(reserved for future test results...)

Great post! It's funny, I just pulled out my Tualatin 1.4-S BX build last night with the intention of doing some benchmarking and now I have some results to compare to. I'm glad you used a Ti 4600 as that is the card that is in my build!

I'm going to try some overclocking on mine also, although i'm very limited in FSB selection 133/140/150. I tried 140 last night and it booted into Windows without any issue what so ever! I will contribute my results once i have them 😀

Awesome work!

Very interesting and thorough results. Thank you for your hard work.

I have a personal preference for the y-axis to start at zero so that differences aren't exaggerated, but that's a personal thing.

Nice! I can imagine how much work went into this, thank you. Impressive showing from SiS, I didnt expect that!

Numbers for the i840 chipset would be interesting: performance-wise it should be better than the i820 and 440BX, and be on par with the Apollo Pro266. Of course, its not widely available, needs adapters to run Tualatins, and only comes--I think--in dual CPU configurations, so I understand why it was not included.

Also, its very likely someone will complain about the graphs and their origin not being zero.

Thanks again for the hard work!

This is a wonderful post!

I've always been a fan of the Apollo Pro 266T. My QDI Advance 12T based on that chipset is the fastest P3 board I've tested, easily beating the other chipsets I've put it up against (440BX, i815, i820, and 694T) clock-for-clock.

But that SiS 635T! Great stock performance and proper AGP support @ 166FSB?! I never knew such a beast existed! I'll have to get my hands on a board based on that chipset...hopefully one with BIOS-based overclocking.

It's interesting to see that your stock-clocked CPU on SiS 635T is beating my overclocked CPU in 3DMark2000. And that's with a much slower video card too!

Here are some of my QDI/Apollo Pro 266T results with a 6800GT.
The CPU is running at 1575MHz/150FSB, and the RAM is running 300MHz with 2-2-2-5 timings. OS is WinXP SP3.
I'd love to redo these benchmarks on a 635T/DDR @ FSB166.

Quake III Arena demo001, 1024x768, all settings maxed:

3DMark99, default settings:

3DMark2000, default settings:

3DMark2001SE, default settings:

[humor]
*scoffs derisively*
Single cpu boards, and the title reads

Fastest Tualatin Chipset / Best Pentium III Motherboard

[/humor]

But in all seriousness, this looks awesome! That 440BX chipset is a absolute beast as is it almost always #1 or #2. A conclusion that really makes sense, when you realize 133mhz is overclocking. I feel, to be more 'fair' i'd exclude it from the results as the overclocking the pci/agp bus makes a big difference in testing.

Just a curiosity, why so much difference between the i815 vs the i820? The d815eea2 I have is looking better than I expected. It's a really great board so long as long as overclocking isn't a big worry, and it was stable as a rock for me.

Did this tests incude VIA tweaks for bank interleaving?

- AGP 1.0 connector does not allow for ANY NEWER cards with 1.5V AGP

Only most cards, not all.

Impressive benchmark!

I'm not so surprised by the SiS 635T performance. From 486 days on, SiS repeatedly made top-notch performing chipsets, let down by bad marketing and worse low-end boards - and a reputation for shitty performance from their integrated VGA solutions. Put a (non-integrated VGA) SiS chipset on a decent board and you can expect really good performance. ECS clearly made exactly the right design choices with the P6S5A(T) and K7S5A - despite being obviously cheap, flimsy boards that retailed for less than any other full-size, non-integrated boards on their platforms at the time of release, they were both also rock solid (at least, after a few embarrassing early BIOS bugs were fixed) and the fastest performers on their respective platforms at that time. Makes you wonder what would have been possible if the high-end board makers had fully embraced SiS as a high-end option afterwards.

Agree with the comments on lack of i840 though - even leaving out dual CPU options (which were also available on i440BX, i820 and Via694 - and was no reason to rule them out), its dual-channel RDRAM memory controller was unique and actually beat the hell out of an i440BX (let alone i820) in reviews from 1999/2000. So - unlike other 'server'-type chipsets that can actually be slower for 'normal' workoads (Serverworks III-HE, I'm looking at you...) this one is a serious contender for fastest-on-platform. I'd love to see it face off vs 635T and a fully OCs i440BX like you have here 😀

Bancho wrote:

Great post! It's funny, I just pulled out my Tualatin 1.4-S BX build last night with the intention of doing some benchmarking and now I have some results to compare to. I'm glad you used a Ti 4600 as that is the card that is in my build!

