TPU's WCG/BOINC Team

[Nordic]

Does anyone on linux mint want to do an experiment? I want you to run a linux benchmark, update your kernal, and run the benchmark again. Shouldn't be too difficult. It does not have to be mint, but I think users running mint might see a large change in performance since the kernal is so old.
This is all apart of some tweaks I am trying to get more performance out of linux. I forgot to do a before everything test, to compare with an after test. So I would really like to see others results. Also, the more results from more people will show if updating really makes a difference. I have done some other tests, but coming from an earlier version of 3.19 gave me a 13% boost.

It should only take 10 minutes of your time, well, not counting the 2 hours waiting for the before and after benchmark to run.

Spoiler: Instructions

This is for intel cpu's and X64 bit systems only!!
Linux mint uses an old kernal of 3.13. You can check by using the command in a terminal.

uname -a

This will download the benchmark, and then run it. 4 lines. The benchmark took almost an hour to run on my system. If you just want to copy, paste, and have linux run all of it at once, look at the spoiler right below the code.

wget http://byte-unixbench.googlecode.com/files/unixbench-5.1.3.tgztar xvf UnixBench-5.1.3.tgz

tar xvf UnixBench-5.1.3.tgz

cd unixbench-5.1.3

./Run

Spoiler

wget http://byte-unixbench.googlecode.com/files/unixbench-5.1.3.tgztar xvf UnixBench-5.1.3.tgz && tar xvf UnixBench-5.1.3.tgz && cd unixbench-5.1.3 && ./Run

The benchmark should give you two scores. A single threaded and a multi threaded. Some example scores from my 2500k system are 2190 single threaded and 5990 multi threaded.

This will update your kernal to 3.19.3-k17-generic. Then reboot your computer. If you just want to copy, paste, and have linux run all of it at once, look at the spoiler right below the code.

wget http://multiclangaming.com/files/linux-headers-3.19.3-k17-generic_3.19.3-k17-generic-10.00.Custom_amd64.deb

wget http://multiclangaming.com/files/linux-image-3.19.3-k17-generic_3.19.3-k17-generic-10.00.Custom_amd64.deb

sudo dpkg -i *.deb

Sudo reboot

Spoiler

wget http://multiclangaming.com/files/linux-headers-3.19.3-k17-generic_3.19.3-k17-generic-10.00.Custom_amd64.deb && wget http://multiclangaming.com/files/linux-image-3.19.3-k17-generic_3.19.3-k17-generic-10.00.Custom_amd64.deb && sudo dpkg -i *.deb && sudo reboot

Then check if the kernal update worked by typing in a terminal:

uname -a

You should see 3.19.3-k17-generic.

Then update your system.

sudo apt-get update && sudo apt-get upgrade

If the update was a success, run the benchmark again.

cd unixbench-5.1.3

./Run

Spoiler

cd unixbench-5.1.3 && ./Run

I don't know if anyone else was interested in trying this too, but my 2500k saw improvements of 13% coming from a recent kernal. My g1620 celeron saw improvements of 50% coming from the really old kernal mint includes. A friend saw improvements of 29% coming from a recent kernal. These are not gains in boinc, but with the benchmark Unixbench.
Remember, the kernal I give instuctions for is for Intel cpu's with 64 bit linux installs.

 

[thebluebumblebee]

I don't know if anyone else was interested in trying this

I'M WORKING ON IT! Finally moving some things around and will be firing up a 2600K, maybe even yet tonight.

 

[Nordic]

Sorry bluebumblebee, I am just impatient!

Out of all my optimizations, that one seems to have the biggest effect so far. It is hard to really know because I am only doing a benchmark, not actually running boinc.

A lot of the optimizations I am going to do, I can't even benchmark for. I have to spool up. TO BAD IT IS A STINKIN HOT EL NINO YEAR.

 

[thebluebumblebee]

not actually running boinc.

I'll try to remember to run the BOINC benchmark before and afterwards. Not sure how to see the results for that though.

 

[Nordic]

I am not sure, it actually shows results does it? If so please do tell. That would save a lot of time for everyone.
EDIT: IT does show benchmark results! Wish I had know this. You can see them in the boinc event viewer in both windows and linux.

Also, instead of making a guide for these optimizations I was thinking of making a "Cruncher's linux tips and tricks" thread (name pending) where we all can share information on making linux crunching faster. That way, it would encourage more to participate, and could be used for more than just my supposed guide. It would also allow me to do more of a running guide, updating as a go so others can test things out with me.
Does that sound like a good idea?


