Old cpus questions.

[cwj717]

Those benchmarks really highlight the inefficiencies of early P4s. Also makes the 3.06 Northwood I have look good,

A 3.0G Northwood is still a respectable chip;If you read well between the CPU chart, you'll see it often takes a 3.2 or even 3.4GHz Prescott P4 to match a 3.0GHz Northwood. The 3.0C Northwood was faster than Prescott (at the same clock)... I don't think Northwood A or B are... 3.06GHz P4s had a 533MHz bus (thats a NW B), did they not?

 

[cwj717]

The performance difference is upwards of 40%, that's non-trivial, might make the difference in playing some video clips for example or be a fairly linear gain in FPS for gaming.

I like the Tualatins for another reason though, you can drop the FSB and core voltage (given a board that suports it) and run them passively while still having a lot more processing power than something like a Via C3 or C7. While I wouldn't buy a combo for this today it makes a reasonable NAS/fileserver/etc on the cheap.

No, the performance difference between the Celeron 1.40 Tualatin and PIII 1000EB is squat, at least in games. I owned both. The PIII works better for its speed due to the 133MHz bus.
I agree with that. I also think that the benchmarks I linked to early in this thread show that as well. Some benchmarks the P3 was faster some the Celeron... it really depends on the benchmark, some prefer a faster core clock but other prefer a faster bus clock.

 

[Crashman]

You're not sure about anything because the VRM was designed for multiple CPU's. That's like saying I'm sure that Coppermine's run 1.50V because my Celeron 566 (Coppermine) ran 1.50V.

I know for certain the VRM for the earliest Celerons allowed voltages at least as low as 1.80V. Possibly lower, but definately as low as 1.80V.

This board, officially supports coppermine Pentium3 CPUs, that run on 1.60-1.75V ; the only problem is they haven't put there the jumper pins to switch voltage and pin function; There are TWO HOLES where the jumper pins should be. Am I clear now :roll:

Woah now, back the train up, I thought you said you were certain the board would attempt to boot your Coppermine Celeron at 2.0V?

Anyway, if the jumper is missing, there's probably something else missing in the board, perhaps the VRM isn't 8.4 (1.30-2.xxV) but instead VRM 8.2 (1.80-3.xxV) You could try soldering the jumper pins onto position, and if it doesn't work you'd know something else was amis.

Which brings us back to the "Lin-Lin" adapter. Get one, set it to 1.80V (a simple pin assignment change that causes the motherboard to use 1.80V), add the Coppermine CPU, and clip it in. 1.80V is only 0.05V more than the highest Intel used, and a common voltage used by overclockers as well.

 

[cwj717]

And its not even worth getting an adapter to try it; most boards don't support tualatin voltages.

Oh gawd, go back through the whole conversation and see how wrong you are. The adapters work on most Coppermine-compatible boards, he's using a Coppermine-compatible board.

The performance difference is so small that it's not worth trying the adapter. I can agree that maybe it works on most S370 boards. I was confused early in the thread because I wasn't sure if we were talking Slot-1 or S370, sorry. But if its a Slot-1 board, its really hit or miss with compatibility. I have had no luck with ANY of my boards and various adapters. EDIT: Typos

 

[Crashman]

Most of the consumer boards supported adapters, most server/workstation boards did not. There have been several reasons why, but some guy came out with a web collection of P2B-D information required to make Slotkets work.

Like I said, the reasons were various. Funny to see a Slotket work on an Intel board (so long as the BIOS was one or two versions prior to the "no go" switch) but not on an Asus board (just server boards).

 

[cwj717]

Maybe thats the difference, most of my boards are Dual processor... although, I did have a few that were single. I still have one Asus, its a P2 99 and I also still have an Intel SE440BX2. I also still have two S370 boards... and IBM one and a Sis one. I might try to find a Tualatin cheap somewhere to try

 

[Crashman]

SE440BX2 should work for Tualatin with the proper BIOS. Upgradeware and Powerleap have documentation on which BIOS you'll need.

For Slot 1, I always advise to use the cheaper Upgradeware adapter for VRM 8.4 boards like your SE440BX2, because it's more reliable and more adjustable than the Powerleap adapter.

