Also check the memory timings. Some BIOS's don't support changing them, so check with CPU-Z what they actually are.

Clock and timings were set to 133Mhz and 3-2-3 on both boards. Memory modules are from Infineon.

AGP is set to 4X on both boards (Intel board doesn't have bios settings for it but HWinfo and Aida64 both report AGP 4X). Could it be AGP Fast Writes? The VIA board has Fast Write enabled by default (witch I keep having to disable because it makes older video cards BSOD), and the Intel board does not support fast writes.

In this case, Fast Writes is enabled on the Apollo because the GF4 Ti supports it and is stable. I benched with it on to make sure the board is at it's full potential.

Something is definitely wrong. There's no way a cheap ECS via board should beat a more expensive intel based Epox.

Another annoying set of issues - the Intel board will sometimes refuse to POST with ANY lan card installed (tried Intel, SMC and Realtek). On the other hand, the VIA board is very picky about sound cards. My Live! 5.1 crackles and chirps in games, and so does my CMI8738. The only card it likes is an old CT4810.

Interesting. I was planning on doing something similar, but you beat me to it 😊

But when I get around to it, we can compare our findings.

kanecvr wrote:

What's the deal? I used fresh Win98SE on both machines and the same nvidia drivers... same HDD type, same soundcard...

Did you install chipset drivers? This might explain the performance difference.

So I reinstalled the chipset drivers - latest ones from intel (had previously installed the ones from the mb's cd) - - 5568 pt - no difference. Phil, I'll be waiting on your results. Maybe I have a bad motherboad (performance-wise).

I only hated the Via Apollo Pro133 on my old system because it always forced any new gpu to run at agp 2x and not 4x 🙁 (like the nvidia geforce fx5500 and the geforce 6200) the only one that runned fine was the ati radeon 7200 that could run at agp 4x

You had a weird board I guess - or immature drivers / firmware. My ECS runs every AGP 4x board I threw at it at 4x. No issues whatsoever (apart from the annoying PCI latency thing and poor quality on board audio).

Do you have some low-level benchmarks for CPU/RAM performance?

Well, I remember I wasn't the only one to have this problem, on the time many foruns related that the chipset could only run at agp 2x with new graphics card.

https://www.wimsbios.com/forum/video-vga-and- … blem-t3188.html

https://forum-en.msi.com/index.php?topic=23277.0

https://forum-en.msi.com/index.php?topic=9773.0

https://forum-en.msi.com/index.php?topic=20060.0

https://forum-en.msi.com/index.php?topic=47048.0
...

Plus the motherboard was a MS-6309, I could force it to run at agp4x but I would get glitches on the screen or it would end into a BSOD 🙁
The only three cards I tested were an ATI radeon 7200 (it worked at agp4x), FX5500 (agp 2x only) geforce 6200 (agp 2x and this gpu was unstable with agp2x). After that annoy problem I just gave up from the board, I updated any drivers and bios that I could, tested many gpu drivers and os from 98 to windows 7, but none of them fixed the problem 🙁 (At least on my board, the onboard sound was actually good)

EDIT: I still remember completing Half Life 2 and episode one on this board *_* (pIII 1GHz, 384mb sdram, ati radeon 7200 ;D) it actually worked fine in most of times (oh that slow loading time)

Yeah, I encountered this problem as well on my ASUS CUVX. Nvidia cards will not run at AGP 4X. Worse off, Aida64 reports the chipset is only capable of AGP 2x when an nvidia card is installed, and forcing 4x from bios does nothing. My 8500 and 9000 work fine at 4x on it. Still, with the GF4 Ti 4200 it scores 5800-5900pts - 500 more then my intel board.

After doing some tests, I noticed that the FPU scores are a bit lower on the intel i815 - ~200pts in FPU Julia for the intel and ~220 for the VIA.

What chipset driver did you use for the 815?

Can you try the 2001 driver from here: http://www.philscomputerlab.com/intel.html

And what driver did you use for VIA?

I would go with via_4in1_443v.zip from here: http://www.philscomputerlab.com/via.html

Chipset software have "target chipsets", so it's sometimes important to go with older, more period correct, drivers for best performance.

