TPU's F@H Team

[Nordic]

Looking at the screenshot, it has all sorts of extra information if someone wants to know. Memory usage, gpu load, volts used in overclock if overclocked etc. The lack of feedback makes me think you guys approve. Still looking for feedback though, check it out here. Maybe specifically about the cpu. Seems people think it is unnecessary. Maybe I will just remove it.

I think the only thing holding me back from opening it up for submissions is understanding of run, clone, and gen. This link here describes in the technical details. I know enough biology and math to understand what they mean. I need to know where they are shown in the client. At the top of the client I did see a PRCG. Where the R, C, and G seem to follow the format of run, gen, clone. Here is a screenshot of my client. Are those the correct run, clone, and gen numbers?


Edit: Should I take all the thanks I am getting on various posts about this as positive feedback? I just want to make sure its done right from the beginning.

 

[mx500torid]

You are correct James. this is a screenie from hfm where it shows it. An OLD screenie.

 

[mx500torid]

newer version

 

[mstenholm]

Run, generation ect WAS important at a point for 13000, as stated some pages back. I don't care if you uses CPU info. I know it is has an impact. Core 17 and 18 uses 100% CPU all the time. Why shouldn't is have an impact? Driver vision is always nice but since I do monitor other fora you can for all that I care leave it out.

Edit: some projects, 1300X comes into mind, has very different 1 % times during the WU. If the 90 % elapsed time is used as suggested (and not the real calculated value) it should be stated. The 90 % time has that advantage that you don't have to calculate TPF and it doesn't take upload time into account but it will not give you the real PPD. How many have set next_unit_percentage to 100 %? Not many, so here you lose 2-22 minutes (1 % of your processing time, a lot more PPD wise). Should be stated as well if you really want to go into details.

 

[Nordic]

Run, generation ect WAS important at a point for 13000, as stated some pages back. I don't care if you uses CPU info. I know it is has an impact. Core 17 and 18 uses 100% CPU all the time. Why shouldn't is have an impact? Driver vision is always nice but since I do monitor other fora you can for all that I care leave it out.

Edit: some projects, 1300X comes into mind, has very different 1 % times during the WU. If the 90 % elapsed time is used as suggested (and not the real calculated value) it should be stated. The 90 % time has that advantage that you don't have to calculate TPF and it doesn't take upload time into account but it will not give you the real PPD. How many have set next_unit_percentage to 100 %? Not many, so here you lose 2-22 minutes (1 % of your processing time, a lot more PPD wise). Should be stated as well if you really want to go into details.

Those are very good points. I think I will keep cpu and driver version then.

Yes, I guess currently I am using estimated ppd. If I am to continue with that method I certainly should mention that. I chose this method because I did not know of an easy way to calculate real ppd. Do you have any suggestions on this?

 

[mstenholm]

Those are very good points. I think I will keep cpu and driver version then.

Yes, I guess currently I am using estimated ppd. If I am to continue with that method I certainly should mention that. I chose this method because I did not know of an easy way to calculate real ppd. Do you have any suggestions on this?

Yes, the easy one that BBB suggested some post back- http://www.linuxforge.net/bonuscalc2.php.
The problem is that all the new projects (91xx) are not in that...

 

[Nordic]

Yes, the easy one that BBB suggested some post back- http://www.linuxforge.net/bonuscalc2.php.
The problem is that all the new projects (91xx) are not in that...

I did check that out so I guess I did know of a way to calculate real ppd. At the time it did not have the project I was looking for. The only way I know to see TPF is on the folding client, which says "Estimated TPF" so this calculator would still be an estimate.

I propose: I will communicate that the numbers given are estimated ppd and why we are using it. I would also provide the link to that calculator with instructions on how to use it so people can input information into that calculator if they want to. Hopefully they will update it.

The purpose of these charts is to give tpu folders the most realistic expectations of ppd. PPD is not a static number. These charts will have enough information inputted to for the tpu folder to get the best estimate possible.



On a separate topic. I am folding on both 750ti's and my 970 right now. My ppd is spiking upward. If I could get a good string of core 17 wu's I may end up having some very nice ppd.

 

[Ahhzz]

.....

Edit: now at 598K PPD. C'mon 600K!

View attachment 61078

 

[BUCK NASTY]

View attachment 61245

It's actually hovers between 601K and 611K now after upping some clocks.

 

[mx500torid]

Kakaostats shut down today

Im gonna miss them.

 

[BUCK NASTY]

Kakaostats shut down today

View attachment 61256

Im gonna miss them.

