BOINC benchmarks

[Assimilator1]

After looking around some old benchmarks for SETI classic ,particularly Reolofs Top 200 for TLC (though their are many TA members in there too ) I'd realised I've seen nothing like that for SETI BOINC & other BOINC projects.(Maybe I haven't looked properly?)

How about we start with posting your BOINC benchmarks here?
I think the BOINC benchmark is not associated to any particular project ,is that right?
Does BOINC version matter?

Btw I found out that Reolofs (SETI Spy author) old site has been virtually totally archived ,including his Top 200 for S@H classic if your interested.

Anyway my XPM 2500 @2.5GHz (178FSB) manages this (v5.4.9).

2305 floating point MIPS (Whetstone) per CPU
3881 integer MIPS (Dhrystone) per CPU

I'd be particularly interested to see how that compares to mid range single core Ath64s
I'll get benchmarks from my Sempron64 3100 @2.5GHz a bit latter too.

I know some of ya have got some awesome rigs ,can't wait to see those figures

Don't forget to state clock speed & cache size where it varies & BOINC version

 

[Fullmetal Chocobo]

Well, I know for a fact that the BOINC benchmark isn't associated with any project, because I can't get any project (just tried TANAPKU or whatever...), and it isn't working. So with no project loaded...

1148 floating point MIPS [Whetstone] per CPU (x 4 = 4592)
1123 integer MIPS [Dhrystone] per CPU (x 4 = 4492)

 

[Assimilator1]

Hi mate ,hows GWF going?

Btw you forgot to mention what CPU you've got
What's the x4 about?

My Sempron @2.5GHz records this (v5.2.13)

2365 floating point MIPS (Whetstone) per CPU
4413 integer MIPS (Dhrystone) per CPU

[edit] updated BOINC to 5.4.9 & the scores are virtually the same
2372
4405

 

[Fullmetal Chocobo]

Originally posted by: Assimilator1
Hi mate ,hows GWF going?

Btw you forgot to mention what CPU you've got
What's the x4 about?

My Sempron @2.5GHz records this (v5.2.13)

2365 floating point MIPS (Whetstone) per CPU
4413 integer MIPS (Dhrystone) per CPU

GW is going pretty good. I finally was able to start playing today until my video card overheated @ 200 degrees (F). Big computer & small room doesn't mix.

Fullmetal Chocobo (my computer): dual Xeon 2.66GHz with HT (so 4 processors total)
Mercury Chocobo (laptop): 1.86GHz Centrino (will post benchmarks later when wife wakes up)

 

[Assimilator1]

Yikes ,that's hot!:Q ,even with a hot room how the heck did it get to 200F?

That Pentium M(?) laptop should give a good score

 

[LANMAN]

Intel Centrino 2.13Ghz

1874 floating points MIPS (Whetstone)
3867 Integer MIPS (Dhrystone)

Update:

Dual CPU - Dual Core Xeon 3.0

1357 Whetstone per CPU
987 Dhrystone per CPU

So I guess if I did the numbers then that would be (?) (8 processors per CPU)
10856 Whetstone
7896 Dhrystone

--LANMAN

 

[Bateluer]

This thread is a good idea, let people know how different CPUs stack up.

9/9/2006 5:28:37 PM||Starting BOINC client version 5.4.9 for windows_intelx86
9/9/2006 5:28:37 PM||Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 4200+
9/9/2006 5:29:47 PM|| Number of CPUs: 2
9/9/2006 5:29:47 PM|| 2082 floating point MIPS (Whetstone) per CPU
9/9/2006 5:29:47 PM|| 2052 integer MIPS (Dhrystone) per CPU


Edit - Odd, my scores seem abnormally low compared to other systems listed in this thread.

I see there is a new client out for windows PCs though, 5.4.11.

Edit 2 - 5.4.11 is a little better. 3880 Dhrystone, similar Whetstone scores.

