Probably the Celeron is best, then the 1.6A, then the 1.8 Willamette. That said, I wouldn't expect very dramatic differences real-world between them.

I decided on using the 1.8GHz and 1.6AGHz P4 CPU's.

Now I need to decide between a Pentium 4 1.7 GHz (256KB) and the Celeron 2.4GHz (128KB).

I'm leaning towards the P4 1.7 unless anyone can suggest otherwise.

Thanks

obobskivich wrote:

Probably the Celeron is best, then the 1.6A, then the 1.8 Willamette. That said, I wouldn't expect very dramatic differences real-world between them.

So the Celeron is better? Even thought it only has 128KB cache?

Totempole wrote:

So the Celeron is better? Even thought it only has 128KB cache?

It's clocked considerably higher though. I vaguely remember a rule of thumb along the lines of doubling cache on a P4 equates to something like 200MHz of clockspeed; that may not be 100% accurate but the gist is that while extra cache helps, it doesn't help universally across the board (it's usually most "felt" with multimedia applications), and there is a point where higher clockspeed will win out. For gaming I would expect the Celeron 2.4 to be superior.

Depending on the stepping, there's a good chance your 1.6A might be highly overclockable. That's a a very low clocked Northwood core, so it's likely it was clocked far below it's actual abilities. If you have any interest to overclock anything, I'd experiment with that one.
Just staying conservative, it would likely run fine at 2.1GHz on a 133/533 bus. That's a standard bus speed so all the PCI/AGP/etc bus clocks should be able to stay in spec.
If overclocking is considered then the 1.6A should definitely be your fastest chip. Even without, it might still be.

The 1.8 is a Willamette, probably near it's limit already. The fastest Willy they sold was 2.0GHz.
The Celeron 2.4 should be a Northwood core. I don't know how it would stack up against the others. The Northwood chips were normally better than Willamette, but with the smaller cache of the Celeron version I don't know how it works out. A cache miss is very expensive on the P4 architecture. Clock speed is significantly higher though, as obobskivich mentioned. For apps/games that don't need much cache, it might be a good performer.
I think the Willamette P4 1.7 would be slower, but if you have both chips, you could benchmark both of them and see what happens. The thing about cache size is that it doesn't affect all applications equally, so you'd have to judge based on what happens in the things you intend to run.

1.6a Should easily overclock.
I'd say that would be the best of the three once its running above 2ghz.

The Celeron would only be faster in specific workloads (media encoding, for example). When it comes to gaming though, the 128K Northwood Celeron is quite the slug. For example, in 3DMark01 the Celeron 2400 only hits 9302 with a Radeon 9800 Pro. P4 1.8 Willamette gets 9178, PIII-S at 1.59 scores 11544, and P4 2.4A/400 hits 12125. The tiny cache hurts gaming performance in a big way.

Out of the three, my choice would be the 1.6A, then overclock it to at least 2.13/533. Those early Northwoods ran cool and were excellent overclockers

i am sure that the overall performance is celeron2.4>p4-1.6a>p4-1.8g
the celeron does have only 128kb cache and about 70% the performance of the northwood p4, but higher clock makes up for it.

Standard Def Steve wrote:

The Celeron would only be faster in specific workloads (media encoding, for example). When it comes to gaming though, the 128K Northwood Celeron is quite the slug. For example, in 3DMark01 the Celeron 2400 only hits 9302 with a Radeon 9800 Pro. P4 1.8 Willamette gets 9178, PIII-S at 1.59 scores 11544, and P4 2.4A/400 hits 12125. The tiny cache hurts gaming performance in a big way.

Out of the three, my choice would be the 1.6A, then overclock it to at least 2.13/533. Those early Northwoods ran cool and were excellent overclockers

Maybe his motherboard will auto-OC the CPU to 2.13/533.

At least mine does and it results 2.40/533,however I have a 1.8A while his is a 1.6A.

These motherboards just have bug in CPU_ID readings, IMHO.
Because the CPU multiplier and the default FSB frequency are hard coded in a CPU. At least for Intel P4s.

I'd say the 1.8Ghz Pentium 4. It has the most cache being 256kb. The Celerons of this era are either 500 or 600 MHz slower at the same clock speed as the P4s.

computergeek92 wrote:

I'd say the 1.8Ghz Pentium 4. It has the most cache being 256kb. The Celerons of this era are either 500 or 600 MHz slower at the same clock speed as the P4s.

Actually I'd say the 1.6A is better. It has 512KB cache.

shamino wrote:

Depending on the stepping, there's a good chance your 1.6A might be highly overclockable. That's a a very low clocked Northwood core, so it's likely it was clocked far below it's actual abilities. If you have any interest to overclock anything, I'd experiment with that one.