Yep, gonna miss them as well. They were my "go to" stats when I started folding, but their up-time has been sporadic over the past year. Much appreciate of the work he did. Good thing we still have Xtreme and Free-DC.

 

[mx500torid]

Ive done some experimenting with my stuff and my GTX 680 folds the same ppd on a c2d @ 2.8 GHZ as it did on a 2600k @ 4.4 GHZ. I tried a 290x @ 3.5 GHZ on a 9201 and then on another 9201 @ 4.4 GHZ and results are virtually the same. I didnt see anything on the forums so I ask on there and will report back when I have an answer. Also on my rigs the cpu will spike at 100% when the WU starts then will drop never to see that 100% until it starts another WU.

This is a screenie of a GTX 780 folding, crunching and I have Everquest 2 running in the background.

 

[mstenholm]

11 threads of WGC and one WU on my 970. 100 % CPU load from start to finish. I wonder why there is this difference between my rigs and yours.

http://img.techpowerup.org/150102/100CPUfolding.jpg[/IMG]']

 

[mx500torid]

They said the same thing i am seeing that 100% when starting a WU then it lowers. Hopefully they will answer my question and maybe we will know. I checked all my rigs, 5 in all, and none are at 100%. You are certainly showing 100% though.

 

[mx500torid]

 

[mx500torid]

I got an answer on the cpu speed thing. https://foldingforum.org/viewtopic.php?f=61&t=27209&sid=3bbf5a172fe5a43df160de490479f9c5

 

[mstenholm]

I got an answer on the cpu speed thing. https://foldingforum.org/viewtopic.php?f=61&t=27209&sid=3bbf5a172fe5a43df160de490479f9c5

I will prove them wrong with two identical GPUs running at same speed but with different CPUs. I don't have my old data for my 7970s but going from 4 GHz to stock 2.67 GHz lost me 5k PPD on the 1300xs.

 

[mstenholm]

4 GHz i7 970 and 4.4 GHz 4790 both running with a 480 GB SSD. I bumped the GPU clock a bit higher on slower CPU just to make a point.

 

[mstenholm]

 

[Steevo]

My i7 machine was stalled between WU's since yesterday sometime for some reason. Back in the saddle now though.

 

[NastyHabits]

On my low power ITX rig - AM1 with a AMD 5350 (Kabini) CPU and a GTX 750 Ti - I only use 25% of the CPU. When I fold on my i3 PC its CPU usage is slightly less. Around 20-21%. Speed of the CPU has no effect. A faster CPU might use a little less CPU time. What does make a difference is what else is going on. A lean and clean OS installation can make a difference with a weak CPU. On my old PC with an E8400 dual core, killing every process and service that wasn't necessary yielded about a 5% increase in PPD with my GTX 660 Ti. I found that overclocking the E8400 had no measurable effect.

 

[Nordic]

It seems that with a strong gpu like a 970 that cpu does matter a bit. It matters so little though I don't see any reason I should keep it in the charts though since I seemed to be the biggest advocate for it.

 

[hat]

Monday I'm getting some new parts (including a new motherboard). I tried folding on my 660ti, but it didn't go well (random bsod). I can't seem to run 3DMARK on it either. I suspect the reason for my problems might be because I have such an old motherboard right now, running PCI-E 1.0/1.1/1.0a (still not sure what the hell it is). The new board supports PCI-E 2.0 I believe so maybe the card will work better.

 

[mstenholm]

Good news https://foldingforum.org/viewtopic.php?f=74&t=27208
No 91xx and 1047x for pre-Maxwell. I did notice a good steady stream of 9201 today. Bad news I guess if 9201 dries out but there must be plenty since even my AMD's are getting them.

Edit: core_15 is still around. Just got one. I know I shouldn't have posted just before finishing a nice 9201

 

[NastyHabits]

Good news https://foldingforum.org/viewtopic.php?f=74&t=27208
No 91xx and 1047x for pre-Maxwell. I did notice a good steady stream of 9201 today. Bad news I guess if 9201 dries out but there must be plenty since even my AMD's are getting them.

Edit: core_15 is still around. Just got one. I know I shouldn't have posted just before finishing a nice 9201

Thank goodness! My 750 Ti gets 50K+ PPD on core 17's, and around 20K PPD on core 18's. Notice the nice goose egg below.

Core 18's totally negate my purpose in building my low cost, lower power box. I built it to produce slightly less PPD than my old setup, but be able to fold 24/7/365 without sending my electric bill through the roof. I could only fold for 2 weeks a month on the old box. Core 18's set me back 180000 PPD per month instead of gaining 660000.