The first thing to note is the fact that the scores increase as the resolution increase! That denotes a huge CPU bottleneck; 7600gt is hold back by the Thunderbird so much that in can not show what it is capable of!
KT266A and nforce2 are only 5% apart, but that is enough to declare nforce2 ultra the victor in 3dmark 2000. KT133A is 18% behind its bigger brother equipped with DDRAM so, again, SDRAM is limiting performance.

18. Next is 3dmark 2001, the de facto standard for testing 3d performance on socket A/462.
1024*768 resolution, default settings
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

19. Next resolution is 1280*1024 maximum details, very common today when you can get 17-19 LCD's for cheap or even for free:
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

20. We reach the final and most demanding settings: 1600*1200 maxed out, including AA:
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

3dmark 2001 results make any kind of doubt about a possible come back for VIA KT133A vanish; it is again last and by large amounts too: KT266A with DDRAM is 15% better and nforce2 ultra is 18% better.
In this test KT266A and nforce2 ultra are neck in neck; the small 2-3% lead that nvidia has over VIA is too smal l to declare a winner!

21. We reached the final 3dmark version:2003. Why this is the final one and not 2005 will become apparent in just a moment. This is the only version to include a sound test.
We start at the good ol'1024*768 resolution with default settings:
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

22. We test the next resolution: 1280*1024
Abit KT7A-RAID
Abit KR7A-RAID
ABIT AN7
There, I finally managed to put the correct pictures and not the same picture twice!

23. We are now at the final and most demanding settings: 1600*1200 maxed out details:
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

Between VIA KT266A and nforce2 ultra there is less than 2% difference, however KT133A is again 5% or more slower than its DDR counterpart!
Again no clear winner in the 3dmark 2003 standard score, only a clear looser: KT133A and SDRAM are bottlenecking the 1200C Thunderbird.

24 &25. As I said 3dmark 2003 also has 3 sound tests:
- no sound
- 24 sounds
- 60 sounds
Only nforce2 APU can do the 60 sounds test so that result is excluded but for the first 2 we got results on all 3 platforms: KT133A, KT266A, nforce2 ultra.
There is a COLOSSAL performance hit on VIA KT133A in both tests with sound; in the first one - no sound - KT266A is faster by over 46% while nforce2 is faster by over 67%.
In the second test - 24 sounds - VIA KT133A is crushed: the DDR equipped KT266A is 70% faster!! while nforce2 is 172% faster!!!
Clearly the southbridge of VIA KT133A is at fault here. This is a known problem with the 686A and 686B southbridges and Creative live cards. I tested with other Live cards: 0060, 0220, 4830 and the problem of low scores remains!

26 & 27 & 28 3DMARK 2005 should be the last version to be tested. However because Thunderbird lacks SSE instruction set when someone try to install it all you get is this warning:
And yes, as you can see in this screenshots, I have tested the 1400C, the most powerful Thunderbird on all 7 boards, I just think it would be far more useful to see the advantages that come with increased FSB on identical clocked chips!

29. TOM2D: the only GUI test in the entire suite shows us one more time that VIA chipsets are much inferior to nforce ones in this case; it is the one and only case where KT133A escapes the last place by overtaking the KT266A!
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

30 & 31. We are now getting to benchmarks that test general system performance and the first one is PCMARK 2002:
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

In PCMARK2002 the only CPU score to stand out negatively is the one obtain by KT133A; nforce2 and KT266A are too close to say we have a winner.
The memory score is abysmal on KT133A, being 30% slower than on KT266A, on the other hand nforce2 ultra is the clear winner thanks to dual channel with a 10% lead over KT266A.

32 & 33 & 34 The last version of PCMARK to be tested is the 2004 one - the motive will become clear right away. This version also has a Graphics score so we will look at that one too.
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

CPU score: Again VIA KT133A finishes last, while KT266A and nforce2 are fighting for the first place with neither of them winning
Memory score: SDRAM platform is crippled by the low bandwidth; the DDR powered KT266A is 37% faster!, while nforce2 ultra is 55% faster and scores a clear victory!
Graphic score: strangled by SDRAM, 7600GT scores poorly; meanwhile nforce2 manages another win - 6% over KT266A!

