* UNIX bench better on 2.2 than 2.4?
@ 2002-03-15 7:36 Jean-Eric Cuendet
0 siblings, 0 replies; 5+ messages in thread
From: Jean-Eric Cuendet @ 2002-03-15 7:36 UTC (permalink / raw)
To: linux-kernel
Hi,
I ran UNIX bench on 4 of our machines. The results are below.
What is strange is that 2.2 makes better than 2.4! Is it a known fact?
And is the UNIX bench better with 2 CPU?
-jec
Results:
Host1: 2xPIII 550MHz / 1Gb RAM / RAID5 SCSI
Result: 164.7
Host2: 2xPIII 866MHz / 1Gb RAM / RAID1 soft IDE
Result: 195.7
Host3: 1xPIII 800MHz / 512Mb RAM / IDE
Result: 208.6
Host4: 1xPIII 600MHz / 256Mb RAM / IDE
Result: 153.6
Host1:
Dhrystone 2 using register variables 116700.0 1139475.0 97.6
Double-Precision Whetstone 55.0 295.2 53.7
Execl Throughput 43.0 697.7 162.3
File Copy 1024 bufsize 2000 maxblocks 3960.0 83438.0 210.7
File Copy 256 bufsize 500 maxblocks 1655.0 38613.0 233.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 121259.0 209.1
Pipe Throughput 12440.0 254518.7 204.6
Process Creation 126.0 1831.7 145.4
Shell Scripts (8 concurrent) 6.0 201.7 336.2
System Call Overhead 15000.0 251716.1 167.8
=========
FINAL SCORE 164.7
Host2:
Dhrystone 2 using register variables 116700.0 1837005.3 157.4
Double-Precision Whetstone 55.0 478.4 87.0
Execl Throughput 43.0 923.8 214.8
File Copy 1024 bufsize 2000 maxblocks 3960.0 54663.0 138.0
File Copy 256 bufsize 500 maxblocks 1655.0 18601.0 112.4
File Copy 4096 bufsize 8000 maxblocks 5800.0 96807.0 166.9
Pipe Throughput 12440.0 399294.5 321.0
Process Creation 126.0 3503.8 278.1
Shell Scripts (8 concurrent) 6.0 277.0 461.7
System Call Overhead 15000.0 393738.7 262.5
=========
FINAL SCORE 195.7
Host3:
Dhrystone 2 using register variables 116700.0 1506306.9 129.1
Double-Precision Whetstone 55.0 364.4 66.3
Execl Throughput 43.0 1103.3 256.6
File Copy 1024 bufsize 2000 maxblocks 3960.0 110403.0 278.8
File Copy 256 bufsize 500 maxblocks 1655.0 53065.0 320.6
File Copy 4096 bufsize 8000 maxblocks 5800.0 151107.0 260.5
Pipe Throughput 12440.0 378679.6 304.4
Pipe-based Context Switching 4000.0 161272.9 403.2
Process Creation 126.0 7144.7 567.0
Shell Scripts (8 concurrent) 6.0 20.0 33.3
System Call Overhead 15000.0 412399.5 274.9
=========
FINAL SCORE 208.6
Host4:
Dhrystone 2 using register variables 116700.0 1099714.7 94.2
Double-Precision Whetstone 55.0 328.1 59.7
Execl Throughput 43.0 692.1 161.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 56029.0 141.5
File Copy 256 bufsize 500 maxblocks 1655.0 18783.0 113.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 96243.0 165.9
Pipe Throughput 12440.0 306228.7 246.2
Process Creation 126.0 3937.3 312.5
Shell Scripts (8 concurrent) 6.0 125.9 209.8
System Call Overhead 15000.0 282470.2 188.3
=========
FINAL SCORE 153.6
--
Jean-Eric Cuendet
Linkvest SA
Av des Baumettes 19, 1020 Renens Switzerland
Tel +41 21 632 9043 Fax +41 21 632 9090
E-mail: jean-eric.cuendet@linkvest.com
http://www.linkvest.com
--------------------------------------------------------
^ permalink raw reply [flat|nested] 5+ messages in thread* UNIX bench better on 2.2 than 2.4?
@ 2002-03-15 7:52 Jean-Eric Cuendet
2002-03-15 10:32 ` Mark Mielke
0 siblings, 1 reply; 5+ messages in thread
From: Jean-Eric Cuendet @ 2002-03-15 7:52 UTC (permalink / raw)
To: linux-kernel
Sorry, I forgot to give kernel versions...
Hi,
I ran UNIX bench on 4 of our machines. The results are below.
What is strange is that 2.2 makes better than 2.4! Is it a known fact?
And is the UNIX bench better with 2 CPU?
