Nice article. However I don't know if you meant it that way, but it suggests the vrm and AGTL changes are related, which I don't believe to be the case. Also, the voltage clamp on some slotkets is for the so-called non vcore voltages, not AGTL. I don't think any slotket including Powerleap changes those.

Coppermine-T has a heatspreader.

https://www.cpu-world.com/CPUs/Pentium- ... ne-T).html

Neato!

An interesting "fact" I seem to recall about Tualatin--but can't remember where on earth I read it--is that it can cache up to 64GB of RAM. Coppermine could only cache up to 4GB. And yet another thing I can't seem to remember is if that extra cacheability only applied to PIII-S, or if all Tualatins could do it.

Tualeron at 2.6GHz? Dayum! Years ago I was able to get a 1400 running at 1.86/133, but it needed a ton of extra voltage to get there. It burned out after just a few hours of benchmarking. Probably wasn't the best sample. 😀

Oetker wrote on 2020-10-02, 21:09:

Nice article. However I don't know if you meant it that way, but it suggests the vrm and AGTL changes are related, which I don't believe to be the case. Also, the voltage clamp on some slotkets is for the so-called non vcore voltages, not AGTL. I don't think any slotket including Powerleap changes those.

With details like this I had the choice of speculating and getting it wrong or leaving it out and potentially not including it at all. Thanks for the clarification - I'll add it in.

Standard Def Steve wrote on 2020-10-02, 22:46:

An interesting "fact" I seem to recall about Tualatin--but can't remember where on earth I read it--is that it can cache up to 64GB of RAM. Coppermine could only cache up to 4GB. And yet another thing I can't seem to remember is if that extra cacheability only applied to PIII-S, or if all Tualatins could do it.

Cheers, Steve, I'll try and find a source for this and include it. I did feel the section on cache could have been more detailed.

Standard Def Steve wrote on 2020-10-02, 22:46:

Tualeron at 2.6GHz? Dayum! Years ago I was able to get a 1400 running at 1.86/133, but it needed a ton of extra voltage to get there. It burned out after just a few hours of benchmarking. Probably wasn't the best sample. 😀

Oof. Lucky Tualatins aren't that rare, then. This guy used liquid nitrogen. Gonna be hard to beat.

debs3759 wrote on 2020-10-02, 21:26:

Coppermine-T has a heatspreader.

https://www.cpu-world.com/CPUs/Pentium- ... ne-T).html

Correct! Making statement number 1 FALSE 😀

Not only Coppermine-T processors, there is at least 1 regular Coppermine that also comes with IHS, the SL5B3

Am386DX-40 wrote on 2020-10-03, 04:22:

Not only Coppermine-T processors, there is at least 1 regular Coppermine that also comes with IHS, the SL5B3

I didn't want to get into making a list of all the versions that did come with a heatspreader. I think as long as people know that it's included on late Coppermines as well then they can be aware of checking more closely what the model actually is.

I found a 'source' that mentions Tualatins being able to cache 64GB of RAM, but it doesn't say where this info came from. The Intel datasheets don't include such info.

EDIT: the same source provides a (supposedly) definitive list of CPUs in the Pentium III family that are known to have the integrated heat spreader (IHS), as included on the FC-PGA2 package, so I'm putting it here to make it easier to find / update.

Celeron (Coppermine)
Celeron 900 - 900MHz (100x9.0) - 1.75v - 128KB L2 cache (Coppermine-128 - FCPGA or FCPGA2)
Celeron 950 - 950MHz (100x9.5) - 1.75v - 128KB L2 cache (Coppermine-128 - FCPGA or FCPGA2)
Celeron 1.0G - 1000MHz (100x10.0) - 1.75v - 128KB L2 cache (Coppermine-128 - FCPGA or FCPGA2)
Celeron 1.1G - 1100MHz (100x11.0) - 1.75v - 128KB L2 cache (Coppermine-128 - FCPGA)

