Tualutin-Celeron **IS** cache crippled compared to P3-S: Proof....

[Falkentyne]

Some proof for you guys:

"<none@ofyourbusiness.com> wrote in message
news:a0efqh$pc6$1@nyheter.chalmers.se...
[snip]

>
>
> Well, any thoughts? I am a programmer, and I have taken some basic
> CPU architecture classes. I have also done some assembler programming
> (M680000, MIPS R* family, Intel 8051), but I guess I'll take a look
> at Intel's interactive tutorial...

Exerpt of email from Intel marketing:

"As far as 1.2 GHz Celeron, it does have different features than the .13
micron Pentium III processor. 133MHz bus vs. 100MHz front side bus for one.
Also .13 micron Pentium III processor has the data prefetch feature, Celeron
1.2 GHz does not. We have targeted these processors for different market
segments with different features for each."

If you go read the data sheets, the PIII doc mentions the prefetch feature,
while the Celeron doc does not...

I would say offhand that either your contact is wrong, or my contact *and*
the data sheets are wrong. Don't know how to test this, however... :-).

Regards,
Dean "


Also, for those who want the data sheets:

"For PIII Tualatin:
http://developer.intel.com/design/pentiumiii/datashts/249765.htm
For Celeron-T: http://developer.intel.com/design/celeron/datashts/298596.htm

Enjoy!

Regards,
Dean"

[UnseenMenace]

I personally do not consider any statement from the company which produces the very product being discussed to be proof of anything for the simple reason, they have to justify charging more for one of the products. Intel is not going to tell people that the AMD processors are faster for a given Mhz even though there is evidence to back this up, and I view this information in the same manner.. The reply obtained is based on Intels business plan rather than facts, perhaps the information supplied is correct but In the same manner I would not take it as being factual or proof due to the fact the reply can effect Intels business

[ol' man]

Originally posted by UnseenMenace
I personally do not consider any statement from the company which produces the very product being discussed to be proof of anything for the simple reason, they have to justify charging more for one of the products. Intel is not going to tell people that the AMD processors are faster for a given Mhz even though there is evidence to back this up, and I view this information in the same manner.. The reply obtained is based on Intels business plan rather than facts, perhaps the information supplied is correct but In the same manner I would not take it as being factual or proof due to the fact the reply can effect Intels business

Yeah you got your story as does everyone. You say AMD chips are faster! At what? Quake maybe???? MPEG encoding??????? In some things actually the cel 1.2 beats the XP if you must know.

Check this out, notice it says the cel 1.2 has data prefetch. Only difference between it and the cumine is it has a cache latency of 1 and the cumine cache is 0. The prefetch allows the IPC to be higher though than the cumine.

[ol' man]

Also a cel 1.2 engineering sample on a 133bus at 1GHz has a higher IPC than a 1GHz PIII cumine. If you do not remeber the IPC of the Athy and the PIII were quite equal. Same here with the XP in most apps! I am sure maybe the XP will beat it but it is not going to be by much, I mean the XP has more L2 cache than the cel or cumine! You would have to expect the IPC's to be close sine I am sure AMD reverse engineered the PIII core to make the athy and simply added more cache and transistors. This is why they get so hot but the PIII and ATHLON IPC's are very close. Do not kid your self. You think the athlon is a totally fresh design with none of the workings of a PIII I have a bridge to sell ya too. AMD has always cloned intel, why would it be different now?

[Pinky]

yeah, common sense told me these chips were mucle-equals when i ran benchmarks and got the results proportional to a P3 at a 1.53ghz clock speed.

[Flu!d]

If you do a Sandra mem benchmark you will notice on the Celeron that the FPU and ALU scores are very close to one another, peform the test on a cumine P3 and you will notice a difference of about 50Mb/s....This indicates that the Celeron Tualatin 'does' in fact have data prefetch. It would want to, it's an integral part of the Tualatin core....

I think Intel, from a marketing point of view, would like to keep the Celeron's data prefetch under wrapps to make the P3, not to mention the 'slow at maths' P4 look better....

The difference really is only the FSB, The Celeron is basically just a 100mhz FSB P3. And these chips fly!

[JaY_III]

Originally posted by ol' man
You would have to expect the IPC's to be close sine I am sure AMD reverse engineered the PIII core to make the athy and simply added more cache and transistors. This is why they get so hot but the PIII and ATHLON IPC's are very close. Do not kid your self. You think the athlon is a totally fresh design with none of the workings of a PIII I have a bridge to sell ya too. AMD has always cloned intel, why would it be different now?

I wouldnt say the Athlon is a reversed enginered P!!! @ all. True that they make clones (and not to much new stuff has come out of AMD historicly)...
Its that AMD hasnet even made thier last few cores.
Ever heard of a company called NextGen ??? Yeah, AMD bought them out and that was the K6 family of cores, a Nextgen product...

About the Athlon though, i would say its the ALPHA in it that makes it so fast/hot. AMD is using the EVE6 bus (the ALPHA BUS)and has some ALPHA engineers now... And that is your athlon... a scaled down 32bit alpha made to run on a X86 platform..... (well sort of)

Dont get me wrong or anything, I am not saying its bad, Intel now owens APLHA, not just a few engineers, but ALPHA and all its assets.... (64BIT home PC's is coming :) )

Think ATI had anything to do with the GAMECUBE and its FX?? No they didnt, they just bought the company out that made the graphics chip.... so in the semiconductor market, that is fair game.

oh yeah, this post is about the Tuatutin CPU's...
Shame Intel has its head to far up its.............. to realize how good these Tualutin CPU's are....
they make the P4's look so bad. Good one intel.

