LubieCipy wrote on 2021-12-10, 10:17:

Thanks for documentation 😀 I will check more information on this

Hey, no problem. Keep digging! 😀👍

LubieCipy wrote on 2021-12-10, 10:17:

There is a simpler explanation. Maybe Data Prefetch Logic is a feature of Pentium 3 tualatin only

I believe "Data Prefetch Logic is a feature of Pentium 3 tualatin only" is correct. But since this is an integral part of Tualatin microarchitecture which can't just be removed that would make "Tualeron" Celerons not Tualatin-based.
'Intel rereleased Coppermine on a newer process with higher clocks under the Celeron brand to compete in the budget segment." is an easy enough explaination for me. 😉 But it's just a hypothesis, maybe there are bigger changes to the core. If anyone knows more please chime in!

The issue was discussed in this old thread.
https://www.overclockers.com/forums/showthrea … ion-Accelerator
When you scroll down to [02-05-02, 12:54 PM] you find a comment by an Intel spokesperson that confirms that 'Celerons' also have data prefetch logic.
The main difference for end users between Tualatin 512/256 and Celeron 256 is that Celerons run on 100 MHz fsb.

PARKE wrote on 2021-12-10, 12:05:

The issue was discussed in this old thread. https://www.overclockers.com/forums/showthrea … ion-Accelerator When you scroll down […]
Show full quote

The issue was discussed in this old thread.
https://www.overclockers.com/forums/showthrea … ion-Accelerator
When you scroll down to [02-05-02, 12:54 PM] you find a comment by an Intel spokesperson that confirms that 'Celerons' also have data prefetch logic.
The main difference for end users between Tualatin 512/256 and Celeron 256 is that Celerons run on 100 MHz fsb.

I have discussed this response by Intel support previously in this thread. Official documentation suggests they were wrong. At least the part where they say

This information is also detailed in the home page for the Intel Celeron Processor.

is wrong because Data Prefetch Logic is not listed as a feature of the "Tualeron" Celerons.

All the proof/documentation they provide relates to Pentium III processors only. They do not refer to any Celeron documentation.
I think their mistake was assuming the "Tualeron" Celerons are based off Tualatin while there is no indication of that (that I know of).

Edit: Actually seeing how they seemingly base their answer on the statement

The Intel(R) Celeron(R) Processor that you are referring to is the same.13 Micron Technology that the Pentium(R) III Processor is.

they seem to be confused about production process vs. micoarchitecture. Possibly they assumed 'same production process = same microarchitecture' and based their response on that, i.e. (possibly) falsely assuming 'same production process as Tualatin = same feature set as Tualatin'.

You may well be right, I don't know.
There is an article online that attempts to prove that the 13 micron Celeron does not support DPL but the way how the benchmarks are set up does not convince me.
https://www.realworldtech.com/data-prefetching/
What I would for example like to see for this particular issue is a comparison between a 1 GHz Tualatin Celeron like SL6CB or SL6JT running overclocked @ 10x133= 1.333GHz and a 1.33 GHz Tualatin 256 like SL6BZ running @ stock. That is imo the only fair way to see if a Tualatin Celeron is indeed inferior to a Tualatin and a comparison between just two cpu's -one of each type- would not necessarily convince me either.

The benchmarks in this very interesting thread are focused on overclocking ability and that is in my book not the same as assessing differences in basic performance. But, opinions can of course differ.

What I would for example like to see for this particular issue is a comparison between a 1 GHz Tualatin Celeron like SL6CB or SL6JT running overclocked @ 10x133= 1.333GHz and a 1.33 GHz Tualatin 256 like SL6BZ running @ stock. That is imo the only fair way to see if a Tualatin Celeron is indeed inferior to a Tualatin and a comparison between just two cpu's -one of each type- would not necessarily convince me either.

Unfortunately there's also L2 cache latency difference. Not so sure that data prefetch is useful in comparisons without active L2.

PARKE wrote on 2021-12-10, 16:08:

That is imo the only fair way to see if a Tualatin Celeron is indeed inferior to a Tualatin and a comparison between just two cpu's -one of each type- would not necessarily convince me either.

