Discussion in 'Folding@Home' started by W2hCYK, Jun 10, 2006.
Welcome to the fun
Thank you very much for joining, phanbuey!
I did notice that you are having a great kick start (7K+ PPD) - Fantastic.
Use this link to check your stats:
Any doubt, we are here to Help.
Welcome aboard PHANBUEY!!! We appreciate you unleashing those 260's for TPU. If you have any questions, let us know.
You will be the first one I contact
Welcome to the team bro
Whoa thanks guys . Ill try to run it 24.7 if i can... 64C is the top temp my gpus get (with fans at 80%)... I also have a spare PCI-X slot, so ill see if i can get a 9600GSO cheap for a PPU/Folding Card...
That E-0 is gonna rock for gaming .
Mine runns bout the same. Those 260's will be fine, the newer cards are rated for alot more than 65c.. i think its upwards to 90-115c... then you need to worry!
By the Way.. great sig.. i love that episode of ATHF
Forced 1.19 and 5800's ?
Copy and paste from F@H forum about Core 1.19 forced update:
Sorry if this turned out to be an unpleasant surprise. Here's the scoop.
Short version: Right now core_13 and core_11 have identical science code, so either one can be used for the production science here. P5748 is now a production project. I think it should have been set for core_11 to make this more clear, but science-wise it's no different right now.
Long version: We need to broadly push out 1.19 in order to have stability on the broadest range of clients. It's clear that 1.18 works for many people, but the extreme protests of the rest made it clear that we need to work this out. Letting people choose cores is problematic from many points of view when we hit the production stage. We need to keep the science consistent, we need to keep the system simple and stable (FAH is complex enough as it is), etc. For us to use FAH to do science, having data generated by the same code is often the most important part, so we can understand the meaning of the results, rather than having data generated by different codes in different parts of a trajectory.
We have rolled out v1.19 pretty gradually in terms of QA, but I can see your point of view in terms of the big roll out now. After seeing that it solved most people's stability issues, it was clear it was time for a broader roll out. Note that once a new core is past QA, if it has new science or better behavior, we will make it the requested core to clean up the older ones and get to a common base to do production science.
We will look to put optimizations back in as long as we can maintain stability. That's on the roadmap, but I'm looking forward to some stability over all the GPU clients for a while before stirring up the pot again. There's been many pleas from donors to make FAH more stable in general, even at the cost of bells and whistles, etc, and I'm trying to balance donors needs. Some stability would be useful in general to push out some more production science, which of course is the real goal of all of this from our point of view.
PS For those calling 1.19 a "POS" -- is that due to PPD or is there some new stability issue?
ALL MY RIGS ARE RUNNING 1.19 NOW.
WTF is Stanford doing? I have a copy of 1.15 on a flashdrive and replaced all the 1.19 cores with the old 1.15. They fold great until i get a 5748 WU(511 pt) and then I get a "Core Outdated" error. Then the client downloads and sticks FAHCORE 11 1.19 right up my ass!!!! I had no stability problems with 1.15 and I have every card O/C'd to the hilt. I now am taking a hit of 10% PPD on the 480 pt WU's. Looks like i'll have to keep pasting the 1.15 over the 1.19
EDIT: It seems that I am having trouble submitting some of the work done with the 1.15 core. Looks like i'm stuck with 1.19
The timing sucks for this. I am getting ready to have another 4x GPU rig online later this week. I'll need the new rig to offset the 10 GPU's I'm losing points on now.:shadedshu
Yeap! I did try the same, replacing the 1.15 back.
But the thing is: If is mandatory, we aren't the only ones going to a lower PPD.
We will have a better idea about the loses on the following days.
I quit folding a few months ago because it's clear to me that Stanford's priorities have shifted from more of a charitable mindset to a business mindset. I'm not going to let someone use my clocks that has only administrative interests at heart.
lookout guys. i'm picking up a 9800gtx+ tomarrow.
going to put up my 4870 for sale too, maybe make enough off it to buy another nvidia based card for the farm.
I'm felling sad...
...my PPD is falling down the hill.
Dammit 1.19 is not my favorite number.
Vijay made mention of adding back in the optimizations. We can only hope that this is true.
I am just catching up on this now and I had a couple of thoughts...
1) The main goal here is to do the science that allows researchers to cure diseases. If the new core gives stability to more people then the overall work into Stanford might go up. Longer to solve for individual machines but more machines able to submit valid results.
2) Since this is rolled out to everyone then everyone will have the same "hit" so the playing field is still level. We just need to reset expectations on points. Like I said it is not points that matter but valid results.
Unless I am missing some bit of information... just my $.02...
I can see a bit of this. In a way, I see folding as the "Livestrong" for benchmarkers. When those bands came out, they eventually became a fashion icon, and every tool at school with his popped coller got one.
Its not bad bad thing, because they sold a crapload of those bands, and the money was there. However, it became a more about the fashion, than the cause.
Just liek folding is becoming. More and more cardmakers are advertising PPD! What the hell? Why does that matter. Charity profit should never cross this. I could see a review company, such as TPU, Tom's Hardware, or others puttin gout the numbers, but the companies themselves should not use charity for their own greed and gain.
I disagree. Greed is the best motivator, why not harness it for good? The more douchebags with popped collars folding the better. Its not the "Charitable Cause" that matters its the end result. When there is a publicly available cure for alzheimer's what does it matter that an Nvidia box says "can fold at blah PPD," << thats amazing. Since i want to donate to folding, I want a card that has a high PPD. Part of the fun is that it combines benchmarking with charity. Just my 2 cents tho. (I was the douchebag that bought the HP livestrong laptop lol. It broke tho since it was a cheap Hp. )
When greed leads to overclocks, overclocks lead to instability, and instability leads to invalid research, the whole system is self-destructing. If the end result is "discovered" due to a glitch someone's overclocked computer introduced, it is an invalid result. Pande Group is doing nothing to ensure valid results.
I bought an 8800 GT because my X800 XL was doing a pathetic job at NFS:MW, not because it's better at folding. I'll say it right now: if you bought hardware specifically to fold, the F@H project has failed.
I mean, look at super computers. They don't have 50,000 of the best processors around--they have 50,000 processors that have the best bang for the buck (usually a generation or two old). It is volumes that count, not individual performance. All they stress now is individual performance and not volumes. Maybe they always did, I don't know but where it became obvious to me that their objectives are in the wrong place is when they released GPU clients which intentional devalued all contributions from CPU users. They made it abundantly clear that they don't want the reliable results of the millions of CPUs out there.
Benchmarking and research should never be complimentary. They have two different, opposing objectives. Benchmarking is the hare (fast and potentionally runs in the wrong direction); research is the tortoise (slow and steady). All good things take time.
the new 9800gtx+ is up folding now guys. whatch me climb back into the top 10( i hope )
now to go take some picks of the 4870 and get that dog sold.
Benchmarking is incentive. Incentive works, expecially for charitable organizations. Would you rather buy from a bake sale or just give your money away? And the benchmark is more suited for how many points you've folded rather than how fast ur system is.
Dont get me wrong, there are going to be knuckleheads overclocking their cards to the point of instability, but there are workarounds that from the other end. If you have 50,000 processors you better believe that at least some of them will be defective/ return invalid results. Every supercomputer has error checking built in, and so should F@H. Even if no one overclocked, from 16,000 people you would still get defective cards that artifact at their stock clocks (ahem Diamond). So the whole argument that overclocking will lead to instability that destroys the system is flawed. Every system has error, its up to the system to control it - overclocking is irrelevant.
And yes. If you bought an 8800GT just for folding then folding is a huge success, and whatever data the researchers need will be available to them sooner. Thats a win for F@H in my book.
Let me give you a better analogy: I am like a Dunkin' Dounuts manager. I made 1,000 dounuts today but only sold 800. Company policy mandates all dounuts be fresh so I must either throw those 200 dounuts away or donate them.
Similarly, I have a 2 x Xeon 5310 system that is using all of 1% CPU cycle for non-intensive but mandatory tasks. Folding @ Home was a way for me to put those 99% unused clock cycles to good use. I donated those clocks to the project. Now, it is abundantly clear they don't care about those donated clocks so instead, I'll save the additional expenditure of power consumed and heat produced by not running the computer at full bore 24/7. It's their loss, anyway.
Benchmarks are standardized tests that do not change under any circumstances. They are intended for the sole purpose of comparing system A to system B. There is nothing "benchmark" about the F@H clients. They're just looking for a reason to produce invalid results.
If it were up to me, I'd have every WU be processed no less than twice on unique hardware. If they don't produce the same result, a third will be brought in to verify the results. If three computers attempted to solve it and failed, the WU is rejected to be addressed by a research team to figure out what is going wrong.
F@H supposedly does have error checking but it sounds to me like it is extremely weak. It in no way addressed common binary-switch errors.
I'm not going to argue specifically about overclocking because it is just an example of a cause. The effect is what is important to me and that is producing invalid data that is treated as sound research.
And an epic fail for science. See how F@H is more and more business like? All they care about is the publicity, and therefore profits, they produce. They don't care about the science anymore.
are you cereal?
Great Forum and Folding Team you have here.
aka: The Commander
(1) Dell XPS 720 2.4 Quad Core, GeForce 8800GTX 2-SMP-MPICH 1-GPU2
(20) Dell Vostro 400's 2.4 Quad GeForce 8800GT 40-SMP-MPICH 20-GPU2
(3) Dell Vostro 410's 2.4 Quad GeForce 8800GT 6-SMP-MPICH 3-GPU2
No doubt that science is our goal.
But without competition dadi_oh, there's not motivation and folding will become BORING.
Just like athletes going to the Olympics, we are looking for gold.
Every announcement of a new member, member upgrading hardware and getting more POINTS - that's where "science" get extra points too.
NO MOTIVATION, NO NEW CHALLENGES - NO FUTURE.
I didn't forget about what you said "Catch me if you can" (check first quote). At that time I had 21.000 Points against 400.000+ you had.
I guess you brought the competition at that time.
All I can say to you is THANK YOU VERY MUCH
That is awesome I was truly hoping you would catch me. It is all for a great cause. I've only got so much to invest and I am really grateful that others have more to contribute. If we can urge others on in a good natured way then it is win-win.
Separate names with a comma.