Why Celerons Suck.

Robert said:

The L2 memory and fsb on Celeron is less if compare to Pentium 4.

Click to expand...

The fsb can be made up for quite easy

Ok, a Northwood 100 FSB (400 system bus) Celeron with 128k cache running at default speed is not that good performancewise. Even all the Intel people admit that. Why would you even compare that to an overclocked A64?

The Prescott Celeron D has 256 cache and 133 FSB (533 system bus) and overclock like crazy. A 2.4 Celly D will easily do 3.6 gig, sometimes on default vcore and with stock cooling.

Lots of guys are hitting 4 gig with good cooling. I'd enjoy seeing one of these demons post their Super Pi score. For a CPU that cost half of what a A64 does, I bet you'd be surprised.

Besides, the budget Celeron was never made to compete with an A64. Additionally, a Super Pi 1M score of 44 seconds isn't exactly something to brag about, I have posted 29 seconds (no patch) with my green signature rig.

Even the old 100fsb 128k Celerons aren't that bad. My media PC is built from the following components:

[email protected]
Abit AI7
2x256 PC2700 (mismatched)
SP-94 w/slow NMB 92mm fan
40GB GXP180, 160GB 7k250
64MB Chaintech 128bit BGA GF4-MX4000 TV-out
Leadtek XP2000TV Expert TV tuner/capture card
FSP350-60PN

This thing encodes video as well as the [email protected] that it replaced. It is snappy beyond all belief on the desktop. It will play games better than anyone would credit it with, and its 1M superPi, while probably its biggest weakness, is 1m 17s. And it is almost entirely silent. While basically twice the time my 2.8c takes, who really sits around waiting on their superPi to complete? This is exactly the type of pursuit that means nothing to the celeron's target audience, and even if it did, it does it well enough.

Anytime people stop their bitching and make the most of what they have, celerons get the job done. OC'ing is making the most of what you have, not so much what you have. The above machine is a joy to use, impressive on paper or not.

Indeed, the Celeron "D" when pumped up does a fine job. 43s for an 80.00 processor. I overclock to get the most for the least and so far this processor makes me think of Celeron 300A @ 472mhz.



I posted this in April:

http://www.ocforums.com/showthread.php?t=359020&highlight=celeron

R

My p4 1.6 willie takes over 2min.

The Prescott Celerons are actually a pleasent surprise, in my opinion. It's still a low end chip, but it's far better than the previous iteration of Cellies.

theres something i really wanna see, a sempron 2800+ or 3100+ at its max oc, like 2.6ghz vs a Celly D 2.4 or 2.6 oced to 4.0ghz. id love to see which one is better. they both cost around the same for a budget setup.

42 secs, woot

Captain Newbie said:

It's still a low end chip, but it's far better than the previous iteration of Cellies.

Click to expand...

A low end chip that takes out P4 3.4 and A64300+ on everything but HT enabled tests and workstation database number crunching?

What constitutes a low-end chip to you? Price or performance? If performance, please show me where the Celeron "D" underperforms or is it low-end because of the non-ht and lighter cache? I have used and abused A64 3000+ and P4 3.4 and find the Celery "D" a worthy low cost solution to High end performance in a comparitive testing procedure. Considering the value of the above A64 and P4 there is no doubt in my eyes that these processors are as good as the Celeron 300A @ 450 that took out P3 @ 450 for one third the cost.

To me a low end chip is the Celeron "D" with inability to be overclocked. Once overclocked it becomes a high end chip. Certainly low end of the high end processor since it competes robustly with P4 and A64 at those (stated) levels.

R

Ugh, this thread should be deleted .... now people are gonna start a Celeron Mulisha!

ropey said:

Considering the value of the above A64 and P4 there is no doubt in my eyes that these processors are as good as the Celeron 300A @ 450 that took out P3 @ 450 for one third the cost.

Click to expand...

As useful as I find Celerons, I cannot abide by this statment. The 300A@450 was clearly outperformed by a P3 at 450MHz in any measure of application performance. Only synthetic tests that fit in the 300a's L2 cache posted higher numbers, and those numbers are misleading and do not make it a superior chip. And of course, P3-450s had a lot more than 450MHz in them, where very few 300a's exceeded 450MHz. This kind of view through rose-colored glasses is just fuel for the fire of those that would disparage the actual capability Celerons represent.

larva said:

As useful as I find Celerons, I cannot abide by this statment. The 300A@450 was clearly outperformed by a P3 at 450MHz in any measure of application performance. Only synthetic tests that fit in the 300a's L2 cache posted higher numbers, and those numbers are misleading and do not make it a superior chip. And of course, P3-450s had a lot more than 450MHz in them, where very few 300a's exceeded 450MHz. This kind of view through rose-colored glasses is just fuel for the fire of those that would disparage the actual capability Celerons represent.

