unfortunately no, i686 did not work for me on K6-2.

I'd be interested to know if my generic FFTW client runs on your K6-2 ...

just started it up, no errors using setiathome-4.07-fftw.i686-pc-linux-gnu ~ yay. i'll let it run a while and note what happens. this is a k62-300. i am also running a P2-300 and an ultrasparc 333 so i can compare any improvements with these similar speed grades.

So Metod's generic i686 client runs on an AMD K6-2 but ours doesn't. Strange.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Please let us know how the i686 seti client compares to other clients you've run.

Ned

Maybe this has been discussed somewhere already :-/
Do you prefer a test result using the reference wu from BOINC download site or is it sufficient for you when I give results from real wu?

rattelschneck

Yes, like Chris said, if you want to run the reference WU on the original berkeley client and my new client, that would be great, but don't worry if you don't as Metod has run some tests on intel processors and posted the times on his website already.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Some status information from me so far:

I'm running the boinc_4.19.3_i686-pc-linux-gnu client along with setiathome_4.07.3-i686-fftw3-static-pc-linux-gnu client successful on these platforms now:

Athlon Thunderbird 900 MHz
Pentium II 450 XEON (Deschutes) 512K cache
Pentium III 550 XEON (Katmai) 2M cache
Pentium III 450 (Katmai)

I'm running the boinc_4.19.3_athlon-xp-pc-linux-gnu client along with setiathome_4.07.3-athlon-xp-fftw3-static-pc-linux-gnu client successful on this box:

Athlon XP 3200+

The speed of the new 4.19.3 client is pretty amazing, this box used to run with Windows XP and standard BOINC client 4.25 before and performed like this.
Now, it's running Linux and the 4.19.3 clients and it performs like this.

It's running much faster now, as you can see :))
I'm really, really impressed.
Thanks so much again, Ned and Chris!

P.S.
I will run the new clients on my stack system as soon as possible. There are still some old K6-2 and K6-3 included. But I will wait for the next _regular_ weekend outage, cause i don't like to reboot the whole stack (but I have to, for running kernel 2.6) which has a uptime of 94 days so far. :-/

rattelschneck

unfortunately no, i686 did not work for me on K6-2.

I'd be interested to know if my generic FFTW client runs on your K6-2 ...

just started it up, no errors using setiathome-4.07-fftw.i686-pc-linux-gnu ~ yay. i'll let it run a while and note what happens. this is a k62-300. i am also running a P2-300 and an ultrasparc 333 so i can compare any improvements with these similar speed grades.

So Metod's generic i686 client runs on an AMD K6-2 but ours doesn't. Strange.

Ned

it looks that way, i don't quite understand. different GCC version or flags? libraries maybe?

It's running much faster now, as you can see :))
I'm really, really impressed.
Thanks so much again, Ned and Chris!

P.S.
I will run the new clients on my stack system as soon as possible. There are still some old K6-2 and K6-3 included. But I will wait for the next _regular_ weekend outage, cause i don't like to reboot the whole stack (but I have to, for running kernel 2.6) which has a uptime of 94 days so far. :-/

rattelschneck

You're welcome :)

BTW, spacemeat has reported the i686 client doesn't work on k6-2 for some reason (I would be nice if you could confirm this). He said that Metod's generic client will run on these older procesors though :)

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

You're welcome :)

BTW, spacemeat has reported the i686 client doesn't work on k6-2 for some reason (I would be nice if you could confirm this). He said that Metod's generic client will run on these older procesors though :)

Ned

the first WU was completed with Metod's client and it's not saying much yet-- no noticeable improvement, but i will give it a few more. this one may have taken over mid-crunch from the previous 4.07 client i made. i can only do about one per day here so ... patience

BTW, spacemeat has reported the i686 client doesn't work on k6-2 for some reason (I would be nice if you could confirm this). He said that Metod's generic client will run on these older procesors though :)

Ned,

your i686 client doesn't run on my K6-2 box neither.
The error message is:

SIGILL: illegal instruction
Exiting...

regards
rattelschneck

I found that my optimised client just speeded up the benchmarks and did not change the workunit times. Just resulted in lower credits :(

Good fun compiling all the code up though :)

I found that my optimised client just speeded up the benchmarks and did not change the workunit times. Just resulted in lower credits :(

The BOINC core client does not not do any WU processing. It does manage the benchmarks. If you want to improve your WU processing times, you'll have to optimize the specific project application you're running.

---

I found that my optimised client just speeded up the benchmarks and did not change the workunit times. Just resulted in lower credits :(

The BOINC core client does not not do any WU processing. It does manage the benchmarks. If you want to improve your WU processing times, you'll have to optimize the specific project application you're running.

Yep, Recompiled the SETI client and took almost precisely the same time as existing client (within 1 sec of original timing)

---

The more things change, the more they stay the same.

Yep, Recompiled the SETI client and took almost precisely the same time as existing client (within 1 sec of original timing)

My original compiles were like that. Then I found -ffast-math and the fftw3 source code made big increases. Everything else in minor in comparison :)

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

Some statistics for an AMP XP2800+ running kernel 2.6.11:

client average CPU time
4.02 11,703
4.07.02 8,775 -25%
4.07.03 7,145 -39%

The averages are over 7 WU's, except for 4.07.02 which is 80 something. I have left out short WU's (

---

Some statistics for an AMP XP2800+ running kernel 2.6.11:

client average CPU time
4.02 11,703
4.07.02 8,775 -25%
4.07.03 7,145 -39%

The averages are over 7 WU's, except for 4.07.02 which is 80 something. I have left out short WU's (

Thanks,

On the reference benchmark WU we are seeing percentage improvements in the high 50's on an Athlon XP and low to mid 60's on AMD64 processors.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***