Folding@Home - Join The Fight!

[Kaktus]

Originally Posted by Titan29

Just joined the team. Using V7.1.52 client. I do not intend to go 24/7 but will fold as much as I could.

Noticed that the GPUs are folding a lot slower than the CPU. Do I only use the CPU for now?

Welcome to the team.
Not sure about your question though, someone will chip in shortly i'm sure.

[iancook221188]

Originally Posted by Titan29

Just joined the team. Using V7.1.52 client. I do not intend to go 24/7 but will fold as much as I could.

Noticed that the GPUs are folding a lot slower than the CPU. Do I only use the CPU for now?

hello and welcome Titan29
gpu's and cpu's fold on completely different clients and therefore have different work unit's that fold and speed, complexity and get different points also most of the cpu's work get a bonus point for quick return time, hope that help

[Psychlone]

Hey Guys, thought I'd jump back in (after a couple years of hiatus) and help out again.

But, so much has changed that I'm not even sure where I should be. The original setup thread has fairly outdated info and is pretty convoluted now, making it even harder to understand.

As of right now, I'm running this:



Looks like it's jumped up a little since those snips, I'm currently lookint at 15100 for PPD and 30200 for estimated points per day.

Is this where I want to be with my rig? It's on 24/7 and serves as my daily driver, running my business during the day and doing nothing but pulling energy during the night.

I'm a *little* concerned about power consumption - not really put off by it, but I'll see what this next month's bill looks like. It's probably cheaper than heating this room during the winter anyway, right? LOL
I normally pull about 160W at idle, but with FAH running like in those screenshots, I'm puling 450W (exactly).
Here's some more specs for you:

(the system has been on for about 5 days straight without a reboot, which is why my memory usage is so high and why there's so many running processes...)
Also, it's a misnomer that my PSU voltages are reading that way - I've multimetered my OCZ GameXtream 1010W, and it's just fine... it's not read correctly by this new x79 board.

So, am I where I'm doing the best that I can?
I do, sometimes, get to a point where I have no choice but to pause FAH while I do some major work, but for the most part, I could theoretically leave it just like this for days on end. Could I do better?? Anything that YOU would change? And if so, what and how??


Thanks for any advice/input.
Psychlone

[iancook221188]

yer your not far off, two things though you mite try just folding on the cpu as the gpu can take up some of it resource and slow it down as the amd client does not give as much ppd as the nvidia client and it save a bit of power.

have you run HFM.net monitoring program before it will give you a more accurate ppd estimate http://code.google.com/p/hfm-net/

were you folding at the time of the screen shot of task mangier cpu usage should be at 100%

[k1net1cs]

Originally Posted by iancook221188

were you folding at the time of the screen shot of task mangier cpu usage should be at 100%

He was only using his GPUs for folding; take a closer look at his F@H GUI screenshots.

CMIIW, but isn't 15K PPD for a 5970 a bit low?
I mean, my 560 Ti can pull off around ~20K PPD.

[Psychlone]

I was folding at the time of the screenshot... it's using 98% of my 5970 Engineering Sample and ~20% of the CPU.



Uploaded with ImageShack.us

So would I be better off using *just* the CPU version? (SMP and no GPU?)
This is all new now - it's very little like it used to be back in the command-line days of folding, so I'm literally a newb at it now.

I have other machines - in fact, I was thinking about building another as a secondary server that I could stuff 2 MSI 8800 Ultras into and use it mostly for folding... just a thought. I've got 3 processors I could use, all are AMD. A 940BE, a 965BE and a 1090T. Seems a waste to use a 1090T in a system that isn't doing anything else, so I think I just might repatriate it to another build and sell it. But the 940BE and the 965BE can be utilized.

Given what you've seen and what I've said, please give suggestions as to what I could be doing better.
I would almost rather run something on my personal rig (signature) that uses less wattage than where I'm at right now, but whatever gives the most performance - I'm all about giving right now. Right now, while folding, I'm pulling 450W... normal idle is < 160W. This room is literally 4*F hotter than the rest of the house and it's almost not acceptable. Not to mention, when I DO get a chance to do some gaming (which isn't very often), I will have to pause folding and likely reboot the computer to get out from under it's cache and heavy memory usage.
I've got 16GB and after 6 days of constant use for my business and several programs - including F@H now, it's using up a hair over 9.5GB of memory.


Anyway, suggestions are definitely welcome. I want to be able to produce as much as I can with as little impact on my personal machine as possible. I'm not all about points, but points and performance literally go hand-in-hand and I understand the higher point value I get, the more work I've done. I just need to fine-tune what I've got (or what I'm about to do?) and get it working as efficiently as possible.

Looking forward to your suggestions.


Thanks!
Psychlone

[iancook221188]

i don't relay know what the amd 5970's are ment to do in ppd wise all i know is at the moment they wont produce as much as there nvidia counterparts and use more cpu cycles that them to. the 3820 should be same ppd or a bit more than a 2600k do you have it overclock? the cpu's at the moment get better ppd per watt and i would not of thought the system would use more that 250w

my 2600k at 4.5 get form 19-38k depending on work unit running at the time i think the system is pulling 180w with an 8600gt gpu (not folding tho)

the 940BE and 965BE would get you around 8-15k each but dependant on clock speed

im not sure why your system is using so much ram, when folding my system does not use more than 4GB if i run any other program it just goes up then back down to normal usage after, that could be something to do with the AMD client for the gpu's but ive never heard anyone have those problem with those symptoms before

nvidia gpu are the best gpu's to run but i think for heat that they will kick out it not worth folding on them they mite do about 4-6k ppd each for the 8800s

[Graek]

I just put a new rig together(yay me) and was looking to download the gpu client but all the link are telling me access fordidden. Anyone else have this problem? If so anyone know another download link?