My benchmarks here may be of interest to you:
Re: Tualatin Celeron vs Williamette Celeron
Five socket 370 CPUs benchmarked at an equal 7.5 x 100 = 750 MHz.

PARKE wrote on 2021-12-10, 16:08:

You may well be right, I don't know.

Dude, I don't kow either! 😅 It's just a hypothesis based on that Intel support representative giving a sort of inconvincing statement once I looked into their sources.

PARKE wrote on 2021-12-10, 16:08:

There is an article online that attempts to prove that the 13 micron Celeron does not support DPL but the way how the benchmarks […]
Show full quote

There is an article online that attempts to prove that the 13 micron Celeron does not support DPL but the way how the benchmarks are set up does not convince me.
https://www.realworldtech.com/data-prefetching/
What I would for example like to see for this particular issue is a comparison between a 1 GHz Tualatin Celeron like SL6CB or SL6JT running overclocked @ 10x133= 1.333GHz and a 1.33 GHz Tualatin 256 like SL6BZ running @ stock. That is imo the only fair way to see if a Tualatin Celeron is indeed inferior to a Tualatin and a comparison between just two cpu's -one of each type- would not necessarily convince me either.

The benchmarks in this very interesting thread are focused on overclocking ability and that is in my book not the same as assessing differences in basic performance. But, opinions can of course differ.

I was first thinking of a similar approach - trying to figure things out through benchmarks @1GHz (also including the 1GHZ 100MHz FSB Coppermine ) - but as others have pointed out there are more variables than Data Prefetch. 🙁

gerwin wrote on 2021-12-10, 17:03:

My benchmarks here may be of interest to you:
Re: Tualatin Celeron vs Williamette Celeron
Five socket 370 CPUs benchmarked at an equal 7.5 x 100 = 750 MHz.

Hey, awesome! Thanks! 😁👍
@7.5x100MHz the "Tualeron" looks awfully similar to the Coppermine Pentium III.
Are there any timing differences or anything else that would explain why they score so similarly IF the "Tualeron" has Data Prefetch Logic?
Or does this benchmark support my hypothesis of the "Tualeron" actually being based off Coppermine?

Doornkaat wrote on 2021-12-10, 18:18:

@7.5x100MHz the "Tualeron" looks awfully similar to the Coppermine Pentium III.
Are there any timing differences or anything else that would explain why they score so similarly IF the "Tualeron" has Data Prefetch Logic?
Or does this benchmark support my hypothesis of the "Tualeron" actually being based off Coppermine?

One of the probems with comparing the Celeron Tualatin cpu with another cpu at fsb 100 is that the performance of data prefetch ability goes hand in hand with the fsb. Positive effects of data prefetch are already challenged at fsb 133 when compared with contemporary Athlon and P4 and fsb 100 is probably making things worse. (I read somewhere that this may be the reason why it was not mentioned in the Tualatin Celeron datasheets - but who knows ?)
In order to see a possible effect more clearly -it seems to me- that comparing a Celeron Tualatin overclocked @ fsb 133, with a Tualatin 256 cpu that runs stock at fsb 133 -and- that is known to use prefetch, could maybe give a better idea. The only difference then occurs in cache latency which will likely only have a very minor effect on any outcome.

If the yes- or no- application of data prefetch is a sufficient criterion to distinguish between Coppermine and 'real' Tualatin seems more a personal choice ?. It ignores some other technical differences: 1.25v AGTL, .13 micron fabrication and the use of copper interconnects, to name a few.
Just my 2 cents.

Looked at some of that here.

Geforce 5900XT, Asus P3B-F, 1GB RAM 2-2-2 settings.

Pentium III-S 1,4GHz (133 FSB) , 3DMark01SE 10200
Celeron 1,3GHz @1612MHz (124 FSB), 3DMark01SE 9300

Pentium III-S @1570MHz (150 FSB), 3DMark01SE 11500

How do disks and IDE devices handle the 200+ FSB? Any disk drives or flash devices seem to handle the high PCI frequency better than others?