FAH Benchmark App

M

musky

[H]ard|DCer of the Year 2012
Joined
Dec 14, 2009
Messages
3,154
Let's try to keep results for various machines in their own threads and just keep feedack on the utility in this thread. This is fairly crude, but it does the job pretty easily and pretty well. The units are just randon units a few of us have captured. The F@H client and A3 core will be downloaded from Stanford when you run the app for the first time.

If anyone has the knowledge and motivation to convert this to a pretty GUI, please PM me to discuss it.

MIBW - it now shows all frames run all the time per your request
10e - your 48 core monster will still break it. You will still need to modify the client.cfg for the -smp 48 flag. Everyone else, as long as you are at or under 32 cores, you will be fine.

http://www.avid-edge.com/fah/FAHbench_v12.zip

Folding@Home Benchmarking Application

HOW TO START:
Double-click on the FAHbench_v11.bat file to start the benchmark. You will be prompted for a
unit to run (6061, 6701, 2684, 2685, 2686, or 6900) and for a number of frames to run (1-9.)
The batch will then spawn a new command window to run a Folding@Home instance and minimize it.
The original window will monitor this process and give you basic information about the run.
The first time you run the batch, you will be prompted to connect to Stanford. This is normal
and is needed to download the A3 core and assign your machine an ID. Subsequent runs will not
need to connect. Once you have completed the number of frames specified, the batch will close
the Folding@Home process and show you the average frame time for the unit you selected.

Alternatively, you can pass in the unit and number of frames as parameters to the batch file.
Either make a shortcut to the batch file and add the parameters there or add them when calling
the batch file from a command window.

Example: FAHbench.bat 2686 5

HOW TO STOP
Everything will stop once you complete the frames specified. If you need to stop before then,
Control-C will stop the batch file. You will also need to manually stop the F@H client that will
be minimized in your tray, also with Control-C. Any needed clean-up will happen the next time you
run the benchmark utility.

RESULTS
The last frame time will display in the monitoring window. The total run time and average frame time
will display once the number of frames you specifiy have been completed. This info is also saved in
a text file in \data\log\log_<unit #>_<date>_<time>.txt.



Q&A
1. Q: Can I finish folding these work units for credit?
A: These units are way past their due date. They have already been completed and turned in
to Stanford. Please do not try to do this. It will harm the project and gain you nothing

2. Q: Can I run the bigadv units on my dual core/quad core/AMD X6 machine that can't run bigadv normally?
A: Yes, you can. The frames will take a long time to run, which is why these types of machines
are not supposed to run bigadv work units. For benchmarking purposes, you will be fine.

3. Q: How can I use the frame times I see in the benchmark utility to calculate potential points per day?
A: http://linuxforge.net/bonuscalc2.php


REFERENCE RESULTS
**Coming Soon**


BUG REPORTS
If you run into any issues, please send a private message to musky at http://hardforum.com or via email
([email protected]).
 
Wahoo. I encourage everyone to run this - if you want to know where you are really at compared to others, this will let you compare apples to apples.

Please post your results in Jebos spreadsheet.

Just a note - the work units are not necessarily what you will see on average. The 6900 is pretty fast compared to the average for example. But the point is really to see a direct folding comparison to another machine.

I would suggest running 4 frames minimum. The 6900 and 6061 are a good place to start. The 2686 unit is the same one I have been using and testing with for months (As Troy Maclure would say: You might remember it from such Threads as the SR2 Optimisation Thread)

BIG thanks to Musky for the hard work to make this a reality.:D

EDIT: - tried on my painfully slow ancient crappy SR2#1 and the new frame times come up perfectly. Awesome! I will test on SR2#3 to see if they eat any CPU. I did get a "Ready to connect to network to download" FAH dialog that I don't remember getting before, but who cares, it downloaded them perfectly.

original.jpg


original.jpg


EDIT 2: ran 4 frame test on v1.12 on SR2C, followed by 4 frame test of the 2686 without your app running as a control

Frame 1: 10 minutes, 59 seconds
Frame 2: 11 minutes, 0 seconds
Frame 3: 11 minutes, 2 seconds
Frame 4: 10 minutes, 54 seconds
Unit: 2686
Frames: 4
Total run time: 43 minutes, 55 seconds
Average frame time: 10 minutes, 58 seconds

4 frame manual test
10:57
10:59
11:00
10:58
average 10:58.

