P4 core for RC5 seems like a pipe dream

[denali]

After reading this article it would seem impossible to make a P4 core that is faster than the P2/P3 core. I don't know if Train is still looking into this or not but thought it might be of interest to some.

 

[Train]

I actually havent worked on it in a while, but i read that article earlier today, there was a thread on it in GH, where i posted some comments.

That article is very innacurate, biased and just plain stupid.

 

[denali]

I'll go over to GH then and read your comments. Guess I'll have to spend more time in the various forums.

 

[BoberFett]

Well, if it could be done you'd think there'd be one by now.

 

[Moose]

We simply do not have access to a P4 computer. Bruce Ford (our fine coder who brought us many MMX cores in the past) said he would look at it while riding on the bus/train to and from work (or something to that effect.)

I'm sure we will get one sometime in the near future...

moose

 

[BoberFett]

I'm not implying that you're putting it off Moose, not at all. But in my [twisted?] mind, I would think Intel would have set somebody on the task. It's no secret that right now they need all the impressive benchmarks they can muster if they want the P4 to take off.

 

[Train]

Intel has open source examples of SSE2 free to download.

http://developer.intel.com/software/opensource/numerics/index.htm these may give you some more confidence in its Mathematical abilities

 

[Moose]

By that logic Intel should be porting every bit of code from games to OS's so it runs better on the P4. It should not be put up to the chip maker to make applications take advantage of new chips. Sure I can see them doing something like FlasK MPEG which was used specifically in a popular review site (tom's hardware guide.) As well I can see them showing companies howto port apps to run better on the P4 for benchmarks. All chip makers do it. Look at Apple and the G4. The Photoshop plugins in their demo were written to use and show off the Alivec unit. Thats great they showed how fast it was and what it could do compaired to a PIII. And don't get me wrong its great. But I'm sure that Intel could and they prob do have applictions and specs that blow the G4 out of the water with the PIII and with apps that were rewritten to take advantage of the PIII.

Its all about public understanding of what the numbers mean. Sure to you and me on RC5 to say that the chip can do 3.7Mkeys/sec on a P4 when optimized where a PIII does 1.5Mkeys/sec Makes sense. To somone who has never heard about RC5 that makes no sense and the person has no reference point.

Of course at this point in time I am just rambuling and I have to hed out the door.

moose

 

[BoberFett]

No, by my logic Intel would port small programs which are frequently used as benchmarks to demo their hardware. I wouldn't expect them to port JoeBob's Donut Shop Management program to every new instruction set.

<< Its all about public understanding of what the numbers mean. Sure to you and me on RC5 to say that the chip can do 3.7Mkeys/sec on a P4 when optimized where a PIII does 1.5Mkeys/sec Makes sense. To somone who has never heard about RC5 that makes no sense and the person has no reference point. >>

Really? Here's some Intel marketing, are you going to try to tell me that those numbers actually mean something to the consumer? No, it's marketing. If Intel said that the P4 scored 400% higher on the BunchaRandomNumbers test than the P3, the average consumer wouldn't know the difference.

 

[Russ]

Interesting article. Even as a big fan of AMD, even I can see that this guy has a major hardon for Intel. Not only is he ranting hysterically, the work is packed with factual errors, and misleading assertions. Example:



<< When that didn't pan out, Intel kept selling its older line of Pentium MMX chips. While running at the same 233, 266, and 300 MHz clock speeds as the Pentium II, the Pentium MMX was based on the older design of the original Pentium >>



Intel did not produce the Pentium MMX in speeds faster then 233, unless they did so in the mobile market?

Another example:



<< Buying a Celeron was like buying an old 486 system, it was that slow. >>



I guess he doesn't realize that people who actually USED the original Celery would be reading his article. While the lack of L2 cache did make it slower then an equivalently clocked P2, it was certainly NOT nearly as slow as a 486. Most work uses L1 cache, so much of the time, L2 isn't even a factor.

He does have a valid point, though: The P4 is overpriced when we consider the very small incremental improvement over the P3.

Russ, NCNE

 

[ViRGE]

Russ, the PMMX went up to 300mhz in the mobile market.

 

[SpaceWalker]

Russ,

I had 2 of the original celeron, the 266 (which ran at 400 like a charm) and the 300, only when the programs where using FPU instruction did you see a speed boost, but the 300 celeron did performed worst than a P233mmx with 512kb L2 at most tasks.....

MOOOOOOOOOOOOOOOOOOOOOooooooooooooooooooooooooo..............

 

[Moose]

As much as I would love to beleive that the world revolves around distributed.net. It doesn't. Sure people in this forum know all about distributed.net and rc5 benchmarks. But from a marketing stand point youhave to give numbers that people that the marketing is geared towards will understand. Frankly rc5 is not one of those things. Just saying that X app is 400 times faster onthe P4 just doesn't do it when you have a large majority of people that don't understand what X app does.

If you can get Intel to produce a core we will take it. If they don't we will produce one as we always have... with a little timeand effort we will produce one....

moose