 

[I]

The performance difference is upwards of 40%, that's non-trivial, might make the difference in playing some video clips for example or be a fairly linear gain in FPS for gaming.

I like the Tualatins for another reason though, you can drop the FSB and core voltage (given a board that suports it) and run them passively while still having a lot more processing power than something like a Via C3 or C7. While I wouldn't buy a combo for this today it makes a reasonable NAS/fileserver/etc on the cheap.

No, the performance difference between the Celeron 1.40 Tualatin and PIII 1000EB is squat, at least in games. I owned both. The PIII works better for its speed due to the 133MHz bus.

In fact, the Tualatin Celeron 1100 at 1466 (overclocked to 133MHz FSB) is around 30% faster in some apps than the Tualatin Celeron 1400. And because of that, I used this configuration.

I also tried underclocking to 66MHz FSB and running passive cooling, using an oversized Pentium MMX passive cooler pulled from an old IBM. It was great, would have made a nice HTPC.

I've benchmarked this plenty of times, you are clearly wrong. There is a very close to linear improvement in FPS in games with the Tualatin Celeron, even on 100MHz FSB so long as the video card is modern enough that it isn't the bottleneck (say GF2 era or newer). While the lower FSB and memory certainly don't help the Tually Celeron at all, they're not yet much of a bottleneck or detriment relative to the CPU clock rate unless you had integrated video instead of a card.

Celeron 1100 @ 1466 is "Maybe" 10% faster on average, not 30% compared to Celeron 1400 unless all you are considering is some benchmark that isolates memory throughput instead of real world uses... and I do have a Tualatin 1.1 @ 1.5 here as well as a P3 9xx-something coppermine. High FSB matters far less than MHz once you get past 66MHz FSB.

http://www.overclockers.com/tips716/

The following link at the bottom is a more direct comparison of Celeron 1.4 vs 1.1 oc to 1.47GHz. The stock 1.4GHz Celeron is only 95% as high a CPU clock as the overclocked 1.47GHz (@ 133 FSB), but notice how if it's given a decent video card (like Radeon 8500 or better) the scores are 7136:7432 or only 4% difference from a 5% higher clocked CPU.

In other words, clock for clock the Celeron 1.1 @ 1.47 is SLOWER than the Celeron 1.4 at stock in gaming, the FSB did not matter.

http://www.geocities.com/_lunchbox/tualatin_performance_chart.html

That pretty much wipes the floor with any P3 1GHz, you have no idea what you're talking about.

 

[I]

The performance difference is upwards of 40%, that's non-trivial, might make the difference in playing some video clips for example or be a fairly linear gain in FPS for gaming.

I like the Tualatins for another reason though, you can drop the FSB and core voltage (given a board that suports it) and run them passively while still having a lot more processing power than something like a Via C3 or C7. While I wouldn't buy a combo for this today it makes a reasonable NAS/fileserver/etc on the cheap.

No, the performance difference between the Celeron 1.40 Tualatin and PIII 1000EB is squat, at least in games. I owned both. The PIII works better for its speed due to the 133MHz bus.
I agree with that. I also think that the benchmarks I linked to early in this thread show that as well. Some benchmarks the P3 was faster some the Celeron... it really depends on the benchmark, some prefer a faster core clock but other prefer a faster bus clock.

With a higher FSB & Memory, you will of course get more performance per CPU clock. That does not change that fact that in the vast majority of cases having a 40% higher CPU clock is still significantly faster than a 33% higher FSB. The issue is not "P3 vs Tualatin Celeron", it is what speed of each we're comparing, and in that context there's very very little a P3 @ 1Ghz can do even 90% as fast as a 1.4GHz Celeron. Then when you consider you can run tighter memory timings on the lower memory bus speed or jump that Tualatin FSB up some, it's no contest.

 

[m25]

I am so glad some ppl remember celerons and know somthing about them

Budget has always been a concern for me, and since AMD CPUs were missing where I live, I've been sitting for a long time on celerons.

 

[Crashman]

No fucknut, you have no idea what YOU'RE talking about. Why don't you come to michigan with some hardware and a face mask. Trust me, you'll need the face mask. I OWNED ALL THIS HARDWARE IN 2002 AND BENCHMARKED ALL OF IT MYSELF AND YOU'RE MAKING A SERIOUS MISTAKE IF YOU THINK YOU CAN CALL ME A LIAR.

