Folding Forum

I have two computers folding, both are stable. One is a dedicated Fodling@Home box, the other is a multi-use model W520 Lenovo laptop. The laptop, which is an i7-2760QM based CPU at 2.4 GHz (stock speed) running only in SMP mode for slot, takes 11+ hours to complete each of these projects(it has done more than eight, that is an average). The other box, an i7-2600k box at 3.4 GHz stock speed, takes about 7+ hours for same project(6+ of them, again an average). Folding treats both i7's as 8-core computers for SMP.

The laptop gets about 3500 points per 7904 WU with the bonus, and the other box gets about 4000 points per 7904 for this project. In the timeframes concerned, both boxes earn much more per hour with other WU projects.

As perspective, the 2600k box got a project 10085 done in about 3 hours earlier today, and earned 4723 points for that project.

Commenting on the difference between WUs using same core, same box for the 10085 vs 7904 compare. Max bonus and perhaps base points might want to be increased for the 7904's to compensate for hugely more time of computing them per 7904 WU. Just a strong suggestion

Hello jdii1215, welcome to the forum.

Let's collect input from a few more people before making any assumptions. For example, how do you know that the points for Project 7904 are too low? Maybe Project 10085 points are too high?

Two things to add: for folding I am now Straight_Man on Team 93. Second, in about 2 months these computers have earned over 1.3 million points(folding started on one computer Feb 4, 2012, then later I added the laptop to compare them-- added it about 2 weeks ago).

Added to reply.... For most projects, other than 7904's I get an average of over 1k per hour. Feel free to gather data. I wanted to bring this up for consideration and am not demanding. On days when I get 2 7904'3s my production drops to below 1k per hour. See EOC stats, the days when my points were unusually low were on days I got 7904's (one to two). Been tracking this for weeks now, not one to jump to conclusions.

Welcome to foldingforum.org, jdii1215.

P7904 is a bonus WU, which means the quicker that you return it, the more points you'll earn. How many hours elapsed from when you downloaded it until the results were uploaded?

You said you are "... now Straight_Man" which probably means you changed your UserName. That means you'll have to complete 10 bonus WUs with that name before your passkey is qualified.

Are you folding it as a SMP assignment or as a Uniprocessor assignment?

Every WU I have gotten as Straight_Man has been a bonus after the first ten. I understood that.

I folded as jdii1215 in 2008-2009. had boht computers die due to overheating down here in Florida, USA.

Have been only SMP folding, no GPU heads to fold with per se-- the Intel CPUs have built-in video which is fine for me as I do not game. SMP on the Public Beta client, and earlier betas starting with 7.1.34, without specing CPUs, yeilds np 8 call to client on each of the computers. Without 2-3 7904s per day across both boxes, my PPD combined for the two boxes is 33-50k depending on how well the WUs are strategized toward SMP, and not GPU folding. With 2-3 7904s, I get 18-30k per day depending on other WUs combined for both boxes. the 2600k box folds in Windows, and has Windows 7 Pro 64 bit on it. The laptop has the same OS on it, literally Windows 7 Pro 64 bit. The laptop CAN dual boot into Fedora Core 16.

OT aside: I started using consumer computers with the Timex Sinclair 1000-- the old Sinclair ZX-81 ala Timex. It kept locking up while I was typing papers on it for college back in the 1980s-- until I opened up the box and stuck a Z80b CPU in it that I persuaded Tech Data to sell me quantity one of at some huge price.

I have hung out on Icrontic for 7 years plus now, and hung out mostly lurking and learning on the old folding-forum site. I am an A+ Computer Tech, per CompTIA. Excuse me, I run on in the wee hours.

Some projects for uniprocessors are QRB projects, but they have relatively long deadlines consistent with a single processor. They can also be assigned to multi-cored CPUs as SMP projects. but without the same bonus structure as those projects with SMP deadlines. I don't understand it well enough to explain it very well.

Ok, so the 7904 and 7903 are long-deadline and intended for Uniprocessors which means each folding processor in my two i-7s is in emulation mode. The older uniprocessors use older graphics gen modes , to be backwards compatible the SMP processors EMULATE those earlier forms of graphics calc-- which is inefficient compared to what they are designed to do now with modern graphics modes they run natively and thus very effectively. That is point number one.

Folks who fold will be discouraged if their machines do not fold well, and stop folding or if real motivated build more advanced boxes(I have personal reasons to fold offset that tendency to stop folding to a large degree). You folks at folding want to have more folks folding, not less as a trend, right??

