- Yield

- Conclusions

Released page
in 1998

Juan Herrerías Rey

Celeron 366 and 400: the accessible speed

Abit BH6-BX6-BM6 and Celeron socket 370

Adapter ABIT of Socket 370 to Slot1

Celeron (Pentium II light) and AMD K6-2 (3D)

What is... the microprocessor?

What is... the cache memory?

principal page
what is...
to update
Contact Us
forum and chat



The distribution of the present pages is prohibited without the express assent of the authors.

All the marks are registered by his respective owners.

For more legal details, pulsate in the following icon


Celeron "A" Mendocino

IT NOTICES: this information is partially out of date, since he contemplates the mikes Celeron in format PPGA or Socket370 not even the Celeron at speeds superior to 333 MHz. To know more on this question, consult in the margin the related topics.


It does a pair of months it seemed that the only news that could arise on the part of Intel on the market of the microprocessors it was that a model of Pentium II was appearing with one more MHz fistful. But, for good or for evil, the things are never so simple with Intel...

If we ask someone moderately informed in these topics what the Celeron thinks of, it is almost sure that he will answer us that it is a question of a quite mediocre chip; and partly it will be right. As we mention already on these pages, any chip that he should lack cache memory of the second level (well in the cartridge SEC like Pentium II, good in the badge as the AMD K6-2) will not be able to obtain good results in "serious" applications like the office computerization, and the only one very good unit of floating comma (for mathematical calculations and 3D) will save him of doing same ridiculously in the games.

So although the mark Celeron is not of most appreciated on the market, Intel does dull ears and there extracts a new model of microprocessor whom also it names Celeron (with one "To" added or the internal name of the project, Mendocino), but whose characteristics have not anything in common with the previous models. The new Celeron yes have cache memory of the second level, what does that they are much better than the ancient ones (?) Celeron.

Curiously, they have the cache memory in the encapsulated proper one of the chip (moreover close that in Pentium II, which have it inside the cartridge but out of the encapsulated one), a system of manufacture that Intel was not using from the times of Pentium Pro. This does that the nucleus of the chip is more voluminous than that of the previous Celeron or Pentium II, what observes easily put that the Celeron lack protective cartridge (we suppose that to low the price of costs):

Celeron Mendocino

In particular they have 128 KB of secondary cache memory, the fourth part of the one that has a Pentium II. But while at Pentium II the above mentioned cache memory is employed to half of the internal speed of the mike (to 150 MHz for a Pentium II to 300 MHz, for example), at the new Celeron it is employed at the same speed as the mike. This supposes that it works to 300 ó 333 MHz, according to the model, what turns out to be almost incredible: it is very difficult to achieve that a few memory chips reach this speed of stable form, as one has seen with the SDRAM chips to only 100 MHz, which incompatibilities forced Intel to define the standard PC100 to obtain a few quality minimums independent from the mark of the memory.

Otherwise, the Celeron has the same nucleus as the most modern Pentium II, the models to 350 MHz or more, for what his unit of floating comma is so good as in them. The conjunction of these factors (technology Pentium II, small but rapid cache memory and good mathematical yield) they do this chip a very good election, only one step below the Pentium II.: How much for below? Well, now we go with that.


Let's see the famous and debatable "index first iCOMP" of the same Intel:

Graph iCOMP of general yield of microprocessors Intel - Source Intel, based on the index iCOMP, that sums diverse tests up and whose copyright belongs to Intel Corporation

In view of the graph, we extract two conclusions:

  • The new Celeron with cache memory (300th and 333) are much better than the ancient ones: nothing that objecting, is absolutely true.
  • Pentium II are approximately 14 % more rapid than the new Celeron of the same speed of clock (the same MHz): since not, not much less. This is a clear consequence of the type of tests that compose the index iCOMP, that do that the differences between processors become huge (so that the people decide to buy the most rapid and expensive models although only they have a few MHz more).