 

[quantumsheep]

No ******, you have no idea what YOU'RE talking about. Why don't you come to michigan with some hardware and a face mask. Trust me, you'll need the face mask. I OWNED ALL THIS HARDWARE IN 2002 AND BENCHMARKED ALL OF IT MYSELF AND YOU'RE MAKING A SERIOUS MISTAKE IF YOU THINK YOU CAN CALL ME A LIAR.

Calling Crashman a liar is almost the same as someone saying i'm not a sheep (either is a blatant falacy).

 

[Crashman]

I remember back then, I was trying to find the fastest solution for running TMPGenc to encode VCD's. I had a Radeon LE DDR TVO overlocked to Radeon DDR speeds, an Asus P3B-F, and 512MB PC-133.

 

[quantumsheep]

I remember back then, I was trying to find the fastest solution for running TMPGenc to encode VCD's. I had a Radeon LE DDR TVO overlocked to Radeon DDR speeds, an Asus P3B-F, and 512MB PC-133.

Back in 2002 I was still running my SLI Voodoo 2s, i couldn't bare to part with them. Can't remember the precise CPU, think it was a PIII 733 @ 800ish.

 

[cwj717]

The performance difference is upwards of 40%, that's non-trivial, might make the difference in playing some video clips for example or be a fairly linear gain in FPS for gaming.

I like the Tualatins for another reason though, you can drop the FSB and core voltage (given a board that suports it) and run them passively while still having a lot more processing power than something like a Via C3 or C7. While I wouldn't buy a combo for this today it makes a reasonable NAS/fileserver/etc on the cheap.

No, the performance difference between the Celeron 1.40 Tualatin and PIII 1000EB is squat, at least in games. I owned both. The PIII works better for its speed due to the 133MHz bus.

In fact, the Tualatin Celeron 1100 at 1466 (overclocked to 133MHz FSB) is around 30% faster in some apps than the Tualatin Celeron 1400. And because of that, I used this configuration.

I also tried underclocking to 66MHz FSB and running passive cooling, using an oversized Pentium MMX passive cooler pulled from an old IBM. It was great, would have made a nice HTPC.

I've benchmarked this plenty of times, you are clearly wrong. There is a very close to linear improvement in FPS in games with the Tualatin Celeron, even on 100MHz FSB so long as the video card is modern enough that it isn't the bottleneck (say GF2 era or newer). While the lower FSB and memory certainly don't help the Tually Celeron at all, they're not yet much of a bottleneck or detriment relative to the CPU clock rate unless you had integrated video instead of a card.

Celeron 1100 @ 1466 is "Maybe" 10% faster on average, not 30% compared to Celeron 1400 unless all you are considering is some benchmark that isolates memory throughput instead of real world uses... and I do have a Tualatin 1.1 @ 1.5 here as well as a P3 9xx-something coppermine. High FSB matters far less than MHz once you get past 66MHz FSB.

http://www.overclockers.com/tips716/

The following link at the bottom is a more direct comparison of Celeron 1.4 vs 1.1 oc to 1.47GHz. The stock 1.4GHz Celeron is only 95% as high a CPU clock as the overclocked 1.47GHz (@ 133 FSB), but notice how if it's given a decent video card (like Radeon 8500 or better) the scores are 7136:7432 or only 4% difference from a 5% higher clocked CPU.

In other words, clock for clock the Celeron 1.1 @ 1.47 is SLOWER than the Celeron 1.4 at stock in gaming, the FSB did not matter.

http://www.geocities.com/_lunchbox/tualatin_performance_chart.html

That pretty much wipes the floor with any P3 1GHz, you have no idea what you're talking about. I went to both links you posted. I didn't see a single case where the Celeron 1.4@100MHz FSB was faster than the 1467MHz Celeron with a 133MHz bus. The OCed Celeron was much faster with a better video card like the TI4200. Besides, 3Dmark and Sandra are synthetic... the only real world test was Quake3. EDIT: I noticed you said clock for clock... no matter, the test was probably mostly GPU bound. With the TI4200 the OCed Celeron was 12.8% faster than the stock 1.4GHz one. Thats better clock for clock.