This is another thing I want to try. http://www.pperry.f2s.com/boinc-compile.htm
It is old, from 2004. They saw some substantial gains. Today we shouldn't see no where near that much, but maybe something still. The question is how much.

 

[thebluebumblebee]

I am not sure, it actually shows results does it?

It shows it in the event log. I just ran it on my 3770K and got: 3778 floating point MIPS (Whetstone) per CPU and 9911 integer MIPS (Dhrystone) per CPU. How that relates to what @TRWOV did a while back (link) I'm not sure - actually I think it's the same thing, just has the decimal in a different place.

Does that sound like a good idea?

Yep.

 

[Nordic]

The unixbench I have been using, and requested in the instructions does include a whetstone and drystone. So the total score it comes up with might truely be useful.

 

[thebluebumblebee]

I can not get unixbench to run. First, your info as to what to type in is messed up. Once I finally got the program downloaded and extracted, it won't run, just keeps giving me an error.

gcc -o ./pgms/arithoh -DTIME -Wall -pedantic -ansi -O2 -fomit-frame-pointer -fforce-addr -ffast-math -Wall -Darithoh ./src/arith.c
./src/arith.c:32:19: fatal error: stdio.h: No such file or directory
#include <stdio.h>
                   ^
compilation terminated.
make: *** [pgms/arithoh] Error 1
Checking distribution of files
./pgms  exists
./src  exists
./testdir  exists
./tmp  exists
./results  exists
gcc -o ./pgms/arithoh -DTIME -Wall -pedantic -ansi -O2 -fomit-frame-pointer -fforce-addr -ffast-math -Wall -Darithoh ./src/arith.c
./src/arith.c:32:19: fatal error: stdio.h: No such file or directory
#include <stdio.h>
                   ^
compilation terminated.
make: *** [pgms/arithoh] Error 1

**********************************************
Run: "make all" failed; aborting

 

[Nordic]

I did edit the code once after I first pasted it. What is in the spoiler now, works for me. My ubuntu server did not, have issues running unixbench, but mint did. I resolved it with:
apt-get install gcc make –y
apt-get install build-essential

 

[thebluebumblebee]

I did edit the code once after I first pasted it. What is in the spoiler now, works for me. My ubuntu server did not, have issues running unixbench, but mint did. I resolved it with:
apt-get install gcc make –y
apt-get install build-essential

Took me a while to remember that I needed to use sudo before those commands.
The second one works just fine, but the first gives me this:

sudo apt-get install gcc make –y
Reading package lists... Done
Building dependency tree     
Reading state information... Done
E: Unable to locate package –y

I don't think I've ever updated this Mint install. Could that be an issue? I didn't want to do it in case it updates the core.

BTW, the BOINC event log also shows the core version.

Do you really need results from Unixbench? Will the BOINC benchmarks be enough?
Initial results: 2600K on a just RMA'd motherboard with a 500GB 2.5" thin HDD
3.13.0-24 Generic
3467 MIPS Whetstone - BOINC
12635 MIPS DHrystone - BOINC
Unixbench: 1161.3/4457.7

Spoiler: log

Benchmark Run: Tue Aug 11 2015 08:57:55 - 09:26:21
8 CPUs in system; running 1 parallel copy of tests

Dhrystone 2 using register variables 41555569.5 lps (10.0 s, 7 samples)
Double-Precision Whetstone 4197.1 MWIPS (11.8 s, 7 samples)
Execl Throughput 1769.6 lps (30.0 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 691423.9 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 182032.8 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 1999326.2 KBps (30.0 s, 2 samples)
Pipe Throughput 961555.0 lps (10.0 s, 7 samples)
Pipe-based Context Switching 147115.6 lps (10.0 s, 7 samples)
Process Creation 10659.4 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 5628.2 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 3382.1 lpm (60.0 s, 2 samples)
System Call Overhead 677614.6 lps (10.0 s, 7 samples)

