1.3Ghz Celeron replacement for 700Mhz PIII

[imported_Skorpio]

How is the performance increase if I buy a 1.3Ghz Celeron (Tualatin) to replace a 700Mhz PIII?

Or would there be any since its a celeron?

 

[Spikesoldier]

big difference

great upgrade with minimal costs instead of buying new cpu/mb/ram

cheap upgrade compared to what you would have to get by moving to a new platform. it will definately breathe new life into that system of yours.

 

[mindless1]

Well if it's still the same system otherwise, your HDD is probably the biggest bottleneck, presuming you've enough memory.

Tualatin Celeron is same as a Coppermine P3 w/100MHz FSB. Coppermine Celerons had only 128K L2 cache but it doubled in the Tualatin. At CPU intensive tasks it will make a very big different, maybe 80% faster, though ultimately with same FSB and memory bus speed you may not notice a much snappier system for everyday use, more like a situation of those times things used to take awhile, now don't take as long.

Keep in mind that the average mobo sold with a P3 700 can't run a Tualatin CPU due to pin changes. There are adapters for about $5

 

[imported_Skorpio]

Thanks for the link mindles...but...already knew it

how does the 1.3Ghz Celeron compare to a 1.2Ghz Pentium III...both are tualatin cores but the PIII is 133FSB compared to a Celeron which is 100Fsb?

 

[MercenaryForHire]

The 1.2GHz P3-T will kill that C-1.3, but will probably also cost twice as much.

Make sure your board supports Tualatins before buying one though.

- M4H

 

[mindless1]

Board doesn't have to support Tualatins if the adapter is used, but "some" boards may not work. I've had luck in the past with Via 694X based, but a Sis (620?) wouldn't even post with one.

P3 Tualatin is a little faster per MHz due to 133FSB, if/when memory sync'd @ 133. Depends on your task, if it's memory bound or CPU, which matters more. Generally, the cost premium for P3 Tualatin isn't worth the performance benefit, but some p3 had the 512K L2 cache. Best bang for buck in the past was taking a ~1.1GHz Tualatin Celery and o'c to 133FSB, with voltage increase to about 1.65V. YMMV, nothing is guaranteed but most Tualatins would hit > 1.35GHz. Then again i haven't priced any Tualatins recently. Frankly if the P3 version is more than $10 more, I'd pass on it. Beyond a certain point it'd be better to just find a cheap board that'll run an Athlon XP Tbred-B.

 

[BCinSC]

Originally posted by: mindless1
Tualatin Celeron is same as a Coppermine P3 w/100MHz FSB. Coppermine Celerons had only 128K L2 cache but it doubled in the Tualatin. At CPU intensive tasks it will make a very big different, maybe 80% faster, though ultimately with same FSB and memory bus speed you may not notice a much snappier system for everyday use, more like a situation of those times things used to take awhile, now don't take as long.

Um, no. Tualatin Celeron and Tualatin P3 are identicle except FSB: 100 vs 133 (with exception of P3-S with 512Kcache). 1.3GHz Celeron nicely overclocks to 1.7GHz at 133FSB. I've got 3 running and a P3-S 1.13.

 

[mindless1]

Originally posted by: BCinSC

Originally posted by: mindless1
Tualatin Celeron is same as a Coppermine P3 w/100MHz FSB. Coppermine Celerons had only 128K L2 cache but it doubled in the Tualatin. At CPU intensive tasks it will make a very big different, maybe 80% faster, though ultimately with same FSB and memory bus speed you may not notice a much snappier system for everyday use, more like a situation of those times things used to take awhile, now don't take as long.

Um, no. Tualatin Celeron and P3 are identicle except FSB: 100 vs 133 (with exception of P3-S with 512Kcache). 1.3GHz Celeron nicely overclocks to 1.7GHz at 133FSB. I've got 3 running and a P3-S 1.13.

"Um, no", what? I was't speaking of process size since OP did't either, but nothing else I wrote is incorrect in context.

As for o'c to 1.7GHz, definitely not. It is extremely rare to find a Tualatin that o'c past 1.6GHz without very exotic cooling, and some won't even do 1.4GHz except at such an extreme voltage ramp that it's not worthwhile. It is possible, but does nobody any good to pretend that's a realistic target.

 

[rogue1979]

If the board supports a 133MHz fsb, just get a 1.0a or a 1.1a.

Both should make the 133MHz fsb overclock and will give you a substantial boost over the 1.3GHz because it runs in a 100MHz fsb.

 

[BCinSC]

Originally posted by: mindless1
"Um, no", what? I was't speaking of process size since OP did't either, but nothing else I wrote is incorrect in context.