 

[m25]

It's 2007 and I don't believe I didi something like this; I took my old rig, the socket 370 celeron 333 and pinmodded it to lower the voltage; the mod worked great, because the Mendocino celeron, with it's 2.0V rating could barely pull up a trashed startup screen and froze there on the 1.675V I had set. Unfortunately, it looked like I had burned up the Coppermine 633 celeron, because it didn't boot at all
Vcore apart, the two celerons are identical in pin function; it was definitely dead. I only hope to find a new one and play with it.

 

[Crashman]

Nah, the Coppermine and Mendicino cores did have different pin assignments, THG even had an article on how to mod your old Celeron Slotket to support newer Coppermine cores.

At any rate, incompatible CPU's will black screen by default, the only way to know if they're fried is to try them in a compatible system.

 

[m25]

Nah, the Coppermine and Mendicino cores did have different pin assignments, THG even had an article on how to mod your old Celeron Slotket to support newer Coppermine cores.

At any rate, incompatible CPU's will black screen by default, the only way to know if they're fried is to try them in a compatible system.

I have followed tutorial on modding a celeron only board which used calerons from 233MHz up to 1.1GHz (both mendocinos and coppermines) to use a P3. The P3 had some differences in pin arrangement but the res was only voltage adjusting. That board officially supports coppermine P3s and slightly after them, coppermine celerons were released, so there's no reason why it should have blacked out. I have tried that celeron before without modding, on 2.0V, and that could have fried it.
I guess the modding article you mention is pretty much about Vcore adjustment, because an OCed coppermine run at most on 1.80-1.85v and mendocinos used 2.0V.

 

[Crashman]

No, it's a different mod, it's just darned near impossible to find articles that old. One of the redundant pins was reassigned a new function, the mod was done to carry a signal from one pin to another.

Read about this old adapter:
http://www.powerleap.ca/Products/Neo-S370.htm

Notice they say both voltage and pin definitions. I still can't find the socket mod, but maybe you'll have better luck finding a "PPGA to FC-PGA mod"

 

[Crashman]

Hah, I found it:

The 'RESET'-pin

Intel was quite humorous when designing the pin layout of Coppermine. To ensure that Coppermine was incompatible to Celeron the 'Reset#'-pin was placed at a new location (AH4) . The old location (X4) known from Celeron is called 'Reset2#', but it has no official function anymore. No processor in the whole world would ever start to work without a signal to 'Reset#' in the moment when you turn on your system. Placing 'Reset#' at a formerly 'reserved' pin of Celeron made sure that Coppermine would never work in Socket370.

Source

And now that you know how, here's a guy who did it to his Socket 370 board, rather than a slotket.

http://www.3feetunder.com/krick/370mod/

 

[apache_lives]

Nah, the Coppermine and Mendicino cores did have different pin assignments, THG even had an article on how to mod your old Celeron Slotket to support newer Coppermine cores.

At any rate, incompatible CPU's will black screen by default, the only way to know if they're fried is to try them in a compatible system.

I have followed tutorial on modding a celeron only board which used calerons from 233MHz up to 1.1GHz (both mendocinos and coppermines) to use a P3. The P3 had some differences in pin arrangement but the res was only voltage adjusting. That board officially supports coppermine P3s and slightly after them, coppermine celerons were released, so there's no reason why it should have blacked out. I have tried that celeron before without modding, on 2.0V, and that could have fried it.
I guess the modding article you mention is pretty much about Vcore adjustment, because an OCed coppermine run at most on 1.80-1.85v and mendocinos used 2.0V.

if the board (unmoded, default) supports mendocino/covington (250nm) cpus and not coppermine (180nm) cpus, the system will not post, and if i remember correctly will not even power the cpu - feel the cpu die, it will remain cold (all my boards did that when i had em).

If i remember correctly, Intel even conciders 2.0v vcore usage safe for there 180nm cpus, THG even commented saying to use with caution and decent cooling, i must find the article.

Modding deschutes/katmai/covington/mendocino (250nm, 2.0v) boards (well, even 350nm 2.8v clamath cpus aka P2, ~40w @ 333mhz! more then a Tualatin 1400!) to support coppermine (180nm, ~1.75v) cpus is easy as - cut one pin on the cpu and soldier a wire across the back of the socket and presto it works.