System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 41555569.5 3560.9
Double-Precision Whetstone 55.0 4197.1 763.1
Execl Throughput 43.0 1769.6 411.5
File Copy 1024 bufsize 2000 maxblocks 3960.0 691423.9 1746.0
File Copy 256 bufsize 500 maxblocks 1655.0 182032.8 1099.9
File Copy 4096 bufsize 8000 maxblocks 5800.0 1999326.2 3447.1
Pipe Throughput 12440.0 961555.0 773.0
Pipe-based Context Switching 4000.0 147115.6 367.8
Process Creation 126.0 10659.4 846.0
Shell Scripts (1 concurrent) 42.4 5628.2 1327.4
Shell Scripts (8 concurrent) 6.0 3382.1 5636.8
System Call Overhead 15000.0 677614.6 451.7
========
System Benchmarks Index Score 1161.3

------------------------------------------------------------------------
Benchmark Run: Tue Aug 11 2015 09:26:21 - 09:54:38
8 CPUs in system; running 8 parallel copies of tests

Dhrystone 2 using register variables 162615450.5 lps (10.0 s, 7 samples)
Double-Precision Whetstone 30007.2 MWIPS (10.0 s, 7 samples)
Execl Throughput 17748.4 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 1067336.3 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 274118.6 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 3351747.9 KBps (30.0 s, 2 samples)
Pipe Throughput 4739817.4 lps (10.0 s, 7 samples)
Pipe-based Context Switching 1135794.3 lps (10.0 s, 7 samples)
Process Creation 48007.8 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 36300.8 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 5363.3 lpm (60.0 s, 2 samples)
System Call Overhead 3615769.7 lps (10.0 s, 7 samples)

System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 162615450.5 13934.5
Double-Precision Whetstone 55.0 30007.2 5455.9
Execl Throughput 43.0 17748.4 4127.5
File Copy 1024 bufsize 2000 maxblocks 3960.0 1067336.3 2695.3
File Copy 256 bufsize 500 maxblocks 1655.0 274118.6 1656.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 3351747.9 5778.9
Pipe Throughput 12440.0 4739817.4 3810.1
Pipe-based Context Switching 4000.0 1135794.3 2839.5
Process Creation 126.0 48007.8 3810.1
Shell Scripts (1 concurrent) 42.4 36300.8 8561.5
Shell Scripts (8 concurrent) 6.0 5363.3 8938.8
System Call Overhead 15000.0 3615769.7 2410.5
========
System Benchmarks Index Score 4457.7

I'm using a stock HSF and saw temps as high as 98 degrees. Not sure if I need to change that and start again.

 

[CAPSLOCKSTUCK]

Thanks for your concern buddies.
Ive been rough..Brain and spinal scan 08.00 Sunday morning. I actually cried when they told me. Ive waited so long.

Im fine though. Im strong.
Hugs to all
Ger

 

[manofthem]

Thanks for your concern buddies.
Ive been rough..Brain and spinal scan 08.00 Sunday morning. I actually cried when they told me. Ive waited so long.

Im fine though. Im strong.
Hugs to all
Ger

Welcome back! So very glad to hear buddy!

I actually started a pm to you earlier today but got tied up at work and never finished

 

[VulkanBros]

Thanks for your concern buddies.
Ive been rough..Brain and spinal scan 08.00 Sunday morning. I actually cried when they told me. Ive waited so long.

Im fine though. Im strong.
Hugs to all
Ger

Brain and spinal scan...sounds serious...or? Hopefully not

 

[Nordic]

Took me a while to remember that I needed to use sudo before those commands.
The second one works just fine, but the first gives me this:

sudo apt-get install gcc make –y
Reading package lists... Done
Building dependency tree   
Reading state information... Done
E: Unable to locate package –y

I don't think I've ever updated this Mint install. Could that be an issue? I didn't want to do it in case it updates the core.

BTW, the BOINC event log also shows the core version.

Do you really need results from Unixbench? Will the BOINC benchmarks be enough?
Initial results: 2600K on a just RMA'd motherboard with a 500GB 2.5" thin HDD
3.13.0-24 Generic
3467 MIPS Whetstone - BOINC
12635 MIPS DHrystone - BOINC
Unixbench: 1161.3/4457.7

Spoiler: log

System Benchmarks Index Score 1161.3
System Benchmarks Index Score 4457.7

I'm using a stock HSF and saw temps as high as 98 degrees. Not sure if I need to change that and start again.

If it works, it works and you don't need the other command that did not work.

I have been using unixbench, which does the same tests boinc and more. I did not know I could actually bench with boinc or else I would have done a before and after with it.
These are my scores in unixbench on my 2500k.