As for o'c to 1.7GHz, definitely not. It is extremely rare to find a Tualatin that o'c past 1.6GHz without very exotic cooling, and some won't even do 1.4GHz except at such an extreme voltage ramp that it's not worthwhile. It is possible, but does nobody any good to pretend that's a realistic target.

You said Tualatin Celeron was same as Coppermine P3and that isn't true. At same clock/fsb, the tualatin is faster - because of the .13 process.

I've got 3 Tualatin 1.3 at 1729MHz using 133fsb and stock cooling. I've even had one go to 1.82GHz Maybe I've been lucky, maybe it just works. If it scares you, wuss out and get a 1.2 and settle for 1596 or the 1/1.1a that Rogue mentions.

 

[blackhawk]

I've run a few 700s and all went to 133 fsb which was a nice bump up.

I've also had 4 tualitan celerons, and still have one 1100 in my daughter's running 1466 for several years. I wouldn't count on getting much overclocking out of a 1300 without heroic efforts. I also had two celeron 900s that both went to 1200/133 although they have only 128 cache.

A 1000 or 1100 would make most sense for an overclock and realize any tualitan you'd buy would be old.

As someone else stated above, you probably have so many bottlenecks elsewhere in a system that old it would be more sensible to upgrade the whole thing and transfer all your data over.

However, if you dont want to do that, I'd think upping your fsb to 133 for 933mz and increasing your ram to 512 plus either wiping and re installing your OS would do you the most good.

Tualitan motherboards are rare, adaptors are iffy although most via chipsets handle them well.

 

[BCinSC]

TUSL2-C was an excellent board for OC. Some new stock tualatin celerons still exist.

 

[mindless1]

Originally posted by: BCinSC

Originally posted by: mindless1
"Um, no", what? I was't speaking of process size since OP did't either, but nothing else I wrote is incorrect in context.

As for o'c to 1.7GHz, definitely not. It is extremely rare to find a Tualatin that o'c past 1.6GHz without very exotic cooling, and some won't even do 1.4GHz except at such an extreme voltage ramp that it's not worthwhile. It is possible, but does nobody any good to pretend that's a realistic target.

You said Tualatin Celeron was same as Coppermine P3and that isn't true. At same clock/fsb, the tualatin is faster - because of the .13 process.

LOL. No, sorry but that's quite wrong. There was one other, mostly trivial optimization in the Tualatins, but @ same clock, FSB rate, and L2 cache size, Tualatin performance is exactly same as P3. A change from .18 to .13 in itself does NOTHING to improve performance, rather it facilitated the other changes (cost-effective manufacturing, lower voltage/heat, (or) higher clock speeds, & more L2 cache on P3-S). Benching a Tualatin Celery @ 1GHz and a P3 1GHz w/100FSB would show same results, at least so close they're within statistical margin of error.

I've got 3 Tualatin 1.3 at 1729MHz using 133fsb and stock cooling. I've even had one go to 1.82GHz Maybe I've been lucky, maybe it just works. If it scares you, wuss out and get a 1.2 and settle for 1596 or the 1/1.1a that Rogue mentions.

You've been more than lucky, you can't even have anything but "golden" samples. Maybe you're not even talking about Tualatin but Pentium-M? Back in that era tons of people overclocked Tualatins. It was common knowledge that to even hit 133 FSB with a Celery one should choose 1.1-1.2GHz default CPU at MOST. Maybe their last stepping improved the design layout such that 1.6GHz was more likely, given enough effort, but today that much effort to only end up with 1.6Ghz seems a waste of time, since one of the primary benefits of Tualatin was it's lower heat... if it's going to be overvolted enough to do 1.6GHz and beyond then suddenly there's nothing positive about Tualatins, a cheap sempron will walk all over one, not to mention that excessive o'c on an old socket 370 has been known to vent capacitors- S370 boards simply weren't designed with even 3/4 that much amperage in mind.

 

[BCinSC]

If the chips are physically identicle and one (celeron) is set to run at 100FSB while the other (Pentium3) is set to run at 133, how does setting Celeron to 133 change the amperage? I've never adjusted my vCore - hell, even my P4/2.4B overclocked to 3.24GHz is stock voltage/cooling. In fact, the Soyo board it's on UNDERvolts the CPU at 1.42v per CPUZ.

My Celery stash came from NewEgg - retail boxes. My 2.4B out of a Dell server.

So, Mindless1, are you an Intel, AMD, IBM, or some similar Engineer? I've been doing PCs since the XT was $5000 new. We used punch card readers to program Fortran in college. I have 40-odd systems here in my house, many overclocked. What's your deal?

