[wEvil]

Got this from vanshardware.com

*******

Sandra Drystone Scores

Gabriel sent in very interesting observations regarding the decline of K6-2 and Duron Drystone scores on Sisoftware's Sandra:

 

Machine K6-2 400@400 (mobo Elpina TX PRO (Aladin IV chipset), aircooled, 6 X 66 Mhz)

Sandra 99  : Drystone Alu: 987 MIPS Drystone FPU: 473 MFLOPS
Sandra2k1se: Drystone Alu: 781 MIPS Drystone FPU: 478 MFLOPS
Sandra2k1te: Drystone Alu: 688 MIPS Drystone FPU: 486 MFLOPS

Machine K6-2 500 @ 600 (mobo Fic PA2013, aircooled, 2.9 V)

Sandra 99  : Drystone Alu: 1533 MIPS Drystone FPU: 725 MFLOPS
Sandra2k1se: Drystone Alu: 1397 MIPS Drystone FPU: 708 MFLOPS
Sandra2k1te: Drustone Alu: 1192 MIPS Drystone FPU: 724 MFLOPS

Machine Duron 700@700 (MSI K7TPro, aircooled 7 X 100 Mhz)

Sandra 99  : Drystone Alu: 2327 MIPS Drystone FPU: 962 MFLOPS
Sandra2k1se: Drystone Alu: 1963 MIPS Drystone FPU: 954 MFLOPS
Sandra2k1te: Drystone Alu: 1945 MIPS Drystone FPU: 960 MFLOPS
 

It seems that all of these platforms decline in score, while they all report the same reference score of 1350 for the Pentium III.  More data points from other chips would be interesting.

*******

------------------
AARGGH!!
My Normals' Flipped!!

[Taristin]

So... you're saying that they confessed to not having data on the durons? Figures...

[wEvil]

Machine Duron 700@700 (MSI K7TPro, aircooled 7 X 100 Mhz)

Sandra 99 : Drystone Alu: 2327 MIPS Drystone FPU: 962 MFLOPS
Sandra2k1se: Drystone Alu: 1963 MIPS Drystone FPU: 954 MFLOPS
Sandra2k1te: Drystone Alu: 1945 MIPS Drystone FPU: 960 MFLOPS

****

eh?



------------------
AARGGH!!
My Normals' Flipped!!

[aldo_14]

Yeah, but MIPs and MFLOPS are a completely useless indicator of performance, anyway.

I don't have my CAD notes here, but I think MFLOPs in completely dependent on the machines instruction set, so it ios near impossible to compare different architecture (at least with native MFLOPS)

Same goes for MIPs - in fact, the MIPs rating can vary inversely to performance, i.e. a machine with specialised Floating Point hardware will run faster but have a lower MIPS.

I remember something about Drystone.... damn, sliiped my mind.

If anyone wants to see what I'm talking about;
 http://www.cs.strath.ac.uk/~dunc/teaching/cad/

(EDIt - may have missed the point of this post completely)

[This message has been edited by aldo_14 (edited 10-20-2001).]

[wEvil]

Yep..the point was that the benchmark(eting) software is giving deliberately misleading results.


------------------
AARGGH!!
My Normals' Flipped!!

[Thorn]

If you want to test a machines performance use something like 3DMark2001...

[aldo_14]

 

Originally posted by wEvil:
Yep..the point was that the benchmark(eting) software is giving deliberately misleading results.

Meaningless Indicator of Performance

 

It's a pretty common problem.... AMD have been wanting an offical benchmark set (as their clock speed is slower than the p4s, but have better performance), whereas Intel have, obviously, been advertising on GHz speed.... hence AMD trying this 'quantispeed thing'.

Of course, you can always set up the procesor to recognise the instrucitons required by the benchmarking software, and optimise it so these give a misleadingly fast performance....

  maybe i'm actually learning something from CAD  

[Vertigo1]

   http://vertigo1-2.freeyellow.com/Bench/CPUMMBench.jpg  http://vertigo1-2.freeyellow.com/Bench/MemBench.jpg
Benchmark via SiSoft Sandra 2000.

Duron 700 @ 700Mhz (air-cooled via stock heatsink)
I rest my case.  

------------------
"You know you've played Freespace too much when you're driving along trying to use countermeasures to shake off pursuing highway police." - Top Gun

Material Defender Studios

[Martinus]

 

Originally posted by aldo_14:
Meaningless Indicator of Performance

Eh? I always thought it was Meaningless Interpretation of Processor Speed

Ohhhh, I can't wait for clockless processors to go into mass production. Right now the clock speed is the biggest limiting factor of processing speed.


------------------
Snack samurai sandwiches!! We slice 'em, we dice 'em, you eat 'em!!
(may contain traces of body parts)

[This message has been edited by Maeglamor (edited 10-20-2001).]

[Carl]

they never have the data you want

[Setekh]

There's something to note. Always something misleading around, which bites hard - I rely on benchmarks for a lot of things.