Interesting thread swaaye, I will be reading this one regularly as I actually like the VIA chipsets you've mentioned.
Can't compare it to contemporary chipsets as I never build any. The second oldest non-VIA systems I have working here is probably one of my P3's.

I never noticed it's 2D GUI performance to be slow, but otoh I was never really paying attention to it either.

VIA does seem to come with a price no matter what it's vintage is, but I still like it for it's passive chipsets.

I actually miss VIA since they abandoned the desktop chipset market, though I have to say the AMD chipsets are at least as good

This may not apply because it's info based on old hardware, but I've certainly seen "general weirdness" such as odd slowdowns when using various different VIA chipset drivers on mobos circa '99-'01. I remember their "ATAPI fix" would slow the mouse movement and general desktop experience down, when using a particular version on a particular board.

If it's possible to successfully uninstall your current chipset driver and try other versions, I would give that a go. ATM, for my testing I'm using "slow" CF drives, but it's handy because I can back them up fairly quickly on my main machine. If any driver installation makes a mess of things, it takes less than 5 minutes to roll back to a previous good back up point. Useful for experiments such as this, where you can try various versions of things quite quickly.

I remember this article on tomshardware. It investigated 2D performance of recent ATI/Nvidia cards and drivers. They found that manufacturers started to not worry about 2D performance and it showed. From what I remember, Nvidia faired much better...

But using the desktop such as opening and closing a large window is important and also noticeable. I wonder if this is not strictly related to a problem with the video card.

Good point. Could be a software / driver issue as well.

I reinstalled XP and tried my 6600GT with the stock XP VIA AGP GART driver. No improvement with that. I ran across a suggestion online to set the VIA VLINK speed to 4x instead of 8x (BIOS setting) and this did make a noticeable improvement but it only lasted for about a minute after boot. It slowed back down before my eyes. Very strange!

I also played around with the BIOS options like Fast Writes, AGP Calibration and some others but nothing helped there.

GeForce 6600 GT = slow
GeForce 5900 U = slow
Radeon 9600 = slow
Radeon 8500 = clearly faster.

Maybe some sort of AGP 8x issue? I'm going to try forcing AGP 4x with Catalyst Control Center. There's no BIOS option for this unfortunately.

Ok I think the problem definitely is related to AGP somehow.

The Radeon 9600 is about as fast in 2D as the Radeon 8500 when I force AGP 4x.

If I go in the BIOS and turn on the AGP read/write 1WS, AGP calibration, fast write, etc this causes 2D to be even slower.

I don't like VIA AGP but I thought it was finally quality at the end. I guess not. 😁

So I guess it's time to get a couple PCI Millenniums from Ebay for my quadcore? 😜

swaaye, are you confident that this issue has nothing to do whatever chipset driver revision you are using? Just wondering...

I tried the stock XP VIA AGP driver and the latest Hyperion pack. Same performance either way...

I found a nice little Windows GDI benchmark at Tom's Hardware. Lets just say the results are illuminating.

Don't take this as conclusive final judgment of these cards because I do think this VIA chipset is influencing the results.

Test box
DFI K8M800-MLV(F) / K8M800 Infinity
1GB PC3200
2.2 GHz Athlon 64
Intel Pro 1000 GT NIC
Audigy 4

Cards - fastest to slowest
1909 - Radeon 8500
1867 - GeForce FX 5900 Ultra
1661 - GeForce 6600 GT
1513 - GeForce 3
657 - G400 MAX

Drivers used
8500 - Catalyst 4.2
GF3 & 5900U - 56.72
6600GT - 275.33
G400MAX - 5.96.004.0

Screenshots of card results with individual test scores

The stretch result definitely makes me think the VIA AGP is up to something. See Tom's stretch scores for comparison.

Tom's Hardware article with lots more test results
http://www.tomshardware.com/reviews/2d-windows-gdi,2547.html

Tom2D Download
http://www.tomshardware.de/download/Tom2D,0301-26150.html

Dug out the Shuttle AN35N nForce 2 Ultra board that I have.

Athlon XP 3200+
nForce 2 Ultra 400
1GB PC3200 dual channel
Intel Pro 1000GT NIC
Promise SATA150 TX2
Audigy 4

GeForce 3 (driver 56.72)

K8M800

nForce 2 Ultra

It does "seem" VIA has a part to play. But like I said, as I've always used VIA post-P3 I never seen the difference with my own eyes.

swaaye wrote:

Yes this is definitely tangibly faster. VIA sucks. 😉

Lol!
But otoh the VIA chipsets at least didn't need active cooling as they didn't run as hot as NF. Btw, I have no idea where SiS would end up within this comparison.

The difference between 2425 and 1513 is so large, that I don't think it's just a hardware issue. If I had that hardware to hand, I would experiment with various VIA 4-in-1 driver revisions, and when doing so, I would only install the bare bones AGP segment, and skip the other features. I would also set up both mobos with nothing else in them, other than the GF3 card. I would be surprised if you are unable to close that curiously large performance gap.

Heh yeah maybe somewhere in their vault of ~18 Hyperion / 4-in-1 packages (post 2004) lies a VIA GART that is amazing. We certainly can't rely on VIA to put out quality software after 5 years of revisions.

But as I said I tried the stock XP VIA AGP GART and the latest Hyperion pack and saw the same results. There's about a 6 year difference in age between those drivers.

It could also be something with this individual board. But the thing is I recently setup a KT880 Athlon XP for someone and I remember thinking that 2D felt strangely sluggish. And there are also other complaints on the web to corroborate this...

Tetrium wrote:

But otoh the VIA chipsets at least didn't need active cooling as they didn't run as hot as NF. Btw, I have no idea where SiS would end up within this comparison.

The annoying tiny active coolers on chipsets back then was just them going cheap and also believing that a fan is more exciting to buyers. None of the chipsets were particularly hot. My Shuttle nForce2 Ultra has a passive heatsink.

I think ASUS and Abit in particular liked those little worthless fansinks. I've worked with the Abit NF7-S (popular nForce 2 board) and the Abit KD7 (KT400) with them and both had the fans die. Both operated fine with the fans removed though.