[iancook221188]

there a new and simpler client called V7 witch will help you set up what you want i.e gpu or cpu client or bothhttps://fah-web.stanford.edu/file-re..._7.2.9_x86.exe

helpful guide to setting it up if you run into problemshttp://folding.stanford.edu/English/WinGuide

have you got a passkey it will be need soon for the gpu client as you do with the cpu client now

http://folding.stanford.edu/English/FAQ-passkey

hope that helps

[Athlonite]

Well after a few months of downtime I'm back into folding again with some new gear for a new year
1: Asus Crosshair IV Formula
2: Phenom II x4 965 @ 3.4GHz HT and NB at 2.4GHz
3: Corsair Vengence LP DDR3-1600MHz 9-9-9-24 (no more touching the HSF with the dominator's heat spreader)

so I'll keep going for as long as I can afford the Power to do so

and hells teeth what happened to the GPU WU's I don't remember them taking so long to fold I used to be able to do 4~5 a day on each GPU now it's 1 a day

[PantherX]

If it is a Fermi or newer GPU then it does take long. However if is is a an older one, then you can still do a couple of them in a day.

[Athlonite]

both clients are ATI HD5770's but I now know why they take so long to fold a single WU it's because the Core_16 does not use my GPU's to to their fullest instead of folding at 98~99% load they sit at around 39~42% so only doing half what they should be....

I'm as yet undecided as to whether or not I'll continue to fold ....

oh and do the environment variables make any difference anymore or are they now made redundant by V7.xx client

[PantherX]

If you think that using 39% - 42% of the GPUs might be costing you more electricity or isn't worth it, then my suggestion is that you can take a temporary break from folding on the AMD GPUs and save some money. You can always resume folding once the new FahCore will be released to the public after undergoing the beta testing so you can keep an eye on that thread to see how far we have progressed.

The environment variables are used by the FahCore_11 and aren't supported in FahCore_16 and I don't think will be even for the future FahCores.

[Athlonite]

any more news on how core_17 is coming along PantherX

[iancook221188]

have you had a look through FAH updates page post 195-197 http://forums.guru3d.com/showthread.php?t=322785&page=8

[yasamoka]

Originally Posted by Athlonite

both clients are ATI HD5770's but I now know why they take so long to fold a single WU it's because the Core_16 does not use my GPU's to to their fullest instead of folding at 98~99% load they sit at around 39~42% so only doing half what they should be....

I'm as yet undecided as to whether or not I'll continue to fold ....

oh and do the environment variables make any difference anymore or are they now made redundant by V7.xx client

Had that issue since 12.11 Betas. Existed since 12.10. Fixed by uninstalling the newest APP SDK that comes with your driver and installing version 2.8 (I think) or simply installing APP SDK from a driver prior to 12.10. That fixed the GPU usage for me, rose back to 94-99% or so.

[PantherX]

Originally Posted by Athlonite

any more news on how core_17 is coming along PantherX

Yep, iancook221188 was referring to this post in particular.

Moreover, we are in Beta Testing another project 7662 and you can view its details here.

A short summary (can change during the Beta testing) is that Kepler GPUs are not as powerful as their Femi GPUs counterpart. In some cases, Nvidia GPUs require a full core. AMD GPUs in the upper end of 7000/6000 Series are performing very well when compared to their Nvidia counterparts. Most of the time, AMD GPUs do not require a full core. In some cases, the GPU is being under-utilized and it seems that the Driver version plays an important role. If you want more details, you can check out the Google Spreadsheet.

[Athlonite]

yeah Catalyst 13.2beta7 sorted out some of the under utilization for me and F@H went from
less than 59% load upto 89% load on the GPU.. But core_17 is working like a charm at 99% load on the GPU and only a 1~2% CPU usage

HD7850 @ 950core and 1250mem is doing just a squidge under 16300ppd

[yasamoka]

Originally Posted by Athlonite

yeah Catalyst 13.2beta7 sorted out some of the under utilization for me and F@H went from
less than 59% load upto 89% load on the GPU.. But core_17 is working like a charm at 99% load on the GPU and only a 1~2% CPU usage

HD7850 @ 950core and 1250mem is doing just a squidge under 16300ppd

WHATTTT? Core_17? 16k ppd on 7850? Low CPU usage? FINALLY!

[PantherX]

Yep, FahCore_17 is a pretty significant step forward for GPU folding! Do note that it is still in Beta so things may change.

[yasamoka]

Originally Posted by PantherX

Yep, FahCore_17 is a pretty significant step forward for GPU folding! Do note that it is still in Beta so things may change.

Ummmm PantherX, seeing my 7970 almost break 40k ppd on stock clocks was enough to make me very, very, very happy. One main reason I got this card, for folding@home potential.
OMG back to watching my baby fold! Cya!

[PantherX]

That's great news yasamoka Fold on

[Mannerheim]

im back.. with better mobo. no more multiplier 7 bull**** cause vrm phase overheating.

[iancook221188]

good to have you back Mannerheim what mobo did you have before

[Mannerheim]

M5A78L-LE and its with sh.itty 4+1 vrm phase.

I love my "new" PC =) stock 3913MHz and 50*c

https://www.dropbox.com/s/corbux1t2b9fu6k/uuskone.JPG

https://www.dropbox.com/s/ihh3wt464b8trgf/uuskone2.JPG