F@H Top Crunchers...

[xbassman]

Originally posted by: Assimilator1

Originally posted by: GLeeM

Originally posted by: Assimilator1
What's 'EUEing F@H' ?

In addition to xbassmans' explanation I will add:

EUE is what it is called in the log.

You can have a stable overclock for everything else but it might be too high for some F@H WUs. I don't know, maybe certain F@H WUs push some parts of the CPU harder than other programs that test stability?

I doubt those people who claim to have a stable overclock have done a 24hr Orthos/Prime 95 test.

GLeeM is right.....
All my boxes are >24hr prime stable, but F@H is more sensitive. F@H is the final stability test as far as I am concerned.

 

[MadMurph]

Originally posted by: Markfw900 . . .

Exactly ! My new Q6600 G0 is perfectly stable@3.6, until you run F@H, then you get EUE's. I had to go down to 3430 to be totally stable with no EUE's.

Hey Mark -- I was having the same problem. I've finally gotten stability at 3.6, vCore 1.44, on air, but at the expense of some slightly under clocked (1000@2.2v) DDR2 1066 (PC2-8500).

BTW, for those of you who might have missed it, I happened to log on to Kakao Stats last week, only to see Mark_F_Williams at a 24 hour mark of just around 18,000 points!!! I bow down.

 

[Assimilator1]

3.6GHz on a quad is awesome ..............damn I forgot I've borrowed a mates Q6600 ! ,had it for a week & done nowt with it! ,I'll play with it tomorrow night.

Originally posted by: Markfw900

Originally posted by: GLeeM

Originally posted by: Assimilator1
What's 'EUEing F@H' ?

In addition to xbassmans' explanation I will add:

EUE is what it is called in the log.

You can have a stable overclock for everything else but it might be too high for some F@H WUs. I don't know, maybe certain F@H WUs push some parts of the CPU harder than other programs that test stability?

And like xbassman said, some WUs normally EUE on their own. Usually you will get partial points for EUEs because they still give the scientists data.

Exactly ! My new Q6600 G0 is perfectly stable@3.6, until you run F@H, then you get EUE's. I had to go down to 3430 to be totally stable with no EUE's.

Had you run Orthos for 24hrs?

xbassman
So you're saying you've done a 24hr Orthos test & yet it's still given more than a 1 off EUE?

 

[xbassman]

xbassman
So you're saying you've done a 24hr Orthos test & yet it's still given more than a 1 off EUE?

Not really sure what you mean by 1 off EUE.....

I have never run Orthos, but Prime95 has run >24hrs on every OC. (The main reason why I have identical processors running @ different speeds.)

Every once in a while I see an EUE. The only times I have ever seen more than one it was due to heat from dust buildup.

The boxes in my herd are getting a little old except for the one I am on right now.....

 

[Insidious]

Originally posted by: xbassman

Originally posted by: Assimilator1

Originally posted by: GLeeM

Originally posted by: Assimilator1
What's 'EUEing F@H' ?

In addition to xbassmans' explanation I will add:

EUE is what it is called in the log.

You can have a stable overclock for everything else but it might be too high for some F@H WUs. I don't know, maybe certain F@H WUs push some parts of the CPU harder than other programs that test stability?

I doubt those people who claim to have a stable overclock have done a 24hr Orthos/Prime 95 test.

GLeeM is right.....
All my boxes are >24hr prime stable, but F@H is more sensitive. F@H is the final stability test as far as I am concerned.

(My opinion... ymmv)

It's not so much the demanding nature of the F@H client that brings about this 'unstability'. Rather it is the (IMO: short sighted) authors of F@H software that have done a poor job of ensuring the application will continue to function when it encounters an unexpected result of an operation. (such as an impossible value returned from RAM, etc.) A 'well written' application will be checking the 'reasonableness' of its own function as it goes and recover/re-check when an error in data occurs. I don't think F@H does (or if it tries.. it is unsucessful). So a computer that is plenty 'stable' enough to run competent software forever, looks 'unstable' when F@H runs.

My own personal solution has been to make sure my computer can run any software I care about 'forever'. If F@H is the only thing that behaves as though my PC is not set up correctly, then that is Stanford's problem.... they can do without my results until they write an application capable of providing them on an otherwise 'stable' PC. I'll be happy to restart the F@H app. when it needs it, but I won't sacrifice the speed I have available for everything else... just to make up for poor programming at Stanford.

-Sid

(before you question my dedication.... make sure you've been folding longer than I have.)

 

[xbassman]

Sid....You're not causing trouble @ the Stanford forums again, are you?

 

[Insidious]

Originally posted by: xbassman
Sid....You're not causing trouble @ the Stanford forums again, are you?