 

[cmdrdredd]

E6400 @2.8Ghz
2628 floating point MIPS (Whetstone) per CPU
5479 integer MIPS (Dhrystone) per CPU


When I did the same test at 3.2Ghz my Whetstone test went up to 3000 but my Dhrystone went down to 4800 which was really odd. Anyway 2.8Ghz is where I keep it at 24/7

 

[Assimilator1]

lol ,you get over 40% higher Integer score than my XPM yet its only clocked 12% faster!
Awesome CPUs the Core2

At 3.2GHz you either pushing the CPU,FSB or cache too hard & thus errors are cropping up which slows things down ,you'd probably find it'd be much faster at 3.1GHz.

With v5.4.11 I get this on my XPM @2.5GHz

2298 FP MIPS
3863 In MIPS

So again no real difference between versions for my rigs.

LANMAN
Is that Pentium M?

Bateluer
Odd about your scores ,maybe you had something hogging CPU power on the 1st run?
The 2nd score is a lot better
Btw am I right in saying an X2 4200 runs at 2.2GHz with 2x512KB L2 cache?

If so then 1 of your cores near enough matches my XPM (except FP a bit) & its clocked about 13% faster.


Interesting reading so far ,lets see some more rigs including older machines

 

[Bateluer]

Originally posted by: Assimilator1
Bateluer
Odd about your scores ,maybe you had something hogging CPU power on the 1st run?
The 2nd score is a lot better
Btw am I right in saying an X2 4200 runs at 2.2GHz with 2x512KB L2 cache?

If so then 1 of your cores near enough matches my XPM (except FP a bit) & its clocked about 13% faster.

Yes. Technically, when I ran the first run, I had firefox and thunderbird open, though minimized and not actively doing anything. I also had a windows explorer window open to a network drive and a paused DivX playing in WMP. Second time around, I closed all the windows.

 

[petrusbroder]

Here are some benchmarks for more unusual processors: PowerMac G4 and G5.
As a comparison I have also included a Intel-based MacBook. All the numbers are for each processor, some of these computers have two processors, then the numbers have to be doubled ...

__ Mac ____ CPU ___ Whet- ____ Dhry- ___ in Seti _____ in Seti
Processor _ MHz ___ stones __ stones __ cred/week __ seconds/credit
__ G4 _____ 400 ____ 269 ______ 599 _____ 359.1 ______ 1684 ____ (768 MByte RAM)
__ G4 ____ 1299 ____ 642 _____ 2013 _____ 821.7 _______ 736 ____ (768 MByte RAM)
__ G4 ____ 1420 ____ 716 _____ 2216 _____ 946.5 _______ 639 ____ (1024 MBytes RAM)
__ G5 ____ 2000 ___ 1067 _____ 3334 ____ 1542.9 _______ 392 ____ (2048 MBytes RAM)

Core Duo _ 1800 ___ 1199 _____ 2126 ____ 2488.9 _______ 243 ____ (dual core; for each core, 1024 MBytes RAM)

All measurments were done in quadruplicate and the average is presented.
Operating system OS X 10.4.7. All running the standard BOINC 5.4.9 or 5.4.11.

One interesting observation:
When the mouse was moved during measurment the Whetstone-score in the Intel based computer dropped by 290 Whetstones. In the G5 the Whetstone score fell by some 20 - 30 Whetstones ...

 

[cmdrdredd]

Originally posted by: Assimilator1
lol ,you get over 40% higher Integer score than my XPM yet its only clocked 12% faster!
Awesome CPUs the Core2

At 3.2GHz you either pushing the CPU,FSB or cache too hard & thus errors are cropping up which slows things down ,you'd probably find it'd be much faster at 3.1GHz.

Yeah it could be. Anyhow I don't see the difference to be that large between 2.8Ghz and 3.2Ghz to warrant the heat differences. I live in Florida and it's generally pretty warm so things tend to heatup.

