Core 2 Quad and up to 3 NVIDIA 8800 GTXs

[Sciurus Winterwolf]

Core 2 Quad and up to 3 NVIDIA 8800 GTXs

I'm Building a dedicated machine for folding.

Using the above, what would an optimum set up be. 8Gig of memory available

[Zagen30]

I'd recommend trying out the v7 client with 1 SMP slot and 3 GPU slots (if you do put all 3 in there, otherwise 1 slot per GPU). Running advanced WUs is up to you; if you're not familiar, setting "client-type: advanced" tells the servers that you're willing to accept more beta-ish WUs that may be less stable than those which have been cleared for release to everyone. These projects aren't usually too buggy (they do go through an internal beta and a closed beta before being put on advanced), but they don't net you better PPD, so it basically comes down to whether you want to run potentially less stable WUs. Most other options aren't really all that necessary.

Do be sure to get and use a passkey. While the science done is the same whether or not you have a passkey, you get additional points for returning WUs before the preferred deadline. Currently only SMP units (and uniprocessor units, but since you probably won't be running them it's a moot point) get a Quick Return Bonus, but GPU units may be getting them sometime in the near future as well. Note it'll take 10 completed SMP Wus before you start getting the bonus points.

[Sciurus Winterwolf]

Ok, thanks for that. I'm using V7. I think I have a grasp of the basics.

Does the amount of memory make a big difference to the outcome? I'm searching the site but have not found anything about this. It's prob there but just have not found it.

One of the boards I have (abit in9 32x max) takes, wait for it, 32 Gb max. It also takes 3-SLI, but I believe SLIing doesn't have any effect on folding

[rs199208]

I am new also but this is what I have found out so far.
Task manager on my windows 7 64 running Linux through vmware player @ 100% load on all for cores (amd @3.9) with a ton of chrome tabs open is only at 1.58gb usage out of 4094 total. it says I have 2466 available 1393 free.
You don't need more then 4gb for that CPU since you are not doing Big advanced.
More memory depends on what else you do. Faster memory is better.
With big advanced and more cores or multiple CPUs the need for more ram goes up.
I am building an I7 2600k soon and would be fine with 6gb but for the price might as well get 8gb.
To run big advance you should have more then 4gb as far as I know.
You don't run the video cards in sli and the memory from each card will be used.

I seen you are new to RatSkep did you know there is a folding thread there also?
Lot of info on that also http://www.rationalskepticism.org/gener ... -1300.html

[Zagen30]

SLI doesn't make a difference one way or the other. Each card's memory will be used even if they're in SLI.

rs199208's mention of bigadv doesn't apply to you. It's only available to machines that have 8 or more cores, which your C2Q does not.

He's right that for regular SMP you don't need that much memory. The a3 core (which most regular SMP WUs run on) uses about 110 MB of memory- it won't use more even if you have it. It is true that faster memory can speed the process up, but on regular SMP the effects are not very noticeable, so if you were thinking of buying new memory I wouldn't bother.

[kiore]

Hi Sciurus Winterwolf, my first folding rig was a quad and 4 x 9800GTs not too dissimilar to your proposed one.
Currently running a SMP, 4 x GTX 260s writing this and task manager indicates I am using 1.02GB of memory in winXP.
So RAM not such a big deal, what will be a big deal is heat, multi gpus is tricky to run in a case, I run mine in the open.
Suggest you try 2 gpus and see how you go.
Foldon, oh and welcome to the team, nice to see some team members over here

[Sciurus Winterwolf]

No, I won't be upgrading the RAM, I'll go with what I have. I'll probably go with my abit board in9-32x-max-wi-fi and use two of the 8800s. I have a Cooler Master HAF so the heat should be ok. I'll probably only be running a browser as well as folding.