-jec
Results:
Host1: 2xPIII 550MHz / 1Gb RAM / RAID5 SCSI / 2.4.6smp + LVM
Result: 164.7
Host2: 2xPIII 866MHz / 1Gb RAM / RAID1 soft IDE / 2.4.16smp + LVM
Result: 195.7
Host3: 1xPIII 800MHz / 512Mb RAM / IDE / 2.2.19 RedHat 6.2
Result: 208.6
Host4: 1xPIII 600MHz / 256Mb RAM / IDE / 2.4.19-pre2-ac4-preempt
Result: 153.6
Host1:
Dhrystone 2 using register variables 116700.0 1139475.0 97.6
Double-Precision Whetstone 55.0 295.2 53.7
Execl Throughput 43.0 697.7 162.3
File Copy 1024 bufsize 2000 maxblocks 3960.0 83438.0 210.7
File Copy 256 bufsize 500 maxblocks 1655.0 38613.0 233.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 121259.0 209.1
Pipe Throughput 12440.0 254518.7 204.6
Process Creation 126.0 1831.7 145.4
Shell Scripts (8 concurrent) 6.0 201.7 336.2
System Call Overhead 15000.0 251716.1 167.8
=========
FINAL SCORE 164.7
Host2:
Dhrystone 2 using register variables 116700.0 1837005.3 157.4
Double-Precision Whetstone 55.0 478.4 87.0
Execl Throughput 43.0 923.8 214.8
File Copy 1024 bufsize 2000 maxblocks 3960.0 54663.0 138.0
File Copy 256 bufsize 500 maxblocks 1655.0 18601.0 112.4
File Copy 4096 bufsize 8000 maxblocks 5800.0 96807.0 166.9
Pipe Throughput 12440.0 399294.5 321.0
Process Creation 126.0 3503.8 278.1
Shell Scripts (8 concurrent) 6.0 277.0 461.7
System Call Overhead 15000.0 393738.7 262.5
=========
FINAL SCORE 195.7
Host3:
Dhrystone 2 using register variables 116700.0 1506306.9 129.1
Double-Precision Whetstone 55.0 364.4 66.3
Execl Throughput 43.0 1103.3 256.6
File Copy 1024 bufsize 2000 maxblocks 3960.0 110403.0 278.8
File Copy 256 bufsize 500 maxblocks 1655.0 53065.0 320.6
File Copy 4096 bufsize 8000 maxblocks 5800.0 151107.0 260.5
Pipe Throughput 12440.0 378679.6 304.4
Pipe-based Context Switching 4000.0 161272.9 403.2
Process Creation 126.0 7144.7 567.0
Shell Scripts (8 concurrent) 6.0 20.0 33.3
System Call Overhead 15000.0 412399.5 274.9
=========
FINAL SCORE 208.6
Host4:
Dhrystone 2 using register variables 116700.0 1099714.7 94.2
Double-Precision Whetstone 55.0 328.1 59.7
Execl Throughput 43.0 692.1 161.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 56029.0 141.5
File Copy 256 bufsize 500 maxblocks 1655.0 18783.0 113.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 96243.0 165.9
Pipe Throughput 12440.0 306228.7 246.2
Process Creation 126.0 3937.3 312.5
Shell Scripts (8 concurrent) 6.0 125.9 209.8
System Call Overhead 15000.0 282470.2 188.3
=========
FINAL SCORE 153.6
--
Jean-Eric Cuendet
Linkvest SA
Av des Baumettes 19, 1020 Renens Switzerland
Tel +41 21 632 9043 Fax +41 21 632 9090
E-mail: jean-eric.cuendet@linkvest.com
http://www.linkvest.com
--------------------------------------------------------
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: UNIX bench better on 2.2 than 2.4?
2002-03-15 7:52 Jean-Eric Cuendet
@ 2002-03-15 10:32 ` Mark Mielke
2002-03-15 10:44 ` Jean-Eric Cuendet
0 siblings, 1 reply; 5+ messages in thread
From: Mark Mielke @ 2002-03-15 10:32 UTC (permalink / raw)
To: Jean-Eric Cuendet; +Cc: linux-kernel
On Fri, Mar 15, 2002 at 08:52:02AM +0100, Jean-Eric Cuendet wrote:
> Results:
> Host1: 2xPIII 550MHz / 1Gb RAM / RAID5 SCSI / 2.4.6smp + LVM
> Result: 164.7
> Host2: 2xPIII 866MHz / 1Gb RAM / RAID1 soft IDE / 2.4.16smp + LVM
> Result: 195.7
> Host3: 1xPIII 800MHz / 512Mb RAM / IDE / 2.2.19 RedHat 6.2
> Result: 208.6
> Host4: 1xPIII 600MHz / 256Mb RAM / IDE / 2.4.19-pre2-ac4-preempt
> Result: 153.6
I dunno what Unix bench is... but isn't 153.6/208.6 close to 600/800
in terms of a fraction?