LOL :laugh:

Nope, I've been a good little folder.

I guess I may be overdue for a browse..... It's never boring. Maybe since it's been a few months, they've had time to write a new post. (I had the old cutnpastes memorized )

-Sid

 

[Thump553]

I'm amazed the difference hardware makes. I have 4 machines (24/7/365) running F@H, sometimes up to six machines, but all are ancient by today's standards-avereage machine age is probably four years. Only one machine is even dual core and I had to limit F@H to one core there because it slowed me down too much. I average 400-650 points a day, MadMurph-the next user to blow by me-is well over 10,000 per day.

 

[mgpaulus]

I have the following:
Home
------------------------------------------------
AMD Opty 148 S939 Single core / XP Pro
AMD 4000+ S939 Single core / XP Home
AMD 3800+ X2 S939 / Debian Etch
AMD 3800+ X2 S939 / Debian Etch
AMD 3800+ X2 S939 / Debian Etch
AMD 3000+ S939 / Win2K

Work
--------------------------------------------------------------
Intel P3 733 / Debian Etch

The thing that kills me about all this is the darn utility bill. My utility bill is running $80-$100/mo consistently. That sucks.
2 of the above X2 are for my HTPC environment. The X2 is my general purpose fileserver and linux playaround box.
The Win2K box is a holdover from my Seti days, and was running Setiq on it. Now it folds.
All these are in the basement.
The 2 XP boxes are upstairs for family use (bookkeeping/email/homework/games/etc/etc).
The work box is a general Linux box for my use. I plan on upgrading that to a P4 1.7 in the very near term, as soon as I get a couple of issues straightened out.
I don't fold on my work laptop, since I feel it's performance strapped as it is, and it's only on 1/3 of the day. I suppose I could fold, and get a bit more of a boost,
but not going to for right now.

 

[MadMurph]

Originally posted by: Thump553
I'm amazed the difference hardware makes. I have 4 machines (24/7/365) running F@H, sometimes up to six machines, but all are ancient by today's standards-avereage machine age is probably four years. Only one machine is even dual core and I had to limit F@H to one core there because it slowed me down too much. I average 400-650 points a day, MadMurph-the next user to blow by me-is well over 10,000 per day.

Here's the fodder to support your point:
My original SMP cruncher was an Asus P5B Deluxe, C2D E6400, 2gb DDR2800, X1950Pro, stock settings. It would crunch a standard 5mb packet in 34 hours, or about 20min/unit.

This same machine, overclocked to 3.0g, HyperX DDR2 800, 23.5 hours, 14.1min/unit.

Some others:

Asus P5N-E Sli, E6750, 2gb DDR2 800, overclocked to 3.1, 23.3 hours, 14min/unit.

Asus Blitz Formula (HTPC), Q6600, HyperX DDR2 800, X2600Pro, overclocked to 2.8g, 17.5 hours, 10.5 min/unit

Asus Blitz Formula, Q6600, Dominator DDR2 8500, X1950Pro, overclocked to 3.6g, 12.6 hours, 7.5 min/unit.

Finally:
I've got two (of the late great "overclockers") Pentium D805's, on ECS throw-aways, pushed to 3.2g with DDR400, they each take just over 53 hours @ 32 min/unit!!!

These are all rock stable, but as Insidious pointed out ^^^ the software does have flaws that make things "hiccup" every now and then, for no apparent reason, though apparently sometimes on purpose. At stock timings, this would probably be to no ill effect, but for these rigs, it usually causes a restart. For the two 805's, they will miss the work unit deadline. However, Insidious should go easy on the Stanford boys, since their project performs remarkably well in comparison to other software authors (MicroBlunder) that have huge staffs, billions of dollars, years of research, yet still can't seem to produce a piece of software (ME II) with a fraction of the stability and functionality the Stanford team has achieved.

 

[Markfw]

Originally posted by: mgpaulus
I have the following:
Home
------------------------------------------------
AMD Opty 148 S939 Single core / XP Pro
AMD 4000+ S939 Single core / XP Home
AMD 3800+ X2 S939 / Debian Etch
AMD 3800+ X2 S939 / Debian Etch
AMD 3800+ X2 S939 / Debian Etch
AMD 3000+ S939 / Win2K

Work
--------------------------------------------------------------
Intel P3 733 / Debian Etch

The thing that kills me about all this is the darn utility bill. My utility bill is running $80-$100/mo consistently. That sucks.
2 of the above X2 are for my HTPC environment. The X2 is my general purpose fileserver and linux playaround box.
The Win2K box is a holdover from my Seti days, and was running Setiq on it. Now it folds.
All these are in the basement.
The 2 XP boxes are upstairs for family use (bookkeeping/email/homework/games/etc/etc).
The work box is a general Linux box for my use. I plan on upgrading that to a P4 1.7 in the very near term, as soon as I get a couple of issues straightened out.
I don't fold on my work laptop, since I feel it's performance strapped as it is, and it's only on 1/3 of the day. I suppose I could fold, and get a bit more of a boost,
but not going to for right now.

