(THIS TUNE - NIGHT.MID, COMPOSER UNKNOWN)

BY TRED

For my first article here at Nude Overclockers I figured I'd go over the history of the latest and greatest overclocking champ, Intel's Celeron.

To understand why the Celeron is so damned overclockable, you really have to take a look at why it was created. This goes back to early '99 (back in the 20th century :) ).

In late 1998 and early 1999, AMD was solidifying their place in the processor market. They had introduced their kick-ass K6-2 with 3d-Now!, their 3d version of Intel's MMX.

While AMD wasn't competing with Intel's top of the line CPUs, the fact of the matter is AMD was kicking their ass all over the place in the sub-$1000 PC market (read: low end market).

It's not that AMDs chips are really that cheaper to produce (and they obviously weren't better), but the only way AMD can sell any chips at all is if they sell 'em cheap, so they mark them down very low - sometimes so low as to actually lose money in the process.

Once they have a user base and get name recognition it's worth it, and AMD was scaring Intel with how much they had sold. In fact, in January 1999 (ancient history now) more AMD based PCs were sold than Intel based ones.

This was the first time since Intel became popular that they actually showed signs of vulnerability. So Intel basically tried to beat them at their own game...by selling chips really cheap. Intel couldn't just lower the price on their flagship Pentium 2, they had to keep up their triple digit profit per chip.

Intel instead decided to strip the Pentium 2 of its cache (In case you don't know, cache is a fast type of memory where frequently used data is stored...very vital in server situations where the same web page is served thousands of times over and over again or something like that; but not really all that important to home users), and sell it as a 'new' CPU - the Celeron.

Initially released in 266 and 300 MHz versions, the Celeron was supposed to be the AMD killer. The problem came when the print media got their hands on evaluation copies. Now smart folks like yourself know to read online sites like our great one to get your information, but most people (the dumb masses) just pick up a magazine. The problem is, the magazines don't know their stuff like we do.

They love to throw around numbers because people can relate to numbers: "Oh that one got a higher number than the other one, I'll get that". Well numbers don't always tell the full story, and they definately did not in the case of the Celeron.

Their take on the cache-less Celeron was not neccisarily false, it scored pretty badly in office applications because of the lack of cache. Because of this, most magazines immediately dubbed it a failure (like I said they're dumbasses). They failed to see two things;

First of all, office performance is really irrelevant in new processors these days. When was the last time you had to wait long enough to worry about how long it takes for Word or Word Perfect (my editor of choice) to save a file, or even to open for that matter. As long as you have a Pentium class system or higher, you're basically set.

Secondly, all office performance aside, it kicked for gaming. While playing games, especially 3D games, you are always coming upon new scenes. For that reason the cache is pretty much useless, because if you always see new things nothing will be stored in there.

Anyway, all big and even most small print magazines decided the Celeron pretty much sucked, and dubbed it a failure. Needless to say this hurt Intel big time. Not only were they selling what were basically Pentium 2s for dirt cheap prices, but almost no one was buying them because the print mags were not giving the full story - not that I'm a big fan of Intel however.

So to make the Celeron really sell they decided they needed it to have strong office performance. What they did was add that cache back onto it, only less of it - 128k instead of the Pentium 2s 512k.

However, they didn't add it back on the same way. Instead of adding cache chips onto the silicon wafer, they actually embedded it onto the chip itself (hence, on-die cache). They also ran this cache at full speed, whereas the P2's run it at half speed. This means that if you have a 400mhz Celeron, the cache runs at 400mhz. If you have a 400mhz P2, the cache runs at 200mhz.

So they added the cache, but kept the same name Celeron. That seems like just a flat out bad marketing decision, but I guess they figured that they could still bring some respect to the name.

This new celeron was released in two flavors; 300 MHz and 333 MHz. This obviously created a conflict, since they already had a 300 MHz 'cacheless' Celeron, so they dubbed the new one the 300a. The magazines got their hands on the new copies, which performed very well - in fact very similar to Pentium 2s at the same clock speed (but of course the print media wouldn't realize that) and called them a good low budget solution.

As far as Intel was concerned it was case closed at that point, and all they had to do was sit back and watch Celeron sales skyrocket. But Intel was wrong. You may be thinking that Intel is dumber than your dead grandmother for releasing a powerful chip without knowing it. Well your wrong, Intel knows what the hell they're doing, they're the ones making the chips for christs sake (when was the last time you shoved 8 million transistors onto a piece of silicon?).

Intel knew exactly what they were doing when my put that 128k of cache on the chip, they were making a chip that is basically capable of being a direct competitor with the Pentium 2. The chip is really a Pentium 2 at heart anyway. Well why doesn't everyone just buy Celerons instead of P2/3's?