 

[The Borg]

Hi all,

Yes, I think this is a good thread. There are so many type of setups that it is difficult to judge systems. Anyway here goes mine:

Windows: All running BOINC 5.5.0

1 x Celeron (Coppermine-T) 950 MHz (@1263.5 MHz)
1453 floating point MIPS
5408 integer MIPS

1 x AMD64 3500+ (Clawhammer) (2.2 GHz @ 2.52 GHz)
3148 floating point MIPS
13343 integer MIPS

6 x Intel P4-630 (3GHz @ stock)
2614 floating point MIPS per CPU (x2 - HT)
5578 integer MIPS per CPU (x2 - HT)

1 x AMD64 X2 4800+ (Toledo) (2.4 GHz @ 2.79 GHz) - TT Bigwater cooling
4630 floating point MIPS per CPU (x2)
15309 integer MIPS per CPU (x2)

Linux: Runnig BOINC 5.4.9

1 x Dual Intel PIII - 600 (600MHz @ stock)
302 floating point MIPS per CPU
514 integer MIPS per CPU

 

[Assimilator1]

There's something very odd about your scores Borg

Look at the ratio of your FP score to Intger ,seems like your Intger scores are out by a factor of about 10 ,lol

Btw v5.5 ,is that a beta client?

Peter
Windows mouse CPU hungry? ,shouldn't be surprised I guess

 

[petrusbroder]

@ Assim: no, not Windows! it is a MacBook, running OS X 10.4.7., with a dual core intel-processor.
That is the surprising observation: a computer using the G5 processor (OS and BOINC being the same) looses much less performance in Whetstones than the same OS and BOINC do in a Mac with an Intel-processor. I think this is a matter of using a code with compromises (usable for PowerProcessors and Inte-processors).

 

[caferace]

Some of my toys.

All running Windows XP and 5.5.0 Client unless indicated otherwise:

Intel Pentium D 805 2.66 Ghz (currently OC'ed to 3.33Ghz - Thermaltake Big Typhoon and runs @ ~90F w/full load)
Measured floating point speed 3707.9 million ops/sec (x2)
Measured integer speed 12021.43 million ops/sec (x2)

AMD FX-55 (slight OC to 2.66Ghz - Stock HSF)
Measured floating point speed 4435.54 million ops/sec
Measured integer speed 14000.35 million ops/sec

Intel P4 2.0 Ghz
Measured floating point speed 2395.64 million ops/sec
Measured integer speed 7245.5 million ops/sec

Intel x86 Family 6 Model 8 Stepping 10 937MHz (dual CPU running 5.2.13 client))
Measured floating point speed 963.28 million ops/sec (x2)
Measured integer speed 2461.81 million ops/sec (x2)

Mobile AMD Athlon XP 2000+
Measured floating point speed 2051.64 million ops/sec
Measured integer speed 8490.14 million ops/sec

AMD Athlon 1700+
Measured floating point speed 1692.92 million ops/sec
Measured integer speed 7554.59 million ops/sec

Intel P4 1.6Ghz
Measured floating point speed 1871.84 million ops/sec
Measured integer speed 5504.27 million ops/sec

Intel x86 Family 6 Model 8 Stepping 1 731MHz (PIII, doubles as a webserver)
Measured floating point speed 906.43 million ops/sec
Measured integer speed 3210.02 million ops/sec

Power Macintosh G4 PowerBook 6,8 (5.4.11 client)
Measured floating point speed 761.14 million ops/sec
Measured integer speed 2389.94 million ops/sec

The Mobile AMD CPU is impressive, but I have to keep air moving or it'll prolly catch on fire. It's been crunching and nothing else for 6 months straight.

I have a whole bunch of these PIII boxes running Windows XPe at work that live double lives. Unfortunately, our configuration needs for XPe don't allow it to work on all BOINC-enabled projects. These are all crunching and doing regular work simultaneously:

Intel x86 Family 6 Model 8 Stepping 10 869MHz (PIII)
Measured floating point speed 768.49 million ops/sec
Measured integer speed 1341.42 million ops/sec

If needs be (like now) I'l co-opt the occasional 3-3.4 HT P4 I see doing nothing important.