I have 9 boxes, and my electric is $180/month, so I know what you mean ! 4 are quads though, so I get more bang/buck out of them.

 

[VirtualLarry]

Originally posted by: xbassman
GLeeM is right.....
All my boxes are >24hr prime stable, but F@H is more sensitive. F@H is the final stability test as far as I am concerned.

Interesting datapoint. I'm interested in ultimate stability for my OCed systems. I guess I should start using the F@H client to test that too. Is there a testing mode, or do I need to sign up for the real deal to begin using it?

 

[xbassman]

Originally posted by: VirtualLarry

Originally posted by: xbassman
GLeeM is right.....
All my boxes are >24hr prime stable, but F@H is more sensitive. F@H is the final stability test as far as I am concerned.

Interesting datapoint. I'm interested in ultimate stability for my OCed systems. I guess I should start using the F@H client to test that too. Is there a testing mode, or do I need to sign up for the real deal to begin using it?

There is no testing mode.....
Just install one of the clients and keep an eye on the log.
F@H is really transparent you can let it run as long as you want and it shouldn't interfere with anything else you are doing with your box.

Keep an eye on the logs and your core temps. (The SMP beta sometimes just shuts off and the only way I can tell is a 10-15C drop in core temp.)

 

[Assimilator1]

Originally posted by: xbassman

xbassman
So you're saying you've done a 24hr Orthos test & yet it's still given more than a 1 off EUE?

Not really sure what you mean by 1 off EUE.....

I have never run Orthos, but Prime95 has run >24hrs on every OC. (The main reason why I have identical processors running @ different speeds.)

Every once in a while I see an EUE. The only times I have ever seen more than one it was due to heat from dust buildup.

The boxes in my herd are getting a little old except for the one I am on right now.....

Orthos is a program which automatically runs 2 Prime 95 clients (before P95 itself was capable).
1 off EUE as in rare occasion.

If you only ran 1 instance of pre v25.3 P95 on a dual core rig then you've only been loading the CPU tp 50% ,or was that on single core rigs?

mgpaulus
I pay £15/mth (approx $30) for 1 rig running 24/7(C2D @3.2GHz + 1950Pro) ,I think you're getting a bargain
Seriously though ,yes it does add up to quite a bit of cash!

 

[caferace]

Finally took a pic last week of the home boxen...

http://www.flickr.com/photos/15887989@N00/2170359743/

-jim

 

[GenHoth]

Originally posted by: caferace
Finally took a pic last week of the home boxen...

http://www.flickr.com/photos/15887989@N00/2170359743/

-jim

Nice Shelby Cobra on the laptop!

I'm glad this got resurrected.

At school I run a Q6600, stock right now but I'll be doing a mild OC when I get back.
A P4, 3GHZ?
at home my dad's computer (That I occasionally hijack) is a P4 2.66GHZ

So 3 computers total!

 

[caferace]

Originally posted by: GenHoth
Nice Shelby Cobra on the laptop!

Thanks for noticing. Enjoy some walpaper.

http://www.flickr.com/photos/1...ets/72157600052445217/

-jim

 

[XJustMeX21]

Just started out and right now i have only 2 machines running.

Q6600@3ghz with 4gbs of ram on vista x64 and a X2 4200+ at 2.7ghz running on win xp.

 

[scottv67]

XJustMeX21 wrote:

Q6600, 4gb dd2800, 8800gts 640mb, Abit ip35 Pro, Tunic Tower 120 and Corsair 620watt psu.

I have three machines in my basement with the Abit ip35 Pro and Q6600s running the F@H SMP client. All three machines overclocked very easily (didn't have to raise the CPU voltage much - if at all). That motherboard and the Q6600 CPU are a good match. I ran the Q6600s in some Abit AW9D-MAX mboards and I had problems with a healthy overclock. I think the AW9D-MAX is better suited for a C2D.

I think you'll be pretty happy with the PPD from your ip35 and Q6600 combo.

-s :beer:

 

[XJustMeX21]

Went to 3ghz without having to push the volatge at all. I'm sure that it could go higher but i know these things run warm to begin with and don't want to push it.