Celeron (Tualatin)
Celeron 900A - 900MHz (100x9.0) - 1.475v - 256KB L2 cache (Tualatin - FCPGA2)
Celeron 1.0A - 1000MHz (100x10.0) - 1.475v or 1.5v - 256KB L2 cache (Tualatin - FCPGA2)
Celeron 1.1A - 1100MHz (100x11.0) - 1.475v or 1.5v - 256KB L2 cache (Tualatin - FCPGA2)
Celeron 1.2G - 1200MHz (100x12.0) - 1.475v or 1.5v - 256KB L2 cache (Tualatin - FCPGA2)
Celeron 1.3G - 1300MHz (100x13.0) - 1.5v - 256KB L2 cache (Tualatin - FCPGA2)
Celeron 1.4G - 1400MHz (100x14.0) - 1.5v - 256KB L2 cache (Tualatin - FCPGA2)

Coppermine / Coppermine-T
Pentium III 866 - 866MHz (133x6.5) - 1.65v or 1.7v or 1.75v - 256KB L2 (socketed Coppermine - FCPGA or FCPGA2)
Pentium III 933 - 933MHz (133x7.0) - 1.65v or 1.7v or 1.75v - 256KB L2 (socketed Coppermine - FCPGA or FCPGA2)
Pentium III 1.0B - 1000MHz (133x7.5) - 1.7v or 1.75v or 1.76v - 256KB L2 (socketed Coppermine - FCPGA or FCPGA2)
Pentium III 1.0B - 1000MHz (133x7.5) - 1.75v - 256KB L2 (Coppermine-T - FCPGA2)
Pentium III 1.13 - 1133MHz (133x8.5) - 1.75v - 256KB L2 (Coppermine-T - FCPGA2)

Tualatin (256KB Cache)
Pentium III 1.0A - 1000MHz (133x7.5) - 1.475v - 256KB L2 (Tualatin - FCPGA2)
Pentium III 1.13A - 1133MHz (133x8.5) - 1.475v or 1.5v - 256KB L2 (Tualatin - FCPGA2)
Pentium III 1.2 - 1200MHz (133x9.0) - 1.475v or 1.5v - 256KB L2 (Tualatin - FCPGA2)
Pentium III 1.33 - 1333MHz (133x10.0) - 1.5v - 256KB L2 (Tualatin - FCPGA2)
Pentium III 1.4 - 1400MHz (133x10.5) - 1.5v - 256KB L2 (Tualatin - FCPGA2)

Tualatin (512KB Cache)
Pentium III-S 700 - 700MHz (100x7.0) - 1.1v - 512KB L2 (Tualatin - FCPGA2)
Pentium III-S 1.13 - 1133MHz (133x8.5) - 1.45v - 512KB L2 (Tualatin - FCPGA2)
Pentium III-S 1.26 - 1266MHz (133x9.5) - 1.45v - 512KB L2 (Tualatin - FCPGA2)
Pentium III-S 1.4 - 1400MHz (133x10.5) - 1.45v - 512KB L2 (Tualatin - FCPGA2)

Apparently there was a 700MHz version of the Pentium III-S. Haven't seen it mentioned anywhere else, so I'm sceptical. AFAIK only mobile versions were clocked that low.

brassicGamer wrote on 2020-10-03, 15:10:

I found a 'source' that mentions Tualatins being able to cache 64GB of RAM, but it doesn't say where this info came from. The Intel datasheets don't include such info.

It is mentioned in this Intel Datasheet
[Intel Pentium III Processor with 512KB L2 Cache at 1.11GHz to 1.40GHz - jun 2002]
24965705.pdf
on page 1:

PARKE wrote on 2020-10-03, 16:03:

It is mentioned in this Intel Datasheet [Intel Pentium III Processor with 512KB L2 Cache at 1.11GHz to 1.40GHz - jun 2002] 24965 […]
Show full quote

brassicGamer wrote on 2020-10-03, 15:10:

I found a 'source' that mentions Tualatins being able to cache 64GB of RAM, but it doesn't say where this info came from. The Intel datasheets don't include such info.

