Pentium and Celeron: what's the difference?

oldwrench · Post by **oldwrench** » Sun Jun 13, 2004 1:05 am

The AMD 64 also put the memory controler on the chip with the processor. 8) This eliminates the bottleneck of running the memory bus through the motherbord chipset, the memory now runs at prossessor speed. Intel tried to match this by putting a huge cache in the P4EE (Extremely Expensive) edition.

the Black Monarch · Post by **the Black Monarch** » Tue Jun 22, 2004 5:08 pm

As much as AMD likes to harp on the Pentium 4s being so inefficient, they rarely like to acknowledge that having a 50% higher clock speed is an equally valid way of achieving higher performance. The 3.6 GHz Pentium 4s (socket 775 only) are finally finding their way into people's systems, and AMD's best are 2.4 ghz. When they both execute 21.6 billion instructions per second, what difference does it make whether that was accomplished through efficiency or brute clock speed? I mean, aside from AMD's new 250MHz system bus and integrated memory controllers and Hypertransport and 64-bit extensions and... okay, AMD is better, but it has nothing to do with being more efficient.

Hey DW, remember when you told me that the two differences between the 64 FX and the non-FX chips was that the FX had a 10% higher clock speed and an integrated memory controller? Well, I got damn near run out of AMDzone.com when I started passing this "info" along. They both have integrated memory controllers, you bitch! The ACTUAL difference (aside from clock speed) is that Socket 754's memory controller is single-channel while 940/939 chips have a dual-channel memory controller.

And I'm still not too happy about AMD claiming that the Athlon 64 is "8th-generation" when it's really just the 7th-gen Athlon XP with an integrated memory controller and 64-bit extentions.

Now I'm going to leave this thread to finish work on my new Dual Socket A system. I'm going to see if I can get these Durons up to 2.4 GHz...

My apartment currently contains a 486, two Pentiums, a Pentium Pro, a K6-II, a Pentium III, two Durons, two low-voltage Athlon XP-Ms, an Athlon 64, and a Pentium 4. The Athlon 64 will be departing soon, to be replaced with an Opteron. I love technology

dwchang · Post by **dwchang** » Tue Jun 22, 2004 6:57 pm

the Black Monarch wrote:As much as AMD likes to harp on the Pentium 4s being so inefficient, they rarely like to acknowledge that having a 50% higher clock speed is an equally valid way of achieving higher performance.

We do eh? I seem to recall me posting that Performance = Speed * instructions per clock. Oh wait, I bet you didn't read that...

Wow you sure got me and my company with your not reading ability.

the Black Monarch wrote:I mean, aside from AMD's new 250MHz system bus

Our what?

the Black Monarch wrote:Hey DW, remember when you told me that the two differences between the 64 FX and the non-FX chips was that the FX had a 10% higher clock speed and an integrated memory controller?

No, I don't recall saying anything about a memory controller. I did say that the FX is slightly faster, but not 10%. In one iteration it was 200 Mhz, so roughly 8% then.

I also recall saying that the Opteron has 3 HT links while the Athlon-64 does not. Also that the FX has a 1 MB L2 cache while the initial Athlon 64's had 256 KB. BTW the 64's are now have *up to* 1 MB, but when I said that, it was true.

I guess what I'm getting at is that you need to read more carefully before trying to "refute" something.

the Black Monarch wrote:And I'm still not too happy about AMD claiming that the Athlon 64 is "8th-generation" when it's really just the 7th-gen Athlon XP with an integrated memory controller and 64-bit extentions.

Yes because redesigning your North-Bridge, IOs, adding 64-bit extensions, etc. is really easy to do and not a big deal. Thank you Mr. Computer Architect.

So where have you been published again?

DaNuKa_SAN · Post by **DaNuKa_SAN** » Wed Jun 23, 2004 1:19 am

We must also not forget that AMD also implements more pipelines in the XP series and athlon...making it possible to have an inferior clockspeed but enough routs in order to make up bringing it equal to a higher core spped processor with fewer pieplines

quick example, AMD XP 2500+...actually has a core speed of 1.83 Ghz...does the same job as a 2.5 Ghz processor with fewer pipelines just about.

dwchang · Post by **dwchang** » Wed Jun 23, 2004 12:11 pm

DaNuKa_SAN wrote:We must also not forget that AMD also implements more pipelines in the XP series and athlon...making it possible to have an inferior clockspeed but enough routs in order to make up bringing it equal to a higher core spped processor with fewer pieplines

quick example, AMD XP 2500+...actually has a core speed of 1.83 Ghz...does the same job as a 2.5 Ghz processor with fewer pipelines just about.

Uhm what?

The Athlon and Athlon XP have identical cores and have for some time now.

Also more pipelines = more clockspeed generally. It's a trick Intel has used to get such high frequencies.