Eager to see if it makes any difference.

I used the latest Intel Chipset driver for the 815 off their website and VIA 4in1 443 for the test. I have a 3rd s370 board, an Asus CUV-X (also via based). It scores 5790 with the 933MHz PIII / same GPU / RAM / HDD as the ECS. It's probably something the ECS board has that gives it an advantage over the others...

Also the EPOX board had bad capacitor rot... check these out:

The ones near the CPU socket were not bloated so I didn't suspect them - until I noticed they leaked underneath... The two 1000uf caps under the AGP slot were also swollen, one 1500uf 6.3v cap near the RAM was poorly replaced since it was loose, and the same person replaced some of the brown caps near the cpu socket with Rubicons. They did a piss-poor job at it too - contact points were not heated enough and the caps were a little loose. I tore out the brown noname caps and replaced them with low-esr panasonic caps, same for the rest.

Not as bad as the ECS had when I got it (all cpu caps were blown, so were the ones under the AGP port - replaced a total of 12 capacitors!) - but still.

I'll retest with the Intel drivers you posted and attach screenshots of benches all 3 motherboards (ASUS CUV-X / ECS P6VXA / Epox EP-3S1A) and I'll also test them with something more period correct like a GF 256 or Geforce 3 Ti (don't own ANY working GF2 cards would you belive it?)

I'm still putting together a video about my project, but I'd like to tell you now, because I know how frustrating it can be if something is bugging you and you can't find out why.

I can confirm your findings. In my tests, the Intel platform performs faster under DOS (VGA and high resolution, only Doom is faster on VIA), but under Windows VIA is faster in every single benchmark. Storage performance is slightly better on Intel, with writes being faster standing out.

I used identical memory, hard-drive, processor, optical drive. Quadro4 980 XGL on both machines, and CPU-Z told me they both running in AGP 4x.

Also, how the heck does one pronounce VIA? 🤣 I pronounced it like veeeeya

The Epox is a hell of a motherboard, it has its own VDIM VRM, the EliteGroup's an ordinary board. But in 2015, i wont think much about performance, but RAM size , 1.5gb vs 5 12mb..., Apollo 133a was a great chip with bad reviews..., another story will be if you had an Epox BX7+100 !!

philscomputerlab wrote:

I'm still putting together a video about my project, but I'd like to tell you now, because I know how frustrating it can be if s […]
Show full quote

I'm still putting together a video about my project, but I'd like to tell you now, because I know how frustrating it can be if something is bugging you and you can't find out why.

I can confirm your findings. In my tests, the Intel platform performs faster under DOS (VGA and high resolution, only Doom is faster on VIA), but under Windows VIA is faster in every single benchmark. Storage performance is slightly better on Intel, with writes being faster standing out.

I used identical memory, hard-drive, processor, optical drive. Quadro4 980 XGL on both machines, and CPU-Z told me they both running in AGP 4x.

Also, how the heck does one pronounce VIA? 🤣 I pronounced it like veeeeya

Weird right? Up until now I've always believed that the Intel chipsets are faster - when benchmarks proved otherwise I was shocked (still am). I've had a thing for VIA chipsets since I was little (my first PC was via based) but I've never tough of them as great performers. I did reach the conclusion that they are all-round the most balanced chipsets for most platforms.

I've tested VIA chips on several platforms - ALi Aladdin V vs MVP3 - MVP3's AGP performance is marginally better and it's less buggy then the ALi chipset witch gives me headaches constantly / i815 vs VIA Apollo Pro 133 - VIA is noticeably faster and there are no stability issues on either platforms - both rock solid / VIA KT880 vs nForce2 - AGP performance slightly favors the nForce but the KT880 is more stable, takes any ram you trow at it, doesn't act wierd with some ATi cards like the nForce, and works great in win98 - the KT880 is the mature socket A platform.

I'll try to get a hold of a socket 478 platform via board and pit it against the mighty i865 witch I love. There has to be a reason VIA is considered a mediocre chip.

