Page 1 of 1

Project 7623, CheY points WAY TOO LOW!

Posted: Fri Feb 13, 2015 5:47 am
by Geoffric
According to the detai; page: "CheY has been studied as part of the cannon of simple proteins. However, it is much larger than proteins previously simulated on FAH, and therefore provides a test of the current limits of FAHs computational power."

I am running 960s and 970s non clocked but they are the super factory overclocked versions. With the 960s I get many jobs worth about 25 - 28 K that I finish in 4.5 to 5.5 hours. Same thing with the 970 but my folding time is in the 3.75 - 5 hour range for a 25 - 28 k project.

On the Project 7623 projects I successfully complete them with either card, and with no errors BUT at a significantly lower point per hour score. The 970 completes this job in 5 hours roughly, the 960s in about 7 hours BUT this project is only bringing me 14000 - 14500 points per completed work unit. I feel that this is a very significant reduction in points, and I am seeing it only on the CheY project. I have recently purchased 2 new Vishera 9590s, (8 cores@5.1 GHz), a 970, a 960, and I just traded my 760s in for 960s, so I expect to get 2 more 960s up within 10 days. I also have 2 1090T cores, and an 1100T that I will be running with older cards: 660Ti and a couple Radeon R7 260s. I am VERY impressed with the 960s, and bang for the buck wise I certainly wont buy another 970 or 980 and plan on getting 3 more 960s to run on the older AMD 6 cores.

Are there any other folders seeing a significant pointe per hour loss on the CheY project? Is it considered cancer, Parkinson's, Alzheimer's etc? and if the point loss continues what folding group preference would I want to avoid in order to avoid this project with the 970? I'll still fold it with the 960s, but with tax, shipping, card cost and interest this card was just under $400 and for some reason its as slow as the 960s on this project. It is for this reason I would like to avoid this project with the 970.

Thanks for all the hints and tips. I just hit 10,000,000 but nowadays that is no big deal. I am hoping to hit 500,000 PPD once I get the other 2 960s back from the step up program and set up.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Fri Feb 13, 2015 6:48 am
by jimerickson
if it is core_15 there is no qrb.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Fri Feb 13, 2015 11:52 pm
by codysluder
jimerickson wrote:if it is core_15 there is no qrb.
Project 7623 says it is OPENMMGPU which is Core_15, so there is no Quick Return Bonus.

Re: Project 7620, CheY points WAY TOO LOW!

Posted: Sun Feb 15, 2015 3:07 pm
by Geoffric
Now I am on 7620, and its a 1.5 day project for only 14093 points, running my evga 970.

This card has been averaging at least 100 k ppd, so I don't understand this core 15, but will read up on it thanks.

Just out of curiosity, why is there no QRB.?

Now I am running 9201 which is cheY protein in description BUT I am getting 35K for 3 hours work with a 970.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sun Feb 15, 2015 3:08 pm
by davidcoton
QRB was introduced after Core15, and never applied to Core15. It's unlikely now since Core15 is announced as EOL.

It does make it difficult to compare points to see how much science a card is doing, because the way in which points reflect science value is different for Core15 compared with Core17 and Core18.

9201 is Core17 (IIRC) so does get QRB, which accounts for the difference in points. You can work out a PPD from the "Base credit" shown in Advanced Control and the time taken for a WU -- this will be comparable between Core15 and other cores.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sat Feb 21, 2015 4:28 pm
by Geoffric
Well for some reason, all 3 of my marines are folding only this work order, and its 7 hours work for 14000, before I was getting 4 hours for 30 k. I cant understand WHY all my machines are getting this but they are 3 identical machines I guess. I have went from 600 - 800 k a day to a now miserable 150 k the last 2 days. I just invested $750 in 3 new strong GPUs and after 1 good week all of a sudden I am making half of what I was on the old, crappy cards,

Very disappointing, almost as bad as the big adv scandal, I'm not smart enough to work on this stuff first hand, all I can do is buy the parts and build the machine. Every time I invest money, then something changes so I am doing worse than before.

