nd22 wrote on 2025-12-30, 09:43:

The scores for Abit NV8 are extraordinary!

I understand your enthusiasm, but in fairness, it is worth assuming that the AGP bus was simply a little outdated by that point.
And I also understand that your preferred Abit did not seem to produce S754 motherboards based on the VIA K8T890 chipset.
Perhaps it would be worth expanding the testing a little, using, for example, the Soltek SL-K890-754G.
Suddenly a miracle will happen, now it's New Year 😀

Abit did produce a board with VIA KT890 chipset just that is has 185 more pins than all socket 754 boards 😀 ! It will feature in the next Battle of the platforms!
As much I would like to test all possible variations of chipsets for socket 754 i can not do it because it is simply: 1 - too expensive; 2 - limited space; 3 - other obligations. I specifically limited myself to Abit only boards from the outset as I already liked them and knew them and it is a feasible goal to accomplish within a life time for what is, essentially, a hobby. So I am sorry if i disappoint you.

Next game is the legendary Half life 2, one of my top ten games. This is the original version released in 2004. The updated steam version, currently the only one available, is far more tougher on the system because of many effects such as HDR added to the game so it is best left for pci-express platfroms when one can put a better CPU and a stronger video card.
This game is very well optimized and requires a balanced system with a good match between the CPU and the GPU. We do not have this as the Athlon 64 3400 is no match for the x1950 pro so, again, we are strongly processor limited - the radeon can easily output over 100 FPS even at 1600*1200 with a better CPU.
Again hardware OC half life tool refused to work so fraps to the rescue. The test is a 60 seconds interval in the first level in front of the train station where Dr Breen is making his infamous speech. I waited 1 minute before pressing F11 and I walked around the monument; after fraps finished i changed the resolution and waited 1 minute before testing again and so on. Repeated 3 times, the last runs are the actual results that I kept. Tested with sound enabled and all details maxed out! Radeon X1950 pro is more than capable of running the original release at max details.

43. HALF LIFE 2: 1024*768

A. NFORCE3
2025-05-14 16:47:53 - hl2
Frames: 4333 - Time: 60000ms - Avg: 72.217 - Min: 51 - Max: 98

B. VIA
2025-05-24 11:57:32 - hl2
Frames: 4808 - Time: 60000ms - Avg: 80.133 - Min: 55 - Max: 113

C. ULI
2025-06-14 15:37:48 - hl2
Frames: 4568 - Time: 60000ms - Avg: 76.133 - Min: 55 - Max: 102

D. NFORCE4
2025-09-24 18:59:39 - hl2
Frames: 4903 - Time: 60000ms - Avg: 81.717 - Min: 22 - Max: 120

44. HALF LIFE 2 : 1280*1024

A. NFORCE3
2025-05-14 16:46:24 - hl2
Frames: 4269 - Time: 60000ms - Avg: 71.150 - Min: 37 - Max: 91

B. VIA
2025-05-24 11:55:59 - hl2
Frames: 4624 - Time: 60000ms - Avg: 77.067 - Min: 60 - Max: 99

C. ULI
2025-06-14 15:36:10 - hl2
Frames: 4325 - Time: 60000ms - Avg: 72.083 - Min: 58 - Max: 90

D. NFORCE4
2025-09-24 18:58:05 - hl2
Frames: 4739 - Time: 60000ms - Avg: 78.983 - Min: 59 - Max: 110

45. HALF LIFE 2: 1600*1200

A. NFORCE3
2025-05-14 16:39:59 - hl2
Frames: 4218 - Time: 60000ms - Avg: 70.300 - Min: 58 - Max: 92

B. VIA
2025-05-24 11:54:30 - hl2
Frames: 4241 - Time: 60000ms - Avg: 70.683 - Min: 53 - Max: 100

C. ULI
2025-06-14 15:34:41 - hl2
Frames: 3964 - Time: 60000ms - Avg: 66.067 - Min: 53 - Max: 86

D. NFORCE4
2025-09-24 18:56:35 - hl2
Frames: 4573 - Time: 60000ms - Avg: 76.217 - Min: 14 - Max: 106

ULI suffers a humiliation as its performance tanks as we increase the resolution; nforce3 cripples the radeon x1950 with poor AGP perfromance; VIA puts up a honorable fight but it can not stop nforce4 for taking this game with the best frame rate at every single resolution.

46. Doom 3 is next. This is an openggl game and unlike with Direct3d games and benchmarks where Abit KU8 has to use omega drivers to get a stable system this is not the case here: 10.2 drivers can be installed and benchmarked with as the system is stable and BSOD free. Despite being a more powerful card than 7800gs, the inherent weakness of ATI hardware in opengl prevent it from truly taking the lead so only a small increase in performance can be observed!

