P2b shoot-out PIII 1ghz -- tualatin 1.3ghz

  • Thread starter Thread starter Guest
  • Start date Start date
G

Guest

Well, finally got time to do this, here are the results obtained:

Tualatin 1.3Ghz

wintune

CPU Integer 4131.134 MIPS
CPU Floating Point 1587.452 MFLOPS

SiSoft Sandra

cpu mips 4286
cpu mflops 1868
integer ram 745
float buffered ram 720

PCMark 2002

cpu 3370
ram 1473

CPUMark

111

Pentium III 1Ghz

wintune

CPU Integer 2955 MIPS
CPU Floating Point 1167 MFLOPS

SiSoft Sandra

cpu mips 3130
cpu mflops 1332
integer ram 739
float buffered ram 738

PCMark 2002

cpu 2687
ram 1601

CPUMark

90.3

comments?? Corrections??

eric
 
comments?? Corrections??
Click to expand...

Specifying the CPUs' FSB would be an idea. Also, pure synthetic
benchmarks are boring. (BTW, Sandra memory benchmarks are usually more
interesting with buffering off.) What about some *real* applications
(including games)?

Stephan
 
Specifying the CPUs' FSB would be an idea.
Click to expand...

The tualatin was run at 105FSb and the PIII at 133Mhz.
Also, pure synthetic
benchmarks are boring. (BTW, Sandra memory benchmarks are usually more
interesting with buffering off.)
Click to expand...

buffering off? I need an explanation. of the term.
What about some *real* applications
(including games)?
Click to expand...

How would i go about doing that , Stephen? I don't play games. How
would I benchmark a real application?

eric
 
On Wed, 24 Mar 2004 14:33:29 +0100, Stephan Grossklass <


buffering off? I need an explanation. of the term.
Click to expand...

Should be in Sandra's help. As I understand it: When you turn buffering
on (it is by default; you can turn it off in the memory benchmark
module's options), larger chunks of data are being read/written with the
help of SSE and the like. That's supposed to give an impression of
memory bandwidth in "multimedia" applications (encoding, games,
whatever). Since the resulting benchmark numbers have always been
notoriously low on the i440BX in spite of the chipset being very fast in
real applications, I don't trust buffered benchmarks.
How would i go about doing that , Stephen? I don't play games. How
would I benchmark a real application?
Click to expand...

Usually with a stopwatch (many digital watches have that functionality)
and some CPU/memory limited task that keeps the PC busy long enough to
allow halfway accurate measurements. Such applications might be
* relaunching a rather heavy app from the disk cache (Mozilla 1.7b on a
P2B-D with Celeron 300A needs 5.7...5.8 s here)
* rendering a local copy of a rather heavy web page like
<http://users.erols.com/chare/elec.htm> on Mozilla
* applying some effect to some big photo, saving it in PNG or JPEG-2000
format (doing this with a freely available source image - which may
have to be enlarged - and, say, a freely available tool like XNView
helps if others want to reproduce the results)

Stephan
 
Well, finally got time to do this, here are the results obtained:

Tualatin 1.3Ghz

wintune

CPU Integer 4131.134 MIPS
CPU Floating Point 1587.452 MFLOPS

SiSoft Sandra

cpu mips 4286
cpu mflops 1868
integer ram 745
float buffered ram 720

PCMark 2002

cpu 3370
ram 1473

CPUMark

111

Pentium III 1Ghz

wintune

CPU Integer 2955 MIPS
CPU Floating Point 1167 MFLOPS

SiSoft Sandra

cpu mips 3130
cpu mflops 1332
integer ram 739
float buffered ram 738

PCMark 2002

cpu 2687
ram 1601

CPUMark

90.3

comments?? Corrections??

eric
Click to expand...

Interesting, and generally about what I would have expected - although
some of the Tualatin 1.3Ghz numbers are suspiciously high given I got
mostly lower numbers for a Tualatin 1.0Ghz running at 1.33Ghz (133Mhz
FSB vs. 105). The P3 1Ghz results OTOH are remarkably consistent with mine.