It isn't really fair to compare different machine configurations with
different OS's, and declare that the OS is the only factor involved.
Make Host3 or Host4 have both versions of the linux kernel in /boot
and try each on the same machine.
mark
--
mark@mielke.cc/markm@ncf.ca/markm@nortelnetworks.com __________________________
. . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder
|\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ |
| | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada
One ring to rule them all, one ring to find them, one ring to bring them all
and in the darkness bind them...
http://mark.mielke.cc/
^ permalink raw reply [flat|nested] 5+ messages in thread* Re: UNIX bench better on 2.2 than 2.4?
2002-03-15 10:32 ` Mark Mielke
@ 2002-03-15 10:44 ` Jean-Eric Cuendet
2002-03-15 10:56 ` Mark Mielke
0 siblings, 1 reply; 5+ messages in thread
From: Jean-Eric Cuendet @ 2002-03-15 10:44 UTC (permalink / raw)
To: Mark Mielke; +Cc: linux-kernel
>
>
>>Results:
>>Host1: 2xPIII 550MHz / 1Gb RAM / RAID5 SCSI / 2.4.6smp + LVM
>> Result: 164.7
>>Host2: 2xPIII 866MHz / 1Gb RAM / RAID1 soft IDE / 2.4.16smp + LVM
>> Result: 195.7
>>Host3: 1xPIII 800MHz / 512Mb RAM / IDE / 2.2.19 RedHat 6.2
>> Result: 208.6
>>Host4: 1xPIII 600MHz / 256Mb RAM / IDE / 2.4.19-pre2-ac4-preempt
>> Result: 153.6
>>
>
>I dunno what Unix bench is... but isn't 153.6/208.6 close to 600/800
>in terms of a fraction?
>
Yes, you are right for this one.
But what impressed me is between Host2/Host3. One is a dual-866 with
2.4.16 and the other is mono-800 with 2.2.19. And the host3 with 2.2.19
is ~7% better than Host2 (Not counting that Host3 has ~8% more CPU
clock, which should have be ~15% better for Host2 vs Host3...)
>Make Host3 or Host4 have both versions of the linux kernel in /boot
>and try each on the same machine.
>
But that's just what I'd like to see. How 2.4 compares vs 2.2
-jec
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: UNIX bench better on 2.2 than 2.4?
2002-03-15 10:44 ` Jean-Eric Cuendet
@ 2002-03-15 10:56 ` Mark Mielke
0 siblings, 0 replies; 5+ messages in thread
From: Mark Mielke @ 2002-03-15 10:56 UTC (permalink / raw)
To: Jean-Eric Cuendet; +Cc: linux-kernel
On Fri, Mar 15, 2002 at 11:44:51AM +0100, Jean-Eric Cuendet wrote:
> Yes, you are right for this one.
> But what impressed me is between Host2/Host3. One is a dual-866 with
> 2.4.16 and the other is mono-800 with 2.2.19. And the host3 with 2.2.19
> is ~7% better than Host2 (Not counting that Host3 has ~8% more CPU
> clock, which should have be ~15% better for Host2 vs Host3...)
There is no guarantee that UNIX bench is SMP-friendly, or even
designed to test an SMP machine. Single programs that make good use of
an SMP machine usually have to be designed very differently than a
traditional program designed for a single processor.
It is probably just luck that the numbers jump around.
Just because Unix bench performs better or worse, or not as better, on
an SMP machine, as a non-SMP machine, does not mean that the numbers
accurately effect the capabilities of a real-life application. Unless
specifically designed for the environment, it is even possible that the
numbers would not even come close.
mark
--
mark@mielke.cc/markm@ncf.ca/markm@nortelnetworks.com __________________________
. . _ ._ . . .__ . . ._. .__ . . . .__ | Neighbourhood Coder
|\/| |_| |_| |/ |_ |\/| | |_ | |/ |_ |
| | | | | \ | \ |__ . | | .|. |__ |__ | \ |__ | Ottawa, Ontario, Canada
One ring to rule them all, one ring to find them, one ring to bring them all
and in the darkness bind them...
http://mark.mielke.cc/
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2002-03-15 11:03 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-03-15 7:36 UNIX bench better on 2.2 than 2.4? Jean-Eric Cuendet
-- strict thread matches above, loose matches on Subject: below --
2002-03-15 7:52 Jean-Eric Cuendet
2002-03-15 10:32 ` Mark Mielke
2002-03-15 10:44 ` Jean-Eric Cuendet
2002-03-15 10:56 ` Mark Mielke
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox