INTRO :

Currently Joining Folding@Home and find that it is a usefull way to spend IDLE processes at home and as well stress your component for humanity and by the mean time check your stability of your RIG other than using Pi or PRIME that only spending time and precesses for nothing

Folding@Home website
http://folding.stanford.edu/

QUESTION :

Computers is actually Calculators that can do more things that just calculations ... so i need a really good idea of building a good calculator to calculate workable Proteins

i find that ... the calculation and rendering of a working unit is far more HARD that just unlimit numbers of Pi ...

I use the SCHOOL computer (2.8GHz P4 ) and it need >6 hours to get 10% of a calculation done .... argh ... using 1M Pi ... it need only 55Sec .... see the difference ?

So slow the comp yet so complicated the calculation ...

So WAHTs MAKEs a BEST CALCULATOR ?

Currently there is war on Bus speed , multiplier , pipeline and Actual GHz ... So which is the most important component for a calculation ? The transfer of data ? The Brain works ?...

Celeron had High Multiplier , P4 had long Pipeline , Conroe and AMD 64 had fast Bus speed .....

Which of the breed is the Calculation master ?

Opetron Wins Xeon in most of the data transfer benchmarks ...

OCed Computer can reach 9Sec in 1M Pi calculations ...

Ah .. still need all the OCkers ideas ...

Current i think of OC my p4630 to 5.4GHz by all means to beat the 9 sec 1M Pi record or should i use PD 950 to OC to 4.8GHz with DDR2 1K RAMS ... Or just get Xeon or Opetron server and get the Calculation done ...

IDEAs OCkers ? thanks

Any HARD CORE OVERCLOCKERs Know How to solve This Problem
Just want my Computer to Calculate not for Gamming

Ah ... One more thing ... How to get shortest time in 1M Pi ?
Should i get those Processor with fast Bus speed or Fast RAM speed ?

u sir, need a tweaked conroe
currently, conroe have proven that its architecture can achieve fast data processing with the combination of high bus speed n big cache even with low multi's, compared to previous architecture of both its predecessor n amd's
n to achieve overall best combination, ram speed n latency need to be taken into account too

This post has been edited by e-jump: Jul 24 2006, 10:19 AM

Folding@Home is totally depends on raw clockspeed, the more Ghz, the faster it folds. SSE and SSE2 also affected the folding time, but I think now most of the CPU has both. But to some point base on experience, arj's 2.7Ghz Opty's beat my P4 3.8Ghz with the same protein!

Relatively, it's hard to benchmark with a total different specs, plus u gotta fold the same protein, class gen etc etc. And again, to some point, Tinker folds faster on AMD, because AMD has a better FPU than Intels. Whereas, QMD only folds on Intels due to compiler restriction, and more point given for the massive RAM usage when folding a QMD.

Well, I don't really know the requirement needed to setup a hardcore Folding@Home crunching machine, and I'm sort of noob with these technical specs that helps it to calculate faster (multiplier, fsb and etc). Anyway, nowadays, I'm more into low power consumptions machine as I have many rigs at home, and it's now becoming a pain in the arse just to pay the bills. And for that, I'm eyeing for a Yonah or Conroe (Preschot is a electricity hogging beast). Why don't you head to folding forum (at folding-community.org), they discuss alot about computing power and stuffs. And I'm sure some sifus are able to answer you.

P/s: Perhaps, should move this to Folding@Malaysia forum, so that we can create a database or something. Or share some thoughts on building the most leet folding beast. By the way, I think in HenryLow's thread there's a database made by him, with a few CPUs (Xeons, P4, Dual-P3 etc etc). Just check that out!

This post has been edited by c9: Jul 24 2006, 02:28 PM

Opty can beat P4 as easy as < whatever you think is easy >
Because of the HTT and high FPU

Haha .... Prescott ? not my concern either when comes to Electric Bills
I use the school's Hostel Electricity for free and as long as they don't know My "extreme Prescott" is running 24/7 ... then .... use the previlage to the max

Anyway , i also a noob in Folding ... So thanks for the weblinks
Just second day of joining FOLDING and More protein to go .....
Just want to make sure that any OCkers here had better Idea of setting a system for folding ... Of course .. it is such a waste to create FOLD only RM6K machine .. just sometimes it might help in my Graphic rendering Project ...

More ideas from diffrent source and be compare easier ....
Just that OCed Computer is cheaper , thats why i am here

Ah ... those who had any more idea .. can still share some juice here
Thanks ...

Ah .. PROVE them MALAYSIA BOLEH !!! We need More LOCAL OCkers

This post has been edited by RIGmaster: Jul 24 2006, 03:32 PM

Well, I have like 5PCs currently at home. Every single PC folding 24/7, 2X Preschot, 1xAMD, 1xNorthwood and 1xAMDX2.

And soon to come, a Yonah or a Conroe system.

Is that what we call a waste? I have 2 more P3s collecting dust, and used to run 8PCs before. Haha!

This post has been edited by c9: Jul 24 2006, 06:34 PM

1st thing is the mobo design...
a good mobo makes ur pc works well...

start from clean power provide to hardware.
good signal, nice layout for cooling.

many things affect it.
so, it determine how many effort u try to improve ur pc.

Ah ... i planing to get a System Running on Intel

Mobo : asus P5WD2 E Primium with i975x chipset
RAM : DDR2 1066 2X1GB
Graphic card : 2 Piece of ATI Radeon 1900XTX in Cross Fire Mode
Processor : Pentium D 950
Sound Card : Sound Blaster X-FI Platinium
Casing Fans : 4 pieces of 3000RPM 12Cm Aluminium Fans
PSU : Cooler Master 550W SLI
Casing : Cooler Master Stacker
HDD : 10,000 RPM Western Digital RAPTOR

Plan On OC ...
PD 950 OC at 4.8GHz or >5GHz

Ah ... Is this System Qualify to do extreme Calculations and Fold without error ?

Not planing to get Xeon or Opteron ... because they had slow ECC Rams .. DDR2 1066 or at least DDR2 800 should be enough

Any Comments ?

Argh .. this RIG cost > RM7K ....

Unless u really into gaming, i dont see why u put a PD 950 with SLI of 1900XTX. If me, i will pick X6800 with low end GPU bcos i hardly play games but X6800 is a great cruncher. BTW, how big is your monitor?

BTW, most opteron in F@H should be the S939 version. It use normal DDR.

My monitor ... ah wide screen 19 inch ... Not planing for 21 inch ... Too big ...

I use the Cross Fire Mode because i saw there in the F@Q and said "does Graphic acceleration hardware affect the Folding" it answer " yes ... it helps allot for the protein to fold " thats why i go for X1900XTX ... but still need more information from you all ... before i spend money for nothing ...

DDR is too slow i think ... i saw DDR2 1066 around in LYN and so i would like to have some of it ... I see RAM speed helps in super Pi calculation ... so i think it help in Folding also ... not too sure .... Since AMD machine still not running DDR2 1066 ... i kinda let go ...

Ah .. thats why i am here .. asking for ideas from fellow OCkers
Which system is the best ... Aim for what folding needed and strike for it
Planing to create a system that can create 100 WU per day ... can i ?

Where did u read that? AFAIK, it slightly affect ur output only if u r running the graphical client. But if u want better performance, you should run console client....

F@H is purely CPU intensive. I dont think ram speed will give more advantage.

What he mean was, most probably higher ram speeds gives extra oomphhs. And about the does Graphic acceleration hardware affect the Folding probably he's saying that running the folding calculation on the gpu itself.

GPU have more Pipelines then CPU
X1900XTX had 16 Pipelines ... 48 Pixel Shadder ... ah ... if run in X1900XTX CROSS FIRE mode ... should help allot
Imagine 16+16 pipeline and 48+48 Pixel shader ...
Imagine >30GB transfer rate and a 512BIT Bus RING ... if can use the X1900XTX for full Folding capability and calculation ..... ah ...
CPU ... just Goyang Kaki saja ... Celeron D enough can ka ...
As i folding ... the Rendered Frame per second kinda slow ...
Even My HD DVD rendering is even faster than that because of AVIVO


Anyone Upgrade the Code for the console
Graphic Console in X1900XTX like running nothing .. no lag ... Text console boring ... Just need to comfirm the Help of OC X1900XTX in the overall performance on folding ... if can .. i upgrade another X1900XTX to Cross Fire and OC it ...

if can use a ATI Physix card to calculate ... haha .... even better ...

Currently not many Support Physix Card ... ehrm .... the Graphic console Kinda OLD in the FOLDING@HOME website .... hope they come out with 64BIT computing soon .... and also the AVIVO support that can take the load from the CPU

Ah .. and also the full SSE4 & EM64T support ... The Console kinda backward ... still using SEE2 ....