-jim

 

[cmdrdredd]

Originally posted by: The Borg
Hi all,

Yes, I think this is a good thread. There are so many type of setups that it is difficult to judge systems. Anyway here goes mine:

Windows: All running BOINC 5.5.0

1 x Celeron (Coppermine-T) 950 MHz (@1263.5 MHz)
1453 floating point MIPS
5408 integer MIPS

1 x AMD64 3500+ (Clawhammer) (2.2 GHz @ 2.52 GHz)
3148 floating point MIPS
13343 integer MIPS

6 x Intel P4-630 (3GHz @ stock)
2614 floating point MIPS per CPU (x2 - HT)
5578 integer MIPS per CPU (x2 - HT)

1 x AMD64 X2 4800+ (Toledo) (2.4 GHz @ 2.79 GHz) - TT Bigwater cooling
4630 floating point MIPS per CPU (x2)
15309 integer MIPS per CPU (x2)

Linux: Runnig BOINC 5.4.9

1 x Dual Intel PIII - 600 (600MHz @ stock)
302 floating point MIPS per CPU
514 integer MIPS per CPU

Your scores are either faked, typoed, or just plain wrong. There is no possible way any of your AMD systems would have better scores than my Core2. Plus look at that celeron score. There's no way possible for it to be that high compared to an A64 or Core2

 

[petrusbroder]

Hmmm, I am not so sure - considering that Borg is running an "optimized" BOINC (5.5.0) the numbers could very well be correct.
I have tested some other BOINC-clients and they gave some very interesting (and surprisingly high) benchmarks ... especially when they are optimized for SSE3 ...
Looking at Borg's PIII running @ 600 MHz and using a standardized BOINC-client the numbers look very OK to me...

If you look at Caferace's numbers they seem to corroborate Borg's numbers, and he uses BOINC 5.5.0 too. This may indeed be more a matter of the BOINC-client than of the processor! Especially the Athlon cores give much more in the benchmarks when an optimized client is used.

 

[cmdrdredd]

Originally posted by: petrusbroder
Hmmm, I am not so sure - considering that Borg is running an "optimized" BOINC (5.5.0) the numbers could very well be correct.
I have tested some other BOINC-clients and they gave some very interesting (and surprisingly high) benchmarks ... especially when they are optimized for SSE3 ...
Looking at Borg's PIII running @ 600 MHz and using a standardized BOINC-client the numbers look very OK to me...

If you look at Caferace's numbers they seem to corroborate Borg's numbers, and he uses BOINC 5.5.0 too. This may indeed be more a matter of the BOINC-client than of the processor! Especially the Athlon cores give much more in the benchmarks when an optimized client is used.

Ok I just downloaded the 5.5 client with SSE3 and here's my score with a C2D at 2.8Ghz
5015 floating point MIPS (Whetstone) per CPU
23156 integer MIPS (Dhrystone) per CPU

Now it makes sense...

 

[petrusbroder]

I just wish I had PCs with those processors ...

 

[caferace]

Originally posted by: petrusbroder


I just wish I had PCs with those processors ...

Until you get the power bill. I think I'm going to take my Pentium D 805 box to work, and use it as a SQA box. Crunch, work, crunch, work... etc.

-jim

 

[Bateluer]

Anyone got a link for the 5.5 client? I can't seem to find one via google.

Edit - I see 5.6.3 at the Alpha page below.

http://isaac.ssl.berkeley.edu/alpha/

 

[petrusbroder]

That is not the version being run by Caferace and Borg!
They are running a BOINC-version originally recompiled by crunch3r, before he quit.
The version you link to abuve is a alpha-test and should be used as such - for testing!

here is one link: http://devel.bluetentacle.co.uk/dc/boinc/

 

[Assimilator1]

Hmm the 5.5 client seems to be throwing off the benchmarks ,at least on the Integer.

I wonder if that's a bug rather than an improvment? ,it does seem an excessive difference.
[edit] the more I look at the 5.5 scores the more I think it's a bug....

cmdrdredd
Saying that a team mate is putting up faked scores is not good team spirit ,it would be better if you removed that word.
Anyway ,I really think those scores look like a bug to me........

Caferace
Nice set of rigs ,but I don't see the XPM listed?
Btw is 'Intel x86 Family 6 Model' PIIIs?

Where are you guys getting v5.5 from anyway?

 

[Bateluer]

Originally posted by: Assimilator1

the more I look at the 5.5 scores the more I think it's a bug....

Where are you guys getting v5.5 from anyway?

I would agree that its a bug. The 5.6.3 client that I'm trying out now gives similar benchs to the 5.4.11 client I had previously.

Edit - I don't see such massive improvements from SSE3 optimizations. I especially don't like that Celeron with less than half my clock speed, and single core'd, outperforming my X2 4200.