Theres more boards that support the mod then not, mind you one or two will not work with it, one example is the ASUS P2B, atleast for me it didnt work at all.

My favourites in there day - ABit BH6 (no FSB133 support 🙁 ) and the AOpen AX6BC - both worked a treat, both Slot1.

The greatest product MSI ever made - the MS6905 slocket adapter, mind you they fcuked it up so many times and made what 4 revisions to support coppermine finally and even tualatin cores? i still have one or two!

 

[m25]

Hah, I found it:

The 'RESET'-pin

Intel was quite humorous when designing the pin layout of Coppermine. To ensure that Coppermine was incompatible to Celeron the 'Reset#'-pin was placed at a new location (AH4) . The old location (X4) known from Celeron is called 'Reset2#', but it has no official function anymore. No processor in the whole world would ever start to work without a signal to 'Reset#' in the moment when you turn on your system. Placing 'Reset#' at a formerly 'reserved' pin of Celeron made sure that Coppermine would never work in Socket370.

Source

And now that you know how, here's a guy who did it to his Socket 370 board, rather than a slotket.

http://www.3feetunder.com/krick/370mod/
Thanks for the articles; I tried everything but that celeron seems to be really dead.

 

[I]

No ******, you have no idea what YOU'RE talking about. Why don't you come to michigan with some hardware and a face mask. Trust me, you'll need the face mask. I OWNED ALL THIS HARDWARE IN 2002 AND BENCHMARKED ALL OF IT MYSELF AND YOU'RE MAKING A SERIOUS MISTAKE IF YOU THINK YOU CAN CALL ME A LIAR.

You're deluded about Tualatin performance. What did you think, nobody else has owned hardware a mere 4 years ago? Did you think nobody bothered to benchmark back then but you are a font of P3 knowledge?

Let me explain something to you plainly. You will end up being pretty badly treated if you don't learn how to act. I don't care what offends you even though I didn't consider you a liar, only incompetent.

As for serious mistake, how about calling you an idiot? I've been punched in the face, child, I don't melt. You obviously hide behind a keyboard because if you acted like that in public you'd already be unable to type, or in prison.

 

[I]

I went to both links you posted. I didn't see a single case where the Celeron 1.4@100MHz FSB was faster than the 1467MHz Celeron with a 133MHz bus. The OCed Celeron was much faster with a better video card like the TI4200. Besides, 3Dmark and Sandra are synthetic... the only real world test was Quake3. EDIT: I noticed you said clock for clock... no matter, the test was probably mostly GPU bound. With the TI4200 the OCed Celeron was 12.8% faster than the stock 1.4GHz one. Thats better clock for clock.

Yes I wrote clock for clock. The subcontext of the thread was that the FSB makes this big difference when it does not. Even with the o'c Celeron on a 33% faster FSB, faster memory, and higher clock speed, it was only ~ 13% faster "sometimes" and even less difference much of the time. We can clearly see from this that FSB rate doesn't have nearly as much of an effect as core rate. The P3 coppermine hasn't a chance at being competitive at the vast majority of applications this system would be useful for.

 

[dasickninja]

No ******, you have no idea what YOU'RE talking about. Why don't you come to michigan with some hardware and a face mask. Trust me, you'll need the face mask. I OWNED ALL THIS HARDWARE IN 2002 AND BENCHMARKED ALL OF IT MYSELF AND YOU'RE MAKING A SERIOUS MISTAKE IF YOU THINK YOU CAN CALL ME A LIAR.

You're deluded about Tualatin performance. What did you think, nobody else has owned hardware a mere 4 years ago? Did you think nobody bothered to benchmark back then but you are a font of P3 knowledge?

Let me explain something to you plainly. You will end up being pretty badly treated if you don't learn how to act. I don't care what offends you even though I didn't consider you a liar, only incompetent.

As for serious mistake, how about calling you an idiot? I've been punched in the face, child, I don't melt. You obviously hide behind a keyboard because if you acted like that in public you'd already be unable to type, or in prison.
You are an idiot. I look forward to Crash totally shutting you up.