Standard Def Steve wrote:

This is off topic, since the OP already chose a Radeon 9800 (excellent choice, BTW). One thing I've noticed with VIA boards and NVIDIA cards is the dreadfully poor 2D GUI performance. All three of the VIA boards I tried (694X, Apollo Pro 266, and even a newer C2D-compatible board) had very poor GUI perfomance with a 6800GT and 7800GS installed. This only affected WinXP; Win7 and its desktop composition seemed to be unaffected.

I ran tests for this years ago. You're right, VIA AGP (or something about their chipsets) affects GUI performance. It's tangible and measurable. I don't believe it's limited to NVIDIA cards.

VIA vs. 2D speed

swaaye wrote:

I found a nice little Windows GDI benchmark at Tom's Hardware. Lets just say the results are illuminating. […]
Show full quote

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

swaaye wrote:

Dug out the Shuttle AN35N nForce 2 Ultra board that I have. […]
Show full quote

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

I suppose that Windows 9x, since lacks most GUI effects, is not affected by this issue in any significant way, isn't it?

Win9x stops scaling with graphics card performance up to a point anyway. Too much is done on the CPU, I think.

swaaye wrote:

I ran tests for this years ago. You're right, VIA AGP (or something about their chipsets) affects GUI performance. It's tangible and measurable. I don't believe it's limited to NVIDIA cards.

What's interesting about NVIDIA on VIA is that 2D performance starts off nice and fast, but after 5-10 minutes of use, it starts to slow down. Eventually, it reaches the point where certain GUI operations look as if 2D acceleration is disabled (it's not; window moves are still accelerated). On my 7800GS card, the fan always kicks into high gear at this point.

I haven't seen this behavior with any of my ATI cards--not even with the newer x800 and x1950. I use a 9800 Pro in my 694X-based Tualatin rig and its 2D GUI performance is consistently lightning quick.

Standard Def Steve wrote:

swaaye wrote:

I ran tests for this years ago. You're right, VIA AGP (or something about their chipsets) affects GUI performance. It's tangible and measurable. I don't believe it's limited to NVIDIA cards.

What's interesting about NVIDIA on VIA is that 2D performance starts off nice and fast, but after 5-10 minutes of use, it starts to slow down. Eventually, it reaches the point where certain GUI operations look as if 2D acceleration is disabled (it's not; window moves are still accelerated). On my 7800GS card, the fan always kicks into high gear at this point.

I haven't seen this behavior with any of my ATI cards--not even with the newer x800 and x1950. I use a 9800 Pro in my 694X-based Tualatin rig and its 2D GUI performance is consistently lightning quick.

Mmm... have you observed this over all VIA chipsets? I mean, not of course every single one VIA chipsets but K6-2/3 and PI/III chipsets, without discerning AMD or Intel based platform.

I don't have any experience with VIA SS7 chipsets, but I can say that GUI performance absolutely SUCKS when an NVIDIA card is used with the following VIA chipsets:

Apollo Pro 133/694X (PIII/PIII-S)
Apollo Pro 266 (PIII with DDR support)
PT880 Ultra (P4 and Core 2)

The two PIII chipsets also had trouble running games with the 6800GT and 7800GS (missing textures, artifacts, crashing back to the desktop). FX5200 didn't have any trouble with games, but its GUI performance was very poor.

Bottom line: if you have a VIA-based AGP motherboard--even an LGA775--do yourself a favour and get an ATI graphics card.

I wouldn't go all universal with NV+VIA=TERRIBLE. I have a KT333 board that's ok with NVIDIA cards.

Think about it-
-VIA was a dominant chipset player in 1997-2004
-VIA was probably the #1 AMD platform from 1999-2004
-NVIDIA was dominant through most of those years. ATi has never been ahead of NVIDIA market share AFAIK.
-People typically buy NVIDIA because of driver reputation and positive past experience

Granted there is certainly something wrong with VIA AGP. As you go back in time to early AGP 4x and 2x, VIA AGP becomes extremely buggy. But not every board sees that degeneration of GUI speed nor general instability.

Actually since you single out 6800 and 7800, it seems possible that those cards are specifically problematic. My KT333 + 6800 is picky about which NVIDIA driver is used. 61.76 BSODs on boot. 77.72 crashes games within minutes. 81.98 seems fine.

On the other hand, some VIA KT266/KT333 boards have problems with ATI Radeon 8500. That includes strange, unpredictable 3D problems. I enjoyed that first hand in 2001. Some boards received BIOS updates to address this.

