testmy.net Folding@Home team

[TriRan]

i went back to using GPU3 and SMP instead of beta, i can actuallly max my gpu's core this way and the cpu stays about the same its about 1-2k pts diff each day which is a lot

[lorne]

Yeah the 17th was a big day! That card really pumps em out. I haven't had a chance to overclock it yet so looking forward to seeing how that goes. I'll do it slowly and make sure it's stable, I will be adding a another rig pretty quick. It's actually a laptop but everything helps Also a Alienware (laptop) being added pretty quick. I have being trying to post stats updates but have been very busy the last week or so, so sorry for not being consistent there.

Crunch on!

[lorne]

Team is still picking up steam with you guys adding more power great work! We are starting to climb back up the rankings at a pretty good speed now. One bonus for me is all my computers are in the basement which can get a bit chilly in the winter. With FAH going 24/7 my feet stay nice and toasty. I don't even have the heaters turned on

Dec. 19th stats

[mudmanc4]

Trying a few things with this server, I'll let these five threads run out and see if there valuable, if not I'll direct them through the CPU by thread/CPU

There is a dual core running just two threads same OS different server obviously, tomorrow I'll see how many threads the CPU can take .

[TriRan]

Team is still picking up steam with you guys adding more power great work! We are starting to climb back up the rankings at a pretty good speed now. One bonus for me is all my computers are in the basement which can get a bit chilly in the winter. With FAH going 24/7 my feet stay nice and toasty. I don't even have the heaters turned on

Dec. 19th stats

lol i know! i've been leaving my window open 24/7 and my computer is generating enough heat to keep me warm while i'm on it!

btw am i not getting early turn in bonuses? i'm getting roughly 500 pts per WU my video card chews out and it doesn't appear to be offering any bonuses of any sort

[Blako]

Mud: I found a quadcore client to out perform 4 single core clients. Hyperthreaded cpus are not recommended unless there are more then 2 physical cores, else double the work gets done in half the time and science slows waiting for your results. I would recommend against running 5 threads on a 4 thread machine as collisions of "is cpu busy? Yes" would happen all the time, unless your already taking this into account. In fact on the hyperthreaded 4 physical 8 logical core i7s, seven cores at full load out perform eight cores at full load.

TriRan: Only the smp client with its large memory usage and high system requirements has an early return bonus. The idea of "early bonus" came from bigadv units that required 20-100 megabyte uploads verses 1mb results file from your radeon. Points FAQ said:

Our benchmarking philosophy tries to balance two elements: keeping a system reasonably simple (both for donors and for the FAH team to calculate) as well as keeping points in alignment with scientific value of a given calculation. Indeed, donors will optimize their machines (eg choice of hardware, choice of clients, etc) based on points, so it is important that points awarded be reflective of the scientific gain

Vijay Pande said, donors really pick up in the winter months while folding machines are turned off in the hot summers.

[mudmanc4]

If you can show me another core that will run on two dual core Xeon 32 bit procs running centos that will outperform what I have been testing , I'm open for suggestions

on a hilarious note , I installed folding on a VPS with unlimited resources per se' , the server load went up to 10.0 and 2700% of the allotted gflops within 10 seconds, lmao Had to kill it.

[TriRan]

Mud: I found a quadcore client to out perform 4 single core clients. Hyperthreaded cpus are not recommended unless there are more then 2 physical cores, else double the work gets done in half the time and science slows waiting for your results. I would recommend against running 5 threads on a 4 thread machine as collisions of "is cpu busy? Yes" would happen all the time, unless your already taking this into account. In fact on the hyperthreaded 4 physical 8 logical core i7s, seven cores at full load out perform eight cores at full load.

TriRan: Only the smp client with its large memory usage and high system requirements has an early return bonus. The idea of "early bonus" came from bigadv units that required 20-100 megabyte uploads verses 1mb results file from your radeon. Points FAQ said:

Vijay Pande said, donors really pick up in the winter months while folding machines are turned off in the hot summers.

oh okay! are there any special -args i need to use to get the bonus points then?