Its almost scandalous that my $400 investment into a 970 is getting me 60% of the points I was getting with my 760.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sat Feb 21, 2015 4:34 pm
by jimerickson
run linux. voila! no more core 15.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sat Feb 21, 2015 5:34 pm
by davidcoton
It's just a matter of how points are allocated. Because the Quick Return Bonus has never been implemented for Core15, you see fewer points. In this case, points do not accurately reflect the scientific value of the work. To do that you need to look at the base points for the newer (Core17 and Core18) WUs and work out the PPD on base points alone. That will give a fair comparison with Core15 projects, and you should find it is nearer in line. It is not that these are projects with low scientific value, it's just an anomaly in the points.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sat Feb 21, 2015 6:02 pm
by Joe_H
You have not posted any information as to your settings, or your log file which would show those settings. So why you are getting Core_15 lately can not be determined. However, as has been posted here before, until nVidia gets its implementation of OpenCL fixed up in the drivers for Maxwell based video cards there is a limited number of projects and WU's that will run on them. So the Assignment Server will give your machines work based on the settings you have used and from what is available at the time they connect to the AS.

Directions for posting your log file can be found here - viewtopic.php?f=19&t=26036.
jimerickson wrote:run linux. voila! no more core 15.
And no Core_18 yet either. If WU's that will run on the Maxwell GPU's happen to be out when your system connects to the AS, then your machine will have to wait until some are available and make no points from the GPU at all during that time.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sat Feb 21, 2015 7:02 pm
by SwingOp3
The simple and unfortunate truth is that there appears to be a shortage of Core17 work units for Windows as my 3 Maxwells have been on a steady diet of 90% Core15 for the last couple of days. I've even received a Core15-P8018 this morning on a GTX970 for a whopping 33,500 ppd.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sun Feb 22, 2015 5:22 am
by PS3EdOlkkola
Similar situation - just one Core 17 wu across four 980's in the last 3 days.

Re: Project 7623, CheY points WAY TOO LOW!

Posted: Sun Feb 22, 2015 11:31 am
by Geoffric
Yes I have been around for maybe 3 years but was inactive the last couple. I am on disability and I build another machine when someone in my family passes (if I can afford it) and the last 3 have almost the same CPU, motherboard, and GPUs.

I was Running GTX 760s, but I did try a 960 when they came out about 3 weeks ago and the score with that card was just phenomenal, so I used EVGA's "step up" program to turn in 2 760s in exchange for 2 960s for only a $10 step up fee and a $9 shipping fee (The step up program is why I buy ONLY EVGA cards now.) Having see such a big difference between the 760s and 960s I took a big leap of faith and bought a single 970 that with shipping, tax, and the card was almost $400. I was kind of expecting the same "step up" with the 970 being better than the 960s but actually only saw about a 20% point difference so bang for buck wise I maybe would have been better buying 2 more 960s instead.

I know how to build different computer rigs by reading on line a lot and using tools like pcpartpicker.com to make sure all my stuff is compatible, but many of the overclocking and fine tuning others seem to do has been hard for me, as I have some pretty big holes and gaps in my knowledge base as I am self taught, and geographically isolated from other enthusiasts. I had one of the first home computers on the market back in the late 70s, and learned z-80 assembly language, but I don't know any of the new stuff and need to get back to school to fill in those holes.

I have figured out that even though I can very easily get a 10 - 15% OC on my CPU and GPU, and it works fine on most games and stress testers, when folding my OC'd GPUs get errors halfway through and I have decided to err on the side of caution. Hopefully as the GTX 960s seem to be very well received there are going to be a lot of you OC gurus playing with the 960s now, and hopefully I can get some settings from you guys and try applying those settings to my GPUs here.

By next Friday I will have a 970, 3 960s, 2 650Ti's, and 2 GTX 460 Hawk Talon Attacks. Even though the 460s are my oldest cards, I kind of keep them for sentimental reasons as they were on my first charity folding rig. All my builds have been AMD except for one Intel E8600 core duo on an older LGA 775.

I am looking for a Radeon MSI R7 260 so I can hook it up with another one I have here as I have never tried a crossfire rig and would like to do one just because I haven't done it yet. Anybody with a used R7 260 that's upgrading to the R9 I would consider buying your used call please drop me a note here or at jeffmartis@yahoo.com.

Thanks for all the help,