Which cpu would be the better performer? (system specs would be the same oterwise)

[oddjob]

I'm reviving an old system and everything works well; but I have 2 cpu's to choose from and was wondering which one would be the better choice. Higher overall clock speed, or higher FSB? It's just a back-up system, and will be run at stock speed.

The cpu's are as follows:
1.) 0.13u Tualatin Celeron 1400/256/100/1.5V
2.) 0.18u Coppermine PIII 933/256/133/1.75V

[HAL9000]

Tough call really.... but if you were to consider overclocking.... the Tualatin will clock up pretty easy.

[Cricket]

Yeah, tough call...I think performance will probably be real close to one another...real close.

Cricket

[glc]

What exact motherboard and ram? That will help with the decision. If the board is overclock friendly (lockable PCI/AGP and easy FSB adjustments) and you have PC133 ram, I'd go with the Tualatin and crank it up a bit.

[oddjob]

Originally Posted by glc What exact motherboard and ram? That will help with the decision. If the board is overclock friendly (lockable PCI/AGP and easy FSB adjustments) and you have PC133 ram, I'd go with the Tualatin and crank it up a bit.

512MB PC133 (1x256, 2x128), Asus TUV4X mobo - VIA chipset w/no lockable PCI/AGP option. Thus, I have no intention to overclock.

[glc]

That clouds it even more! The Via chipset can run async - you can run the FSB at 100 and the ram at 133. If you possess both processors, why not try each one and run full benchmarks and see what you get? I'd be interested in the results. I don't think there will be a stability difference between either one with the ram at 133. I'm almost thinking that the Tualatin may win simply due to the considerably higher processor speed and core efficiency. That core was a real sleeper - it makes early P4's look sick. I think the only reason they even called it a Celeron was the 100 FSB.

[Snod Blatter]

I'm going to say the P3 would be the faster one in everyday applications, just because a VIA chipset running at 100MHz is going to be a real bottleneck. If it was an Intel chipset then it would be even harder to call...

[Tin]

Originally Posted by glc What exact motherboard and ram? That will help with the decision. If the board is overclock friendly (lockable PCI/AGP and easy FSB adjustments) and you have PC133 ram, I'd go with the Tualatin and crank it up a bit.

Did motherboards even have a lockable AGP/PCI host clock back then? Back on topic...its a tough call as far as those 2 processors go. The Celeron has the higher clock speed, while the PIII has the higher bus speed. Tough call there.

[glc]

Good point - the Via and 815 based boards had a switchover point around 124 FSB where it would change dividers from 1/3 to 1/4 and 2/3 to 1/2. Later BX boards did too, but only the PCI would switch, the AGP was fixed 2/3.

[oddjob]

Well, I did a quick benchmarking of the two cpus using PCMark2002 and these are the results:


0.18u Coppermine PIII 933/256/133/1.75V - FSB/RAM: 1/1 (133Mhz)



0.13u Tualatin Celeron 1400/256/100/1.5V - FSB/RAM: 3/4 (100/133Mhz)

[blue60007]

Interesting...that makes sense, the Coppermine has a higher FSB so the memory score is higher (more bandwidth) while the Tualatin has a higher speed...so more on the CPU score.

Thanks for the update

[glc]

Actually it looks like the ram is running at 133 in both situations........that's very interesting. I'm guessing it's the pisspoor way Via handles the memory running asynchronous. Try running the Tualatin again with the FSB/Ram at 1/1.

[oddjob]

FSB/RAM: 1/1 (100Mhz)

Not much difference:
CPU Score: 3267
Memory score: 1324
HDD score: 195

Out of curiosity, I ran it again at 100/133 immediately afterwards:
CPU Score: 3269
Memory score: 1361
HDD score: 196

I ran it 1/1 and tried oc'ing it; scores were better, but of course it was also ramping up the pci bus accordingly.

@ 1.596GHz (114x14)
CPU Score: 3804
Memory score: 1591
HDD score: 191

115x14 made no difference, 116x14 was unstable and froze almost immediately.

[glc]

Now try overclocking to a speed where the PCI/AGP shift dividers. If you use the DIP switches, try 124/31. The bios may offer finer steps than the switches, the switches jump from 120/40 to 124/31.

Then again - if it's stable at 114, leave it - that's totally blowing the 933 away. Most PCI devices can handle 37 or 38 fairly well.