TPU's F@H Team

[Nordic]

TPF=Time Per Frame - given in the client

I am looking at my two 750 ti's right now, and it says 5 minutes 27 seconds. Is that a performance metric of some sort?

 

[thebluebumblebee]

Just how long it takes to finish a frame, which is 1%

 

[Nordic]

Just how long it takes to finish a frame, which is 1%

That makes sense knowing what a frame is now.

I also just moved both 750 ti's to one computer, and now under my name and not kreij. TPF is also now 4.5 minutes. I think it was still estimating.

 

[BUCK NASTY]

That makes sense knowing what a frame is now.

I also just moved both 750 ti's to one computer, and now under my name and not kreij. TPF is also now 4.5 minutes. I think it was still estimating.

Work unit needs to run for 3 frames(or 3 percentage points) to get a accurate measure of PPD. What work unit are you folding?

 

[Steevo]

Whats up with the unit 13000, perhaps its just slow on this 5870, had a 1901 that was fast.

Also got the 7970 to have under water, so it should be folding by the new year.

 

[Nordic]

Work unit needs to run for 3 frames(or 3 percentage points) to get a accurate measure of PPD. What work unit are you folding?

I was at about 2% when I made that post. I am folding FahCore 0x17 project 9201. Estimated ppd for both gpu's is at 141,00 at the moment. About 70k ppd each.

 

[flmatter]

nice bump in ppd for both cards running at sametime! Great many thanks to everyone!! Thank You!!

 

[Steevo]

Got the CPU upgrades too, should have at least one of them done up by the first of the year, see if the PSU will handle it and the 5870.


I will have a 5870 with a full coverage block to give away for the cost of shipping, and a couple water blocks to fit other cards is anyone is interested. Dunno if I should put this here or in the BST thread.

 

[Nordic]

Turns out I have more free time and access to a computer than I thought I would. I am currently making the charts as I said I would. I need some help though.

What would be the best way to get a realistic completed PPD for a given WU?
For the fields in the chart would the following be worthwhile: |Run, Clone, Gen|FahCore|Peak Output: Single Precision (GFlops)|Yield Factor (smaller is better) ?

I don't even know what run, clone, and gen are.

Do you guys think I should require a screenshot with gpuz and the FAH client for verification. I think it would be nice to have verification but it is more work. On the other hand, these are tpu specific results, its for science not a competition, so we don't really need to verify anything since we all trust each other right?

Here is the first draft of the thread. Take a look, tell me what you think.

 

[mstenholm]

There was a time when early and later run of 13000´s mattered. I now in the believe that other than boost clock (maybe base and boost clock) only CPU clock matters. Driver version and so on is OK, not all watch the FAH forum. The yield factor.......forget it

Mine: old and optimal drivers 11.8 and 344.16 Hhttps://docs.google.com/spreadsheets/d/1W-PlatA9CgOOTA3qcvTxhFm_YoedvoHmXHNgZqczY4g/edit#gid=0ere

 

[mx500torid]

gflops and yield factor are not shown on advanced control. Nice chart there ms I think thats the format we should use.

 

[BUCK NASTY]

gflops and yield factor are not shown on advanced control. Nice chart there ms I think thats the format we should use.

Agreed. Good chart, but I was wondering how to extrapolate GFLOPS and Yeild as well.

 

[mstenholm]

gflops and yield factor are not shown on advanced control. Nice chart there ms I think thats the format we should use.

But including driver version. It might matter, some time does.

 

[Nordic]

All the fields used currently I took from the overclock.net chart or I added. I would also rather err on the side of too much information. I do admit some look unnecessary so that is why I want your opinions.

 

[mx500torid]

Grandson knocked 1 of my folders off the desk and its broke. 2600k, 290x, and 7770 is of unknown condition at the moment. Ill look at it tomorrow, kinda too pissed right now.

 

[[Ion]]

Grandson knocked 1 of my folders off the desk and its broke. 2600k, 290x, and 7770 is of unknown condition at the moment. Ill look at it tomorrow, kinda too pissed right now.

I sure hope it's mostly OK still!

 

[stinger608]

Grandson knocked 1 of my folders off the desk and its broke. 2600k, 290x, and 7770 is of unknown condition at the moment. Ill look at it tomorrow, kinda too pissed right now.

Oh man, that's a huge bummer man!!!
That is exactly why my grandchildren are not allowed in the computer room.

