How do I keep from getting 50 point gromac

Well, had all five computers setup to run over 2,000 points a day, then they pulled the QMDS from 512mg ram systems. I am now getting almost entirely 50 or so point Gromacs. You know, 748, 747, etc. How do I stop them? I have tried removing advmethods as well as a number of other settings to no avail. Stanford is saying that it is the luck of the draw so to speak on the larger gromacs however my experience is given 5 computers and 9 clients running at the same time that this just ain't so. Either that or I am truly one unlucky fellow in that I am defying the odds. I am about as upset as I can be about this having bought, put together, etc. five systems along with router, 8 port switch, cables, etc. If I can not find a solution to this problem then I am going to pull them all down save for possibly the xeons. I can not see the purpose in folding 9 clients with 15Ghz or so to only achieve 600 to 700 points a day and a $200 a month electric bill. I am very upset about the whole thing. I tried moving files amongst the computers only to have Stanford thank me very much for the results then not credit 2,000 points.

If you have a suggestion that might work then do say so prior to me shutting them down. If you know of a project that might make much better use of the cpu cycles then do let me know. It's been fun (till rescently).

Edit: We are cooking now. Finished the 51 point Gromac and am now working on a 44 point Gromac. That one will work out to a whopping 45PPD/Ghz. Now really, someone tell me again how important getting these things back to Stanford is. Do tell me the meaning of the switches again and what they are supposed to do. Maybe someone can explain why there seems to be a ton of the little bitty ones that run slow and are worth nothing while the large hard to do ones just can't seem to be found. Maybe the little ones are meant to give us something to do while Stanford gets off its' lame butt. I'm pulling them all down till they can get it together.

Here's to thinking your doing something.

Edit: As of 04:10 July 22, 2005 all of my folders are off line. They will not return online doing FAH until/if ever Stanford gets their act together. Till then, if you know of a good working project do let me know. Sure is quite in here now.

All of the proteins stanford sends out are important. Some aren't worth that many points but they're all important to the research they're doing. Those of us that have been folding for a while aren't really into the whole points thing even though it does spur on competition to fold more. We're here for the research purposes that the project is based on. Every protein is important in it's own way, there's no way to tell whether it will be a 600 point gromacs or a 44 pointer that will give them the information they need that will lead to a cure for a disease.

The bonus WUs have spoiled most of us. My experience with the 74x series is that on an XP-M, 200MHz FSB, they produce about 55ppd/GHz. Which is average for normal Gromacs. The highest normal gromac I've seen was produced @ 60ppd/GHz and the lowest @ 47ppd/GHz. So when bigpackets run out and Tinkers are in short supply, if you don't get a p74x, you're going to get something else that produces very near the same ppd.
There is a strong tendancy for the assignment system to give you a new WU from the same server as the WU turned in. That's why you keep getting p74x. One approach to breaking the cycle would be finish the current WU with the -oneunit flag. That should break the "connection" with server .130. Change the -advmethods flag and restart. Eliminating bigpackets will probably get you out of the loop but won't accomplish what you really want, which is a return of big WUs in masse.

Try clearing out all the FAH logs and the work folders then restart the clients one at a time to see what they get.

I heard if you delete wu's stanford will remove your account. is that true? i sure hear a lot about people doing it. it must really hurt the project.

Welcome to AMD land Silver. If we ain't gettin' 364-600 point BP's or 240/241 point Tinkers, all we get are those little Gromacs, Ambers, and Tiny Tinkers. Just keep them folding, but don't do it 24/7. Do it for 8-10 hours a day, and tey'll get finished, and by then, you should get a new, better WU.

dwschoon said:

I heard if you delete wu's stanford will remove your account. is that true? i sure hear a lot about people doing it. it must really hurt the project.

Click to expand...

Not that I know of. The only reason I've heard of for Stanford to delete accounts is for the buttheads that have released stealth installers on p2p networks, borging people's computers without permission. And they don't delete the accounts AKAIK, they just reset the accounts to 0 points and don't allow any more points accumulations on that username.