I don't need the whole log. At the end you will see two numbers (1161.3/4457.7). Were those before or after the kernel update? Your numbers look really low for what I would expect from a 2600k. My 2500k is at 4ghz though, and had a substantially newer kernel to begin with.

 

[thebluebumblebee]

Were those before or after the kernel update? Your numbers look really low for what I would expect from a 2600k.

I listed the kernel used, 3.13.0-24 Generic, which is what comes with Mint 17. I too was surprised by how low those numbers are. That install originally had a 3770K and was my best performer. Maybe there's something wrong with the RMA motherboard. I haven't moved on with the kernel update because I'm trying to figure out what I want to do next.

 

[Nordic]

I listed the kernel used, 3.13.0-24 Generic, which is what comes with Mint 17. I too was surprised by how low those numbers are. That install originally had a 3770K and was my best performer. Maybe there's something wrong with the RMA motherboard. I haven't moved on with the kernel update because I'm trying to figure out what I want to do next.

If you update the kernal, which is very quick and easy you should see at least a 10% boost as I did with my 2500k. My g1620 celeron saw a 50% boost overall.

 

[mstenholm]

I'm sure other people have noticed this but I don't recall reading anything here about it. Since we can't get statistics by project and device except by looking at results status on the WCG site I can't say this definitely but it seems to me that the OET project only awards about 10-12 points per hour of work while UGM seems to award between 20 and 25. Any idea why there is such a disparity or am I just not looking carefully enough?

I know that hours spent is only one factor in determining points but I'm seeing this difference on the same machines pretty much across the board.

The points for OET has been (still is?) all over the place. My 4790 some days got 50-60 points/hour and my i7 970 got 10-20 for WUs in the same serie. I have seen points as low as 0,8/hour and as high as 65. My normal average points for these two systems are 40 and 24. Sitting in Koh Lanta waiting for the rain to stop

 

[Arjai]

Well, The Mad Cruncher is having some issues.

The Corsair that was in there? Bad, wouldn't post on either stick.

The new Mushkins, 1 gig-ers, are fine, as expected. The Crucial, 512's, passed the POST test but, I was getting some errors in the memory test, I can only assume it is the older set of Crusial's.

Meanwhile, I tried installing Mint, 17 Rafaela and 12 Lisa, but it kept hanging, during the "Configuring Hardware" step of the install.

I'm wondering about two things. One, it has a GE 5500 AGP GPU with 256MB! 128megs I still think it's bad, the upper screen edge was always a bit funky Could not even tell what was up there!. Two, it has that old Crucial.

I know it will run with the two gig Mushkins, so I should just dump the Crucials, right?

The problem with the 5500, well, I have a GT 6800 256MB Dual DVI, in working condition, in This Ole Comp.

Catch is, my monitor is a single DVI input, single D-Sub. I guess I will give this a switch out and just deal with it. I'll install Teamviewer, and eventually get a better wifi solution, from my 3G Hotspot.

That is, if I can manage to get an OS to run on it!! I even tried Ubuntu12.04 and 14.04, all the way back to 10!!

OK, on with the assembly and disassembly and plug in and try again!!

All for the Love of WCG...I need a new hobby.

P.S. @stinger608 , didn't you say you had a bike you were gonna photograph, for me? I'd like to see it, and then bring them to a local resto shop, see what they can do for me.

 

[stinger608]

P.S. @stinger608 , didn't you say you had a bike you were gonna photograph, for me? I'd like to see it, and then bring them to a local resto shop, see what they can do for me.

Ah yes. Thanks for reminding me bro. I'll get some pictures this week and post em up

 

[Arjai]

Well, The Mad Cruncher is having some issues.

The Corsair that was in there? Bad, wouldn't post on either stick.

The new Mushkins, 1 gig-ers, are fine, as expected. The Crucial, 512's, passed the POST test but, I was getting some errors in the memory test, I can only assume it is the older set of Crusial's.

Meanwhile, I tried installing Mint, 17 Rafaela and 12 Lisa, but it kept hanging, during the "Configuring Hardware" step of the install.

I'm wondering about two things. One, it has a GE 5500 AGP GPU with 256MB! 128megs I still think it's bad, the upper screen edge was always a bit funky Could not even tell what was up there!. Two, it has that old Crucial.

I know it will run with the two gig Mushkins, so I should just dump the Crucials, right?

The problem with the 5500, well, I have a GT 6800 256MB Dual DVI, in working condition, in This Ole Comp.

