Processor clock speed project

[Soral 3.0]

Hi all!
I am doing a presentation over Intel's potential new naming system for their CPU's. Instead of Intel naming the processor according to clock speed, they now have a funky numbering system. ex. Instead of Pentium 3.2 Ghz, it's gonna be Pentium 500.
Superficially, this is a flawed decision that will confuse the uninitiated consumer, because they believe that clock speed is the only benchmark for CPU speed.
What is the best approach to presenting CPU architecture so that I can inform an A+ certification class that clock speed isn't the sole measure of speed?
Thanks
-Soral 3.0

[Catweazle]

Soral, if you can't pull that off the top of your head, you've got problems instore for you!

Clockspeed means nothing because it can only compare one model of a particular processor to another model of the same processor. Use a different type of processor, even if it's the same brand, and the internal architecture will be different and it will use a different clockspeed to get the same result.

Want a really good demonstration? Get 3 notebooks which all give comparable scores from benchmark tests. Make sure one uses an Athlon64, the second a Pentium 4 mobile, and the third is a Centrino notebook with a PentiumM processor. Sit a sign beside each showing the processor clockspeed.

The Pentium 4 Mobile will have the highest clockspped, the Athlon64 will have a considerably lower clockspeed, and the PentiumM will be almost half the clockspeed of the Pentium 4 Mobile!

Same amount of work done, different clockspeeds.


The decision to move to a new system of numbering is a responsible one, not a flawed one, because comparing clockspeeds is the stupidest, most meaningless method of comparison that exists. Those 'uninitiated consumers' have fallen prey to the stupidity because Intel foisted it on them in the first place, and the fact that Intel has finally made a move away from the practice is to be applauded.


And, of course, you might want to make mention that the change is mostly motivated by the fact that processor development has basically 'hit a wall' at around the 3 to 4 GHz mark, which is a point beyond which current technology can't go, and that future processors will use different architecture and use less clock cycles to perform more work.

'Big Blue' indeed!

[NiTeHaWk4787]

For instance, a Pentium 3.4ghz not OC'ed may not be as good as an AMD 3500+ at 2.2ghz because AMD's do more work per process, which generates less heat and doesnt require a huge amount of ghz's. Pentiums do less work per process which is why they have to have a faster process rate therefore generating more heat and the need for throttling. It would be nice to have throttling on all CPU's for added safety but unless your a complete moron, you wont let your computer CPU rise high enough to do damage.. well unless your heatsink falls off lol. And about the Pentium vs AMD thing I just said, some parts of it may be a bit off so correct me if I said anything completely off the chart there.

[Soral 3.0]

Thank you guys for reinforcing my understanding of RISC and CISC architecture.
I guess that I was trying to see if anyone believe that it was necessary to talk about cache also -- and demonstrate how that profoundly affects the CPU's ability to process (i.e why celerons have a decent clock speed but are still bad processors).

-Soral 3.0

P.S.
Catweazle, I see why Intel is changing the name, and I believe it is a good move for them, But people who use clock speed to gauge the worth of a processor think it's the stupidest thing since the solar-powered flashlight.

[Catweazle]

But people who use clock speed to gauge the worth of a processor think it's the stupidest thing since the solar-powered flashlight.

All the more reason to hit 'em in the face with a funky demonstration! :D

Seriously, it makes no difference whatsoever using logic. You got to make people doubt what they've believed previously by shocking them out of it. Telling them why won't change opinions. You've got to get them ASKING why.

[DuncanIdaho]

It'll all be well and good if any Pentium over Pentium 500 runs faster than the previous. If Intel does stupid things like come out with a Pentium 510a that is actually slower than the 500, it'll suck, and lead to great confusion. This is my only concern on the change to the CPU numbering system, it has happened with other products, (Nvidia's graphics cards come to mind, you needed to do a LOT of research before buying an Nvidia card a few years ago, when GF2's, GF3's, GF4's, and MX's were all floating about at the same time...GF4MX's often could be beaten easily by GF3's, even some GF2's could outperform some of them!).

Here's hoping that Intel's not opening a kettle full of worms, or worse, cockroaches.

[alc6379]

It'll all be well and good if any Pentium over Pentium 500 runs faster than the previous. If Intel does stupid things like come out with a Pentium 510a that is actually slower than the 500, it'll suck, and lead to great confusion. This is my only concern on the change to the CPU numbering system, it has happened with other products, (Nvidia's graphics cards come to mind, you needed to do a LOT of research before buying an Nvidia card a few years ago, when GF2's, GF3's, GF4's, and MX's were all floating about at the same time...GF4MX's often could be beaten easily by GF3's, even some GF2's could outperform some of them!).

Here's hoping that Intel's not opening a kettle full of worms, or worse, cockroaches.