So the good news is your app is still undetectable in terms of CPU.:D Now all I have to do is work out why this rig has lost 8 seconds of speed in the last week!:eek:
 
Last edited:
10e - your 48 core monster will still break it.
Click to expand...

All I can think of is the scene in Grandma's Boy when Jeff beats the high score in Dance Dance Revolution and yells, "Did I break it?" What's high score mean.

I will give this a try with my X6. Thanks.
 
Just started a 5 frame 2686 run on my E5530s @ Stock....should be a good baseline.
 
I added p6900 results from normal folding, but I just ran it again and running a 2684 to see how fast this baby smashes it.

So I did the following:

1) Made a temporary folding directory in C:\FAH_SMP5
2) Copied the client.cfg file to it
3) Ran the [email protected] executable with the -configonly switch
4) Hit return a bunch of times until I got to the "extra parameters" section and changed -smp to -smp 48
5) Backed up the client.cfg in the FAHBench directory with a .BAK1 extension and copied the modded client.cfg back
6) Ran the FAHBench batch file in a DOS prompt with "run as administrator" privileges
7) Set it to run 5 frames of 2684.


So it came back with a "% was unexpected at this time" error in the original DOS box, as the "spawned" DOS box didn't spawn, BUT a FAHLOG.txt file is being created AND the proper WU is running (2684, R4, C24, G24) this time, and the CPUs are loaded at 100%.

If this works I'll be happy because I've been dying/crying to run a 2684 on this slutsky! Crossing fingers now!

It's working! Fingers STILL CROSSED.
 
Last edited:
10e - I'll PM you a link for a client.cfg file and some instructions this evening.
 
musky said:
10e - I'll PM you a link for a client.cfg file and some instructions this evening.
Click to expand...

Cool thanks musky.

UPDATE!

The box running 48 cores did the first frame of the 2684 at 14:05 and the second frame at 14:01.

This means it can finish any current BigAdv unit in 24 hours or less. JOY!

I'll wait until all 5 frames are done prior to posting the final tally on the spreadsheet.

YayMD
 
ok, musky disregard that IRC message :D

ill give this a go tonight and wet your appities with what a Thuban can do :D .... truth be told i want to know just to know :D
 
I'll run the bench tomorrow after my current WU drops. I know this will be very helpful to get the last 5% of juice out of my system.
Thanks for all the effort!
 
Ok, successfully installed and run on all 3 rigs now. I had one instance of a crash when starting again, so I tried again for a while just starting and cancelling it to see if I could reproduce, but could not in 30 tries. I haven't tried to run it from a command window like you suggested. (sorry, I know that would help but I am simply too lazy a typist to run it that way all the time ;) )

But this is probably a one in 50 event, so may not be worth fixing.

If I wanted to run a batch of say:

FAHbench_v12.bat 6061 4
FAHbench_v12.bat 2686 4

What lines in your lovely script do I add a : in front of to disable the confirmation/press enter part without killing it dead?

How long since I have publicly thanked you for making this? Too long... THANK YOU. :D
 
may i be worthy said:
Ok, successfully installed and run on all 3 rigs now. I had one instance of a crash when starting again, so I tried again for a while just starting and cancelling it to see if I could reproduce, but could not in 30 tries. I haven't tried to run it from a command window like you suggested. (sorry, I know that would help but I am simply too lazy a typist to run it that way all the time ;) )

But this is probably a one in 50 event, so may not be worth fixing.

If I wanted to run a batch of say:

FAHbench_v12.bat 6061 4
FAHbench_v12.bat 2686 4

What lines in your lovely script do I add a : in front of to disable the confirmation/press enter part without killing it dead?

How long since I have publicly thanked you for making this? Too long... THANK YOU. :D
Click to expand...

You want two colons in front of the line to comment it

line 66
::set /p unit=Going to run %unit% benchmark, %numfrm% frames. Press enter to continue (1 to cancel):

You will also want to break the validation part at the end and force a restart by commenting out lines 205 and 206
::set /p again=Press enter to exit, 1 to run again :
::if %again% neq 1 goto :exit

I think that will work for you.
 
just tested it out. runs pretty damn well. will have to run it later when im not running any programs using my cpu.
 