[mudmanc4]

run -configonly and set the WU size to " [big] " Vs small or normal

[TriRan]

won't those just take longer to complete and not neccessarily offer any more ppd?

btw i just added 2 boxes to fold, a Phenom II 945 X4 at 3.4GHz and a Phenom II 1090T BE X6 at 3.7GHz ontop of my X4 at 4.1GHz ^^

[lorne]

won't those just take longer to complete and not neccessarily offer any more ppd?

btw i just added 2 boxes to fold, a Phenom II 945 X4 at 3.4GHz and a Phenom II 1090T BE X6 at 3.7GHz ontop of my X4 at 4.1GHz ^^

Awesome! Well I enjoyed a few days of the top producer guess its time to hand over the keys to the kingdom !!

[Blako]

Mud, would Linux 64-bit SMP2 client work for you? Client Choices

Only smp clients qualify for early bonus. If your running phenom with "big jobs" allowed you might get a unit that takes twice as long but is worth more then twice the points as the increased memory/hard disk/network usage is proportional to an increased value.

Version 7 client is settling down and chugging away. I think I installed "for all users" which stored the data in a different file folder instead of under my windows documents. Sigh, its still reporting a very exaggerated 100,000 ppd until it has more progress to take the average of.

Dang Lorne nice job, my sources tell me your running 4 clients. Whats under your hood?

[lorne]

Well I tried 7 on my main rig and didn't really like it :/ Lack of features and options on install and it did not want to utilize my card at all, detected it but would not actually use it. Maybe I just been using the console version for so long 7 seems really weird to me. Think I'll give it a bit more time before I try again.

Under the hood is my main computer which is a phenom quad (only 1.8ghz) with that fermi card. Than there is the Dell Power Edge 1u server (2x dual core opterons at 2ghz i believe). That server only runs during the day as the girlfriend doesn't like the noise at night. Than there is a lappy not even sure what it has besides being a dual core.

I have been planning a build for feb-march and had originally planned to go sandybridge and water for OCing but changed my mind and going to go with a FAH twist that will ensure I can fold the big adv WUs post february when the requirements change to a much more commercial style. Once I have it solid I'll post but from other people I have talked to running this setup I expect aibout 280k-300k ppd. That will give us a heck of a boost and it turns out actually being a bit cheaper than the 2k I originally planned on spending

[mudmanc4]

Hey lorne, I also have one of those old dell 1650 power edge 1U's w/ dual 1.6 or 1.8 Id'e have to look.

What core do you run and how is it performing. I can install most any OS other then 7 to make sure I get the best core for the hardware.

And you think that little thing is noisy , you should hear this HP , honestly my house is an 1800 sqft ranch , and at the other end of the house you can make out the sound of the 12 or 16 fan beast hahaha

[nanobot]

Where do I sign up for this Folding stuff? I may as well try it.

Thanks,

EBrown

[mudmanc4]

Download the core for your hardware at http://folding.stanford.edu/English/Download

You will be shown the best candidates for your hardware on the download page. There are also beta and high performance client http://folding.stanford.edu/English/DownloadWinOther

I'm sure we'll all help with anything we can , glad you decided to jump into folding. Society thanks you for doing a little to assist in finding answers to medical cures of all kinds

[nanobot]

Download the core for your hardware at http://folding.stanf...nglish/Download

You will be shown the best candidates for your hardware on the download page. There are also beta and high performance client http://folding.stanf...ownloadWinOther

I'm sure we'll all help with anything we can , glad you decided to jump into folding. Society thanks you for doing a little to assist in finding answers to medical cures of all kinds

I downloaded the console one. (Considering I have enough problematic windows.) I might do the High Performance one later. (Too bad they don't have one to do RAM operations, I have 16GB of that.)

