I had Slot 1 Celeron 300A, it did overclock to 450MHz.
If I remember correctly, it was important which factory did manufacture the exact CPU (Malaysia/etc). Allegedly that specific factory had the line used to manufacture other faster CPUs, so Celeron 300A made on that line came out with much greater overclocking potential.

vstrakh wrote on 2022-07-10, 11:12:

I had Slot 1 Celeron 300A, it did overclock to 450MHz.
If I remember correctly, it was important which factory did manufacture the exact CPU (Malaysia/etc). Allegedly that specific factory had the line used to manufacture other faster CPUs, so Celeron 300A made on that line came out with much greater overclocking potential.

I only ask because I have been given a 370 300A SL36A I was curious if its essentially the same legendary 300A just in a different package, it should manage 450.

Both apparently capable of 450 - https://arstechnica.com/features/1999/04/cele … %20Socket%20370

JidaiGeki wrote on 2022-07-10, 11:24:

Both apparently capable of 450 - https://arstechnica.com/features/1999/04/cele … %20Socket%20370

It should be a fun little experiment seeing how high I can get it with the gear I have here, got a few freaking massive copper 370/462 coolers here that should handle the heat 😀

TrashPanda wrote on 2022-07-10, 11:00:

Quick question, is the Socket 370 version of the 300A the same chip that is in the Slo1 package ?

yes, altho in 1998/1999 I never saw a 370 version of 300A while working at national parts distributor in Europe. The slowest 370 chips we ever sold were celerons 333 (very bad at 100MHz).

TrashPanda wrote on 2022-07-10, 11:00:

Am I right in thinking it was the 370 version that was the famous overclocker ?

no, its same chip

vstrakh wrote on 2022-07-10, 11:12:

If I remember correctly, it was important which factory did manufacture the exact CPU (Malaysia/etc). Allegedly that specific factory had the line used to manufacture other faster CPUs, so Celeron 300A made on that line came out with much greater overclocking potential.

Afaik packaging plants - Malaysia, the Philippines and Costa Rica - were not correlated with cpu die manufacturing fab source (texas, california, oregon, mexico, ireland, one of those wasnt doing 0.25 micron at the time, but I dont know which). Looking for special packaged chip was a meme/red herring/old wifes tale/baseless rumor.

Slot 1 was pretty popular when the 300A came out. All of the overclocking posts I remember were SL32A and SL2WM. I saw a Chinese seller on eBay with a tray of SL36As for $5 each a few months ago, so I bought a couple of them for fun.

I'd say slap the PPGA version into a slotket and go HAM. Have fun.

rasz_pl wrote on 2022-07-10, 19:53:

Looking for special packaged chip was a meme/red herring/old wifes tale/baseless rumor.

We were young, the Google did not exist yet, buying cheap and rationalizing "it's more cost effective" was a thing 😀

A buddy and I bought into the SL2W8 300MHz Deschutes P2 back in late 1998, a bit after the Celeron 300A craze I think. I replaced a Klamath 233 with it. They had the cache chips of the 450 and so usually went to 450 or better. I can't remember what we paid for them. I remember it being at 504 (112x4.5) on the Abit BH6.

That board had a cold boot issue usually necessitating a reset to POST that was pretty irritating. Bring on the BF6. 😀

Hehe fun with overclocking bargains.

swaaye wrote on 2022-07-11, 21:08:

A buddy and I bought into the SL2W8 300MHz Deschutes P2 back in late 1998, a bit after the Celeron 300A craze I think. I replaced a Klamath 233 with it. They had the cache chips of the 450 and so usually went to 450 or better. I can't remember what we paid for them. I remember it being at 504 (112x4.5) on the Abit BH6.

That board had a cold boot issue usually necessitating a reset to POST that was pretty irritating. Bring on the BF6. 😀

Hehe fun with overclocking bargains.

Heh I have a nice board that supports both slo1 and 370 celerons which I’ll use, got a PII 300 so I might test them against each other for period games.

At the very least it’ll be a fun distraction from my HX board adventures and it not being able to power a Parhelia correctly.

As far as I know, my used one managed to go up to 105 MHz with a tomato mainboard at that time. Nearly all Celeron 300A's can go 100 MHz bus speed without problem.

I bought one of those cheap 370 300A's a while back from China and was able to get it to run at 450MHz pretty easily. It was on an Intel i810e Asus motherboard.

Problem is that there are a ton of cheap and faster 370 Pentium 3's and 100FSB Celerons so other than playing around with it, there's really no use for the 300a (outside of nostalgia).

I always wanted one back in the day though. Although a friend of mine bought the famous 300A/Abit mobo combo back then and could only get it to run at 83MHz FSB. It wouldn't even post at 100MHz. So not everyone was successful.

TrashPanda wrote:

Quick question, is the Socket 370 version of the 300A the same chip that is in the Slo1 package ?

Newer stepping, so possibly better oc.

TrashPanda wrote on 2022-07-10, 11:00:

Quick question, is the Socket 370 version of the 300A the same chip that is in the Slo1 package ?

Am I right in thinking it was the 370 version that was the famous overclocker ?

Back in the days (Mid-1998) it was the Slot 1-Celeron 300A that was used for overclocking on an Abit BH6 (and later on other boards). Pretty much all my friends and myself had one.
The next 'big thing' (early 2000) was the S370-Celeron 566@850 with Slotket on the same board.