Very well, if the index iCOMP is so debatable at the time of comparing the Pentium II with the Mendocino: what other tests can we use? Since for example the Winstone 98 of Ziff-Davis, which measures the yield in typical applications like Word, Lotus 123, Netscape Navigator, CorelDraw! and others:

Yield graph in ofimáticas applications - Sources Intel and ZDNet based on the test Winstone 98 which copyright belongs to Ziff-Davis Inc. - It notices: the values given by ZDNet are computers averages published in September, 1998

Under these circumstances, the yield of Pentium II is only 1 % top and enclosed the 333 MHz Mendocino proves between 4 and 7 % top to Pentium II to 300 MHz, turned out perhaps slightly high but that seem closer to the reality (I do not know you, but I believe that the people work much more with Word than with the iCOMP). For that it seems strange to them that we use ZDNet information instead of Intel, to say that it is due to the fact that Intel realizes these tests in very different conditions for two types of microprocessor (it tackles EX-and disc IDE for the Celeron against badge BX and disc SCSI for Pentium II); nevertheless the Intel information is included for Pentium II as comparative element.

We might compare tests 3D, but I save them the effort: the results of the Mendocino and of Pentium II are almost identical in games 3D and only a little low in more serious applications like CAD (logician, since in these cases the cache memory quantity is not so important as the speed of the unit mathematical - identical - and the proper presence or not of the cache memory).


I dress everything previous, and leaving us of tests and tecnicismos: the slower is a Mendocino opposite to a Pentium II? Since very little, the equivalent to scarcely a fistful of MHz, between 10 and 40 MHz according to the applications that we use. Let's say that a Celeron Mendocino to 333 MHz is worse than a Pentium II to 333 MHz, but slightly better than a Pentium II to 300 MHz; with 300 MHz that (eye, "To", not the ancient one without cache memory) the same happens.

Then: why is it so cheap (relatively) opposite to Pentium II? For two motives:

  • it is interesting to Intel to be substituting of the market Pentium II with badge bus 66 MHz (the models up to 333 MHz) for the Celeron, leaving the models with bus to 100 MHz (to 350 MHz and more) like a professional and more expensive option (although in fact the final yield is not so different as it seems for his price);
  • Intel does not want to lose the market of economic mikes opposite to AMD, whose K6-2 are having a very good reception, on the other hand deserved. If for it he needs to sell cheaper it will do it, one can allow it (it is like the negotiation in a bazaar: although the final price is 50 % of initial, does not doubt that the seller keeps on gaining an interesting percentage).

Certainly: how is it of good opposite to an AMD K6-2? Since seemed to opposite to a Pentium II: slightly slower in almost quite, but very little. Also, since he does not need that the games or drivers of screen they are optimized for 3DNow! not that the DirectX 6 uses, his yield in ancient games is better that that of the K6-2 (although if there exist the above mentioned optimizations the K6-2 it is faster). Certainly, I recommend to the users of K6-2 to spend for the web sites of the manufacturers of these games 3D type Quake II to him and for those of the manufacturers of videocards, periodically there appear patches and updates that make use of the characteristics 3DNow! of his mikes.

Then: is it worth it? Yes, of course, although bearing in mind that:

  • Never buy it in a badge with chipset EX-. This chipset, cut away version of the already ancient LX, admits few slots PCI, few sockets of memory and bus only of 66 MHz (that is to say, that we will not be able to extend to Pentium II to 350 MHz or any more). The porqué Intel precisely recommends that to be bought looks like sadly a skylight: he makes sure that in the future we should have to buy another badge (that one of his chipsets will take... or will be quite his) or we buy to him the most expensive OverDrive. And if it is not for this, me like the beard.
  • Preferably buy it in a badge BX and with 100 MHz memory. Thus it will be able to extend Pentium II to 450 MHz or perhaps even to more than one very simple form; do not doubt that in less than one year these mikes will be very cheap and inside two years will not even exist.
  • If it finds an even more cheap K6-2, buy it. Especially if it is cheaper enough or if they offer him more memory (remember that for the K6-2 it must be 100 MHz, what will facilitate the future enlargement) or a monitor of 15 " instead of one of 14 ".

And to end at once, a curious note: it was doing years that a such trouble did not see with the name of a mike: some call it "Celeron 300th" (O.K)., other "Celeron Mendocino" (O.K)., other "Celeron 333" (without the "A"; in agreement, this way it calls it Intel, since 333 MHz that does not exist without cache memory), but other "Mendecino" (!), "Mendocinio" (!) and enclosed "Mendeccino" (!!!). I do not know, I believe that it is Mendocino, but if so much trouble creates that they call him "Celeron with cache memory" and we will understand each other all although we do not know Italian...


To return - Sitemap - Contact us Best website of 2009 & 2010 voted by us are:
, , , , , ,,