 

[MechEng]

Originally posted by: xpblackdragon
I fold 24/7 on my c2 quad Q6600 @3500 it generates 3500ppd or 2 WUs per day

WOW!!!
That'ss....like... amazing.... :Q


My old P4 is really, really, really overdue for an up-to-date replacement... I think it lies at 200-300 PPD or so... haven't run it long run enough this time after a rather long break from folding to say for sure.

I already planned on getting myself a q6600, and it seems that will really kick some serious behind in f@h, compared to my current CPU...

WOW again....

 

[Gravity]

Well, I've been the #1 guy for Team AT for many years. At times I had as many as 35 boxes folding for the team. Now, some of those boxes are outdated and won't run the new smp client. I'm in the process of upgrading or replacing all boxes.

Like Murph, I have a 20 pc lab that works for the team at all times. They don't do heavy lifting but in the fall, they will all be Q6600 or better and contributing to the count.

I only expect to be #1 for a few more months even though I am in the process of purchasing 4-6 new q6600's per month through summer. Perhaps by the end of the year I can once again report decent production.

I have a hard time keeping up with the changes at FAH as well as the seemingly random point assignments. For example, a q6600 running ubuntu can score well. As much as I want to get the highest points, I am not going to invest time I don't have into learning a new operating system.

Even the SMP client is a bit difficult for me. It won't run as a service nad running two instances on my q6600's doesn't work that well for me. Things were simpler in the old days.

So, I will estimate my fleet of folding boxes at about 40, give or take a few and give or take the random box that reports work for anonymous. I like the concept of folding and love the work we help the university do for science.

Keep folding,

Gravity

 

[Insidious]

I'm with you Gravity on the need for a service install and operation of the SMP client. I don't think it would lend itself well to a lab as it operates now (needs babysitting, has to be restarted when the machine is restarted, sensitive to network hiccups, etc..)

IF you do end up running the SMP client, we've found that if you use the Affinity Changer utility to help windows figure out the processor assignments, two instances of the SMP client will produce very nearly the same on a windowsXP box as on a LInux box. (I get >4400 PPD from each of my Q6600s (@ 3.5GHz) this way)

but even if you are running the regular text only clients (CLI) as a service, the sheer number of machines you have at your disposal is going to have a nice impact on your amazing contribution to F@H.

The saddest part is that I setup and maintain over 60 vehicle simulation stations for various software development groups at work and can't put F@H on a one of em...

-Sid

 

[Markfw]

Gravity, sorry about what appears to be my taking over number one, what can I say. I will continue to help you in your quest with advice on the SMP client and ubuntu. And with what you are adding, I am sure it will be short lived.

BTW, I get 4400 on windows on a Q6600@3510, and 4800 on a Q6600 @ 3550 in ubuntu, so with the affinity program, its 10% gain. I have 2 Q6600's both @ 3200, and there its 4000 vs 4300, so again, show 8% gain. So windows isn't too bad.

 

[Gravity]

Originally posted by: Markfw900
Gravity, sorry about what appears to be my taking over number one, what can I say. I will continue to help you in your quest with advice on the SMP client and ubuntu. And with what you are adding, I am sure it will be short lived.

BTW, I get 4400 on windows on a Q6600@3510, and 4800 on a Q6600 @ 3550 in ubuntu, so with the affinity program, its 10% gain. I have 2 Q6600's both @ 3200, and there its 4000 vs 4300, so again, show 8% gain. So windows isn't too bad.

In some other competition, perhaps basketball, I might be a bit sad about losing. However, when you figure the winner is not the top cruncher, but the people that will be served by the science that comes to their aid as a result.

Mark, I welcome your victory, whenever it might come and will follow you closely. Perhaps had you not been climbing so quickly I would be aloof as to the folding status of my boxes. Your initiative has caused me to examine each box and to ensure that what little they do is being done. Bravo Mate!

It was the same years ago when RustyNale and I set sights on spot # 12 and when poopflingers did some intense crunching.

When you consider the social aspect, regardless of whether or not it was intentional, this is an ideal way to have people contribute their computing power to the cause. Competing for points pricks the male ego and can inspire us to make great efforts and significant contributions.

I'm 44 now so winning or losing isn't that big of a deal to me. I see the bigger picture and applaud all the folders that pony their spare cycles for the cause. I'm also type A to some degree which means I will use my own spare cycles (some of them) to improve my efforts, regardless of the current FAH configuring and point standings.

That's a ton of blathering but I don't want you or anyone to feel poorly about taking over the pole position. The key is that we're pulling together in a common cause and when we push each other, we get to the cure faster.

So, dartanion, push forward with your best effort and set a NEW pace that can be admired, envied and eventually conquered by someone with a few more spare cycles than you have. Because as you know, you can only be number one for so long.

Guilelessly,

Gravity