Folding@Home

Thanks to you Bozo, everyone who visits this thread for next 100 posts will be greeted with a typo. How sweet.

Don't worry, we all know that it's the intention which counts. ;-)

What kind of CPUs will you add? 24x7 or day shift only?
 
CougTek said:
[looking at our daily production]

So...we're giving up, aren't we?

I mean, less than 9000 points per day won't get us anywhere.

Pradeep was forced to slack down, Tea has stopped (probably for Aussy summer), Eric too. I seem to have pissed off Miksmi by not knowing who he was (which I'm sorry about).

We still have a bunch of important contributors, but we need to add some meat to the bone to avoid oblivion.
Click to expand...

I'm not giving up. I'm just gonna keep plugging away. I may not have had the potential of a ddrueding or a Pradeep but I have longevity on my side.
 
This time of year, my computers are nice room heaters, so I can rationalize the electricity, and that I might help save myself from cancer, since my mother had a brain tumor, dad died of cancer, as did my favorite aunt, and uncle.

s
 
Well, I think I just might have added another P4 to the mix... runtime depends on whether my mother keeps her system on or finds that little "moon button" too irresistible. However, it will plug along whenever it's on... (yes I did tell her it is there, heh). So, perhaps if we get 1/2 time out of it we should get a few more points out of it...

And we have two cancer survivors in my family so I do so hope this research does go somewhere... (we are partial to the "rare" varieties too... figures).
 
Folding QMD's on AMD machines:

I got The following instructions for running QMD's on AMD machines from the Stanford folding forums. Since the instructions directly violate EULA's from both Intel and F@H I suspect that the thread will not remain. With that in mind, I'm posting them here in addition to the direct link.

RM said:
After a bit of thinking and experimentation, a very easy way to make a Windows console F@h client running on an AMD64 CPU download QMD WUs was found.

I hope I am not telling the obvious stuff, so if some of you guys knew this - sorry for wasting your attention.

Note: by using the method below you WILL violate the Stanford EULA, so please consider carefully whether you want to do this or not. Due to the same EULA I cannot post modified client or core executables. HOWEVER I believe that posting instructions about such modification does NOT constitute such violation, and hope this information won't be censored away from the forum.

So, the instruction(s).

Just open the FAH5??Console.exe in your favourite HEX-editor. In this file, find the string "GenuineIntel" (w/o quotes) and change it to "AuthenticAMD" (luckily, they are of the same length). Also, find the string "AuthenticAMD" (another one) and change it to "GenuineIntel".

In other words, there is one occurence of "GenuineIntel" and one of "AuthenticAMD". Swap them. For your convenience, "GenuineIntel" is "47 65 6e 75 69 6e 65 49 6e 74 65 6c" in hex.

Now the F@h client will report your AMD CPU as being Intel's (and vice versa). After that, enable "big WUs" option in client config AND launch it with the "-advmethods" switch. You will instantly get a QMD WU (I just got 3 of them on my Athlon64).

That's all for now. Stay tuned for the next part, in which we will try to patch QMD core to actually use SSE2 on AMD64 CPUs.
Click to expand...

RM said:
The second part is as quick and painless as the first. In this part, we (probably) violate some Intel's EULA, so please consider carefully, blah blah blah.

Head over to the "Naughty Intel" website, proceed into its "patches" section and download the script named "intel_check_executable_patch.gz".

Here comes the most difficult part: you need to have access to some machine where Perl is installed (almost any Linux machine has it), or install a Windows version of Perl. Then, have your FahCore_96.exe with you on the Perl-having machine and run the script:

Usage: intel_check_executable_patch.pl [options] <filename> [<filename>...]

Options:
-b <suffix> Use the supplied suffix to name the backup file. The
default is '~', and setting this to '' disables backups.
-v Be more verbose.