Does it hurt the project? I would imagine that is does a bit. However, they would be total fools (which I don't think they are) if they only sent 1 copy of a wu out on a DC project like this. I wouldn't be surprised if each wu was sent to 10 different computers, so they have a decent chance of the wu being processed in a reasonable amount of time, plus giving a cross check of the results in case a machine isn't totally stable and returns erronous results.

EDIT: Also, Stanford can also take a good portion of the blame for any wu's deleted, due to the terrible logic of their assignment servers. I don't know how many posts and threads I've seen over the years where people have posted up about high powered machines with lots of ram getting wu's that require 5-25 MB ram and a 5 day deadline, while some poor old Cellie 500 with 256 MB ram gets a 600 point bigWU that will take it 3 months to complete.

Given the concept of deleting WU's, turning Folding into a sport rather than remember what it is, I wouldn't mind Stanford for knocking people off if they kill WU's.

There's no real way of doing it, but come people, this is folding, for a cause.. not a friggin' pinball machine!

/Flix

I agree flix, too may of us are sounding like gamers rather than folders.

dwschoon said:

I heard if you delete wu's stanford will remove your account. is that true? i sure hear a lot about people doing it. it must really hurt the project.

Click to expand...

You don't delete the small WUs but put them aside in other directories to fold in HT mode. That way maybe he can hook up with a better server and get the bigger ones and then fold the small ones at the same time. Since he's quitting, I'm trying to salvage the situation and keep him folding.

I got a boat load of those small slow Gromacs too, but they get folded.

muddocktor said:

Does it hurt the project? I would imagine that is does a bit. However, they would be total fools (which I don't think they are) if they only sent 1 copy of a wu out on a DC project like this. I wouldn't be surprised if each wu was sent to 10 different computers, so they have a decent chance of the wu being processed in a reasonable amount of time, plus giving a cross check of the results in case a machine isn't totally stable and returns erronous results.

Click to expand...

Actually, Stanford rarely sends out duplicate WUs. On the return of a WU, the server produces the next generation. If you delete a WU, Stanford has no way to know it isn't going to be returned by you. The server doesn't reassign the work unit you deleted until the preferred deadline has passed. When you delete a p746 for instance, you delay the production of the next gerneration of that run/clone by 40 days. The deletion of WUs is highly detrimental to the science.

How do I keep from getting CANCER?

TC said:

Those of you that know me are probably aware that I've been a hard-core member of our Seti team since day 1. I was never opposed to Folding - I was just more interested in pure science, such as physics and astronomy. All of that changed when my mother was diagnosed with cancer in July of 2001. As her treatments progressed I became very frustrated with large pharma. Her doctors were either unwilling or unable to try emerging treatments due to the expensive costs of experimental drugs. Her primary oncologist expressed great frustration with the state of the drug industry in general, and all of those factors combined to make me feel that Folding was simply a way of giving large pharma free R&D that would never benefit the average person for decades.

Well today I got the news that cancer has spread to my mom's brain. They think they can treat it, but it's just another blow in a 4 year roller coaster ride. It got me thinking that even if Folding's discoveries don't immediately trickle down to help the average person, some day they will. Perhaps by helping now I can save my future wife, children, or maybe even myself. So the bottom line is I've decided to devote more of my equipment to Folding. It won't be anything earth shattering, but every bit helps right.

Click to expand...

nahmus said:

My mom was diagnosed with pancreatic cancer in 2001. This cancer is fast moving, painful and uncurable. Thankfully the company I was working for went out of business and i was able to spend the last 6 mos taking care of her until her passing. It was so frustrating watching what was happening to her and not being able to do anything about it.
Soon after her passing I found folding. It probably won't help me if i ever needed it but hopefully somebody somewhere at sometime wont loose someone they care about.

thats why i fold.

My prayers are with you and your mom. hopefully she will have a speedy recovery

Click to expand...

Joe Camel said:

so sorry to hear of your mom's situation TC.

my mother was diagnosed with lung cancer. after surgery, radiation and chemo, a follow up x-ray showed that the "treatments" had done nothing; the cancer had spread to just about every major organ and worst in the bones of her lower back...

needless to say she passed within weeks of that checkup. (i think she gave-up the fight @ that point)

please tell your mom to NEVER GIVE UP!!!

i fold in loving memory of my mother and in hopes that we can keep others from having to go through what i and my family went through.

Click to expand...

......

Silver said:

Edit: As of 04:10 July 22, 2005 all of my folders are off line. They will not return online doing FAH until/if ever Stanford gets their act together. Till then, if you know of a good working project do let me know. Sure is quite in here now.

Click to expand...

Silver, please turn you rigs back on.



quotes from: THIS thread

.

I think we're all seeing what Silver is seeing, lots of little Gromacs. Like Silver, I've spent a lot of $ on folding and specifically on folding big WUs. It hurts a bit not to have it fully utilized. The assignment server has a few quirks but, in my experience on 60+ cpus, it is pretty much luck of the draw. Yes a particular rig does seem to get stuck on one server from time to time. The thing is it's just as likely to get stuck on a big WU server as on a regular server. It all balances out. The folks with AMD big WU rigs are still seeing the ripples from the QMD fiasco. When QMDs got suspended, the assignment server suddenly had thousands of P4s and Xeons looking for big WUs. Of course these were already in short supply and so the AS had to send out something else. Since most of the WUs available are small Gromacs, that's what most of us got. With QMDs now being assigned only to rigs with 1000MB (soon to be 900MB) of ram, there are still thousands more rigs trying to get the quite limited number of big WUs than there were two weeks ago. Were just in a cycle where the demand for big WUs far exceeds the supply.

You can temporarily store those small WUs and fold 'em into the mix with designated -oneunit clients. Sometimes that helps to break a locked pattern with a small WU assignment server.

I took me three tries to get my P4c hooked into the QMD server. Those two gromacs I put aside will get folded in between QMD units.

I just filled out 8 folders on big gromacs, first try to 134 on the xeon box. AMDS will not under any condition, no matter how many trys/settings hook up to anything but site 130. Why? Only two differences here, 256mg on the 754 amds vs 1gb on the xeons and one system is amd, the other is Intel. I doubt it is the processors so I am left with the ram. Seems right given that stanford changed the requirements on the qmds. Given that advmethods is required this would have effected the amds that have less than 512/1gb?. Not sure as to whether it will pick up on 512 (I doubt it as the celly and P$ are both getting small units with 512) but again Stanford appears to have changed the requirements on a couple others without letting anyone know. A week ago, no problem. This week nothing but little gromacs. Funny thing is that everyone keeps telling me about the rotation of sites however the xeons have no problem what so ever hooking up first time to get big gromacs while the 256/512 systems can't get them for the life of them. Seems moving to one gig on the ram effected all large units, including large gromacs. Very underhanded of Stanford as they could have announced that not only qmds but all large gromacs would henceforth require 1gb. After all spending another $400 on ram should not be a big deal for most folders and Stanford believes they can justify their requirements on the public folders to their finance commity.

Nothing has changed in the assignment of big Gromacs. There just aren't many of them. My AMD rigs have a mix of p1477, p114x, p74x, and p140x.

Are you running 256mg of ram on them?

When big WUs were released, the minimum ram requirement was set at 240 MB. I've not read anywhere where that has changed. Systems with integrated graphics would have to be set to share as little ram as possible with the video subsystem. I'll keep checking to see if I can find a later reference to ram requirements for big WUs.

I did pull down some 1481s and 82s today. The Intel will pull 1477s at will. Very strange. No real response yet at Stanford. Any idea when/if qmds will be available to 512mb? First work since I shut down and wiped everything was turned in this morning.

I haven't heard but I'm sure pressure will be brought to bear to do it sooner rather than later.