When is a faster clock speed a slower result?

[Guest]

I do a lot of numerical analysis involving pseudorandom number generation. This tends to be floating point processor-intensive work. We recently looked at upgrading our 1GHz machine with a 1.7GHz machine and found that there was no pick-up in processing speed for our key application. The program we ran was a MS Visual C++ program we wrote and compiled on a third machine. Our result was that the Compaq solved the problem in 18.4 hours versus 18.3 hours for the much faster new Dell. This was a bit of a surprise.

Our old computer is a Compaq we bought in 11/2000 with the following specs:
2x1GHz PIII Xeon (256K L2 cache, 133 MHz bus)
1.5GB of ECC PC800 RDRAM
Intel 840 chipset
Running NT 4.0

The new machine was a Dell with these specs:
2x1.7GHz P4 Xeon (256K L2 cache, 400 MHz bus)
4.0GB of PC800 RDRAM
Not sure what chipset
Running Windows 2000

Now, let me give you two additional pieces of information. First, neither computer appears to access the hard drive when the program is run, and indeed, we designed the program to not use excessive amount or RAM. And secondly, the newer machine does run Microsoft products, such as Excel, at twice the speed, which is what we expected for our C++ program.

I have requested that the C++ program be compiled on each of the machines and that the natively compiled program be run on each to see what difference that makes. But my question is: Has anyone ever encountered this sort of problem, and/or do you have any ideas why this would be happening?

Thanks!

 

[MeTaLrOcKeR]

All this realyl means is that your new Dell Machine doesnt have SSE2 optimization codes...therefore your old system would give or take be very comparable to the new one.......

Since you need (needed) a computer that has LOTS of FPU power, You should have gone for a Dual AMD Athlon MP 2000+ system, no P4 or any Intel product can touch it in the majority of things, and ESPECIALLY in FPU Intensive aplications, the Intel based machines just cant hold a candle to the FPU of the Athlon....

-MeTaL RoCkEr

My <font color=red>Z28</font color=red> can take your <font color=blue>P4</font color=blue> off the line!

 

[eden]

If you ever read the articles on this website, or anywhere for that matter, you should know by now, P4s are extremly weak in FPU.
They miss one FPU unit, and work much less for those. This means your old system is as good as the new one. Indeed opt for a Dual AthlonMP 2000, you will notice dramatic changes.

--
For the first time, Hookers are hooked on Phonics!!

 

[Raystonn]

I would suggest you grab the Intel C/C++ Compiler and try it out. There is a 30-day free trial for it. It should improve performance dramatically.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =

 

[eden]

...then again there are always hidden freebies on Intel's website, like the Application Accelerator, which I've yet to see some benchs showing some improvement, or at least have it go more public if it's so good!

--
For the first time, Hookers are hooked on Phonics!!

 

[AmdMELTDOWN]

"If you ever read the articles on this website, or anywhere for that matter, you should know by now, P4s are extremly weak in FPU."

make sure he reads the one where the mp <A HREF="http://www6.tomshardware.com/cpu/01q3/010917/index.html" target="_new">melts down</A> and all the problem posts all over the internet regarding AMD.

"<b>AMD/VIA!</b>...you are <i>still</i> the weakest link, good bye!"

 

[FatBurger]

They miss one FPU unit, and work much less for those.

Please show me where a P4 performs "much" worse than 66% per clock against an Athlon, as per your claim above.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!

 

[FatBurger]

Our result was that the Compaq solved the problem in 18.4 hours versus 18.3 hours for the much faster new Dell.

There's your problem. It's like throwing two turds in a lake and seeing which comes back up to the surface first :tongue:

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!

 

[AMD_Man]

The Sandra synthetic benchmark without SSE2 and the AXP without 3DNow or SSE.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor

 

[FatBurger]

Hmm...how would I do that, use an extremely old version?

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!

 

[AMD_Man]

Umm, use 2001. It shows both before and after SSE2. I don't think 2001 improves AMD FPU performance with any instructions.

AMD technology + Intel technology = Intel/AMD Pentathlon IV; the <b>ULTIMATE</b> PC processor

 

[IIB]

yes... but consider it would be the best version to elustrate - todays preformnce it terms of FPU. sence i doubt much code is SSE optimized anyway...


This post is best viewed with common sense enabled

 

[FatBurger]

That's right, I forgot it does that.

IIB, there's quite of list of software that's SSE2 optimized, let alone SSE optimized.

<font color=orange>Quarter</font color=orange> <font color=blue>Pounder</font color=blue> <font color=orange>Inside</font color=orange>
Don't step in the sarcasm!

 

[kief]

I tend to agree with (most) of the opinions here saying the P4 1700 isnt much faster then the P3 1000. The Athlon MP 2000+ is far faster then the dual P4. The P4 does far less per CPU cycle then wither the P3 or Athlon MP. So while there are more cycles when you take cycles * work done you get a lower number with P4 =(

Jesus saves, but Mario scores!!!