Well, Intel has a massive amount of power. They actually bear a striking resemblence to Microsoft. They can sway the whole market by whatever they say. Well, they say that the Celeron is for low-end machines so they must be. They justify this statement two ways.

First of all, they're cheaper than P2/3's. Cheaper means they aren't as good right? Secondly, they keep the bus speed at 66mhz. That way they have some numbers (remember what I said about consumers loving numbers) to back up their claim. 66 is smaller than 100 (the FSB of the P2/3's), so it must be slower, right?

Well now we're getting to the good stuff. We still haven't brought up why the Celeron is so damned overclockable. You've probably heard that as of quite a while ago, all the new Intel chips are what's refered to as 'Multiplier Locked'.

What this means is the multiplier (multiplier x bus speed == cpu speed) can not be changed at all, leaving the overclockers stuck out in the mud. Intel claims they have nothing against overclockers, and that they do this to protect against remarked chips.

This is fairly valid, considering that they (at least claim to) lose millions every year due to chips being marked as being a higher speed (say a 400mhz marked at 450mhz or even higher) and then sold at a higher price (the 400mhz chip would be reboxed and sold for the price of a 450mhz chip) - so people would technicly be 'overclocking' (but without knowing it of course).

This is done a lot more in Europe than in the States, by the way, which is why it may not sound farmiliar to you. I don't think Intel is a very big fan of Overclockers either, and figured they could kill two birds with one stone by doing this.

So we're really getting into the good stuff now. As I mentioned, all Celerons run at a 66Mhz Front Side Bus speed. This really doesn't hurt performance all that much, at least as far as gaming is concerned (as proven by many benchmarks done by a number of people).

So, like I said, when Jon Q User walks into Ye Ol' CompUSA to buy a computer, and he decides he NEEDS top of the line he can easily walk right by the Celeron CPUs, because they cost less. But lets say Jon stops for a second and says, "this is 400 Mhz, that is 400Mhz, whats the difference?" Fred Q Reseller can easily say, "Well Jon, that 400Mhz Celeron chip runs at a mere 66Mhz Front Side Bus, whereas the new Pentium 2 Chip runs at a blazing fast 100MHz bus with the new PC100 SD-Ram that goes in the next generation motherboards".

Now Fred Q Reseller is obviously using big words to try and impress Jon so he can get a nice commission on that sweet $2,500 system, instead of the $999 one. The sad thing is, it probably worked. So Intel wins, and the Celeron is for the cheap-o systems.

Well, they didn't quite win. Suddenly some people (most notably Dr Tom Pabst from Toms Hardware) noticed exactly what was going on, and ran both chips through some benchmarks. Surprise! The Celeron performs equal to (sometimes actualy greater than) a P2 of the same clock speed. Tom put up a big CPU review, and suddenly the net went crazy with people realizing what was going on here. But that wasn't the only thing that people figured out.

The biggest thing people found out was that not only was the Celeron as good as a Pentium 2, but because it was made from the same batch as the Pentium 2 450 (technically because it was .25 die size, and Intel got good yields, but I'm simplifying it a bit here) almost every Celeron chip that came out was capable of performing at around 450MHz.

Problem was, the multiplier was locked - how were people going to get it to 450? Well, 4.5 x 66 = 300, and 4.5 x 100 = 450, and there we have the hype of the Celeron 300A. Since it was only the multiplier that was locked, if you put the Celeron into a motherboard that supported the 100mhz front side bus, you were ready to go. About 90% of all Celeron 300as can be overclocked to 450Mhz.

Another thing people commonly overlook is that those original 266 and 300mhz varients were also made from pretty much the same die. The 300mhz wasn't able to pull off 450mhz very easily at all, but the 266 Mhz (4 x 66) could get 400Mhz (4 x 100) or more, pretty much without a hitch.

The hype of the 266Mhz celeron was pretty much limited to the Overclocking community, but once the 300a came out there was an explosion, and the overclocking comunity was no longer a few tightly knit web pages, but just about EVERY SITE involved with hardware or even software was talking about the 300a and overclocking.

Everything calmed down for a while, until Intel started releasing higher and higher speed Celerons. Most of them were met with pretty grim results, none would replace the 300a as the overclocking champ it seemed. However, as time went on, the 366 started to overclock better. Pretty soon, the new batches of 366s were almost routinely hitting 550mhz. Now the big craze is 366@550s. They'll be replaced someday too, I hope :).

I know I overlooked many a thing here. First of all, Toms article didn't cause the end of the world as I'd have you think, but it was pretty damned close. Also, there was overclocking mayhem with the cacheless celeron ala the 266. It just didn't come near the craze involved with the 300a, which is why I illustrated it as such. I think so far I've been pretty error free, but to err is human so if you want to correct my ass mail me. Otherwise, until next time... Later.

-TRED

MAIL THE NUDE OVERCLOCKERS