[Kingslayer]

All AMD chips are reverse engineered Intel chips. There is a reason for this. Intel patents everything that they do, therefore AMD can't make exact duplicates. So, AMD reverse engineers the chips and has to do things different ways. Sometimes it works good (Athlon's) sometimes it doesn't work so good (AMD MP).

Nothing wrong with reverse engineering. It's been happening ever since the first of something was made.

[ol' man]

Originally posted by Kingslayer
All AMD chips are reverse engineered Intel chips. There is a reason for this. Intel patents everything that they do, therefore AMD can't make exact duplicates. So, AMD reverse engineers the chips and has to do things different ways. Sometimes it works good (Athlon's) sometimes it doesn't work so good (AMD MP).

Nothing wrong with reverse engineering. It's been happening ever since the first of something was made.

I still think the athlon has some of the PIII workings in it. I mean look at the IPC of the two chips! They are very close. What if you gave the athlon the cache of the PIII? I bet they would almost be equal. Even the PIII has a dual bus like the athlon! Who had it first? I bet it was Intel.

[BigRed]

I still think the athlon has some of the PIII workings in it. I mean look at the IPC of the two chips! They are very close. What if you gave the athlon the cache of the PIII? I bet they would almost be equal. Even the PIII has a dual bus like the athlon! Who had it first? I bet it was Intel.
i would say neither
nexgen started designing and working on the K7 processor at the same time intel started work on the next generation p2s (p3). when the p2 came out intel already started desiging the p3. this was the time that amd took over nexgen who was desiging the K7.
actually i *think* (not 100% sure about this) that nexgen actually started designing the K7 before intel started on the p3
because i remember a mag article about processors a long while ago that said that after NexGen finished work on the Nx686 they started devlopment on a new project (i forget the funky codename they used). but thats when amd bought out the company and slapped a new name on the Nx686, aka K6. and amd shrouded the previous companys projects from public eyes.
but because intels R&D funds are far far far greater than amd they managed to finish their projects far ahead of amd :p

[minoukat]

Originally posted by ol' man
I am sure maybe the XP will beat it but it is not going to be by much, I mean the XP has more L2 cache than the cel or cumine!


actually, I think you'll need to clear your sources (no offense, as there isn't one meant), but the Athlon (Thunderbird *AND* Palomino (athlonXP)) have the same amount of L2 cache as the Celeron. They both have 256KB, just the L1 cache is bigger (128KB if I'm not wrong)

EDIT : just to clarify a little, the Athlon model that had more L2 cache was the Athlon Classic, which had 512KB, but that were at about half the speed of the CPU, as that cache wasn't on the die. The Classic was still made out of .25 micron, so there wasn't enough space left inside the core to add these new transistors. Then, with the Thunderbird, that was made of .18micron process, they added the 256KB of L2 cache on the die, running at full processor speed

[ol' man]

Originally posted by ol' man


I still think the athlon has some of the PIII workings in it. I mean look at the IPC of the two chips! They are very close. What if you gave the athlon the cache of the PIII? I bet they would almost be equal. Even the PIII has a dual bus like the athlon! Who had it first? I bet it was Intel.

Actually the PPRO core is basically what the PII and PIII used with a few minor reworkings and ide shrinks. I would think nexgen would have ripped aprt the PPRO to find out what made it tick. It was a very good chip in its day! You think nexgen started from scratch. I highly doubt it! The two chips in the way they operate are too close together.

As for the cache BS, yeah I screwed up and I meant total cache come on you had to have known what I meant. The total on die cache of the athlon is higher than the PIII's total cache. Thing is the PIII-s chips have more on die cache than the athlons or XP's. Must possibly be why they beat them out in many apps!

Everyone has been ripping off the PPRO core design over the last few years.

[Flu!d]

One of the main reasons why the early slot 'A' Athlons and slot '1' P3 Katmais and P2's used 512 k 'roughly half' speed cache as compared to the 256k full speed cache of the current chips is because we coulden't make cache fast enough to handle the high speeds of full speed cache back then.

[ol' man]

Originally posted by Flu!d
One of the main reasons why the early slot 'A' Athlons and slot '1' P3 Katmais and P2's used 512 k 'roughly half' speed cache as compared to the 256k full speed cache of the current chips is because we coulden't make cache fast enough to handle the high speeds of full speed cache back then.

Actually they could! They used it in the XEONS though. Why all of a sudden did AMD start using on die cache a little later than intel did? I am going to guess it was later too! The cel 300a had on die L2 cache before any of them remember! It was reverse engineer time again for Nexgen, AMD or whatever!

Ironically AMD did not release a chip with on die cache until two years after the first cel 300a was released which had on die full speed cache! Hmmm, you think AMD even bothered to buy a cel 300a and tear it apart:D Nah:D

[JaY_III]

"Hmmm, you think AMD even bothered to buy a cel 300a and tear it apart Nah"

No but intel sure did.
They didnt even know on-die cache would work as well as it did.
thats why P!!!'s has off-die cache to start with....

Not till people started OC'in the 300A's and 333a's to 450-500 did intel start to get the idea of how much better on die cache was (celeron a's overclocked to a 100FSB outpreformed PII's on 100FSB at the same speed)...
And that you didnt need as much (as the latency was lower speed faster) P2's has 512 L2 off-die as did early P3's

See not even intel has a clue, they just make it up as they go along.... its just Intel makes it up 1st.

[ol' man]

I erred folks! The K6III had on die 256k cache!

[Flu!d]

Actually they could! They used it in the XEONS though

Yeah....True, but how much was a Xeon back then!! The reason Xeon's are so expensive is because of their vast amount of fast cache....

[ol' man]

Originally posted by Flu!d


Yeah....True, but how much was a Xeon back then!! The reason Xeon's are so expensive is because of their vast amount of fast cache....

Well I didn't put a price on it though! You want it back then, you had to pay for it!