1024*768 resolution is first. AA is still disabled.

A. nforce3
B. VIA
C. ULI
D. nforce4

47. 1280*1024 resolution is next:
A. nforce3
B. VIA
C. ULI
D. nforce4

48. ... and finally 1600*1200:
A. nforce3
B. VIA
C. ULI
D. nforce4

For the first time ULI takes the lead with very good performance in this opengl game with nforce4 close behind!

59 & 50 & 51. Quake 4 is the second opengl game to be tested!
A. nforce3
B. VIA

I forgot to take the screenshot with the ULI system so here are the results saved by HardwareOC tool in html format:

1024*768:
Quake 4 Benchmark

The benchmark started at 21.12.2025 21:06:20
System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) 64 Processor 3400+
CPU speed: 2200 MHz
Sound system: Realtek AC97 Audio
VGA Information
Graphics card:
Memory: 512.0 MB
Driver version: 6.14.0010.6755 (English)
Benchmark Information
Demo: HOCdemo.demo
Quality: Ultra
Symmetric MultiProcessing (SMP): Disabled
Aspect Ratio: [4:3]
Antialiasing: None
Anisotropic filtering: 16×
Force Ambient Light: Disabled
________________________________________

Resolution: 1024×768
Score = 36 FPS

1280*1024:
Quake 4 Benchmark

The benchmark started at 21.12.2025 21:10:51
System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) 64 Processor 3400+
CPU speed: 2200 MHz
Sound system: Realtek AC97 Audio
VGA Information
Graphics card: Radeon X1950 Pro (Omega 3.8.442)
Memory: 512.0 MB
Driver version: 6.14.0010.6755 (English)
Benchmark Information
Demo: HOCdemo.demo
Quality: Ultra
Symmetric MultiProcessing (SMP): Disabled
Aspect Ratio: [4:3]
Antialiasing: None
Anisotropic filtering: 16×
Force Ambient Light: Disabled
________________________________________

Resolution: 1280×1024
Score = 32 FPS

1600*1200:
Quake 4 Benchmark

The benchmark started at 21.12.2025 18:31:17
System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) 64 Processor 3400+
CPU speed: 2200 MHz
Sound system: Realtek AC97 Audio
VGA Information
Graphics card:
Memory: 512.0 MB
Driver version: 6.14.0010.6755 (English)
Benchmark Information
Demo: HOCdemo.demo
Quality: Ultra
Symmetric MultiProcessing (SMP): Disabled
Aspect Ratio: [4:3]
Antialiasing: None
Anisotropic filtering: 16×
Force Ambient Light: Disabled
________________________________________

Resolution: 1600×1200
Score = 27 FPS

And finally nforce4:

Technically ULI has the highest performance however the game is still unplayable! Very good performance in opengl games does not save the ULI M1689 chipset who is at the bottom of the charts in most of the benchmarks!

54. FEAR is the last game to be tested. Unfortunately the game is so GPU-dependent that we are not going to see any difference between all 4 platforms. Even with an Athlon 64 2800 - the weakest model in the lineup for socket 754 - you are going to get the exact same frame rate with the radeon x1950 pro. Note that soft shadows is still disabled otherwise the frame rate would be cut in half!

1024*768

A. nforce3
B. VIA
C. ULI
D. nforce4

55. FEAR - 1280*1024

A. nforce3
B. VIA
C. ULI
D. nforce4

56. FEAR - 1600*1200

A. nforce3
B. VIA
C. ULI
D. nforce4

Nothing to see here, moving right along...

Does the kv8 pro support Newark at all? its the only reason i kept my dfi nf3 250gb and msi neo's. Awesome work by the way . big s754 fan but i stop at agp based boards though.

So Radeon X1950 PRO, despite being a significant upgrade from 7800 GS still does not deliver satisfactory results in Quake 4 and FEAR. Compromise would have to be made in resolution/details. It's a decent choice if one can get it cheap in AGP version. For PCIe there are much better alternatives.

For PCIe ABIT board, one should consider at least slim NVIDIA GeForce 9800 GT (to avoid blocking SATA ports). It is 1.5 years newer and a significant upgrade from Radeon X1950 PRO. GeForce GTX 2xx is a safe bet, but they are large and won't fit well on some 754/939 boards due to blocking SATA ports. They also run hot and need repasting/cleaning/BIOS fan curve adjustment to live long.