Oh and I pronounce it like an acronym - V.I.A. 😁

RDRAM wrote:

The Epox is a hell of a motherboard, it has its own VDIM VRM, the EliteGroup's an ordinary board. But in 2015, i wont think much about performance, but RAM size , 1.5gb vs 5 12mb..., Apollo 133a was a great chip with bad reviews..., another story will be if you had an Epox BX7+100 !!

Well the BX7+ is 440BX based and only does AGP2x so I think the Apollo would eat it alive... I have a Soyo 440BX based socket 370 board, and It doesn't support CPUs faster than 750MHz...

kanecvr wrote:

Weird right? Up until now I've always believed that the Intel chipsets are faster - when benchmarks proved otherwise I was shock […]
Show full quote

Weird right? Up until now I've always believed that the Intel chipsets are faster - when benchmarks proved otherwise I was shocked (still am). I've had a thing for VIA chipsets since I was little (my first PC was via based) but I've never tough of them as great performers. I did reach the conclusion that they are all-round the most balanced chipsets for most platforms.

I've tested VIA chips on several platforms - ALi Aladdin V vs MVP3 - MVP3's AGP performance is marginally better and it's less buggy then the ALi chipset witch gives me headaches constantly / i815 vs VIA Apollo Pro 133 - VIA is noticeably faster and there are no stability issues on either platforms - both rock solid / VIA KT880 vs nForce2 - AGP performance slightly favors the nForce but the KT880 is more stable, takes any ram you trow at it, doesn't act wierd with some ATi cards like the nForce, and works great in win98 - the KT880 is the mature socket A platform.

I'll try to get a hold of a socket 478 platform via board and pit it against the mighty i865 witch I love. There has to be a reason VIA is considered a mediocre chip.

...

Well the BX7+ is 440BX based and only does AGP2x so I think the Apollo would eat it alive... I have a Soyo 440BX based socket 370 board, and It doesn't support CPUs faster than 750MHz...

Makes me happy that my GA-6VTXE is VIA based. Anyway there are no 440BX boartds with official 133MHz or Tualatin support and nearly no 815 boards with ISA.
I think the bad image might stem from AGP instabilities (if you don't have the correct AGP drivers installed) and the SBLive troubles.
Also, isn't 440BX supposed to faster than 815 with FSB100? (At least RAM and PCI performance, not regarding AGP and IDE.) 440BX should be faster than VIA 694(T) at FSB100.

GA-6VTXE is what I used as well 😀

It was in a bad state when I got it. Last year I replaced the caps with top of the range panasonic caps. Works like a charm now 😀

I did not have a single hang / blue screen / lockup during the entire benchmarking process.

Because I don't have any decent coolers for the Tualatin, I use fairly small Socket 7 coolers, and then just place a 120 mm fan on top. That got me around 45c at idle and worked well.

440BX can run at tighter timings. Pretty much any PC133 SDRAM can do CL2 at 100 MHz. That gives it a nice boost. However I did look into this comparison (440BX vs 815) in my Voodoo 2 scaling project and found it to be only a small lead for the 440BX. However, I believe I used CL3 memory timing, so more speed can be had.

You can read about this in this paper on page 35: http://www.philscomputerlab.com/uploads/3/7/2 … ing_project.pdf

But because still have the two Tualatin systems on my desk, I might revisit this, and test both with the tightest memory timings and compare them.

AFAIK VIA chipsets only got left behind on LGA775. PCIe compatibility issues (some PCIe 2.0 graphics cards make it bluescreen every now and then), SATAII compatibility issues (need to drop HDDs down to SATAI through firmware or jumpers), limited to 1066MHz and DDR2667, finicky overclockability, many of the chipsets do not support Dual-Channel and so on. Still, they work fine once you deal with the compatibility issues.

kanecvr wrote:

ALi Aladdin V vs MVP3 - MVP3's AGP performance is marginally better and it's less buggy then the ALi chipset witch gives me headaches constantly

For me is the other way around,ali aladdin V is the chipset i prefer,the mvp3 on 2 different boards as soon as i enable dma in control panel in win98se,the system starts randomly locking up,sometimes even before it boots into windows,i tried different drivers no luck.
The ali mb has absolutly no problems of this sort,plus it has support for udma 66 compared with the udma 33 on the mvp3.