BOINC benchmark: post results

[[XC] serlv]

Q6600 @ 3266 MHz ~ XP Pro x64 Edition SP2 ~ BOINC v5.10.45 64-bit

3105 floating point MIPS (Whetstone) per CPU
9326 integer MIPS (Dhrystone) per CPU

[D_A]

This is what we are talking about.

D_A, you wrote XP home every time XP x64 comes up, might wanna fix that
Since there is only XP x64 Professional

Also, what is up with that Desktop/Server classification? Any Xeon CPU is a server CPU, Sossaman as well as the Clovers, Harpers and Opterons. However, Q6600 should be Desktop.

The desktop/server refers to the OS not the chip. For example: Ubuntu 8.04 is available in both desktop and server configurations, as are many other Linux variants.

[D_A]

Where is this spreadsheet/list? Is there a URL for it, maybe?
Can it be put on the Google app? Would that do us any good?

"The List" HERE

It's a spreadsheet in .xls (MS Excel) format.

I don't care what people do with it but I maintain it from my home system, updating with info I get from more than one forum.

[D_A]

Ahh. TY, jcool.

Here is another result...

Q9550 ~ XP 64-bit SP2 ~ 3926.1 MHz (8.5 x 461.9 MHz)

10/25/2008 5:39:30 PM|| 3988 floating point MIPS (Whetstone) per CPU
10/25/2008 5:39:30 PM|| 11540 integer MIPS (Dhrystone) per CPU

... and which version of BOINC are you running on this one?

[[XC] serlv]

Sorry,

5.10.45 ~ 64-bit

[[XC] serlv]

OK, now it's 6.2.19 ~ 64-bit.

4009 floating point MIPS (Whetstone) per CPU
11596 integer MIPS (Dhrystone) per CPU

[D_A]

Thanks, updated.

[jcool]

And something more for D_A

AMD A64 3800+ X2 "Toledo" at 2,7Ghz (270x10). Must be the worst X2 ever, cause I'm giving all the Vcore Bios offers (~1,52V load) and that sucker won't stay stable any higher :/

BOINC 6.2.18, Win XP Pro 32 SP2:

FPU 2758
Integer 5040

Need a 64bit OS on that one to boost the score, but let's see what it can do.

[D_A]

Actually that's the second best score for that chip I've seen. The only one better was running Vista 64bit AND 300MHz higher clock.

While I don't even attempt to track it, there are a lot of little things you can do to boost your BOINC benchmark. Killing off unnecessary service can make a big difference for some people. On my own headless machines I gut the OS to the point that it only has just what it needs to do what I want and not a single service/daemon/library more. It takes some time (and I've had casualties) but I think it's worth it.

[jcool]

Well, going from a 32bit to 64bit OS does what, 10-15% increase in ppd?
I usually don't run many services, but since this is the HTPC, I don't get around the X-fi, tv card etc.

I don't think most services take up a lot of time anyway. I have set the windooze taskmgr to show CPU times for each process, and after weeks or even months of uptime I never saw one out of line. Maybe 2-3 seconds CPU in a few weeks, 10 seconds after 2 months. I can live with that...

[D_A]

I suspect killing off services has more of an effect on older machines that take longer to process those transient services. Having said that I even go to the trouble of shutting off devices and ports in the BIOS that I don't use.

[jcool]

I suspect killing off services has more of an effect on older machines that take longer to process those transient services. Having said that I even go to the trouble of shutting off devices and ports in the BIOS that I don't use.

Yes, I still have that habit too, turning off unused onboard stuff used to improve OC performance a LOT back in the day. I guess some things just stick with you

And I agree, the effect of some background services running being so marginal might be due to the 8-core, 8GB mem, dedicated storage subsystem nature of said machine

Need to check on older machines.

[D_A]

Yes, I still have that habit too, turning off unused onboard stuff used to improve OC performance a LOT back in the day. I guess some things just stick with you

Wasted cycles are wasted cycles I figure. Better to put 'em to good use.

And I agree, the effect of some background services running being so marginal might be due to the 8-core, 8GB mem, dedicated storage subsystem nature of said machine
...

Suddenly I dislike you a great deal. (where's that frakkin' 'jealous' smilie )





I just finished doing the same kernel optimisations to my old P4 that I used on the XP2400 and got a 300 point rise in the Dhrystone benchmark. That might not sound like much to you, but its about a 15% improvement over before.