I guess I'll find out how well I'm doing after I complete this first work unit.

I only decided to start this since I ordered two Dell PowerEdge 2850's the other day. When I get them I will be dumping some Linux OS on and using them for this. (May as well help since they have no other purpose. Lol)

Thanks,

EBrown

[mudmanc4]

Those are good units. I would say they compare with the HP DL380's here. Although 32 bit you can do a lot with them. Hell I have one running virtualmin w/ 4 servers and a couple sites per server no problems.

You would think sooner then later they'll be getting a ram client going. I havent kept up w/ the latest in more then a year or so. Maybe there working on it ?

[nanobot]

Those are good units. I would say they compare with the HP DL380's here. Although 32 bit you can do a lot with them. Hell I have one running virtualmin w/ 4 servers and a couple sites per server no problems.

You would think sooner then later they'll be getting a ram client going. I havent kept up w/ the latest in more then a year or so. Maybe there working on it ?

Yeah, I got them at $100/pop with Dual Xeon 3.0 or 3.2, I can't remember.

They'll never make a RAM client. Lol RAM only stores data, it doesn't do any math.

Thanks,

EBrown

[mudmanc4]

Right at the same time it could be useful in holding larger equations while the cpu works out the rest. Dunno , just a thought

And thats about the going rate right now for those machines. You did good ! Any drives w/ them ? Or at the least a few each ? The sleds ?

[nanobot]

Right at the same time it could be useful in holding larger equations while the cpu works out the rest. Dunno , just a thought

And thats about the going rate right now for those machines. You did good ! Any drives w/ them ? Or at the least a few each ? The sleds ?

They each are Dual Xeon 3.2Ghz, 4GB RAM, 4x146GB SCSI, DVD, DRAC/PERC, Dual PSU. No bezels.

Looks like it's going to take two days to complete this Work Unit.

Thanks,

EBrown

[lorne]

Awesome welcome to the team EBrown! Those 2 2850 should crunch up WUs pretty good. Look forward to seeing you come up on the stats

[nanobot]

Awesome welcome to the team EBrown! Those 2 2850 should crunch up WUs pretty good. Look forward to seeing you come up on the stats

I should be when my GPU finishes this first one. I decided to run the GPU3 Client for my card. Working pretty well so far, by my math it will only take 1 hour and 40 minutes total for this job, and it's already at 50%. (I've been getting each percent in about 58-60 seconds. Lol)

I'm going to run an SMP client on each of those 2850's when I get them.

Thanks,

EBrown

[Blako]

Cheers EBrown, unless you got a small job I would estimate you got a 400 or 500 series gtx. Some of the big (multi-core) smp jobs require half a GB of memory per core. So unless you got a multi-socketed board your not going to use 16GB ram

Mmm, used server market, interesting. I guess there is some folding potential in a site that caters to computer geeks users. Well I'm glad mud was able to answer your questions.

[nanobot]

Cheers EBrown, unless you got a small job I would estimate you got a 400 or 500 series gtx. Some of the big (multi-core) smp jobs require half a GB of memory per core. So unless you got a multi-socketed board your not going to use 16GB ram

Mmm, used server market, interesting. I guess there is some folding potential in a site that caters to computer geeks users. Well I'm glad mud was able to answer your questions.

I've been getting small jobs apparently then. Because I have a 9800GT. (Albeit overclocked by 100Mhz Core, and like 150 or 200 Memory.)

I got the same project twice in a row, lol.

[17:45:33] Project: 5766 (Run 1, Clone 287, Gen 824)

[19:36:28] Project: 5766 (Run 8, Clone 97, Gen 2724)

Both are running extremely quickly.

Do you have any tips for optimizing my SMP process when I get those running? I would like to get them pumping folds as quickly as possible.

I have been using this GPU Folding to overclock it, using the actual folding process in the median. (Giving it 45-60 minutes between clock modifications.)

Thanks,

EBrown