As for performance advantage of .13 over .18 process - a ~28% decrease in distance each and every electron has to travel has NO impact? Sorry, I call bullsh|t. I've got Katmai, Coppermine and Tualatins galore. I can assure you, the Tualatins show improvement clock for clock. Just as the Coppermine's did over Katmai going from .25 to .18 - also a 28%+ shrink.

 

[BCinSC]

Oh, and those Celerons produce almost as well as P4/2.53 in Find-A-Drug, Seti and Folding@Home. If I didn't overclock my AthlonXP 2500+s to 2.4GHz+, they'd be in the same bracket.

 

[mindless1]

Originally posted by: BCinSC
If the chips are physically identicle and one (celeron) is set to run at 100FSB while the other (Pentium3) is set to run at 133, how does setting Celeron to 133 change the amperage? I've never adjusted my vCore - hell, even my P4/2.4B overclocked to 3.24GHz is stock voltage/cooling. In fact, the Soyo board it's on UNDERvolts the CPU at 1.42v per CPUZ.

Increasing the clock speed of the Tualatin by increasing the FSB speed, will cause a linear increase in amperage too. That is inescapable at same voltage (vcore).

You have not ran Tualatin Celerons at 1.7GHz with no vcore increase. Previously I might've chalked your experience up to luck, but that (1.7GHz @ stock vcore) is simply impossible. Do you know what the default vcore is on a Tualatin? I suspect your board wasn't fully supportive and was overvolting them already.

My Celery stash came from NewEgg - retail boxes. My 2.4B out of a Dell server.

So, Mindless1, are you an Intel, AMD, IBM, or some similar Engineer? I've been doing PCs since the XT was $5000 new. We used punch card readers to program Fortran in college. I have 40-odd systems here in my house, many overclocked. What's your deal?

Most of my Tuallys came from Newegg also. You might be very competent at PCs, or you might not. It is irrelevant what else you do or don't, have or haven't done. This specific scenario has no ties to punch cards. If you want to reminisce, I too used punch-cards. My father was admin for xxxxxx Telco Info Systems and I was the lucky one stuck with feeding 'em in for conversion to spools in the early days of leased IBM mainframes. It has no bearing on Tualatins.

As for performance advantage of .13 over .18 process - a ~28% decrease in distance each and every electron has to travel has NO impact? Sorry, I call bullsh|t. I've got Katmai, Coppermine and Tualatins galore. I can assure you, the Tualatins show improvement clock for clock. Just as the Coppermine's did over Katmai going from .25 to .18 - also a 28%+ shrink.

With all due respect, your ego is far larger than your knowledge about Tualatins. Differences in OTHER core designs like SSE support, on or off-ide L2 cache, and different cache sizes, played a role in performance differences "per MHz" if we ignore FSB speed. That is not the case with Tualatin, it had no significant advantage over prior P3 IF running at same FSB, with same L2 cache size, same operational frequency.

In other words, your idea about .13 vs .18 "decrease in distance" is nonsense. You have nothing to cite which even begins to support this, you're pulling it out of thin air. Considering this and your misinformation about vcore and Tualatin o'c, I feel we should consider you a troll.

 

[BCinSC]

At 2 years longer on these boards and 50% more posts than you, I would say you are the troll, not me (I was on these boards even before 10/01, but conversion lost prior history). If you used real Tualatin motherboards and not crappy convertors, you'd have better luck overclocking. I have no reason to believe my ASUS boards are overvolting, doesn't say so in BIOS or CPUZ, but if you have another means of telling, I'll check it. The decrease in distance theory comes from Intel's own documentation. I assure you, my Tualatins are running happily at 1729MHz.

 

[CraigRT]

The "tualeron" was IMO the only good Celeron really... so it should be an OK upgrade of a PIII 700.. go for it.

 

[BCinSC]

Overall, CraigRT, I agree, though in it's time, the 300A was pretty decent - easy overclock to 450 and faster than a PII/450 because of on-die full-speed cache, versus off-die half-speed.

 

[mindless1]

Originally posted by: BCinSC
At 2 years longer on these boards and 50% more posts than you, I would say you are the troll, not me (I was on these boards even before 10/01, but conversion lost prior history).

You really are mixed up aren't you?
Length of time makes you not a troll?
Perhaps you're just grossly misinformed.
This IS NOT APPLICABLE to Tualatins nor your misinformation, nor the "bullsh!t" attitude you take when confronted with the facts.

If you used real Tualatin motherboards and not crappy convertors, you'd have better luck overclocking.

