You knew it was coming, how far can the next-generation Celeron's be overclocked?  If you recall, the limiting factor in the overclocking of Pentium II's was determined to be the L2 cache.  According to that theory, adding L2 cache to a Celeron would diminish its overclocking capabilities and therefore ruin its potential as a high end processor,  however there is a key difference between the new Celeron's and the Pentium II's: the placement of the L2 cache.  

While the Pentium II keeps its cache a safe distance away on the processor card, cooling the L2 cache becomes a bit of a problem since your heatsink/fan combo never actually makes contact with the surface of the cache.   In essence, all you are doing is cooling the casing around the Pentium II's processor card.  With the Celeron, the L2 cache is included on the die, so you are not only cooling the processor with your fan, but you're also cooling the integrated 128KB of L2 cache.  The modified theory here is that if the processor can be pushed to a certain limit...why can't the cache be pushed that high too?

The Limits

With BX boards flooding the market it is often difficult to remember that there are other motherboards based on older chipsets in the hands of a large population of tweakers.  If you have anything below a Pentium II 333, and happen to be looking for a cheap processor upgrade, then this next section is definitely for you. 

The Celeron 300A became the choice processor for the AnandTech Overclocking Test Bed, as the 333MHz part is still an unproven soul it will take at least another week or two before production on the 333 reaches a point where it becomes a better overclocking alternative than the 300A.  In any case, with the 300A, you are pretty much restricted by the clock multiplier lock as far as overclocking goes.  The best case scenario is that you have a motherboard with the 100MHz FSB setting, however not everyone has that luck, in which case you just need a board with a working 75 or 83MHz FSB setting.  It would be best to run some stability tests on your motherboard, if possible, at those two overclocked speeds before pursuing any Celeron 300A upgrade with the intentions of overclocking.

It's time for the shocker, how far can the Celeron 300A be pushed?  Using two different chips, one purchased from Azzo Computers and the other from TC Computers, the highest achievable speed was a whopping 450MHz running each processor at the 100MHz FSB x 4.5 at the standard 2.0v core voltage setting.  Of course the intermediate settings, 75 x 4.5 and 83 x 4.5, were tackled without a problem, but the 450MHz limit of the chip seems to be a comfortable level for a chip that retails for under $200.  Running both chips at 112MHz x 4.5 proved to be unsuccessful as Winstone would often crash during its initial test run, upping the core voltage selection to 2.2v didn't enhance stability to a noticable degree, so for now the chip seems to be our favorite at 450MHz.  As far as stability is concerned, the Celeron 300A system ran rock solid at 450MHz for a continuous 24 hour day running Business Winstone 98 in a single loop(with the Task Switching tests disabled as they do cause a known crash with Windows 98 on all systems).  Is that enough proof?  For purposes of authenticity, the markings on the two Celeron CPU's used are as follows:

Azzo Computers 300A

98340078- SL2WM
300A/66 MALAY

TC Computers 300A

98340071- SL2WM
300A/66 MALAY

Celeron Markings

Going back to the ability to run the processors at 75MHz x 4.5 and 83MHz x 4.5, this allows the processors to be used in high performance low end systems as well.  Considering Micro ATX boards based on the 440EX chipset are retailing for well under $100 now, most of which do have 83MHz FSB settings, you can get an easy 375MHz Pentium II level of performance for under $300 provided that your peripherals work at 83MHz.  Not too shabby, but let's take a look at some comparative benchmarks to make the final decision.

Chomp on this Celery Stick Performance


View All Comments

  • Hulk - Thursday, April 26, 2012 - link

    I remember this review fondly. This was the review that prompted me to build my first computer. My 300a went to 450 just like almost every one. Reply
  • dananski - Tuesday, February 24, 2015 - link

    This was sort of before my time (the day after my 11th birthday) but I wish I had known these things in such detail when I first came to buying and building computers. And I do miss Anand's excellent writing now he has left. Reply
  • Kepe - Sunday, March 1, 2015 - link

    Hehe, the WR overclock for this processor was broken yesterday: 721 MHz. Amazing! :D
  • StrangerGuy - Monday, April 1, 2019 - link

    I wasn't into PCs in 1998/99, but looking back there were countless people I know who got ripped off by buying the somewhat-to-insanely overpriced P2/P3s, or the cheap-but-slow K6 variants after the incredible Celly 300A hit the streets. I guess the Internet wasn't that mainstream then, let alone PC review websites like AT.

    It's funny now people spend big bucks elking out a mere 10% OC out of their i7/i9s, while 20 years ago overclocking the 300A by 50% on a 440BX is free in the truest sense of the word.
  • PandaBear - Monday, July 6, 2020 - link

    Back then people expect performance doubling about every 18-24 months. These days you get 10% improvement every 18-24 months because of monopoly.

    Also other than GPU, SSD, and maybe up to 16GB of DDR4, there really isn't anything worth upgrading anymore.
  • bunnyfubbles - Tuesday, March 30, 2021 - link

    these days we get 10% not because of monopoly (you could have made that argument after Sandy Bridge launch, but not now with AMD producing viable competition thanks to Zen) but because we're running into the physical limits of silicon. Hell, we knew even as far back as ~2006 that clock rate ramping would be unsustainable. Reply

Log in

Don't have an account? Sign up now