That's basically what they've done, though. I forget the numbering system off the top of my head, but the rightmost signifigant digit indicates the chip "series", and the 10's and 1's digits actually compare the performance and "features" relative to the other chips in the series. An example:

The 500 and the 600 series of P4s aren't comparable.

in the 500, there's 515, 525, and 535. 525 should outperform the 515, and 535 should outperform both of the lower numbers. But...

Then, there's the 600 series. Say, 615, 625, and 635. Same as before , 625 outperforms 615, 635 outperforms 'em both. But here's the funny stuff....

Does a 615 outperform a 515? Maybe... maybe not. The numerical designations are only relative to the specific series they're a part of, kind of how an AMD XP2500+ with a Barton core can outperform an XP2600+ with a T-Bred core. Intel's not going to guarantee that even a 635 will outdo a 515-- that's not how they designed things. Kind of like how a 1.8ghz Duron will not outperform the 1.53ghz Athlon XP 1800+.

Measuring things by clockspeed has been dumb for quite some time now. What's the XP3200 at? 2.2ghz? Yet it performs up there with Intel's 3.2ghz P4? That's 1000mhz of clockspeed difference! Thank you guys for reinforcing my understanding of RISC and CISC architecture.
I guess that I was trying to see if anyone believe that it was necessary to talk about cache also -- and demonstrate how that profoundly affects the CPU's ability to process (i.e why celerons have a decent clock speed but are still bad processors).

RISC and CISC doesn't come into play too much here. 32 vs 64 bits does, but you've made a good point about cache size. That's what Intel's trying to push; maybe, for example an Intel 525 chip might be 3.8ghz running with 512k of L2 cache, but the 535 might still be 3.8ghz, but have 1 or 2MB of cache, which makes a difference. That's what caused Intel to put the "performance and features" qualifier on the numbering scheme, so people aren't biased by the clock numbers.Disclaimer: I don't know many, if any, of the specs on the new Intel processors. The clock speeds and cache sizes were just pulled out of thin air, meant to be examples. Don't go out and buy one of the processors because of what I posted!

[Soral 3.0]

Hey thanks for answering my question about cache alc6379.
But what do you mean RISC and CISC doesn't play a part here?
I always thought that Intel was going for RISC and AMD was basing its CPU tech over CISC....
Of course, I learned that from word-of-mouth so I could be misinformed...
Could you clear me up on that?

-Soral 3.0

[alc6379]

Hey thanks for answering my question about cache alc6379. But what do you mean RISC and CISC doesn't play a part here? I always thought that Intel was going for RISC and AMD was basing its CPU tech over CISC.... Of course, I learned that from word-of-mouth so I could be misinformed... Could you clear me up on that?

-Soral 3.0

Well... here's what it boils down to, as well as I know it.

The reason why RISC and CISC don't come into play here is because it's all CISC. I think that the Itanium processors are RISC, but they're generally not compared in the same leagues as AMD-64s or the Pentium 4 chips. Intel offers both RISC and CISC chips, but in the desktop (not workstation) sectors, they're pushing their 32-bit CISC Pentium 4s and Xeons.

Where it would matter was if you were comparing Apple, Sun, or IBM hardware. Apple's G5 Chips, Sun's SPARC chips, and IBM's POWER chips are all RISC, and it would make a difference when comparing them to your standard x86 architecture. RISC stands forReduced Instruction Set Computer, and CISC stands Complex Instruction Set Computer. Basically, the RISC processor is supposed to have a smaller command set that you can issue it, which is supposed to mean that you do more with less, and speed things up. That's why an 800mhz HyperSPARC processor is "faster" than an 800mhz Pentium 3 processor. (well, only partially why. The HyperSPARC has WAY more cache memory, too)

But, in the case of Intel vs. AMD processors on the desktop/x86 compatible market, you're really going to be best served by looking at benchmarks. It's been like this for a good while now-- Since there's a big divergence in clock speeds between AMD and Intel, it's apparent that something else needs to be used as a measuring stick-- there are many things you can do to speed up a chip, like increasing the FSB speed, adding more L2/L3 cache memory, etc, and those are starting to make more of a difference than whether a chip is RISC,CISC, 32/64 bit. or whatever.

[Soral 3.0]

Thank you for clarifying that before I started to make a fool of myself by reffering to RISC and CISC in my presentation. Your info was very helpful!
(btw, my presentation went well, and I was complimented by my instructor :-))
-Soral 3.0

[alc6379]

Thank you for clarifying that before I started to make a fool of myself by reffering to RISC and CISC in my presentation. Your info was very helpful! (btw, my presentation went well, and I was complimented by my instructor :-)) -Soral 3.0

Awesome!

Since I kind of work "in the industry", I got to see a presentation on the whole deal. I wish they'd've made the numbers indicate relative performance across all lines, but go figure.

For more information, check out http://www.intel.com/products/processor_number .

[Soral 3.0]

Thanks again all!
-Soral 3.0