I believe the P4 has ~30 - 40 stages (depending on if it's an FP instruction) while the Athlon has 12 or 13. With 30 - 40, you can have speeds of 3+ Ghz since you're not doing as much per stage (efficiency) while with 12 - 13, you're doing more per cycle, but at a slower speed.

There isn't this internal speed difference between Athlons and Athlon XPs. They're effectively the same die with the same number of stages.

DaNuKa_SAN · Post by **DaNuKa_SAN** » Wed Jun 23, 2004 3:59 pm

Oops...yur rite...

i must have been thinking of both of them but it came out wrong..

of course they have the same cores...

though higher number of PLs doesnt necessarily mean more clockspeed...

the Black Monarch · Post by **the Black Monarch** » Wed Jun 30, 2004 4:23 pm

dwchang wrote:We do eh? I seem to recall me posting that Performance = Speed * instructions per clock. Oh wait, I bet you didn't read that...

I did. But what I've never read is AMD saying "Yes, the Pentium 4 gets way higher clock speeds than anything we've done, and that's how they barely manage to keep up with us."

dwchang wrote:Our what?

Haha, I wish I could have seen the look on your face when you read that. Anyway, never mind, I just learned that unlike Front-Side Bus speeds, your Hypertransport bus speeds have absolutely no relation to the system bus speeds. Pretend I didn't say anything about it.

Not that I can't overclock it to 250mhz if I want to

dwchang wrote: No, I don't recall saying anything about a memory controller. I did say that the FX is slightly faster, but not 10%.

I'm glad I don't delete this stuff:

"Athlon 64 FX - Higher frequency 2.2 Ghz I believe... and 128-bit memory controller
-Athlon 64 - 2.0 ghz"

Correct me if I'm wrong, but isn't 2.2 ghz exactly 10% faster than 2.0 ghz?

dwchang wrote: In one iteration it was 200 Mhz, so roughly 8% then.

Wow, only 200mhz? That's one shitty FX... what is it 8% faster than? A 486?

dwchang wrote:I also recall saying that the Opteron has 3 HT links while the Athlon-64 does not.

Fortunately, that part turned out to be true.

dwchang wrote:Also that the FX has a 1 MB L2 cache while the initial Athlon 64's had 256 KB. BTW the 64's are now have *up to* 1 MB, but when I said that, it was true.

Actually, you never told me that the FX would have 1MB and the regular 64s would have 256kb. In fact, you never gave me any specific numbers for the caches. What you said was that the Opteron would have a larger cache than the other two, which turned out not quite true

BTW, I ditched my socket 754 chip and board for an Opteron 146 a while back. It is not simply a superior chip for servers, it's a superior chip for everything.

dwchang wrote:I guess what I'm getting at is that you need to read more carefully before trying to "refute" something.

And what I'm getting at is that not only do I read better, but I remember better, too

dwchang wrote:Yes because redesigning your North-Bridge, IOs, adding 64-bit extensions, etc. is really easy to do and not a big deal. Thank you Mr. Computer Architect.

Compared to designing a completely new architecture from the ground up, it IS easy to do and it's NOT a big deal. Certain changes were made between the Pentium Pro, PII, and PIII, but they're still the same generation of architecture.

For the record, the older Pentium 4s have a 20-stage pieline and the Prescott has a 31-stage pipeline. Not 40. And what's really funny is that all these extra stages didn't do Intel a damn bit of good, since the pipeline wasn't what was holding clock speeds back. Intel even had trouble getting Prescott to reach the same speeds that Northwood had been at for a while. It's so much fun to laugh at Intel pathetically stumbling all over its own stupidity...

DaNuKa_SAN · Post by **DaNuKa_SAN** » Thu Jul 01, 2004 1:48 am

*lookin at wat the man above said*

AMEN!

the Black Monarch · Post by **the Black Monarch** » Tue Jul 13, 2004 2:48 pm

I know, it's kind of funny to see DW get bitchsmacked in an argument related to CPUs, isn't it?

I love my Peltier unit.

dwchang · Post by **dwchang** » Tue Jul 13, 2004 3:25 pm

the Black Monarch wrote:I know, it's kind of funny to see DW get bitchsmacked in an argument related to CPUs, isn't it?

I love my Peltier unit.

First off, I won't even begin a rebuttal since it doesn't do any good regardless of facts and whatnot. Just think you're 100% right and live in your own little world. You seem to do a pretty good job of it. I guess the simplest way of putting it is, you're not worth it. Oh wait, I'm replying again so I guess I sorta broke that

And actually you've *never* bitchslapped me and I've learned in time you're just as retarded as Hatter in terms of *trying* to argue and being a lurking troll.

I think 99% of the .org already sorta knew this and if you haven't noticed most people tend to ignore you or just make fun of you...so with that:

STFU

I think it'd be best for me (and the rest of the .org) to just ignore you since you don't bring any real value to the world and most people have come to the same conclusion. Just took me awhile...