It is mentioned in this Intel Datasheet
[Intel Pentium III Processor with 512KB L2 Cache at 1.11GHz to 1.40GHz - jun 2002]
24965705.pdf
on page 1:
tully.JPG

I'm glad my glasses will be ready for collection in a week. Awesome work, thank you 😀

My suspicion is that the cacheability of 64GB is more theoretical than practical but I am out of my depth on this issue. It is however noteworthy that the s370 Coppermines are in their datasheet (see attached; Introduction on page eight) also mentioned to have this property.

Thanks ! Great info.
But am curious about how the average vintage person will benefit from all this ? Being that P3 motherboards are mostly chipset and BIOS dependent on what cpu they support how would this info help them ?
Just curious... been a long day so maybe am not thinking clearly....

According to https://www.cpu-world.com/CPUs/Pentium-III/In … 30KZ001512.html a 1GHz 512k part existed .. But apparently, very rare, if actually existing https://www.cpu-world.com/forum/viewtopic.php … highlight=sl5ps

Here a russian ad with photo:
https://secondpc.ru/komplektuyushchie-dlya-pk … tualatin-1.html

edit:

PARKE wrote on 2020-10-04, 11:59:

Here a russian ad with photo:
https://secondpc.ru/komplektuyushchie-dlya-pk … tualatin-1.html

Nice!

1000/512/133/1.75V ??
That voltage is in line with a Coppermine but the L2 cache is 512KB.
Maybe Intel was trying to squeeze a little more performance out of their 0.18u CPUs, but didn't end up needing to.

I wonder if it was a prototype Tualatin design at 0.18u, or a legacy Coppermine design with a bigger cache.

Yes, I noticed that discrepancy too. But there is a listing of this cpu at X86 CPU's Guide with reference of three different owners that mentions 1.45 volt and 0.13 manufacturing process:
http://www.x86-guide.net/en/cpu/Intel-Pentium … cpu-no3010.html
So another possibility is that the sticker on this cpu is wrong ?

brassicGamer wrote on 2020-10-02, 20:38:

The Tualatin has historically been a bit of a hot topic round these parts and reading up on other people's builds is what introd […]
Show full quote

The Tualatin has historically been a bit of a hot topic round these parts and reading up on other people's builds is what introduced me to this legend. It became a bit of an obsession and so I decided to try and get as much as possible into one place regarding this special CPU. I hope that some of you will find it an interesting read.

The (Almost) Definitive Pentium III Tualatin Article

Obviously the highest priority in any such an endeavour is accuracy, and I've done my absolute best to ensure that everything I've written is factual with over 60 articles referenced. Obviously if anyone finds something they disagree with then I'd welcome a discussion about it. 😀 Topics covered include:

• changes introduced over Coppermine,
• how to identify visually,
• getting it working with BX motherboards,
• how it performs against the Pentium 4.

Here's an extract from the beginning of the article: a game of...

TRUE or FALSE?

1. You can easily tell the difference between a Coppermine and a Tualatin because there's no integrated heat spreader on a Coppermine.
2. Although released 6 months after the Pentium 4, Tualatin was the first Intel CPU to use the 0.13µm manufacturing process.
3. Tualatin CPUs cannot be used on standard Pentium III motherboard because they use AGTL+ signalling instead of AGTL.
4. Intel's controversial Processor Serial Number feature, introduced with the Pentium III, was not implemented in Tualatin CPUs following privacy concerns.
5. Intel won the GHz war, by releasing their 1GHz Pentium III a month before AMD's Athlon.
6. The Coppermine CPU is so-called because it uses copper interconnects instead of aluminium.
7. The 1.4GHz Tualatin-based Celeron was so good at overclocking, you could get it to run at 2.6GHz.
8. Although not in the Xeon family, the Tualatin could be used in configurations of up to 8 CPUs.

Iirc I found it easier to distinguish Tualatin from pre-Tualatin by looking at the bottom as Tualatin looked very much different from the ~1.75v s370 CPUs that preceded it.
Only major difference between CPUs (of the same type) was that the tiny resistors on the bottom seemed to be in different positions between different CPU models at least, but never got around to see what the exact logic behind this was.