I'm going to try some overclocking on mine also, although i'm very limited in FSB selection 133/140/150. I tried 140 last night and it booted into Windows without any issue what so ever! I will contribute my results once i have them 😀

Good! Thank you for your involvement. I am waiting for your test results 😀 What is the brand of your BX MoBo?

fitzpatr wrote:

Awesome work!

Very interesting and thorough results. Thank you for your hard work.

I have a personal preference for the y-axis to start at zero so that differences aren't exaggerated, but that's a personal thing.

Thanks!
No problem for me to make all charts with y-axis starting from 0, but what is the point? The differences in tests would be almost unnoticeable...

slivercr wrote:

Nice! I can imagine how much work went into this, thank you. Impressive showing from SiS, I didnt expect that!

Thank you for your kind words. Indeed it took many hours to finish this project (the first part of it...) I was also shocked to see the results by SiS!

When it comes to testing the I840 I don't really mind putting the tests results in this project but I simply do not have any motherboard with that chipset... Get me one and I will do it ;P The same applies to ServerWorks ServerSet chipsets 😉

Standard Def Steve wrote:

This is a wonderful post!

I've always been a fan of the Apollo Pro 266T. My QDI Advance 12T based on that chipset is the fastest P3 board I've tested, easily beating the other chipsets I've put it up against (440BX, i815, i820, and 694T) clock-for-clock.

But that SiS 635T! Great stock performance and proper AGP support @ 166FSB?! I never knew such a beast existed! I'll have to get my hands on a board based on that chipset...hopefully one with BIOS-based overclocking.

Great job! I appreciate you involvement in the project 😉 Good results, but they are done on Windows XP 😜 In the future tests I am also planning to test some cards on Windows XP 😀

SiS 635T is really impressive... I was able to get stable 150 MHz with my Tualatin 1400MHz, but no lack so far with this CPU at 166 MHz on any motherboards... So I will have to look after some more batches of SL6BY 😉

I managed to get it working at 166 FSB with 2 of my Coppermine 1GHz processors, I mentioned about it in this thread

luckybob wrote:

[humor]
*scoffs derisively*
Single cpu boards, and the title reads

@luckybob, I knew you gonna be disappointed with the lack of dual CPUs motherboards test results here 😜

Yeah, I know that testing 440BX at 133 FSB and comparing it with other is a little bit unfair... but this is how this chipset works... and I could not do anything about it, so in the future projects there will be also test with Tualarin Celeron at 100 MHz 😉

The Serpent Rider wrote:

Did this tests incude VIA tweaks for bank interleaving?

Yes, they were very helpful on VIA 693A, did not help at all with the other two VIA chipsets...

The Serpent Rider wrote:

- AGP 1.0 connector does not allow for ANY NEWER cards with 1.5V AGP

Only most cards, not all.

I knew this part will be confusing 😜 By that statement I meant that it is not even physically possible to install AGP 1.5V Card in a AGP 3.3V Motherboard Slot 😀

dionb wrote:

Impressive benchmark!

I'm not so surprised by the SiS 635T performance.

Thanks. I really think this chipset deserves more attention today. That is why I will make a separate topic to discuss its weak and strong points. In this thread I want to focus on test results and comparisons of different chipsets instead on focusing on one only 😉

VooDooMan wrote:

The Serpent Rider wrote:

- AGP 1.0 connector does not allow for ANY NEWER cards with 1.5V AGP

Only most cards, not all.

I knew this part will be confusing 😜 By that statement I meant that it is not even physically possible to install AGP 1.5V Card in a AGP 3.3V Motherboard Slot 😀

You'd think so, but this is where universal AGP will bite you in the ass. I have had or seen, 3.3V AND 1.5V only boards with universal slots as well as 3.3V AND 1.5V only cards with universal keying. While they are not overly common (anymore, for obvious reasons) they do exist and you can come across those. NEVER rely on the physical keying.

Holy shit with the thread.... you have done a amazing job bro.

Can I ask you a question about the BE6 II? When you run the P3 1400s with stock settings including the 133MHz FSB, what is the agp frequecy? Is it 66 or 88MHz?

I have a BH6 and I fried several radeons running at 133MHz FSB and I believe this happened because the agp divider was 2/3

Does the same thing apply for the BE6-II or does it have a correct AGP 1/2 divider for 133MHz FSB?

Koltoroc wrote:

You'd think so, but this is where universal AGP will bite you in the ass. I have had or seen, 3.3V AND 1.5V only boards with universal slots as well as 3.3V AND 1.5V only cards with universal keying. While they are not overly common (anymore, for obvious reasons) they do exist and you can come across those. NEVER rely on the physical keying.