I don't suppose you kept a record of the Wintune memory numbers? I will
add your results to my benchmark page when I get time, and would like to
have a complete set of results.

What were your memory timings set to in the BIOS?

P2B

http://tipperlinne.com/benchmark.htm
 
Interesting, and generally about what I would have expected - although
some of the Tualatin 1.3Ghz numbers are suspiciously high given I got
mostly lower numbers for a Tualatin 1.0Ghz running at 1.33Ghz (133Mhz
FSB vs. 105). The P3 1Ghz results OTOH are remarkably consistent with mine.

I don't suppose you kept a record of the Wintune memory numbers? I will
add your results to my benchmark page when I get time, and would like to
have a complete set of results.

What were your memory timings set to in the BIOS?

P2B

http://tipperlinne.com/benchmark.htm
Click to expand...

Right, sorry, I forgot to post the memory numbers from Wintune

Tualatin: Memory: 3244.008 MB/s

PIII 1Ghz Memory: 2396.555 MB/S

funny about those Tualatin numbers, Due to your comments I've run them
several times and I get variations of 1 or 2 points and that's all. So
I'm pretty sure they're correct. The 1,3ghz was runninng at 105Mhz
during all these tests on a slot-T.

This was using a recent install of WinXP, on my 3rd computer and as
such has nothing else installed at start up, like I usually have. No
Office, no nothing. A bare install. I always find that these recent
installs are pretty frisky, even with a Celeron 300Mhz. Really! Then
I begin to add startup programs like an anti-virus or mesesenger or
cpu idle etc and it always feels slower.

Memory settings in the Bios were 7ns 143Mhz. At first I tried the SPD
settings (cas3 etc) cause I was having problems with the boot of the
PIII and thought it was a result of having too tight timings, The
results for the tests were noticeably lower. Then finally after
managing to boot smoothly, I shifted to the 7ns settings, everything
went smoothly and these are the final results. I think I must have had
the FSB jumpers manually set wrong or something at first go.

The mobo is a Rev1.10, vanilla p2b.

The memory (384mb) is a combo of Mushkin REV 2 and Crucial Cas2. The
PIII appears to run at 140Mhz smoothly, but without stressing it or
anything, tested with softfsb.

eric
 
Right, sorry, I forgot to post the memory numbers from Wintune

Tualatin: Memory: 3244.008 MB/s

PIII 1Ghz Memory: 2396.555 MB/S
Click to expand...

Looks like the test runs entirely in the L2 cache, thus scales with the
core clock.

Stephan
 
Right, sorry, I forgot to post the memory numbers from Wintune

Tualatin: Memory: 3244.008 MB/s

PIII 1Ghz Memory: 2396.555 MB/S
Click to expand...

Thanks :-)
funny about those Tualatin numbers, Due to your comments I've run them
several times and I get variations of 1 or 2 points and that's all. So
I'm pretty sure they're correct. The 1,3ghz was runninng at 105Mhz
during all these tests on a slot-T.

This was using a recent install of WinXP, on my 3rd computer and as
such has nothing else installed at start up, like I usually have. No
Office, no nothing. A bare install. I always find that these recent
installs are pretty frisky, even with a Celeron 300Mhz. Really! Then
I begin to add startup programs like an anti-virus or mesesenger or
cpu idle etc and it always feels slower.

Memory settings in the Bios were 7ns 143Mhz. At first I tried the SPD
settings (cas3 etc) cause I was having problems with the boot of the
PIII and thought it was a result of having too tight timings, The
results for the tests were noticeably lower. Then finally after
managing to boot smoothly, I shifted to the 7ns settings, everything
went smoothly and these are the final results. I think I must have had
the FSB jumpers manually set wrong or something at first go.

The mobo is a Rev1.10, vanilla p2b.

The memory (384mb) is a combo of Mushkin REV 2 and Crucial Cas2. The
PIII appears to run at 140Mhz smoothly, but without stressing it or
anything, tested with softfsb.

eric
Click to expand...

