Folding@Home

[Clocker]

Clocker, it was just a typo. A cocker is a nice little dog, so you shouldn't be obfuscated.

No problem. I stopped taking anything you type seriously a long time ago. It's been so long that I forgot who you were for a moment there.

C

 

[Gilbo]

Yeah but I only got 1GB of memory.
I'm going to upgrade it hopefully soon to 2GB but I think this means I'll be using all four memory slots (4x512MB). I'm a little leary about doing that.

As I understand it, although you need a solid amount of memory (512MB for any type of large WU), it's not the quantity that matters so much as how much you can pump through the pipe between the processor and the RAM per unit time. I wouldn't be surprised if there's no speed difference between 1GB & 2GB of RAM for this task.

 

[Handruin]

Thanks Mark! I am a happy man now!

 Folding@Home QMD Core
 Version 1.04 (Apr 7, 2005)

 Preparing to commence simulation
 - Looking at optimizations...
 - Created dyn
 - Files status OK
 - Expanded 260892 -> 826209 (decompressed 316.6 percent)

 Project: 1912 (Run 51, Clone 4, Gen 68)

 Writing local files
 Entering QMD...

Now onto part two...

 

[Handruin]

Woot!

 *------------------------------*
 Folding@Home QMD Core
 Version 1.04 (Apr 7, 2005)

 Preparing to commence simulation
 - Looking at optimizations...
 - Files status OK
 - Expanded 260892 -> 826209 (decompressed 316.6 perce

 Project: 1912 (Run 51, Clone 4, Gen 68)

 Writing local files
 Extra SSE2 boost OK.
 Entering QMD...

SSE2 boost is on.

 

[Handruin]

I'm working on a script that will check if the FahCore_96.exe core has changed based on SHA file checksum values. I'll see if I can work in automatic SSE2 optimization if it has changed.

 

[Gilbo]

Well, I think we can safely assume that QMDs on A64s are not a waste of time Will's last couple days haven't been so bad.

I suppose I'll have to get around to trying this out myself at some point...

 

[Will Rickards]

They seem to be a little less than 2 days per WU.
So about with each at 450 points, I'll probably average 500 or 450 points a day and my graph will look odd with two work units submitted one day and none the next and two the day after that.

 

[CougTek]

Unfortunately, that's only half the speed of a Pentium 4. I think a P4 3.0GHz can finish a single QMD in a little less than a day. Mine is clocked at 3GHz/650MHz fsb and it does one QMD per ~26 hours if I don't use my computer too much. If my fsb was up to 800MHz, then I'm almost sure I could finish a QMD in less than 24 hours. And I have no hyper-threading.

So an Athlon 64 X2, even with QMD on, is still slower than a Pentium 4 for folding :-(

 

[Will Rickards]

Let's see your comparing a single core processor at 3Ghz to a dual core processor at 2Ghz. Mine completes 2 work units in less than 2 days. So divide by 2 = 1 work unit in less than one day. Not sure how this is slower? Slower per core maybe. And we are probably talking about a f@H core optimized for those intel chips.

 

[P5-133XL]

Hey folks, you really need to be giving it some time ( A week) to get a reasonable estimate. You can do better with a log file, but just looking at returned results, one needs more time to remove the variability of the start point/end point, machine usage, and stuff like that.

 

[Clocker]

My computer does a little over one QMD per day.

C

 

[Clocker]

....and I have been playing CoD on it at night for a couple hours at a time.

 

[Handruin]

After roughly 26.5 hours I've completed 1632 out of 2065 steps (79%) on my first QMD. Are all QMD's roughly the same size? This is on my A64 3200+ (winchester core) with 2GB RAM.

 

[P5-133XL]

Here's someone that is trying to create a database of machines and their folding capabilities: FAHInfo.org. Should be interesting...

 

[Handruin]

Almost 34 hours and completed 2065 out of 2065 steps (100)

I looked at that database and found a similar config with my work unit:

1912|QMD|450|AMD|Athlon 64|Winchester|2400.00|426.00|177.50|Mattus

They averaged 426 PPD and I'm way behind that at roughly 300 PPD. I guess mine isnt that much faster.

 

[Handruin]

Could my configuration of 4x 512MB DDR400 on the board be causing a problem? We determined this CPU is not a Venice in another thread but rather a winchester core. How bad of a hit do I take with four sticks of ram in the motherboard (gigabyte nforce3).

 

[CougTek]

I'm not sure, but I think Winchester might drop to 333MHz (from 400MHz) with four sticks installed. LiamC posted about this in another thread.

 

[P5-133XL]

The easy way to find out is to pull two sticks and compare frame-times

 

[Bozo]

Quote: "So about with each at 450 points, I'll probably average 500 or 450 points a day and my graph will look odd with two work units submitted one day and none the next and two the day after that."


Are you running two instant's of folding on the same machine?
I thought you were only supposed to run one when running QMDs.

Bozo :mrgrn:

 

[Will Rickards]

Are you running two instant's of folding on the same machine?
I thought you were only supposed to run one when running QMDs.

Yes, I'm running two instances. With a P4 and hyperthreading you should probably only run one, but a dual core is actually two cpus so I figure why not. This is also why I want to upgrade the RAM from 1GB to 2GB.

 

[Handruin]

Looking at CPU-Z, it does say the frequency of the ram is 167MHz (or 333 double pumped?). Now I'm curious how much of a difference that makes for folding.