Some versions of the Intel Fortran Compiler produce code which checks whether the CPU is made by Intel or not, and disables MMX/SSE/SSE2 code if it isn't.
This program patches executables produced by the Intel compiler to remove this check, so that e.g. programs compiled with -axW will use SSE2 code on Athlon 64 chips.

This script was tested against ifc versions 7.1.040, 8.1.028 and 9.0.024 and should work on other version 7 to 9 releases. The patch process will almost certainly work with executables compiled with the Intel C compiler as well, and with Windows executables as well as Linux binaries, but has not been extensively tested on these, so be careful! If you do test this on icc-compiled executables, then please let me know how it goes!

Patched files are backed up (by default to 'filename~'), see the -b flag to
change this.

This program is provided with NO WARRANTY: please check that the patched executables behaves as you would expect! I can't really foresee any problems that could reasonably be caused by the use of this patch, but you never know .
Click to expand...

Run "./intel_check_executable_patch.pl FahCore_96.exe". It will say:

Patching FahCore_96.exe in 9 places...
Click to expand...

And that's all. Copy the patched FahCore_96.exe in place of the original one inside the F@h folder. Run the client, and... touchdown!

[03:09:50] Writing local files
[03:09:50] Extra SSE2/SSE3 boost OK.
[03:09:50] Entering QMD...
Click to expand...
...
Click to expand...
 
Don't forget that if and when Stanford updates the FahCore96.exe, you will need to repeat the process. Updates are automaticly downloaded, so you will need to monitor that.
 
Should an 800Mhz FSB P4 2.6C be producing only ~150pts/day with -advmethods and large WUs? I'm pretty sure I read earlier in this thread that it should be doing significantly better than that. I've checked that everything is configured correctly and it seems to be. For the life of me I can't figure out why it's behaving the way it is.

You can see the P4 isolated from Dec. 28th to Jan. 2nd.

It's running 24/7, 1 client, -advmethods, -forceasm, large WU's. Anything else I need?
 
How much RAM does it have? QMD's will only appear if there is 512+MB and from what I'm seeing it is not getting these. QMD's are what makes P4's perform so well. To get QMD's you need large WU's enabled, -advmethods, and 512MB of RAM or more.

My history with P4's is that once configured properly, I get continious QMD's
 
I'm running two P4 3.0Ghz and they both get 450 points per day. Both computers have 1GB of RAM or more though.

Bozo :mrgrn:
 
Hmm. I just checked the RAM in the System dialog and it says the system has 480MB. There are definitely 2 256MB DIMMs in it though. Integrated graphics reserving 32MB perhaps? That's irritating...

No QMD's for me I suppose :(.
 
Check the CMOS Setup for any ROM shadowing being turned on and then turn it off. However, Integrated graphics that shares system RAM can definately do this too.
 
Zx must have received his parts ; he's giving me more trouble than he should.

I decided to overclock my puny Pentium 4 J505 this afternoon. It is now clocked at 2.94GHz, with 367MHz RAM. My Asus P5P800's BIOS doesn't allow me to set the RAM at 400MHz unless I push the CPU to 4GHz, which it isn't able to operate at. I had weird results : the system post and even boot into Memtest86 (only to fail on the very first test) at 4GHz CPU/400MHz RAM, but it doesn't even POST at 3.2GHz CPU/333MHz RAM. It can't be the RAM because it works at 400MHz with aggressive timings on other systems. The CPU post at 4GHz, but not at 3.2GHz. No idea why. I haven't played with voltages.

Previously, I needed some 29 hours to complete a 450 points unit. I'll see how many minutes (or perhaps even hours) the 10% overclock will shave.
 
Good luck. Folding seems to bring out the worst of overclocking. Where everything else seems stable -- It won't be to F@H. Check you logs, because the WU's may be finishing with errors (giving you little to no points). I wouldn't want you to lose more than you ever will be able to gain.
 
Coug, take it like a man. ZX is going to walk right by you, just like you did me. I accepted my fate without resorting to over-clocking. ;)

I may try Mark's nifty find on the QMD's for AMD. Then I might have a chance some day.
 