Ah .. still confuse with the calculation Style .. i currently study the ALU unit in the processor ... looks like all calculation takes LOOPs to run ...
Instruction can be give in PARALEL MODE if your Processor had more PIPELINE ... So i had a Question ... which Processor Nowadays Had Most Pipelines

Is the Calculation in the Folding is given as one BIG instruction and long chain of Processes ... or Many Small Instruction with pulses of Processes ... Kinda confuse here .... And the Most important ... for each calculation ... how many MB of Memory must be acessed

If MB of Memory acesses per instruction is >2MB ... i think i better get those 4MB L2 Cache Processor .. ah very expensive .. but fast can up to 50GB to 100GB/s Peak ... ... if the Instruction is <1MB ... then a DDR2 533 is enough . ... and if the instruction is >>10MB per Pulse .. then i need to go for DDR2 1066 RAMs to boost the Calculation .... Any minnor Delay in calculation will delay the whole thing Thats why .. i ask ... is Memory helps in my calculation of folding

OCkers outhere ... DDR2 1066 good or not

Ah ... Anyway .. thanks for the info given expecially uzairi
Kinda talk allot in PM

Ah .... keep the Folding Spirit on ..

WHATs MAKEs a BEST CALCULATOR !!!!

Well, GPU doesn't affect much on Folding@Home, unless client for GPU is out. But then, how it gonna perform (though, some has tested the BETA client, and it's like triple faster than current P4. And that was on a X800 or sumting if I'm not mistaken) or is it able to run in Xfire, the stability, we are not really sure..

Still, they're in the process of making a GPU folding client. The developer team at Stanford only consider what really IN and decide based on majority, when the demand is there, definitely there will a client for it. And I'd suggest you, get a Conroe and single XTX is enough already. All and all, it is still CPU that really makes the difference..

For me, I'd consider what the best for Folding@Home, my daily use and of course budget wise too.

This post has been edited by c9: Jul 27 2006, 03:45 PM

Did HenryLow owns a suppercomputer ?
haha ... the graph looks very steep

Anyway how to download the extreme version and the latest Version of those client ?

Since Old method may only overstress the CPU while NEW method spread the Load effectively ....

Ah ... BETA client .. looking for it but still cant find ...
only 5.03 version ...
is BETA client X64 Bit supported

One OCed XTX and One PHYSIX ... Plus OCed Conroe ... is this fast enough ?
Thanks


Ah .. Conroe can't OC to 5GHz sadly .... because if it is so ... your Bus will be extremely high and your Northbridge will be Overloaded

I think its better to use "No nonsense" text-only console because it wont interfere with the gc at first place. Let the proc itself do the calculation, so its better not to use the graphical client. The latest Text only console were 5.04 beta and i think they did improve something.

RECENT CHANGES (Client):

5.04 (Console only):
- Allows configuration of the amount of memory to indicate available,
especially useful when running multiple "big unit" clients on a single
machine.
- Allows for the setting of -advmethods as a persistent option
through the Configuration.

There will be more optimizations and enhancement made to the client to make use the Ghz clock efficiently. Just fold and make something usefull to the community.

You need to sort of register with Stanford to be a BETA tester.

x64 client? I don't there's one now. As far as I'm concerned, client released was based on their benchmark rig, which is P4 2.8Ghz. And I think they haven't update their CPU ID just yet, where PM Dothan is detected as P3 and AMD64 as normal K7. But then, in the near future there will be a 64bit client to go with Vista and stuffs and as everything is now moving forward to 64Bit.

The latest client released for general use is the 5.04. In addition to what uzairi has pointed out, with CPU the only one stressing and do the calculation, you'll have a lower power consumption where you GPU will be idle most of the time, thus lower power consumption.

Fast enough? For games YES, but not for F@H because there's no client that can utilize F@H for that hardwares at the moment. So I don't see why that kind of hardware would make your folding folds faster. Let's just wait for the client, or perhaps you can make a proposal to the developer team at Stanford, I'm sure they'll look into that. And again, they need demand, if the demand is low there's no point on building a client for that particular hardware.

YES Conroe can't easily get to 5Ghz, but have you seen how fast it folds? Well, it have been a long time since I checked the folding community forum, I think I should now. Anyway, I think Conroe folds real fast. Oh and I'm in the process of getting a Conroe, that should be around this weekend or next week.

This post has been edited by c9: Jul 27 2006, 08:58 PM

Maybe all the X1800 and X1900 series user should make a proposal to the Stanford University ? By that way we can speed up the client/console RnD process.

I think they're already on the final stage of the client, let's just wait for the release..

Can't wait to run it on my R520.