 

[CougTek]

I can't say for sure for an Athlon 64, but for my Pentium 4, boosting the RAM to 375MHz from 333MHz gave me a 9% improvement on the average time per frame for QMD units. The improvement has been linear with the memory frequency increase.

If the same rule applies to the Athlon 64, we are talking about a 17% improvement with 400MHz memory vs 333MHz.

 

[P5-133XL]

Are you running two instant's of folding on the same machine?
I thought you were only supposed to run one when running QMDs.

Yes, I'm running two instances. With a P4 and hyperthreading you should probably only run one, but a dual core is actually two cpus so I figure why not. This is also why I want to upgrade the RAM from 1GB to 2GB.

Because QMD's are RAM bandwidth limited, not CPU bound. Both CPU's use the same bus to access RAM, so you gain little with the second instance. To optimize dual CPU's, it is probably best to have one doing QMD's and the other a non big WU (A non QMD).

 

[P5-133XL]

Are you running two instant's of folding on the same machine?
I thought you were only supposed to run one when running QMDs.

Yes, I'm running two instances. With a P4 and hyperthreading you should probably only run one, but a dual core is actually two cpus so I figure why not. This is also why I want to upgrade the RAM from 1GB to 2GB.

Because QMD's are RAM bandwidth limited, not CPU bound. Both CPU's use the same bus to access RAM, so you gain little with the second instance. To optimize dual CPU's, it is probably best to have one doing QMD's and the other a non big WU (A non QMD).

Or in the case of AMD X2 that are being switched over to QMD's one QMD and one non-QMD big WU (One instance switched to QMD's and one normal large WU instance)

 

[Handruin]

Any info on freeing up memory on my Dell SC420 to hopefully get a QMD? Right now the FAH client is reporting:
"Memory, in MB, to indicate (494 available) [494]?" From what I've read, I believe if I can get it to 504MB, I might get QMD's.

I guess the correct solution is to buy more memory such as this.

 

[CougTek]

You probably have onboard graphics and that's eating a chunk of your memory. Buy a separated graphic card, graphic card, or graphic card and disable the one onboard. Spend less than for 1GB of RAM. Enjoy clearer display and DVI output.

 

[fb]

I had to take the Xeon 2.8 offline a few days ago, but I'll see if I can't get some more points/day out of the P4 3.0.

Anyway, I'm currently folding away with:
1 x P4 3.0GHz (one client)
1 x P4 1.8GHz (one client)
1 x X2 3800+ @2200MHz (two clients)

 

[Will Rickards]

Or in the case of AMD X2 that are being switched over to QMD's one QMD and one non-QMD big WU (One instance switched to QMD's and one normal large WU instance)

After this week I'll switch one instance over and see.
I just want to get a baseline week with QMD's to base performance numbers against.

 

[CougTek]

Well, I think I just might have added another P4 to the mix...

I don't see you in our team. Are you sure you've config the client correctly? Our team number is 10047.

 

[Anathor21]

Well, I think I just might have added another P4 to the mix...

I don't see you in our team. Are you sure you've config the client correctly? Our team number is 10047.

I've been there for quite a while... just not under the same username. Care to try and guess what username I am on folding or should I just spill the beans?

 

[CougTek]

I would have thought about i, but with his production, he cannot have even a single P4, so you're not i.

Maybe Clocker, but from his last post in this thread, he wouldn't have care to reply to my question.

ddrueding and Pradeep must not have a single P4 in their folding farms. JoJo and Tea either.

Mourn and Bozo...nah.

Zx...nope. With the systems he has listed before, he cannot add another P4 or his daily amount would be around 800 points.

LiamC is far too proud of his new sport car avatar to post under a new handle.

Gilbo wouldn't have started a thread like this one as he is far more likely to know the answer. Same for Fush. and Mark.

Can't be Erik...there was no typo in your post.

Greg? No. There would have been something else in your post. I don't know what, but something for sure.

The bartender is too busy behind the bar.

F.Bostrom? The only one I can't scratch for sure.

 

[Groltz]

F.Bostrom? The only one I can't scratch for sure.

We have a forum member named "fb" Coug.

 

[CougTek]

I know, but Anathor has to be the handle for someone else.

 

[fb]

Anathor==ErikHenrikson?

 

[CougTek]

if iti swas erik it swould loke more limek thisd.

 

[CougTek]

Hey, wait a second. EBB's name isn't Henrikson. Ahh, ok. I assumed Erik was Eric. Now that makes sense.

Thanks fb.

 

[Will Rickards]

My googlefoo on Anathor21 and erikhenrikson reveals:
a) someone with three kids in MA USA
b) someone with an IP from comcast in Milford, Massachusetts
c) someone who contributed to seti@home
d) there are Henrikson's in the phone book for MA near milford

So I don't think I'm any closer to identifying anathor21 but I had fun trying.

 

[LiamC]

LiamC is far too proud of his new sport car avatar to post under a new handle.

I wish it was my new sports car!

Only if I win the lottery tonight

 

[CougTek]

Doug is in MA...but I don't think he has three kids.

 

[Anathor21]

Better end this before someone posts my SSN, phone number and notifies my first girlfriend where I am

The winner is indeed fb - too bad I don't have a prize to offer. Guess bragging rights are worth something (otherwise why would we become so obsessive about folding?).