In that case your BIOS memory settings were the same as I used. I also
tested on a fresh OS install, but I used XP Pro on a P2B-S rev 1.04 with
256MB of Corsair PC150 (one DIMM). I'm not sure why these differences in
the test system would affect the Tualatin results, but apparently they did.

P2B
 
In that case your BIOS memory settings were the same as I used. I also
tested on a fresh OS install, but I used XP Pro on a P2B-S rev 1.04 with
256MB of Corsair PC150 (one DIMM). I'm not sure why these differences in
the test system would affect the Tualatin results, but apparently they did.

P2B
Click to expand...

Here's a thought. We all know the p2b in almost all its flavours
went through various revisions up to 1.12. Each rev change must mean
hardware modification of some sort. ¿NO?

I confess my ignorance. To tell you the truth, the only change I am
aware of, or the only one that impinges on my consciousness as
important, is the voltage regulator chip thingie. Perhaps they added
or subtracted a PCI slot or an ISA slot, added a RAM slot...

So yeah, I can run a tualatin on my rev 1.10, and/or use a coppermine
CPU with a slotket without having to adjust the voltage... but there
were obviously other changes made.

Perhaps those changes affect(ed) performance in some way??

eric
 
Exscuse me for butting in but can anyone tell me what the latest BIOS
is for this board? So far I've found BXDS1013 and 1014DS03, I suspect
the latter is the latest but the BXDS has thrown me off a bit. I want
to update to enable recognition of a 126Gb HD, which I'm told the
update will do. TIA

Regards
angie
 
Exscuse me for butting in but can anyone tell me what the latest BIOS
is for this board? So far I've found BXDS1013 and 1014DS03, I suspect
the latter is the latest but the BXDS has thrown me off a bit. I want
to update to enable recognition of a 126Gb HD, which I'm told the
update will do. TIA

Regards
angie
Click to expand...

1013 was the last 'release' BIOS issued for the board, 1014beta3 is
almost certainly the 'last' BIOS Asus will ever publish.

I have been running 1014beta3 on several P2B-DS systems for quite some
time, and have found it to be as reliable and stable as any other
version. IMHO there is no reason not to upgrade to 1014beta3, and you
must to so to run IDE drives larger than 64GB or Tualatin processors.

HTH

P2B
 
Here's a thought. We all know the p2b in almost all its flavours
went through various revisions up to 1.12. Each rev change must mean
hardware modification of some sort. ¿NO?

I confess my ignorance. To tell you the truth, the only change I am
aware of, or the only one that impinges on my consciousness as
important, is the voltage regulator chip thingie. Perhaps they added
or subtracted a PCI slot or an ISA slot, added a RAM slot...

So yeah, I can run a tualatin on my rev 1.10, and/or use a coppermine
CPU with a slotket without having to adjust the voltage... but there
were obviously other changes made.

Perhaps those changes affect(ed) performance in some way??

eric
Click to expand...

AFAIK the differences between hardware revisions were limited to:

- updated voltage regulators
- updated clock generators
- fixes for W2K ACPI issue, USB voltage issue, photoshop bug etc.

Nothing I'm aware of that would affect benchmark results. In addition,
your P2B rev 1.10 and my P2B-S rev 1.04 are fairly equivalent in terms
of the above updates.

When I get time I'll re-run a few of my tests with two DIMMS installed -
seems to me that is more likely to affect benchmark results than any
other difference in configuration we've identified.

P2B
 
1013 was the last 'release' BIOS issued for the board, 1014beta3 is
almost certainly the 'last' BIOS Asus will ever publish.

I have been running 1014beta3 on several P2B-DS systems for quite some
time, and have found it to be as reliable and stable as any other
version. IMHO there is no reason not to upgrade to 1014beta3, and you
must to so to run IDE drives larger than 64GB or Tualatin processors.
Click to expand...
That's excellent, I shall be flashing today, :) Thanks very much!

Regards
angie
 
Back
Top