So, why discourage them by giving them real inefficient WUs compared to what their boxes can do at best (and prove they can do by performing work for folding)??? You may say you need the data, but folks who compete(competition is endemic in Folding) in folding and build newer and newer tech boxes to fold faster FOR FOLDING and more points do not want old-style, tiny point per hour WUs mixed in with highly effective in PPH WUs to a large degree, or to have radically ineffective in PPH terms WUs tie up their boxes for huge amounts of time (like less points per hour by a factor of two to three compared to WUs which are very effective in their computers).

Note that minimum points per folding hour is in essence folding's goal apparently, but if work gets done faster then more data trends become obvious sooner so folding benefits if WUs are more tuned to boxes.

Some fixing suggestions-- for now, given more and more SMP computers fold and will fold faster if they get WUs that are natively calculable for each of them, give more points to SMP boxes if they are fed and complete Uniprocessor WUs by increasing points to SMP boxes or base and bonus points to all boxes until such time as number of cores per folding machine and time per WU from download per machine to completion are data-mineable. Strong suggestion that will patch for a while, like perhaps several years.

Trying to help with this thread, NOT criticize, but this kind of situation will rapidly get worse rather than better as more and more SMP boxes come online and more and more Uniprocessor boxes are retired-- unless something is done.

I don't understand your concern about graphics emulation. FAH's software consists of several parts. The calculations are done by one of several FahCores, some of which use a GPU, others which use one processor, and others which are multithreaded. Those FahCores do not interact with the screen graphics or with the internet. A separate component, FAH's client, uploads results, downloads new assignments, invokes a FahCore, and displays the status messages on the screen. Uniprocessor cores, SMP cores, and GPU cores use no graphics, whatsoever, so there is no possible way to use graphics emulation.

If you're genuinely concerned about graphics efficiency, avoid using the FAH Viewer. It uses considerable graphics resources to display images of proteins. Those images can be very impressive, but displaying them is completely optional. The folding calculations whether they're done by a single processor, a SMP collection of processors or a GPU proceed whether you display the results or not. In fact, they're calculated faster when you're not using resources to display them. Nevertheless, the viewer images are probably more useful in attracting new folks to FAH than a few extra points. What is considered "folding well" does depend on the point of view of the Donor.

I have started using new hardware, namely FX-8120 oc'd to 3898mhz. But my PPD with project 7903 is only 254.21. My old dual core AMD 5600+ got around 1k ppd.
I mean seriously what the hell is up with that?

I got WU 7646 and 7647 on 2 different folding comps and a quick glance shows both machines have lost about 2k ppd with these 2+ day folding time units.
I'm not really complaining, but right now our group is having a contest and it is putting a hurt on my scores.
Right now 1 of the comps that regularly displays 19-20k ppd is down to mid 17k ppd, not near the other atm but I noticed it had a big drop when I looked at it about 6 hours ago.

I guess we will see if the ppd increases when the WU's get close to being done.

dev_enjoys wrote:I have started using new hardware, namely FX-8120 oc'd to 3898mhz. But my PPD with project 7903 is only 254.21. My old dual core AMD 5600+ got around 1k ppd.
I mean seriously what the hell is up with that?

Well a reduction from 20k to 17k is still ok compared to 250 ppd on a 8 core machine, while previously a dual core was giving 1k+ ppd easily!

Also getting a terrible points hit on 7904, 18-22k down to 6-8k with this wu.

Also around 10k ppd on a 2600k clocked @4.7ghz, thats good huh!!

OC'ed Q6600 has never had less than 5.5K ppd now getting <4K on a 7905. Been running for over five years so seems unlikely all the other projects were over-rewarding. This consistent under-rewarding of new units is getting very tired. I don't mind that new faster hardware gets more points, but I do object when existing hardware starts getting lower points.

This project has been running for more than a year by this announcement... http://foldingforum.org/viewtopic.php?f=24&t=19421 It may be a little late to get anything changed, but I let the researcher know...

Hi All,

Thanks for bringing this to our attention. We are looking into these units to ensure they are consistent with current benchmarking practices. We will post here and in the project thread with any updates. Please note this might not mean a change in points -- we will check that nothing went wrong in setting the points for these WUs, and if an error is discovered we'll fix it. Let me know if anyone has any questions!

Thanks,

TJ