1.6A is a B0 stepping ,according to CPU-World. Same like my 1.8A.

Stay away from the Celeron. They are utter shit. I used to have the exact one, that's why I know. 128KB was just too low for NetBurst and it really hits the performance hard. The rectified themselves when they released the Celeron D which was miles better!

Super Pi scores of a few Netburst's I've had in my machine recently, system is a Gigabyte P4 Titan 533(intel 845E chipset), 1 gig DDR @ 266mhz.
As you can see from the numbers the Celly's are quite a bit slower, northwood 1.8 P4 is 15 seconds faster than the northwood celly @ 2.2ghz.
Pentium 4 2.4B Northwood 1m 18.672s
Pentium 4 1.8A Northwood 1m 35.157s
Celeron Northwood 2.2 Ghz 1m 50,097s
Celeron 1.7 (willamette) 2m 09.172s

Games struggle even worse, Need for speed underground 2 is perfectly playable and smooth on the P4 1.8A, but a hopeless mess on either of the Celerons, the cache chop hurt the Celly really badly, netburst really does need a decent size to perform well. I'd say until the Celeron D with 256k cache and the later (and best of the bunch)Cedar Mills with 512k the celly is a no go for gaming.

Thanks for the info everyone. I opted for the two P4 chips in the end. The boards only support 400FSB CPU's.

Something interesting though: When I got the board, one of them had a Pentium 4 2.8GHz/800/512 CPU, which was obviously only running at 1.4GHz due to the FSB restriction.
That got me thinking.... Would the P4 2.8 running at only 1.4 outperform the P4 1.6A? The 1.6A has a 200MHz advantage in theory.

What do you guys think?

I obviously won't put the 2.8 P4 back in there, since it's wasted on that board, but I'm just curious.

Standard Def Steve wrote:

The Celeron would only be faster in specific workloads (media encoding, for example). When it comes to gaming though, the 128K Northwood Celeron is quite the slug. For example, in 3DMark01 the Celeron 2400 only hits 9302 with a Radeon 9800 Pro. P4 1.8 Willamette gets 9178, PIII-S at 1.59 scores 11544, and P4 2.4A/400 hits 12125. The tiny cache hurts gaming performance in a big way.

Out of the three, my choice would be the 1.6A, then overclock it to at least 2.13/533. Those early Northwoods ran cool and were excellent overclockers

He's using the 1.6A on one of two boards, but the more recent question was between a Willamette P4 1.7GHz and a Celeron 2.4GHz. A Willamette 1.8GHz was also mentioned originally, so I'm not sure what happened to that option.
Your numbers show that in 3DMark, the Celeron 2.4 is slightly faster than a Willamette 1.8, and of course there'd be slightly more advantage against a 1.7. If this is a worst case scenario for Celeron, then I think it wins.

BSA Starfire wrote:

Super Pi scores of a few Netburst's I've had in my machine recently, system is a Gigabyte P4 Titan 533(intel 845E chipset), 1 gi […]
Show full quote

Super Pi scores of a few Netburst's I've had in my machine recently, system is a Gigabyte P4 Titan 533(intel 845E chipset), 1 gig DDR @ 266mhz.
As you can see from the numbers the Celly's are quite a bit slower, northwood 1.8 P4 is 15 seconds faster than the northwood celly @ 2.2ghz.
Pentium 4 2.4B Northwood 1m 18.672s
Pentium 4 1.8A Northwood 1m 35.157s
Celeron Northwood 2.2 Ghz 1m 50,097s
Celeron 1.7 (willamette) 2m 09.172s

Subtract 200MHz from the P4 Northwood 1.8A to get his P4 1.6A, and add 200MHz to the Celeron 2.2 to get his Celeron 2.4. At those clocks they might be close. So his Celeron 2.4 might even be close to his Northwood 1.6A, given that he can't overclock it.
Against a Willamette P4, I'd take the Celeron 2.4. Willamette and Celeron are both dogs, but it looks like the Celeron 2.4 comes out a bit better between them.

Totempole wrote:

Thanks for the info everyone. I opted for the two P4 chips in the end. The boards only support 400FSB CPU's. Something interesti […]
Show full quote

Thanks for the info everyone. I opted for the two P4 chips in the end. The boards only support 400FSB CPU's.
Something interesting though: When I got the board, one of them had a Pentium 4 2.8GHz/800/512 CPU, which was obviously only running at 1.4GHz due to the FSB restriction.
That got me thinking.... Would the P4 2.8 running at only 1.4 outperform the P4 1.6A? The 1.6A has a 200MHz advantage in theory.

What do you guys think?

I obviously won't put the 2.8 P4 back in there, since it's wasted on that board, but I'm just curious.