 

[Norton]

Grandson knocked 1 of my folders off the desk and its broke. 2600k, 290x, and 7770 is of unknown condition at the moment. Ill look at it tomorrow, kinda too pissed right now.

Had a similar experience when my son knocked over the 4P a few weeks ago Hope everything is ok with it [crosses][/fingers]

 

[Nordic]

There was a time when early and later run of 13000´s mattered. I now in the believe that other than boost clock (maybe base and boost clock) only CPU clock matters. Driver version and so on is OK, not all watch the FAH forum. The yield factor.......forget it

Mine: old and optimal drivers 11.8 and 344.16 Hhttps://docs.google.com/spreadsheets/d/1W-PlatA9CgOOTA3qcvTxhFm_YoedvoHmXHNgZqczY4g/edit#gid=0ere

gflops and yield factor are not shown on advanced control. Nice chart there ms I think thats the format we should use.

Agreed. Good chart, but I was wondering how to extrapolate GFLOPS and Yeild as well.

Grandson knocked 1 of my folders off the desk and its broke. 2600k, 290x, and 7770 is of unknown condition at the moment. Ill look at it tomorrow, kinda too pissed right now.

So I will removed GFLOPS and Yield. They were only included because of overclock.net includes them in their chart. Do you guys want to include the date too? I figured if we included the driver version that would approximate time frame well enough. It appears run, clone, and gen is valuable information?
If we are all in agreement here, then I think the fields will be finalized then.

Do you guys think screenshot verification would be necessary. Such as a screenshot including FAH client information such as ppd and gpuz for other information.
I think it would be nice to have verification but it is more work. On the other hand, these are tpu specific results, its for science not a competition, so we don't really need to verify anything since we all trust each other right?

That only leaves how to consistently measure for realistic ppd. I could ask to wait till it is at <95% under what the owner thinks are good running conditions, such as it was not interrupted or what not, and use the estimated ppd. Or can we do better than that. Bluebummblebee had some ideas.

That's a tough one. The surest way is to go to one of the stat sites and write down what the final points given was. (that sounds bad, but I have a cold and my brain's not working right) HFM.net might also collect this info in its benchmark data. BTW, you can take the points given for a WU and reverse "engineer" it with this tool to get the TPF. http://www.linuxforge.net/bonuscalc2.php?gpu=1

 

[mx500torid]

Screenshot verification is a necessity. GPUZ, the folding@home client control for expert users. As far as I have seen 13000 is the only one that pumps on me. I would think the screenie would be posted in the thread so if any questions it would be easier to answer.

BTW The only thing thats toast in my accident is the case. So I need a roomy case with good cooling. The 290x is folding but not enough cooling for the 7770 to go in there. Pretty happy now the damage could have been much worse.

 

[Nordic]

Made minor adjustments to charts thread with regards to feedback. Now I just need to make directions for how to post. It looks pretty good right now I think.

 

[BUCK NASTY]

I'm working on having the 4x GTX 970 rig up and folding by the 2nd weekend of January. Guess what Mobo will power it??.....None other than the venerable MSI K9A2 Platinum that I bought in 2008. I still have this board, although it's been Frankenstien'd over the years. Didn't want to spend $100+ on another 4x PCI-e 8/16 board if this one still works. Hopefully it cranks up and shows some signs of stability. Gonna run a Phenom x4 9850BE in it. I hope I still got some DDR2 ram laying around .

 

[mx500torid]

Buck Nasty

mx500torid

 

[Steevo]

I'm working on having the 4x GTX 970 rig up and folding by the 2nd weekend of January. Guess what Mobo will power it??.....None other than the venerable MSI K9A2 Platinum that I bought in 2008. I still have this board, although it's been Frankenstien'd over the years. Didn't want to spend $100+ on another 4x PCI-e 8/16 board if this one still works. Hopefully it cranks up and shows some signs of stability. Gonna run a Phenom x4 9850BE in it. I hope I still got some DDR2 ram laying around .

What I read was is you are planning on heating your house with 140W heater that also allows some folding to happen.

 

[BUCK NASTY]

LOL, you crack me up mx500. Just look at those guns pushing that boat sideways thru the water. WOW.

What I read was is you are planning on heating your house with 140W heater that also allows some folding to happen.

Yes, but If I have issues with the 9850BE, then I will substitute a PII x4 B93 for the 50w heat/energy savings.