Strange CPUMark results

I am (finally) upgrading my PC and testing with each step to see what
improvements I get. 2 motherboards - intel and amd, but same RAM, harddisk,
PSU, GFX, for all tests. With CPUMark2.1, I get strange results! Here is
what I am seeing:

Test 1 Test 2 Test
3 Final Score
Athlon 2400 2.00GHz (133MHz) 969.6 840.3 3601.4 = 3941.3
Athlon 2400 2.13GHz (142MHz) 1050.4 900.4 4201.7 = 4501.8
Pentium 4 3.06GHz (Model 519) 1575.6 208.3 3601.4 = 3678.1
Core 2 Duo e6400 (2.13GHz) 1096.1 484.8 8403.4 = 7549.0

Very confusing - What's going on!

GT

On Thu, 19 Apr 2007 11:21:34 +0100, "GT"
<ContactGT_remove_@hotmail.com> wrote:

>I am (finally) upgrading my PC and testing with each step to see what
>improvements I get. 2 motherboards - intel and amd, but same RAM, harddisk,
>PSU, GFX, for all tests. With CPUMark2.1, I get strange results! Here is
>what I am seeing:
>
> Test 1 Test 2 Test
>3 Final Score
>Athlon 2400 2.00GHz (133MHz) 969.6 840.3 3601.4 = 3941.3
>Athlon 2400 2.13GHz (142MHz) 1050.4 900.4 4201.7 = 4501.8
>Pentium 4 3.06GHz (Model 519) 1575.6 208.3 3601.4 = 3678.1
>Core 2 Duo e6400 (2.13GHz) 1096.1 484.8 8403.4 = 7549.0
>
>Very confusing - What's going on!
>
>GT
>

Different CPU architectures perform differently at different
tasks. IIRC, test 1 is a registry test of some sort so I
also wonder if a more bloated windows installation could
have a consequence, but generally the numbers don't look too
far off to me (but admittedly I'm too lazy to look them up
at the moment).

Most significant is to recognize that just as these vary, so
will performance on different applications and/or versions
of some applications. Athlon XP will look better on legacy
apps than it does here, and Core2Duo worse. P4 is still
reasonably strong on last generations mainstream software
doing linearized tasks.

You shouldn't need same parts for these tests though,
particularly on such a synthetic test it should isolate them
moreso than real world uses. Unfortunately being synthetic
it is also prone to accentuate differences by the simplicity
and singularity of the task.

>>With CPUMark2.1, I get strange results! Here is what I get:
>>
>> Test 1 Test 2
>> Test
>>3 Final Score
>>Athlon 2400 2.00GHz (133MHz) 969.6 840.3 3601.4 = 3941.3
>>Athlon 2400 2.13GHz (142MHz) 1050.4 900.4 4201.7 = 4501.8
>>Pentium 4 3.06GHz (Model 519) 1575.6 208.3 3601.4 = 3678.1
>>Core 2 Duo e6400 (2.13GHz) 1096.1 484.8 8403.4 = 7549.0
>
> Different CPU architectures perform differently at different
> tasks.
>
> Most significant is to recognize that just as these vary, so
> will performance on different applications and/or versions
> of some applications. Athlon XP will look better on legacy
> apps than it does here, and Core2Duo worse. P4 is still
> reasonably strong on last generations mainstream software
> doing linearized tasks.

Yes, the tests are: test 1 - Registry operations, test 2 - Floating-point
operations and test 3 - Integer operations test.

I don't think test 1 actually proves anything for me as it is the same
installation of XP.
The overall score of 7549 looks about right and the integer operations test
looked good also - e6400 being >2x faster than athlon and P4 3ghz just a tad
faster.

What surprises me is the floating point operations - 840 on the athlon, 480
on the e6400 and 208 on the P4. I expected a similar pattern to the integer
tests, certainly not a core 2 duo scoring half the rating of an athlon and a
P4 scoring a quarter!. Even if these tests examine particular performace
niches, I didn't expect the latest dual core CPU to be half the speed of a 4
yr old athlon at anything!

Once my WD5000KS arrives, my half-way house upgrade is complete. I've
decided to use a motherboard that can take my existing 1.5GB DDR333 RAM and
AGP Radeon8500 graphics card for now - Asrock 4CoreDual-VSTA. The m/b has
DDR & DDR2 slots and AGP + PCIe, so can upgrade gfx when I find a bargain -
upgrade RAM (and maybe board) later. Board + CPU - £140(ish) + £90 for the
drive. Should keep me going a bit longer.