Catch is, my monitor is a single DVI input, single D-Sub. I guess I will give this a switch out and just deal with it. I'll install Teamviewer, and eventually get a better wifi solution, from my 3G Hotspot.

That is, if I can manage to get an OS to run on it!! I even tried Ubuntu12.04 and 14.04, all the way back to 10!!

OK, on with the assembly and disassembly and plug in and try again!!

All for the Love of WCG...I need a new hobby.

P.S. @stinger608 , didn't you say you had a bike you were gonna photograph, for me? I'd like to see it, and then bring them to a local resto shop, see what they can do for me.

I do not get it. Once again, after removing suspect Memory sticks AND replacing GPU, I get hung up in the same place!!

Always, 3/4 of the way through "Install", it goes to 'Configuring System.' Green bar goes to 3/4's across....and Hangs.

I am now stuck. I repasted the CPU, Athlon 64x2 3800 from 2001.
Put NEW pair of 1GBx2 Mushkin Silverline DDR.
Switched out a GeForce FX5500 for an AMD GT6800.
The Hard drive is a fresh RMA from WD.

This system ran fine until the HDD took a dump, I got the RMA from @Norton some time ago, before my jaunt camping on the river. The PSU has not given me anything scary, it's old but seems OK. It is a 220W from Enermax?( I think, seen it so many times lately ) MSI board, solid, no blown caps....Bios loads fine, I get the POST beep and can load up Mint UI but can't get through the loading.

I need help with this one!! I hate seeing this dual core sit here unused!! Someday, I will laugh at this crap as I chuck it off the roof! Until then, I wanna use it.

Any suggestions, welcomed. EDIT: Please reply to THIS POST in THIS Forum v
http://www.techpowerup.com/forums/t...-discussion-thread.214610/page-2#post-3329709

Thanks, Sorry for the mix-up in posting places, won't happen again. I PROMISE!!

 

[t_ski]

I rebooted one of the rigs today for Windows Updates but I forgot to restart BOINC That rig has been doing 5-6K PPD and only gets 1500 for the day

The good news is that Windows 10 seems to have increased the PPD by about 30% on my two rigs (over Win7). Either that or WCG is just handing out points like they're candy...

 

[Arjai]

I rebooted one of the rigs today for Windows Updates but I forgot to restart BOINC That rig has been doing 5-6K PPD and only gets 1500 for the day

The good news is that Windows 10 seems to have increased the PPD by about 30% on my two rigs (over Win7). Either that or WCG is just handing out points like they're candy...

Really?

You're the first person I've seen claim a jump in Crunching points. How many days have you been 10-ing it, and are you sure that it's that big of a jump?

 

[Caring1]

I've been running W10 on my laptop for a couple of weeks at least, and can't see any noticeable gain from it, looking at my device profile on WCG shows too much variance over the last month to even try to work it out.
One day it returns 4 results, the next can be 19, I suppose it depends on the W.U.s

 

[t_ski]

Really?

You're the first person I've seen claim a jump in Crunching points. How many days have you been 10-ing it, and are you sure that it's that big of a jump?

Been about a week and a half for one rig, and since Sunday on the other. First rig was getting about 4k PPD consistently, now is getting 5-6K PPD. Other rig was getting about 12-14K PPD, now is getting 18-19K PPD.

Again, these were on Win7 before, which I've read is not as good at crunching as Win8. Win10 is supposed to be optimized even better than 8.

Anyone else got before and after numbers? I could be just smoking crack...

 

[Arjai]

Been about a week and a half for one rig, and since Sunday on the other. First rig was getting about 4k PPD consistently, now is getting 5-6K PPD. Other rig was getting about 12-14K PPD, now is getting 18-19K PPD.

Again, these were on Win7 before, which I've read is not as good at crunching as Win8. Win10 is supposed to be optimized even better than 8.

Anyone else got before and after numbers? I could be just smoking crack...

IDK, I might switch "Karen" over, this weekend. I believe it will mostly benefit my HDD space, and if it boosts my crunching...Why not.

My biggest fear is having WIN10 mess with my 7850 and the Catalyst I have, finally gotten to work. But, it might also be good enough to allow me OpenCL, which I've been unable to get working in WIN7, which will boost my Folding.

Any troubles with GPU Drivers? Any ATI's in the two rigs you switched? Got any Heroine?