Click to expand...

I have personally had a Pentium3 Deshutes 450 and Celeron Mendocino 300A on an ABIT BH6 board. The only time the Pentium3 Deshutes handled better than the Celeron Mendocino 300A @ 450 was when there were multiple cache calls for over 128kb and this is seldom the case unless one is operating heavy (such as database) number crunching and/or scientific algorithmic command structures such as astronomical programs. For the bulk of computer use the Celeron Mendocino 300A was faster as the cache was built on chip and was at full bore.

In other words, the cache of the Celeron Mendocino 300A @ 450 ran at 450MHz and the Pentium3 Deshutes 450 ran at 225mhz. Thus for any real cache calls of 128kb (512bytes/sector burst mode) or less, the Celeron Mendocino core @ 450mhz clearly would outperform Pentium3 Deshutes core @ 450mhz. At 512bytes per sector, there would need to be 256 pipeline calls for a flush of the cache where the Pentium3 Deshutes would not have the flush until 1000 calls. How often are there in concurrency 256 cache calls of burst mode? Not many! Thus the Celeron Mendocino core was clearly the better perfomer than the Pentium3 Deshutes core thus the 512kb (Off Chip) cache of the Pentium 3 was clearly less able except where burst mode of greater than 128kb calls were made and these were a small percentage in normal computer usage. Now the Pentium 3 Coppermine was a different beast but of course the Coppermine P3 was not created at the 450MHz level and the Deshutes was the only P3 of that flavor except for some esoteric OEM's and notebook systems.

The 0.25 micron Pentium3 Deshutes processor had 512KB of off die L2 cache. The 0.18 micron Celeron Mendocino 300 "A" processor has 128K of on die L2 cache. Both processors have the same clock counts as far as latency but clearly you can see that the lower cache calls of P3 Deshutes must be slower than the higher cache calls of Celeron A. Clearly the Celeron Mendocino 300A "A" was a cache disabled Coppermine processor and any Coppermine processor would MHz for Mhz take out the Deshutes in raw core power ability. The only exception to this was more rare functions of large 128kb+ burst mode cache calls where the Pentium3 Deshutes's 512kb would lead. This was seen in the benchmarks that addressed such schematas and often benchmarks were (as opposed to your view) skewed in the favor of the Pentium3. Skewed because they measured Celeron Mendocino against Coppermine. I tested Mendocino against Deshutes as that was a more realistic platform at that time.

Let me be clear. My opinion is that for the price of the Celeron 300A and the price of the Pentium3 Deshutes @ 450 the dollar for dollar value in a comparitive analysis (In My Opinion) the P3 was taken out. Just as in my opinion the Celeron "D" takes out the previous mentioned processors. Of course there are arenas where the higher price processors do outperform but Dollar for Dollar, I believe that the Celeron "D"'s overclocked are better performers. Hopefully my personal view is clear as such, however rather than argue with you on what you can or can not "Abide" I will say that I agree to disagree with your (imo) concrete black and white appraisal.

R

Please, if you feel 128K cache is not limiting the 300A, toss in a P4 coppermine at the same clock and see just how big a factor the L2 cache size is as its size is reduced below 256K.

I had countless C300As, C333s, and four P3-450's when the chips were current. I used them on every BX board sold in numbers in this country. The P3 is the clear winner for each and every application. Achieved fps in nearly any game will drive this point home. And of course, the best of my P3-450s did 600MHz, further widening the gap. I understand 300A's were cheaper, but they were in no way better.

I used a celeron on overclocking and it got 4.78 GHZ!!!!! This is a 60 dollar chip!
Oh and it plays HL2 on top settings with a geforce with no overclocking and I get good fps. Not in the hundreds but 30 is good enough for me.

Some guy bought a $8000+ Alienware with a 20 inch screen for HL2.

I bought a $600+ celeron custom system and play with my 42 inch LCD.
= I win because I get playable FPS on top setting

Guess who has more fun?

larva said:

Please, if you feel 128K cache is not limiting the 300A, toss in a P4 coppermine at the same clock and see just how big a factor the L2 cache size is as its size is reduced below 256K.

I had countless C300As, C333s, and four P3-450's when the chips were current. I used them on every BX board sold in numbers in this country. The P3 is the clear winner for each and every application. Achieved fps in nearly any game will drive this point home. And of course, the best of my P3-450s did 600MHz, further widening the gap. I understand 300A's were cheaper, but they were in no way better.

