AlessandroB wrote on 2025-11-30, 14:57:

The Celeron 266 is an interesting CPU, almost a misdirection (later corrected with the legendary Mendocino). The 266, being cache-free, remains an official CPU on which to experiment with the successful Pentium III architecture in its strangest iteration.

Let's talk about it…

Post scriptum Don't say that the same thing can be achieved by disabling it in other CPUs because here I'm only talking about what has historically and officially happened

Personally I have always stayed away from that CPU, unlike the Mendocino that I used both in the Slot1 and S.370 versions, I often saw those first Celerons on sale, both in 266 and 300 MHz versions, and from what I know, they are able to work without major problems at 400 and 450 MHz, but the mistake of making CPU without L2 cache, was soon evident, they ran to the shelter by churning out the 300A Mendocino, which on the contrary was excellent, in short if they had wanted to make CPU to save money, turn a PII into Celeron, removing the cache chips, well it had not succeeded.
He made the same mistake several years later, VIA with the C3 on S.370, I tried a 600 MHz years ago, it was far below expectations, roughly similar to an Intel 200-233 MHz, I haven't tried that CPU anymore, for me it's terrible 😢, a MII-300 is much better at this point.

PC@LIVE wrote on 2025-11-30, 15:28:

AlessandroB wrote on 2025-11-30, 14:57:

The Celeron 266 is an interesting CPU, almost a misdirection (later corrected with the legendary Mendocino). The 266, being cache-free, remains an official CPU on which to experiment with the successful Pentium III architecture in its strangest iteration.

Let's talk about it…

Post scriptum Don't say that the same thing can be achieved by disabling it in other CPUs because here I'm only talking about what has historically and officially happened

Personally I have always stayed away from that CPU, unlike the Mendocino that I used both in the Slot1 and S.370 versions, I often saw those first Celerons on sale, both in 266 and 300 MHz versions, and from what I know, they are able to work without major problems at 400 and 450 MHz, but the mistake of making CPU without L2 cache, was soon evident, they ran to the shelter by churning out the 300A Mendocino, which on the contrary was excellent, in short if they had wanted to make CPU to save money, turn a PII into Celeron, removing the cache chips, well it had not succeeded.
He made the same mistake several years later, VIA with the C3 on S.370, I tried a 600 MHz years ago, it was far below expectations, roughly similar to an Intel 200-233 MHz, I haven't tried that CPU anymore, for me it's terrible 😢, a MII-300 is much better at this point.

I know what you're talking about, I also had a 300A. Nowadays, when you can buy any type of computer for very little money, in my opinion, the research and study of these strange and "badly made" devices is one of the most intriguing and curious things.

PC@LIVE wrote on 2025-11-30, 15:28:

AlessandroB wrote on 2025-11-30, 14:57:

The Celeron 266 is an interesting CPU, almost a misdirection (later corrected with the legendary Mendocino). The 266, being cache-free, remains an official CPU on which to experiment with the successful Pentium III architecture in its strangest iteration.

Let's talk about it…

Post scriptum Don't say that the same thing can be achieved by disabling it in other CPUs because here I'm only talking about what has historically and officially happened

Personally I have always stayed away from that CPU, unlike the Mendocino that I used both in the Slot1 and S.370 versions, I often saw those first Celerons on sale, both in 266 and 300 MHz versions, and from what I know, they are able to work without major problems at 400 and 450 MHz, but the mistake of making CPU without L2 cache, was soon evident, they ran to the shelter by churning out the 300A Mendocino, which on the contrary was excellent, in short if they had wanted to make CPU to save money, turn a PII into Celeron, removing the cache chips, well it had not succeeded.
He made the same mistake several years later, VIA with the C3 on S.370, I tried a 600 MHz years ago, it was far below expectations, roughly similar to an Intel 200-233 MHz, I haven't tried that CPU anymore, for me it's terrible 😢, a MII-300 is much better at this point.

Not all C3's are the same though. Nehemiah's are quite good. Most will run at 1.4-1.5 on stock voltage. And at that clock they often beat even the highest clocked Coppermine Celeron's.

I had a Celeron 333 @ 400 Mhz in the day.. it was capable of running Quake 3 at over 100 FPS ... though Quake wasn't a huge stress on the CPU.

The 440BX (and similar) chipsets, SD-RAM, AGP was a big jump over last gen. Anything you wanted to run ran fast because it was optimised and the chipsets were fast.