35 & 36 & 37 Pcmark 2005 is another test that requires the presence of the SSE instruction set and therefore can not run on any of the systems! And no, if you copy over the installation files it will not run!

38&39&40. Finally we reach the first game: Farcry 1! From 2004 it is still one of my favorites; I have been trying to finish it on realistic difficulty for the last 20 years and I still have not succeeded 😀

First under the microscope is Abit KT7A-raid/VIA KT133A. This game needs lots of RAM - around 1.3 - 1.4 gb commit charge in task manager can be observed when running it - but the 1.5gb of SDRAM on the Abit board proved sufficient even if it is too close for comfort! All details maxed out as geforce 7600 GT is more than capable of mastering the graphics in this game. I used HardwareOC bench tool to provide accurate and repeatable testing conditions. Because the results are in HTML format I do not have screen shots saved but I will put all of them in one archive for anyone interested.

1024*768 resolution:

The benchmark started at 09.09.2024 20:56:21

System Information
Operating system: Microsoft Windows XP
System memory: 1,5 GB
CPU: AMD Athlon(tm) processor
CPU speed: 1200 MHz
Sound system: : SB Live! Wave Device
VGA Information
Graphics card: NVIDIA GeForce 7600 GT
Memory: 256.0 MB
Current GPU speed: 560 MHz
Current memory speed: 700 MHz
Driver version: 6.14.0010.9371 (English)

Resolution: 1024×768
Ultra quality option, Direct3D renderer
Level: Volcano, demo: hocvolcano.tmd
Pixel shader: default model
Antialising: 8×
Anisotropic filtering: 16×
HDR: disabled
Geometry Instancing: disabled
Normal-maps compression: disabled

Score = 31,71 FPS (Run 1)
Score = 31,16 FPS (Run 2)
Score = 31,48 FPS (Run 3)
Average score = 31,44 FPS

Copyright 2002 - 2006 Zoltan Nemeth - Roadside

We increase the resolution to 1280*1024:

The benchmark started at 10.09.2024 05:40:41

System Information
Operating system: Microsoft Windows XP
System memory: 1,5 GB
CPU: AMD Athlon(tm) processor
CPU speed: 1200 MHz
Sound system: : SB Live! Wave Device
VGA Information
Graphics card: NVIDIA GeForce 7600 GT
Memory: 256.0 MB
Current GPU speed: 560 MHz
Current memory speed: 700 MHz
Driver version: 6.14.0010.9371 (English)

Resolution: 1280×1024
Ultra quality option, Direct3D renderer
Level: Volcano, demo: hocvolcano.tmd
Pixel shader: default model
Antialising: 8×
Anisotropic filtering: 16×
HDR: disabled
Geometry Instancing: disabled
Normal-maps compression: disabled

Score = 31,06 FPS (Run 1)
Score = 30,76 FPS (Run 2)
Score = 31,13 FPS (Run 3)
Average score = 30,98 FPS

And the most demanding settings: 1600*1200 with max details

The benchmark started at 10.09.2024 06:02:27

System Information
Operating system: Microsoft Windows XP
System memory: 1,5 GB
CPU: AMD Athlon(tm) processor
CPU speed: 1200 MHz
Sound system: : SB Live! Wave Device
VGA Information
Graphics card: NVIDIA GeForce 7600 GT
Memory: 256.0 MB
Current GPU speed: 560 MHz
Current memory speed: 700 MHz
Driver version: 6.14.0010.9371 (English)

Resolution: 1600×1200
Ultra quality option, Direct3D renderer
Level: Volcano, demo: hocvolcano.tmd
Pixel shader: default model
Antialising: 8×
Anisotropic filtering: 16×
HDR: disabled
Geometry Instancing: disabled
Normal-maps compression: disabled

Score = 31,40 FPS (Run 1)
Score = 30,94 FPS (Run 2)
Score = 31,36 FPS (Run 3)
Average score = 31,23 FPS

Copyright 2002 - 2006 Zoltan Nemeth - Roadside