PcBytes wrote on 2022-06-02, 14:15:

PARKE wrote on 2022-06-02, 13:37:

It was just meant as illustration for flaws in early boards, not as a 1 on 1 comparison. What type of frequency generator is your Soyo accomodated with ?

Cypress W40S01-04 clockgen, ITE 8671 SuperIO.

According to the datasheet that is a ram buffer chip with 48 legs.
On the online photos of the board the frequency generator has 28 legs - located on the right of the upper pci sot.

My bad.

The clockgen is still a Cypress though, model W196G.

Did some quick tests just to compare cpus.

Ran 3DMark2001 and Quake III.
My old 850MHz 100fsb P3 got the results, 6280 and 74.5fps.
Then I ran the tualatin 1133MHz 133fsb at 100fsb so it was clocked just under 850MHz, the result where 6927 and 90fps.

Thats quite the increase!
Thought that was knterestind, so the cache does matter atleast for 3D mark and Quake.

Nice!
Found a more involved piece of info on comparison between Tualatin with 256 and 512 cache here:
https://www.overclockers.com/a-case-study-on- … pu-performance/
There is a difference between Coppermine and Tualatin cache management that -may- explain (a small) part of the performance increase you notice - it is called data prefetch, you can google it.
Vogons member pshipkov found a difference between Coppermine vs Tualatin in Quake here:
Re: 3 (+3 more) retro battle stations

Got my 1400MHz P3 tualatin so I ran 3DMark2001 and Quake III with it aswell.
Here are my results (C=Coppermine, T=Tualatin, P3=Pentium III, Ce=Celeron):

1CPU                      3DMark2001   Quake III
2C P3  850/100@100        6280          74
3T P3 1133/133@100= 850   6927          90
4T P3 1400/133@100=1050   7654         101.5
5T P3 1400/133@112=1150   8339         115
6T Ce 1200/100@100        7477          94.5
7T Ce 1400/100@100        7922         100.8

I would say just from these two tests that the cache seems to be worth it, especially with Quake III.
If I can only overclock it a little bit it would outperform the Celeron 1400 in 3DMark2001.
Also it performs better then the Celeron 1200 even when its underclocked, if I can get a stable OC then I'll have the best CPU I could on this setup.
What would be the prefered stability test to run if/when I try to overclock it?

EDIT
Updated the list with a small OC, from 100fsb to 112fsb, seems to be running stable.
And yeah, it outperforms the Celeron at this point.

My setup is;
ABIT BX6 rev 2
384MB PC133 RAM
Geforce4 TI 4280
WinME
2 ISA sound cards
PCI sound card
network card.
HDD is a sata 120GB SSD with a generic adapter.

Did some more OC, the furthest I can push this CPU seems to be 124fsb before I get disk read errors.
So thats where I left it, 1300MHz 124fsb.
Can post exact scores tomorrow but it was closer to 9000 in 3DMark01 if I remember correctly.

I tested stability by first running Prime95 for ~8h, then CPU Burn In i think it was called for ~8h.
No problems so I played Morrowind at stable 30fps highest rendering distance for a couple of hours, all seemed good! Before Morrowind played at a quarter of rendering distance at ~18fps so thats a great improvement in this game.
Im happy!

Here are the score for the Tualatin 1.4;

1CPU                      3DMark2001   Quake III
2T P3 1400/133@100=1050   7654         101.5
3T P3 1400/133@112=1150   8339         115
4T P3 1400/133@117=1200   8573         118
5T P3 1400/133@124=1300   8927         124.2

Interesting.
How far can you overclock the Celeron 1400 ?

Here are the Celeron 1.4GHz results:

1CPU                      3DMark2001   Quake III
2T Ce 1400/100@100        7922         100.8
3T Ce 1400/100@112=1570   8564         113.5
4T Ce 1400/100@117=1640   8834         115.5   Note: Was unstable at default voltages so I needed to bump it up by 0.5V EDIT; after playing games it still was unstable

That was as far as it went.
At 117fsb it was unstable at first, got a bunch of crashes and read errors but after increasing the voltage by 0.5V that went away.
When I tried to OC it further all I got was disk errors at startup, raising the voltage more didn't change that.
So even if the 1.4GHz Celeron is overclocked and the P3 Tualatin is underclocked it outperforms it.
If the Celeron would be stable at 124fsb I guess it would beat the P3 Tualatin but at least in this system that wasn't possible.

BTW, the CPUs are the models:
Celeron = SL64V
P3 =SL5XL

So my conclusion is that I'll be using the P3 in my system.

EDIT
I did these test and checked stability with Prime95 and CPU Burn In for 8h each but the Celeron at 117fsb with bumped up voltages that passed these test still became unstable when I was playing some Icewind Dale II today.
Was playing for about 2h and then I started to get dll errors, same errors I have seen with other failed OCs on this system, and then the system froze.

Well , you gathered useful information for future explorers. As already noted earlier in this thread it seems that the ABIT BX6 rev 2 motherboard is likely the weak link in this particular setup. If you ever get a more 'advanced' BX board you may succeed in running the 1.4Ghz Tualatin in its full glory.