There is no reason to get a Covington Celeron.
Just get a Pentium II and disable L2 cache.

Disruptor wrote on 2025-11-30, 19:55:

There is no reason to get a Covington Celeron.

Except for a troll build 🤣
Celeron 266 is the perfect match for Windows ME - castrated CPU running a castrated OS, probably the most retarded combo in PC history 🤣

Grzyb wrote on 2025-11-30, 20:27:

Disruptor wrote on 2025-11-30, 19:55:

There is no reason to get a Covington Celeron.

Except for a troll build 🤣
Celeron 266 is the perfect match for Windows ME - castrated CPU running a castrated OS, probably the most retarded combo in PC history 🤣

At least they are good overclockers.
Like Mendocinos.

AlessandroB wrote on 2025-11-30, 15:46:

PC@LIVE wrote on 2025-11-30, 15:28:

AlessandroB wrote on 2025-11-30, 14:57:

The Celeron 266 is an interesting CPU, almost a misdirection (later corrected with the legendary Mendocino). The 266, being cache-free, remains an official CPU on which to experiment with the successful Pentium III architecture in its strangest iteration.

Let's talk about it…

Post scriptum Don't say that the same thing can be achieved by disabling it in other CPUs because here I'm only talking about what has historically and officially happened

Personally I have always stayed away from that CPU, unlike the Mendocino that I used both in the Slot1 and S.370 versions, I often saw those first Celerons on sale, both in 266 and 300 MHz versions, and from what I know, they are able to work without major problems at 400 and 450 MHz, but the mistake of making CPU without L2 cache, was soon evident, they ran to the shelter by churning out the 300A Mendocino, which on the contrary was excellent, in short if they had wanted to make CPU to save money, turn a PII into Celeron, removing the cache chips, well it had not succeeded.
He made the same mistake several years later, VIA with the C3 on S.370, I tried a 600 MHz years ago, it was far below expectations, roughly similar to an Intel 200-233 MHz, I haven't tried that CPU anymore, for me it's terrible 😢, a MII-300 is much better at this point.

I know what you're talking about, I also had a 300A. Nowadays, when you can buy any type of computer for very little money, in my opinion, the research and study of these strange and "badly made" devices is one of the most intriguing and curious things.

Do they have the capability of SMP like the Mendocinos?
I just remember it was possible to have dual Mendocinos 300 @450 in a server board.

Disruptor wrote on 2025-11-30, 19:55:

There is no reason to get a Covington Celeron.
Just get a Pentium II and disable L2 cache.

There is one, and it is not power.. It's the love for the computer world that we celebrate here at Vogons.

You can record the sound of a 12-cylinder engine and hear it recorded or you can hear it in real life, even if your ears hear the exact same sound timbre, your heart will hear something different.

For me it's an essential part of the retro experience, for everything else there's emulation.

IIRC, the Covington Celerons were originally marketed as "Pentium 2 Celeron" really misguiding the consumers?

Well surely many if they don't use them collect them, those Celeron without L2 cache, they have that something less, which certainly penalizes them against their peers, PII and Mendocino, the strangest thing 🤔 is that, from the mistake of removing the cache, we went on to integrate only 128 KB, which however work at full frequency, and this balanced the cache at half frequency of 512 KB of the PII, however all this was a bit of the competition's fault, if AMD hadn't been successful, they probably wouldn't have abandoned the S.7 and the P.MMX would have been developed with faster versions, it's also true that the Slot1, was a solution maybe designed to last several years, but then at some point, those cheap chips became useless since the cache was integrated, and then the socket reappeared, this time with 370 PINs.
Exactly I don't know how much slower they are, you could check with a Celeron 300 MHz, from BIOS you should deactivate the L2 cache, and do the benches (from Phil's for example), then if you want to do a further comparison test, you could make a comparison with overclock at 450 MHz, in this case I imagine that the difference should grow, between the one without L2 cache and the one with 128 KB integrated.
I'm currently working on some 486, I'm reconfiguring a couple, but in addition to the S.7, the Slot1 are among my favorites 😻, the BX especially.

The Celeron with no external cache was not competing against Pentium II - it was competing against Cyrix M2, WinChip, and K6-2.

For comparison, if you disable the external cache on a K6-2 its performance tanks to near 468 levels - the Celeron was never that bad!