You know, I wonder if VIA's AMD chipsets have a (slightly) better AGP implementation than their Intel offerings. I mean, even PT880 Ultra, one of the last AGPsets VIA made for Intel processors, is slow when paired with an NVIDIA GPU. Yet, I remember reading very positive reviews for VIA's K8T800 Athlon 64 chipset back in the day, and I'm sure that at least a few FX5950s made it into those machines. 😀

The only semi-decent VIA-based AMD board I ever got my hands on was an Asus A7V8X (KT400) a few years ago. I ran my 7800GS on that board and didn't notice any real GUI speed deficiency. Looking back though, I don't believe I ever ran WinXP on that machine--I used it mainly as a Win7 RC evaluation box. The compositing window manager Win7 uses seems to mask a system's poor 2D performance.

The GF3 results seem to show that small operations are slower while transfer intense operations as bitblt are actually faster with via.
So are there some latency issues present? Do both AGP setups run with the same features? (SBA, FW)
Don't look at the final score of Tom2D, the single results are much more interesting.

From my experience VIA can be tuned to performance but it requires manual adjustments. For every chipset a specific 4in1 driver version is optimal, nvidia driver installation may drop to agp 1x speed, where you require to reinstall via drivers etc.

Although it's currently on the testbed, and although I haven't spent many hours testing it, for the moment my GeForce3 Ti 200 is working fine with Windows 98 in a DFI K6VX3+/66 with a VIA MVP3 chipset. I've only test Unreal Torunament so far, which was the game I had more at hand.

VIA may have been the #1 AMD platform, but that's certainly not saying much about AMD platforms. It it weren't for VIA, I could have enjoyed an AMD system instead of holding out with an i815 an Celeron 1200 for all those years. I'm glad VIA finally kicked the bucket...best thing to happen to AMD in recent memory.

Yeah there's no doubt VIA was not building a platform anywhere near the quality of an Intel setup. My experience has been that AMD platforms finally got high quality after AMD pushed SB700 out. The very last NV chipsets (with AHCI) are fine too.

elianda wrote:

The GF3 results seem to show that small operations are slower while transfer intense operations as bitblt are actually faster wi […]
Show full quote

The GF3 results seem to show that small operations are slower while transfer intense operations as bitblt are actually faster with via.
So are there some latency issues present? Do both AGP setups run with the same features? (SBA, FW)
Don't look at the final score of Tom2D, the single results are much more interesting.

From my experience VIA can be tuned to performance but it requires manual adjustments. For every chipset a specific 4in1 driver version is optimal, nvidia driver installation may drop to agp 1x speed, where you require to reinstall via drivers etc.

The scores seem practical though because I did these tests to investigate tangible GUI slowness compared to nforce2. The thought of exploring all 30 or so 4in1 and Hyperion driver releases, in addition to BIOS settings, is terrifying . 😀

I politely dissent. SiS made a decent chipset for the P3 Tualatins and nVidia chipsets for AMD were way more than just decent from the nForce2 onwards. On comparison, AMD's own chipsets for the K6 and the first Athlons were lacking in both features and performance.

On Intel both nVidia and ATI got some raisins but also interesting offerings. Personally, I consider the present monopoly that Intel has over its own platform as harmful to the user and to innovation although AMD hasn't got its own matters sorted in a different way.

Oh some nforce chipsets are loaded with nasty bugs. nforce4 is a good example. It has Ethernet, PCI and SATA problems. You can read Linux source and find out about everything NV kept quiet.

nforce2 is OK but for whatever reason NV kept pushing a IDE driver that caused endless problems with optical drives. Intel had their Application Accelerator that was trouble too though.

That's quite true. I never installed the IDE driver of nVidia chipsets all the way up to nForce3 and the nForce4 issues are in part responsible for the decision of CPU manufacturers to keep tighter control over their platform.

However, even if I think that third parties would make the PC market more diverse and competitive, it's also quite true that the focus is being put now on the mobile and tablet markets rather than desktop offerings.

elianda wrote:

From my experience VIA can be tuned to performance but it requires manual adjustments.

Completely agree with this. Properly tuned 694X is just as fast as i815, plus it supports far more memory. I'd have to say it's my favorite PIII chipset.

Anonymous Coward wrote:

I'm glad VIA finally kicked the bucket...best thing to happen to AMD in recent memory.

PCI-E also really helped AMD and the rest of the PC industry. It's one of the few technologies I can think of that actually worked properly from the get-go.

VLB: had its quirks
PCI: Problematic at first, and had issues with certain peripherals (eg sound cards) years after its introduction.
AGP: Not even worth using at first.
USB: Barely worked at all, and not much faster than the legacy technologies it was trying to replace.
PCI-E: It Just Works. Bidirectionally. 😁

I still have a few first-gen PCIe systems (i915 and nForce4 SLI), and they're just extremely reliable machines, even with a ton of newer hardware plugged into them.

I am formerly own VIA AGP motherboard (Biostar P4M800Pro-M7) with MX4000 installed, but I am never experiencing about GUI problem (I'm not sure because I am forgotten due no longer using AGP).