Those are both Northwoods, though the 2.8 might be a later stepping. I don't know of any reason that it would be faster, I think it would just be a direct clock speed comparison, so 1.6GHz would be faster than 1.4. But I could be wrong if there's anything different about the later stepping that makes it more efficient in some way.

shamino wrote:

He's using the 1.6A on one of two boards, but the more recent question was between a Willamette P4 1.7GHz and a Celeron 2.4GHz. […]
Show full quote

Standard Def Steve wrote:

The Celeron would only be faster in specific workloads (media encoding, for example). When it comes to gaming though, the 128K Northwood Celeron is quite the slug. For example, in 3DMark01 the Celeron 2400 only hits 9302 with a Radeon 9800 Pro. P4 1.8 Willamette gets 9178, PIII-S at 1.59 scores 11544, and P4 2.4A/400 hits 12125. The tiny cache hurts gaming performance in a big way.

Out of the three, my choice would be the 1.6A, then overclock it to at least 2.13/533. Those early Northwoods ran cool and were excellent overclockers

He's using the 1.6A on one of two boards, but the more recent question was between a Willamette P4 1.7GHz and a Celeron 2.4GHz. A Willamette 1.8GHz was also mentioned originally, so I'm not sure what happened to that option.
Your numbers show that in 3DMark, the Celeron 2.4 is slightly faster than a Willamette 1.8, and of course there'd be slightly more advantage against a 1.7. If this is a worst case scenario for Celeron, then I think it wins.

BSA Starfire wrote:

Super Pi scores of a few Netburst's I've had in my machine recently, system is a Gigabyte P4 Titan 533(intel 845E chipset), 1 gi […]
Show full quote

Super Pi scores of a few Netburst's I've had in my machine recently, system is a Gigabyte P4 Titan 533(intel 845E chipset), 1 gig DDR @ 266mhz.
As you can see from the numbers the Celly's are quite a bit slower, northwood 1.8 P4 is 15 seconds faster than the northwood celly @ 2.2ghz.
Pentium 4 2.4B Northwood 1m 18.672s
Pentium 4 1.8A Northwood 1m 35.157s
Celeron Northwood 2.2 Ghz 1m 50,097s
Celeron 1.7 (willamette) 2m 09.172s

Subtract 200MHz from the P4 Northwood 1.8A to get his P4 1.6A, and add 200MHz to the Celeron 2.2 to get his Celeron 2.4. At those clocks they might be close. So his Celeron 2.4 might even be close to his Northwood 1.6A, given that he can't overclock it.
Against a Willamette P4, I'd take the Celeron 2.4. Willamette and Celeron are both dogs, but it looks like the Celeron 2.4 comes out a bit better between them.

Totempole wrote:

Thanks for the info everyone. I opted for the two P4 chips in the end. The boards only support 400FSB CPU's. Something interesti […]
Show full quote

Thanks for the info everyone. I opted for the two P4 chips in the end. The boards only support 400FSB CPU's.
Something interesting though: When I got the board, one of them had a Pentium 4 2.8GHz/800/512 CPU, which was obviously only running at 1.4GHz due to the FSB restriction.
That got me thinking.... Would the P4 2.8 running at only 1.4 outperform the P4 1.6A? The 1.6A has a 200MHz advantage in theory.

What do you guys think?

I obviously won't put the 2.8 P4 back in there, since it's wasted on that board, but I'm just curious.

Those are both Northwoods, though the 2.8 might be a later stepping. I don't know of any reason that it would be faster, I think it would just be a direct clock speed comparison, so 1.6GHz would be faster than 1.4. But I could be wrong if there's anything different about the later stepping that makes it more efficient in some way.

His 2.4A is either a C0 or E0,depending on the SL number. Same for the 2.8,except it's D1 or M0 instead of C0 and E0.

A list for him to identify what stepping he has:
400FSB 2.4 Northwood

B0 - SL67Z, SL684, SL6D7, SL6EU)
C1 - QMU6, SL6DV, SL6EF, SL6RZ, SL6SH
D1 - SL6PC, SL6Q8

800 FSB 2.8 Northwood:

D1 - QWA5, QWP6, SL6WJ, SL6WT, SL78Y
M0 - SL6Z5.

Here's a cool Celeron D review from Anandtech. It features a great comparison with northwood celerons. Seriously, 128KB was just too small for Netburst. Had Intel made it 192KB instead of 128KB, I think this part would have been a much more serious contender. Here's another in-depth review of the Celeron D from XbitLabs!

F2bnp wrote:

Here's a cool Celeron D review from Anandtech. It features a great comparison with northwood celerons. Seriously, 128KB was just too small for Netburst. Had Intel made it 192KB instead of 128KB, I think this part would have been a much more serious contender. Here's another in-depth review of the Celeron D from XbitLabs!

Here's the problem with Celeron D CPUs though : they run too damn hot.