But the easiest way to visually distinguish between these groups of CPUs is by simply reading the part number, which basically makes all other ways to distinguish between the main groups of CPU mostly redundant.
I reckon that reading the voltage printed in the part number would be the easiest way, unless one is looking at some very bad pics of somebody else or if someone has very VERY poor eyesight (which is not a really good trait anyway to have when building computers).

I remember the 1GHz war. In the end Intel basically claimed they were the first with a 1GHz part that was nowhere to be ordered. So I'd rate this as either "untrue" or "questionable". Unless you trust Intel to have told the truth here when they finally got some real competition.
Iirc Coppermine did not have copper interconnects.
Tualatin CPUs can be used on pre-Tualatin boards either by using an adapter or by modifying the CPU pins. Intel probably made them incompatible on purpose to help boost sales of their newest P3 chipsets.
Performance between P3 and P4 should be well known/established by now. Depending on the software one uses, one rule of thumb could be set to a 1.4GHz P3 being roughly as speedy as a 2GHz P4 (but this is far from set in stone, it's just a rough estimate making it easy to remember).

Tualatin was used more often in servers, so multi-CPU support was a feature that was more important for Tualatin then for Coppermine as during the Coppermine era Intel wanted their Xeons to be the server CPUs of choice. However, I am not a server expert so I'd advice to read up on it as information like this should be readily available. Strange to read 60 articles and not find any info on this particular subject. This info should definitely be kinda set in stone 😜 But I wouldn't know why else Intel would have done this or why else Tualatin CPUs have been so commonly used in servers, making them widely available to people like us here.

PARKE wrote on 2020-10-04, 11:59:

Here a russian ad with photo:
https://secondpc.ru/komplektuyushchie-dlya-pk … tualatin-1.html

edit:processors-370intelpentium-3sl5ps-500x500.jpg

Odd, I never heard of this particular CPU. Interesting.

But as been mentioned before, it could be a false label.

Only way to know for sure would be to take a look at the die.

Interesting stuff, but Coppermine did not use copper interconnects, that was marketing fuzz from intel, AMD were first with that using the process they co developed with Motorola in 1998 and used in the Athlon. Intel were way behind in that party and the coppermine didn't use it despite the name.
I was around back then, and the Athlon or more sensibly the Duron was the only smart choice. The coppermines would have been fine enough "if" you could have bought them, you couldn't. Outside shit OEM boxes from compaq and Dell et al(where I reckon 98% of coppermines intel could make went, at least in the UK). Coppermines weren't easy to get until it no longer mattered.
The original Willamette P4@ 1.5 GHz was a bloody joke, I know I had one from work in 2000(a Compaq box, one of the very first about(long gone to the recycle gods)). My cheap ass(built as budget at home crap) Duron was as fast at 650 MHz(and still have and going strong!). But we got those P4's because P III couldn't be bought at the time at high-ish clocks and the IT robots had to buy intel.......(at the time I worked as a journalist, and these boxes were word processors and handled our crappy email, intranet, feeds etc, that's about it..) .
Anyway, as far as PC's went, the coppermines would have been fine, but they weren't available (at least the decent ones ) when it mattered, the P4 was as shit as we now know it is and tulatins I never saw in the eco sysytem at all, they might very well have been sat in the backroom servers(god I hope so, the P4's would never have coped! 🤣).
So at least here in the UK in MY experience, the PIII was never a thing in my field of work. More of us used Winchip C6 computers(they were a real favourite with win 3.1 and Word, locoscript and wordstar even) because you could have a reliable and very quiet box when you were working at home and spend the difference on "expenses"! 🤣(I still have mine)
( I should add that many, many UK journalists and authors hung onto their Amstrad PCW computers, they knew them well, they just worked and were silent(and actually came with a manual!) Back even at that time despite being Z80 based machines from the 80's.(mine has long gone to the landfill sadly)