Benchmark tool worked great with my X6 today (4 frames a piece for each WU). It looks like my X6 can do a 2684 just under the deadline which makes me happy. Really curious to see what W.Feather gets on his.
 
yep those x6's are beasts. now if only they had triple channel DDR3 they would probably give the i7's a run for their money.
 
just ran the 6061 bench no gpu's and scored 2k PPD higher without the GPU's. didnt think i lost that much PPD from them. but at least they are worth 5300 PPD each so its not to bad i guess.
 
great app Musky. thanks for putting it together.

I hope some review sites will get their hands on this and integrate it into their CPU reviews.

I would love it if you would design a GPU app too. If you need a GPU to test with I can probably scrape one up for you.
 
opticalserenity said:
Any chance we'll see this for OS X? There are some of us with access to some strong OS X machines.
Click to expand...

Unfortunately, I know nothing about OS X, let alone scripting for it. I'll probably port this over to Linux eventually once bigadv moves there and I have some Linux boxes to test on.
 
musky said:
Unfortunately, I know nothing about OS X, let alone scripting for it. I'll probably port this over to Linux eventually once bigadv moves there and I have some Linux boxes to test on.
Click to expand...

depending on how complex the batch file is.. I may port it over to shell for linux...
I have no osx machines so no way to see if the shell script would work for them....
 
New core just released - so how do we bench it? :)

EDIT dropped in new core b4, renamed FAH6.34-win32-SMP.exe to FAHbenchmarkApp.exe, deleted core A3, but it just downloads another a3...

EDIT - stuff it - updated all machines, folding live units.
 
Last edited:
mmmmmdonuts said:
Any chance we can get this thread stickied so we can get more results for the database:)
Click to expand...


lol thats like asking for gold to come out of the sky.. nothing ever gets stickied when its requested.. been asking to have the TPF database stickied since day one of it being posted..


jebo_4jc said:
great app Musky. thanks for putting it together.

I hope some review sites will get their hands on this and integrate it into their CPU reviews.

I would love it if you would design a GPU app too. If you need a GPU to test with I can probably scrape one up for you.
Click to expand...

i'm wondering if we should make a request to kyle or one of the other cpu reviewers to see if they would be willing to try it out on a couple reviews. especially when the bulldozer and ivy bridge come out.
 
may i be worthy said:
New core just released - so how do we bench it? :)

EDIT dropped in new core b4, renamed FAH6.34-win32-SMP.exe to FAHbenchmarkApp.exe, deleted core A3, but it just downloads another a3...

EDIT - stuff it - updated all machines, folding live units.
Click to expand...

Unless I misunderstood the FF post, the new client wil run either A3 or A5 cores, and the WUs themselves determine which one. Capture a 6901 and we will see.
 
musky said:
Unless I misunderstood the FF post, the new client wil run either A3 or A5 cores, and the WUs themselves determine which one. Capture a 6901 and we will see.
Click to expand...

Done. I am uploading to dropbox now, will try a cleaned up version first with just the *.dat files at 31mb, if that doesn't work I can do the whole folder with all temp files at about 220mb.... links to follow in PM.
 
I did get the benchmark app to run it, but it wasn't pretty. I still need to do some clean-up before i re-release the next version.

BTW, I should be able to give you some 6901 numbers for a C2D @ 2.5 GHz - in couple days... :)

Edit: I forgot to kill the client on the C2D - for those curious, frame times on a 6901 were 2 hours, 19 minutes. A C2Q at 3 or 3.5 GHz would probably make the deadline on 6901 units.
 
Last edited:
jebo_4jc said:
Musky, is this currently hosted anywhere?
Click to expand...

Doubtful...I haven't looked at this forever. I can move it somewhere accessible if you want. I thought Haitch's version was being maintained, though.
 
musky said:
Doubtful...I haven't looked at this forever. I can move it somewhere accessible if you want. I thought Haitch's version was being maintained, though.
Click to expand...

His version can't do a 3 frame benchmark at this point. It only does live monitoring.
 
You must log in or register to reply here.
Back
Top