OK, I know what you mean, but I meant THIS:

and I am aware that there are instances of sawing a notch to make it work like on Asus V9999... but only few cards can me "modded" this way.

kaiser77_1982 wrote:

Holy shit with the thread.... you have done a amazing job bro.

Thanks a lot! You helped me with finding one mobo for this project so you are also a part of it 😉 Thanks for that!

Johnylyr wrote:

Can I ask you a question about the BE6 II? When you run the P3 1400s with stock settings including the 133MHz FSB, what is the agp frequecy? Is it 66 or 88MHz?

I have a BH6 and I fried several radeons running at 133MHz FSB and I believe this happened because the agp divider was 2/3

Does the same thing apply for the BE6-II or does it have a correct AGP 1/2 divider for 133MHz FSB?

ALL intel i440BX are already overclocked at 133 MHz FSB. The AGP bus speed is always 89 MHz in this case... There is no correct divider, it is a chipset limitation and you cannot do anything about it 🙁 That is why most |ATI Radeon 9*** cards will not even boot on such a system...

I'm not talking about modding, I'm talking about cards and boards with universal keying despite being fixed voltage. They were sold that way. I recently lost a "universal" AGP board, because despite having the proper slot keying and the chipset supporting 3.3V and 1.5v, the board was designed to only run with 1.5v

The SIS 300 series (300, 305, 315) are notorious for that, virtually all are universal keyed but only do 3.3v, I have had (had being the important word here) at least 2 TNT 2 cards that were fixed voltage despite being universal capable and universal keyed (others had a selector jumper) I had a geforce 2 that for some reason was 1.5v only despite it being universal capable and keyed and I have seen universally keyed geforce 5200 and 6200 despite those not working with 3.3v.

None of them were modded, they were sold that way. They were unbranded and possibly OEM models, (except for the SIS, I have never seen one with correct keying) but that is no excuse for breaking the standard.

that's... a pretty big difference on the PCI cards, I'm a little surprised and kind of feel bad about my V4 4500 running on the 440bx now 🤣

VooDooMan wrote:

Good! Thank you for your involvement. I am waiting for your test results 😀 What is the brand of your BX MoBo?

The board I am using is a DFI ITOX CB60-BX Rev C. I have a feeling it is a very late BX Board. Its Socket 370 with 3 ISA slots. It fitted my criteria perfectly.

I went for broke and Selected 150mhz in the Bios. To my surprise the system booted into windows without problem. I was able to carry out some SiSoft 99 Benchmarks and got the results. The Mem Scores look good at 150mhz! Unfortunately the system would not complete a 3Dmark 2000 pass and would freeze. Probably due to the high AGP bus speed.

I did manage to do a 3D mark pass at 140mhz FSB but never caputured the score. That was with a Ti 4600.

Nice job Bancho! Can you also make the same tests without overclocking the CPU? Maybe you can also run some tests with your Ti 4600? 3D Mark 2001 and Qauke is what i care the most 😉 I want to see the differences between my Abit and your DFI 😀 PLEASE USE THE SAME DRIVERS for the video card and chipset, ok? 😁 THANKS A LOT!!!

P.S. I'veI just noticed you have a UNIVERSAL AGP SLOT on this mobo 😜 this is interesting... suits perfectly the issues with universal keying and AGP voltages on motherboards/cards that Koltoroc is talking about 😉 DFI did the same think that ABIT did with their VA6 mobo - The AGP port look like UNIVERSAL AGP (version 2 - 3,3V/1,5V), but it is really version 1(supports only 3,3 volts)

VooDooMan wrote:

Nice job Bancho! Can you also make the same tests without overclocking the CPU? Maybe you can also run some tests with your Ti 4600? 3D Mark 2001 and Qauke is what i care the most 😉 I want to see the differences between my Abit and your DFI 😀 PLEASE USE THE SAME DRIVERS for the video card and chipset, ok? 😁 THANKS A LOT!!!

P.S. I'veI just noticed you have a UNIVERSAL AGP SLOT on this mobo 😜 this is interesting... suits perfectly the issues with universal keying and AGP voltages on motherboards/cards that Koltoroc is talking about 😉 DFI did the same think that ABIT did with their VA6 mobo - The AGP port look like UNIVERSAL AGP (version 2 - 3,3V/1,5V), but it is really version 1(supports only 3,3 volts)

Thanks. It was a quick benchmark run to capture those results 🤣 I was interested to see if the board would post at 150mhz FSB, let alone complete the benchmarks! Its interesting to see it is so close to the VIA DDR board in the Memory benchmark you did.

I will try and get the benchmarks does with the same drivers as you and post the results.