically two GTX260s in one card (albeit with as many shaders as two GTX280s, so even better).

Yeah, I definitely have plenty of room in my case. The only thing I'd be worried about is my 350w psu. Do you think that would be sufficient for a while until I can budget a new one?

Also, does anyone know about how much the new GTX295 is going to cost?

Also, just out of curiosity, with 2 GPU clients, should I still run my 2 standard clients on the CPU, or would I be taking away ppd from the GPU? Thanks for all the help and advice everybody :D
 
Yeah, I definitely have plenty of room in my case. The only thing I'd be worried about is my 350w psu. Do you think that would be sufficient for a while until I can budget a new one?

No. You *might* be able to run a 8800GT/9600GT type card on that if it's a high quality power supply, but even then it's pushing it. Otherwise, you're going to have to upgrade. Never skimp on the PSU, especially if you're going to fold.

If you do end up going with a 9800GX2 or GTX 295, I'd say you'll probably want to go with at least a 750W PSU.
 
Back
Top