Mark, I'm always monitoring my logs. If I ever see an error, I'll back off.

Zx hasn't walk by me, he's been in front of me since last Spring (IIRC). I'm just trying to regain my due. I still haven't sent my current unit, but so far so good despite the slight overclock.
 
I thought he just passed you for some reason, my bad. You'll soon have to worry about Erik and LiamC.

Doesn't Mercutio have a ton of machines at his place? Help us out man!
 
Nah, I can't keep up the pace.

I currently have 2 X2's (4 cores), S754, P4 1.6, XP1800+, P!!! 600 and the P2 400 Smoothy box folding.

One of the X2's is a build for a friend, and when he gets around to bringing his K6-III 400 around, I'll transfer files and say farewell to the box.

Someone has also expressed interest in the S754 box and the XP1800+ looks like it may end up at the mother-in-laws place.

So I'll be back to the X2, Smoothy and P4 1.6 (my work box). The P!!! really doesn't do enough to justify running it either.
 
I noticed Clocker's unit yesterday too, but with only a 40 point unit, I thought he simply forgot to shut down the client on a very old machine (P2-like) and that it finally achieved a unit began a year ago.

If Cocker is back for real, then it's great for us.
 
P5-133XL said:
It is my understanding that Windows has to report at least 504MB.
Click to expand...
You're right, I took a look at the computer today and it hasn't been folding QMD's for a while now. :( I think I'll have to buy a graphics adapter...
 
Ah yes... i received everything. I now have :

- Dual Athlon MP 2600
- Athlon XP 2800 (That's the one I added recently)
- Athlon 1000
- Pentium M 1.5 (laptop, so it's folding 80 % of the time)
- Athlon64 3200 (a friend's laptop... folding about 6-8h/day)

Do you think a Pentium II 233 could help? It's my parent's computer and it's already on 24/7.
 
zx said:
Do you think a Pentium II 233 could help? It's my parent's computer and it's already on 24/7.
Click to expand...

zx, I ran the client on a Pentium Pro 200 for a month or so. It's not worth it. Really. It took about a month non-stop to complete one work unit, and it may have had an impact on the system's overall usability, because let's face it, down at that speed, even an application that's trying to play nicely can have an impact on system performance. You don't want to risk that with anyone's parents, let alone your own.
 
Will Rickards said:
Will Rickards said:
WooHoo got a QMD, now to apply the second part!
Click to expand...
Muhahaha, the second part worked as well!
Hopefully this puts me closer to 600 pts per day.
Click to expand...
It will be very interesting to see how quickly A64's can finish QMDs relative to P4's. P5-133XL said they were memory bandwidth limited, if so, it seems reasonable to assume that 939-pin machines should be outstanding at them.
 
zx said:
Ah yes... i received everything. I now have :

- Dual Athlon MP 2600
- Athlon XP 2800 (That's the one I added recently)
- Athlon 1000
- Pentium M 1.5 (laptop, so it's folding 80 % of the time)
- Athlon64 3200 (a friend's laptop... folding about 6-8h/day)
Click to expand...
Now, how am I supposed to compete with that? I updated my BIOS to version 1401 (on my P5P800) and I successfully bumped the frequency to 3GHz/375MHz RAM. That's the highest I can go. It shaved almost two minutes per frame compared to the default frequency of 2.66GHz/333Mhz RAM. I doubt it will be enough to match Zx though. I'll have to find a way to add another system. My 2 CPUs against Zx's 6 won't compete.

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

Windsor-TaylorRebecca-Jan2006.jpg


That dog would be really nice with a GM logo on its shirt, wouldn't it? ;-)
 
Gilbo said:
It will be very interesting to see how quickly A64's can finish QMDs relative to P4's. P5-133XL said they were memory bandwidth limited, if so, it seems reasonable to assume that 939-pin machines should be outstanding at them.
Click to expand...

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.
 
You must log in or register to reply here.
Back
Top