[illidan]

q9400 @ 3.2GHz
5.10.45 32bit

10/29/2008 9:34:56 AM||Running CPU benchmarks
10/29/2008 9:34:57 AM||Suspending computation - running CPU benchmarks
10/29/2008 9:35:27 AM||Benchmark results:
10/29/2008 9:35:27 AM|| Number of CPUs: 4
10/29/2008 9:35:27 AM|| 3348 floating point MIPS (Whetstone) per CPU
10/29/2008 9:35:27 AM|| 7034 integer MIPS (Dhrystone) per CPU
10/29/2008 9:35:28 AM||Resuming computation

[D_A]

What operating system and version are you running, illidan?

[illidan]

windows xp pro with sp2 32bit

[jcool]

Dammit, 32bit OS sucks especially with Integer scores.

E8300 at 3,95Ghz (465x8,5) on Win XP Pro 32 SP3, BOINC 6.3.12:

FPU 4040
Integer 7586



Same machine would do 4050/11500 on XP x64...

[illidan]

next os will be x64 for sure

[D_A]

Thanks guys, list updated.

[64dragon]

i'll add a few old ones that i've not posted

Xeon X3330 (same as q9400)

clocked @ 3 ghz
Vista Ultimate 32bit
Boinc 6.2.19
3151 floating point MIPS (Whetstone) per CPU
6561 integer MIPS (Dhrystone) per CPU

clocked @ 3.3ghz
Vista Ultimate 32bit
Boinc 6.2.19
3471 floating point MIPS (Whetstone) per CPU
7105 integer MIPS (Dhrystone) per CPU

clocked @ 3.5ghz
Vista Ultimate 32bit
Boinc 6.2.19
3666 floating point MIPS (Whetstone) per CPU
7458 integer MIPS (Dhrystone) per CPU

i posted this result back on pg 17 but i'll post it again to show the direct comparison between 32bit and 64bit at the same clock
clocked @ 3.5ghz - looks like it'll do 2500 Boinc ppd
Vista Ultimate 64bit
Boinc 6.3.14 (had issues with this version crashing)
3561 floating point MIPS (Whetstone) per CPU
10367 integer MIPS (Dhrystone) per CPU

this result was not posted yet; changing Boinc version
clocked @ 3.5ghz
Vista Ultimate 64bit
Boinc 6.2.19
3555 floating point MIPS (Whetstone) per CPU
10374 integer MIPS (Dhrystone) per CPU

with 32bit Vista @ 3.5ghz i averaged about 2100 Boinc points per day (ppd)
with 64bit Vista @ 3.5ghz it seems i'll average about 2500 Boinc ppd, i dont have enough good days on 64bit yet to be conclusive but a 400 ppd increase is nice, over the period of a week thats 2800 points and over the period of a year thats 145,600 extra points
it seems with 32bit, i had slightly higher Whetstone scores but 64bit more than makes up for that with Dhrystone

[D_A]

Those last two are on that X3330 as well?

[D_A]

NM, I found the original. Don't know how I missed those two.

Here is my latest...
QX9770 @ 4.2
XSOS x64bit
5.10.45

Ok, I've been here a little while but I'm not familiar wit "XSOS". There's a custom linux distro or something floating around here?

edit: Ok, some searching turns up it's a custom Windows variant. I'd still like a little more info, just out of curiosity.

[64dragon]

Those last two are on that X3330 as well?

yes, they are all on the X3330

[jcool]

Dammit, 32bit OS sucks especially with Integer scores.

E8300 at 3,95Ghz (465x8,5) on Win XP Pro 32 SP3, BOINC 6.3.12:

FPU 4040
Integer 7586



Same machine would do 4050/11500 on XP x64...

Well, it seems that there is only a big theoretical difference between 32 and 64bit, lol. Machine finishes HCC WUs in about 2:37 average, that's maybe 2-3mins above 64bit completion times. Seems it doesn't care about the lower integer at all...

[SiGfever]

NM, I found the original. Don't know how I missed those two.



Ok, I've been here a little while but I'm not familiar wit "XSOS". There's a custom linux distro or something floating around here?

edit: Ok, some searching turns up it's a custom Windows variant. I'd still like a little more info, just out of curiosity.

It is a custom rework by nn-step that performs very well.

Here is a screen shot of the time it takes to do a work unit.