The interesting difference is the Celeron's use of out-of-order (OoO) instructions. You can (and Intel did) legitimately argue it doesn't need a L2 cache.

The unexpected battle was OoO vs 3DNow, and 3DNow definitely had a marketing edge that Intel needed to counter.

Arguably, countering by adding on-die L2 cache to the Celeron was overkill.

I was gradually assembling a PC at the time and decided to buy the motherboard last (for O/C reasons) and I started with a second hand P2 266MHz chip. That was never used. By the time I finished assembly the 500MHz Celeron was out, and that O/C'd to 1Ghz. Overnight it was goodnight Pentium II, humiliation for Pentium III, and bad news for Xeon too!

Who torpedoes their own products? Probably AMD. Who torpedoes their own products and puts a depth charge under them? Intel.

MattRocks wrote on 2025-12-01, 00:06:

The Celeron with no external cache was not competing against Pentium II - it was competing against Cyrix M2 and K6-2!

For comparison, if you disable the external cache on a K6-2 its performance tanks to near 468 levels - the Celeron was never that bad!

The interesting difference is the Celeron's use of out-of-order instructions. You can (and Intel did) legitimately argue it doesn't need a L2 cache.

Well in my opinion, it was just a competition at the economic level, because removing two L2 cache chips is not that it would make the price collapse so much, obviously they thought that maybe later, disappointed 🫤 by the performance of the Celeron, their loyal customers would buy the PII CPU, as an upgrade.
But anyway, even if what you wrote is true, both the K6-2 and the Cyrix M II, were actually competitors of the P.MMX, and the 2 or II was a comparison in Office applications with the PII, then as far as the Cyrix is concerned it was very efficient on integers, and very poor on decimals, also because it had a lower frequency of 66 MHz (233 for the 300GP).

PC@LIVE wrote on 2025-12-01, 00:21:

.. even if what you wrote is true, both the K6-2 and the Cyrix M II, were actually competitors of the P.MMX...

You have confused Socket 7 with Super Socket 7.

Socket 7: Pentium MMX competed directly with AMD K6 and Cyrix MX - these all ran on Intel VX and TX motherboards with 66MHz FSB and SIMM memory banks.

Super Socket 7: K6-2 is clearly a later generation with 100MHz FSB and DIMM memory banks - there was no matching Intel motherboard and the P.MMX had been discontinued.

Socket 7 chipset designs: Led by Intel.

Super Socket 7 chipset designs: Led by AMD (implemented by VIA, ALI, SiS).

Upgrade path: You could use all Socket 7 era CPUs and expansion cards on a Super Socket 7 motherboard, but comparing K6-2 3DNow to P.MMX makes no historical sense because they launched one year apart and on different platforms with very different instruction sets and different performance envelopes.

Grzyb wrote on 2025-11-30, 20:27:

Disruptor wrote on 2025-11-30, 19:55:

There is no reason to get a Covington Celeron.

Except for a troll build 🤣
Celeron 266 is the perfect match for Windows ME - castrated CPU running a castrated OS, probably the most retarded combo in PC history 🤣

i have a celeron 266 with a PCChips M726, dont make me do it.

Living wrote on 2025-12-01, 01:42:

i have a celeron 266 with a PCChips M726, dont make me do it.

And remember to install Microsoft Bob on top of it!

I ran some gaming benchmarks, and a cache-less Celeron needs to reach 450mhz just to equal the performance of a Celeron 300A at stock, so yeah, it was bad. But for some reason it's not in demand the same way other historical flops are, like a fx5800 or a 2900xt.

Disruptor wrote on 2025-11-30, 20:35:

At least they are good overclockers.
Like Mendocinos.

Oh yeah, it seems that the 1st generation Celerons were actually double-castrated...
The lack of L2 cache is obvious.
But they were probably also underclocked - I believe the technology was designed for 100 MHz FSB, but they were artifically specced as 66 MHz FSB.

MattRocks wrote on 2025-12-01, 00:06:

...
For comparison, if you disable the external cache on a K6-2 its performance tanks to near 468 levels - the Celeron was never that bad!

This is definitely not true. You have confused the internal and external caches of the K6-2.
The original K6-2 had no internal L2 cache only internal L1 cache. The L2 cache was on the motherboard and disabling this external L2 cache resulted in maximum ~20-50% performance loss. Only disabling the internal L1 cache resulted in 486 performance levels but it was also true for Celerons.
Re: to disable Cache or not to