If you had a clue what motherboards I used your comment might be more appropriate. That doesn't even consider the gross evidence all over the 'net for several years. When one hits 1.7GHz, it is noteworthy, an exception rather than the typical result. Most people hoped for 1.5GHz, and some weren't able to hit that.

I have no reason to believe my ASUS boards are overvolting, doesn't say so in BIOS or CPUZ, but if you have another means of telling, I'll check it. The decrease in distance theory comes from Intel's own documentation. I assure you, my Tualatins are running happily at 1729MHz.

Generally one takes a multimeter and measures the VRM circuit output. You might have a rare tualatin that'll do 1729. That is not impossible. Default vcore is 1.475-1.50(?) for the 1.3GHz Celeron parts, it generally takes about 1.65-1.75V to hit over 1.6GHz. BTW, that Tualatin optimization that they added over Coppermnes was Data Prefetch Logic. It's not of much benefit, one can easily see that a 1.3GHz PIII Tualatin has only a linear performance increase over Coppermine P3 1GHz in benchmarks all over the web, like this one.

DO provide us with that Intel documentation, a vague reference doesn't quite cut it. Real benchmarks all point out the same thing, that the move to Tualatin was move of a frequency increase measure than anything else, excepting that the smaller process size allows 512K L2 on P3-S.

 

[BCinSC]

Interesting quote from the article you mis-linked (fixed linky)
"You can probably guess that the 0.13-micron manufacturing process opens up the potential for the Pentium III to be a great overclocker. Our retail 1.2GHz processor was able to hit 1.44GHz (9 x 160MHz), unfortunately the board we tested on (ASUS TUSL2-C) would not allow voltage adjustments thus preventing us from getting a reliable set of benchmarks at 1.44GHz. With voltage adjustments you should be able to hit 1.35 ? 1.5GHz pretty easily with the 1.2GHz Pentium III."

 

[mindless1]

Yes, 1.35-1.5 IS the typical range.
What you mentioned previously about adapters, is a potential issue for getting highest overclock. Since OP is looking at reusing the motherboard it could be a limit, either in o'c or lifespan due to current capability.

 

[BCinSC]

My point was that the motherboards I have don't allow voltage adjustments and oc just fine. Your point is well taken - OP doesn't care what board I have, it's what his board can use.

 

[mindless1]

There are ways around that, boards without voltage settings can still relatively easily be modded. For example, the following pic is a Shuttle MN-31 that had a 10K Ohm POT added and wired to the voltage regulator controller (to the left of the AGP slot). Other more conventional mods include CPU pin wire-wraps or insulation or soldering on the socket pins on the back of the board.

If one uses the adapter I linked previously (pic), the jumpers in it's socket-well can be manipulated to different voltages too. That information is unfortunately not included in the poorly translated "manual" but to anyone who finds use in the info (use at own risk);


NOTE - Original Jumper labels are hard to read, following table called the left column of 9 pins to be "A" and the right column of 9 to be "B".

? No Jumpers a1 through b3 :: 1.30V

? a8-a9 :: 1.35V
? a2-a3 :: 1.40V
? a2-a3 a8-a9 :: 1.45V
? a5-a6 :: 1.50V
? a5-a6 a8-a9 :: 1.55V
? a2-a3 a5-a6 :: 1.60V
? a2-a3 a5-a6 a8-a9 :: 1.65V
? b2-b3 :: 1.70V
? a8-a9 b2-b3 :: 1.75V
? a2-a3 b2-b3 :: 1.80V
? a2-a3 a8-a9 b2-b3 :: 1.85V
? a4-a5 b2-b3 :: 1.90V
? a8-a9 b2-b3 :: 1.95V
? a2-a3 a5-a6 b2-b3 :: 2.00V
? a2-a3 a5-a6 a8-a9 b2-b3 :: 2.05V
-------------------------------------------------
? b5-b6 b8-b9 :: 66MHz
? b8-b9 :: 100MHz
? No Jumpers b4 through b9 :: 133MHz

 

[Zap]

Originally posted by: mindless1
BTW, that Tualatin optimization that they added over Coppermnes was Data Prefetch Logic. It's not of much benefit, one can easily see that a 1.3GHz PIII Tualatin has only a linear performance increase over Coppermine P3 1GHz in benchmarks all over the web, like this one.

Hey now, stop bickering.

Same multiplier and FSB, Tualatin same or faster than Coppermine = TRUE
Speed increase came from smaller process = negligible
Speed increase came from data prefetch = TRUE

Also, neither of you are trolls. You don't even come close. FelixDeKat was a troll (WTF happened to that frisky critter anyways?). The original troll was Nowhereman. In fact, he was such a troll that he issued licenses to troll on these forums to other trolls.