As a side note to this thread, I just today saw that one of my favourite old tech seller here in Australia had a slot1 Celeron 300A with an Abit BH6, 128mb Ram and TNT2 for sale at a reasonable price so I snapped it up, its confirmed to run stable at 450@100 FSB so I'm curious to see how the S370 300A fairs against it.

Can even use the same board for testing both as I have a spare Cerlery S370 slotket.

Never had a BH6 or 300A back in the day as I jumped from SS7 to 462 Athlon instead, I briefly had a PII 450 but I never owned that system ..kinda wish I had as it was a great PC for the time.

Its not that 300A was especially overclockable. All Mendocino celerons started with same 450-500MHz max clock ceiling, but 300A:
- was cheapest
- had lucky 4.5x multiplier close to the max possible reachable clock at 100MHz FSB
- cheapest

Other commonly available FSB options on even the cheapest boards were 75 and 83MHz. 75MHz was weaksauce, and 83 overclocked PCI to 41MHz which was often problematic. If you wanted something more fancy with many more in-betweens you were paying close to double the price (Asus, Abit) of the cheap boards (something like absolutely rock solid Zida ZX98).

Around release of 433 or 466 versions in middle of 1999 Intel tuned manufacturing process to the point most Celerons 366 achieved same optimal level and were able to run at 5.5x100MHz with minimal voltage bump.

Same thing happened with Coppermines and 533A. In fact Intel knew it would happen and tried to hide it 😮 https://www.anandtech.com/show/533/2 533A was great because
- cheapest
- again lucky 8.5x multiplier putting it smack at the edge of reachable clock range with 100MHz FSB
- cheapest

This time around second Coppermine celeron speed bump in November 2000 was the golden ticket in form of 700 733 766 models all being able to work stable at 10.5 11 and 11.5 x100MHz, often with no voltage modification. This was a great financial deal combined with cheapest 440BX board and a cheap slotket. I build and sold tons of this kind of systems at the time, 2-5 a week.
For comparison Duron Spitfire topped out around 950MHz, and required prone to failure overclocking methods - who remembers filling tiny laser cuts with pencil (conductive graphite filling the holes and closing L1 bridges)? https://www.overclockersclub.com/guides/tbird/

Party stopped with Tualatin, for a long time the only way to run it was expensive PowerLeap or expensive 815e boards. Made no financial sense. AMD wasnt better with Morgan Durons barely 200-300MHz faster than overclocked Coppermines.

It was long two years of nothing cheap and good until AMD finally released Applebred Durons. Bonus points for linking L2 bridges and modding Duron into a full Athlon 😀 This was also the peak of Intel paying off system integrators to not stock AMD systems (Dell alone $1Billion a year in bribes), so AMD was forced to price products really cheap to attract clients.

rasz_pl wrote on 2022-11-24, 06:30:

Its not that 300A was especially overclockable. All Mendocino celerons started with same 450-500MHz max clock ceiling, but 300A: […]
Show full quote

Its not that 300A was especially overclockable. All Mendocino celerons started with same 450-500MHz max clock ceiling, but 300A:
- was cheapest
- had lucky 4.5x multiplier close to the max possible reachable clock at 100MHz FSB
- cheapest

Other commonly available FSB options on even the cheapest boards were 75 and 83MHz. 75MHz was weaksauce, and 83 overclocked PCI to 41MHz which was often problematic. If you wanted something more fancy with many more in-betweens you were paying close to double the price (Asus, Abit) of the cheap boards (something like absolutely rock solid Zida ZX98).

Around release of 433 or 466 versions in middle of 1999 Intel tuned manufacturing process to the point most Celerons 366 achieved same optimal level and were able to run at 5.5x100MHz with minimal voltage bump.

I would argue these are exactly the reasons why it was special 😀
The cheapest CPU competed with the fastest and most expensive CPU at the time (P2-450)

And at the time of the 300A there wasn't another option. The 333 wasn't 'guaranteed' to run at 100 FSB and at 83 FSB it's way slower than the 300A. All other Celerons came later.

Well, the Covington easily could be overclocked too, but it still was a Covington.
The problem with the overclocking was in the discrete L2 cache chips.

Disruptor wrote on 2022-11-24, 21:38:

Well, the Covington easily could be overclocked too, but it still was a Covington.
The problem with the overclocking was in the discrete L2 cache chips.

the lack of L2 chips 😀 The release gap between 266 Covington and 300A Mendocino was barely 4 months, and only two months for 300 Covington! The difference in price in April 1998 between 300 Covington ($112) and 300A Mendocino ($149) was $37. I dont think we (national hardware distributor with direct Intel contract) stocked Covington at all after first 300A shipment.

Even overclocked Covington barely scraped in slow Pentium 2 266-300 territory, while 300A was actually beating full Pentium 2 at clock for clock comparisons (faster cache). It seems to me the people with super limited budgets who might pick Covington went with bottom of the barrel AMD systems instead, even cheaper and they were mostly uninformed about performance ramifications or too inelastic financially. K6 and K6-2 used older 66MHz S7 boards. Those were the type of clients getting excited on the prospect of being offered all in one PCChips M592 with graphics on board!

Ah yes, the days when overclocking was worth it (and fun). 333 cerlerons with huge heatsinks.