Click to expand...

Somehow you do not get what I am saying. Dollar for Dollar is what I am saying. An 80 dollar Celeron D taken to 4GHz is DOLLAR FOR DOLLAR better than a 3.4GHz Pentium4 and this is in my opinion. Clearly not yours, but please allow me my view. Just as 512MB of ram is clearly more robust in system performance in comparison to 256MB and the difference is less noticeable with the move to 1 GB or 512mb to 2GB. Thus there is not nearly as much difference between a 128kb full bore on chip and 256kb on chip vs 128kb full bore on chip and 512kb half bore off chip cache. Thus your statement to compare a different architecture holds no water in analysis.

And you still do not address my personal view. So I will reiterate for you.

Ropey said:

Let me be clear. My opinion is that for the price of the Celeron 300A and the price of the Pentium3 Deshutes @ 450 the dollar for dollar value in a comparitive analysis (In My Opinion) the P3 was taken out. Just as in my opinion the Celeron "D" takes out the previous mentioned processors. Of course there are arenas where the higher price processors do outperform but Dollar for Dollar, I believe that the Celeron "D"'s overclocked are better performers. Hopefully my personal view is clear as such, however rather than argue with you on what you can or can not "Abide" I will say that I agree to disagree with your (imo) concrete black and white appraisal.

Click to expand...

postaldudeleo said:

I used a celeron on overclocking and it got 4.78 GHZ!!!!! This is a 60 dollar chip!
Oh and it plays HL2 on top settings with a geforce with no overclocking and I get good fps. Not in the hundreds but 30 is good enough for me.

Some guy bought a $8000+ Alienware with a 20 inch screen for HL2.

I bought a $600+ celeron custom system and play with my 42 inch LCD.
= I win because I get playable FPS on top setting

Guess who has more fun?

Click to expand...

You get what I am saying

R

Though they suck in maya 6 though. But I got a fx-53 system for that.

ropey said:

Somehow you do not get what I am saying.

Click to expand...

That is because what you are saying is untrue. Since you seem incable of recalling your own assertion, and shift the focus to a more defensible point, let me reiterate:

ropey said:

I have personally had a Pentium3 Deshutes 450 and Celeron Mendocino 300A on an ABIT BH6 board. The only time the Pentium3 Deshutes handled better than the Celeron Mendocino 300A @ 450 was when there were multiple cache calls for over 128kb and this is seldom the case unless one is operating heavy (such as database) number crunching and/or scientific algorithmic command structures such as astronomical programs. For the bulk of computer use the Celeron Mendocino 300A was faster as the cache was built on chip and was at full bore.

Click to expand...

Notice the absolute terms used. None of the 'dollar for dollar' stuff. These statements are simply false, at equal clock the P3-450 beats the c300a@450MHz in all but the rarest circumstances, mainly sandra and the like. Application performance is better with the P3 almost uniformly, and certainly accross the 'bulk of computer use'. And a simple comparison of application performance between a celemine and coppermine of equal clock and fsb will handily demonstrate that you have (completely) mis-characterized the magnitude of the performance impediment that the 128KB cache is.

what's the point of you guys arguing over ooolldd cpus..stop it..please, it's killing the sanity

ropey said:

You are right Larva. I am wrong. Excuse me for the temerity to say something you disagree with.

Click to expand...

Yeah, I guess I forgot no one could say something to disagree with you

ropey said:

Somehow you do not get what I am saying. Dollar for Dollar is what I am saying. ... And you still do not address my personal view. So I will reiterate for you.

Click to expand...

You are the one that forced me to clarify my statements. I got what you were saying, no need for the "So I will reiterate" crap. I simply feel your statements saying that a c300a at 450MHz runs applications faster than a P3-450 is wrong. The "problem" you created is not that I don't know what you said, but that I disagree.

Believe it or not I tested this pair against each other ad nauseum when the chips were current, and not from the perspective of justifying saving the money the celeron did over the P3. The P3 was quicker at 450MHz in all but the rarest and non-representative applications, and of course a good P3-450 did 600MHz. So even when you try to shift the focus to a "dollar for dollar" argument, you must compare apples to apples. When you OC the celeron, oc'ing the P3 is fair and must be done to create a valid comparison.

I stated my opinion and was content to leave it at that. You are the one that goaded me into a response by saying that I just didn't know what you were saying. I knew exactly, and still know. I just do not agree and will say so here or anywhere else, whether I am a moderator or not. If I were inclined to abuse my powers I would certainly remove your little tirade, but its wording and content is more damning than anything I could do via my powers as moderator.