We switch the platform and the chipset to KR7A-raid/VIA KT266A:
As I had enough time I tested all 3 resolutions in one shot:

Far Cry Benchmark

The benchmark started at 19.02.2025 16:41:33

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 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 = 36,00 FPS (Run 1)
Score = 35,32 FPS (Run 2)
Score = 35,73 FPS (Run 3)
Average score = 35,68 FPS

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 = 36,26 FPS (Run 1)
Score = 35,74 FPS (Run 2)
Score = 36,15 FPS (Run 3)
Average score = 36,04 FPS

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 = 35,76 FPS (Run 1)
Score = 34,85 FPS (Run 2)
Score = 35,17 FPS (Run 3)
Average score = 35,25 FPS

Copyright 2002 - 2006 Zoltan Nemeth - Roadside

And now the Abit AN7/norce2 ultra:

Far Cry Benchmark

The benchmark started at 23.02.2025 05:09:54

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) processor
CPU speed: 1200 MHz
Sound system: : NVIDIA(R) nForce(TM) Audio
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 = 39,27 FPS (Run 1)
Score = 38,21 FPS (Run 2)
Score = 38,86 FPS (Run 3)
Average score = 38,78 FPS

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 = 39,31 FPS (Run 1)
Score = 38,54 FPS (Run 2)
Score = 38,96 FPS (Run 3)
Average score = 38,93 FPS

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 = 38,50 FPS (Run 1)
Score = 37,32 FPS (Run 2)
Score = 38,34 FPS (Run 3)
Average score = 38,05 FPS

Copyright 2002 - 2006 Zoltan Nemeth - Roadside

Farcry 1 is best played on a nforce2 board as Abit AN7 is faster than KT266A across all resolutions by 8%! Meanwhile KT133A is again last with disappointing results.
Regardless of the rankings the fps obtained on all 3 systems shows a huge bottleneck. Geforce 7600GT can actually play Farcry 1 at max details just fine at much higher FPS!

41&42&43 We switch to Half life 2, one of my favorite FPS that I still play today! It is a very optimized game that requires a balanced system. This is the original 2004 version and not the Steam version. I tested right in front of the train station where Dr Breen is making his speech using max details as the geforce 7600gt is more than capable of handling this game.
I used Fraps 1.9D because HardwareOC bench tool refused to work. I am sorry for the lack of screenshots - Fraps 1.9D outputs the results to a .TXT file.

First is the system with the KT133A chipset represented by KT7A-raid

1024*768
2025-02-18 05:37:41 - hl2
Frames: 1350 - Time: 56281ms - Avg: 23.986 - Min: 13 - Max: 32

1280*1024
2025-02-18 05:42:51 - hl2
Frames: 1651 - Time: 78002ms - Avg: 21.166 - Min: 7 - Max: 28

1600*1200
2025-02-18 05:44:51 - hl2
Frames: 1315 - Time: 65934ms - Avg: 19.944 - Min: 6 - Max: 29

The framerates are quite bad; the game is not actually playable!

Second is the system based on KT266A chipset represented by KR7A-raid:

1024*768
2025-02-20 05:51:42 - hl2
Frames: 2034 - Time: 72234ms - Avg: 28.158 - Min: 21 - Max: 36

1280*1024
2025-02-20 05:54:22 - hl2
Frames: 1947 - Time: 72203ms - Avg: 26.965 - Min: 20 - Max: 37

1600*1200
2025-02-20 05:56:10 - hl2
Frames: 1999 - Time: 72174ms - Avg: 27.696 - Min: 21 - Max: 38

Again you can not actually play the game at these settings !

Third is the nforce2 ultra system with the AN7:

1024*768
2025-02-23 06:18:49 - hl2
Frames: 1668 - Time: 64282ms - Avg: 25.948 - Min: 6 - Max: 35

1280*1024
2025-02-23 06:20:43 - hl2
Frames: 2513 - Time: 82172ms - Avg: 30.582 - Min: 25 - Max: 39

1600*1200
2025-02-23 06:22:28 - hl2
Frames: 2531 - Time: 78031ms - Avg: 32.435 - Min: 27 - Max: 44

Despite an average framerat over 30 fps there are lots of stutters and pauses that make the game unplayable!

Half life 2: for the first time nforce2 ultra manages to beat a board based on a VIA chipset by a small amount! Incredible as it may seem but until now VIA chipsets had the upper hand in Half life 2.
KT7A-RAID is far behind, 20% or more behind its DDR cousin!

44 & 45 & 46. The final game should be Doom 3. The game installs just fine on a system with a Thunderbird CPU, however as I started installing patches - 1.2 patch or the RoE expansion - the game required SSE instruction set. And no, if you copy the game over without installing it and try to start Doom 3 1.3 all you get is the error below. Because modifying the testing conditions would not make for repeatable and accurate results i decided to left out Doom 3, therefore no results for this game!

47. Only 2 tests remain and the second to last one is SUPERPI. A shorter time to calculate 1 million decimals is actually better!

ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

Nforce2 manages to score another convincing win while KT133A with SDRAM is much slower than everyone else!

48. We reached the final test: Cinebench 2003. This is the oldest version I could find when i started testing 7 years ago. Meanwhile, a Youtuber, known as Fully Buffered, managed to found an even older version as seen in this video:
https://www.youtube.com/watch?v=bPBLHSZd6Hg
Because changing version would brake accuracy and compatibility of results I sticked to the version I have been using which is not compatible with Thunderbird because it requires SSE instruction set!
When starting cinebench 2003 all you get is this error message and it will not tell you that it needs SSE unlike 3dmark 2005!

All tests are finished! Time to rank the participants. Despite not showing the screenshots I also tested 1200C in the other boards:
1. Abit AT7 – MAX with VIA KT333 chipset
2. Abit AT7 – MAX2 with VIA KT400 chipset
3. Abit KV7 with VIA KT600 chipset
4. Abit NV7 – 133 RAID with nforce 415 – D chipset
All of them recognize just fine the CPU and ran all tests minus those that require SSE.

When comparing only the VIA KT133A, KT266A and nforce2 ultra, whose screenshots can be seen in this topic on the previous pages, the standings are as follows (KT266A is awarded a 100% score):
1. ABIT AN7 nforce2 ultra: 105.72%
2. ABIT KR7A-RAID Via KT266A: 100%
3. Abit KT7A-RAID Via KT133A: 85.09%%
The KT133A is really showing that matching SDRAM and any Athlon with a 266 MHz FSB - Thunderbird, Palomino, Thoroughbred - is a really bad choice as it is limiting performance a lot! Athlon needs DDR to shine and one more time nforce2 ultra is number 1, being 5.72% faster than KT266A.
Note that when I tested the Athlon XP Thoroughbred 2000+ the difference between the VIA KT266A and nforce2 ultra was exactly 5.7% ! So it doesn't matter if all you got is a Thunderbird 1000C; it needs 266 Mhz DDR to really shine.
Overall for any Athlon with a 266 MHz FSB the best chipset is: NFORCE2 ULTRA!
One more time NVIDIA wins!

Only one more battle remains to be fought on socket A!

I will now mention a few things about each system that can not be seen in the screenshots.
Abit AN7 with Thunderbird 1200C does not feel slow in any way; despite the low frequency Windows XP is still snappy and responsive; when you minimize or maximize a windows there are no slowdowns of any kind.
Abit KR7A-raid with Thunderbird 1200C feels equally fast in Windows as the nforce2 system!
Abit KT7A-raid with Thunderbird 1200C is the only system where XP feels slow; startup and shutdown take a long time, most desktop activity is fast, however loading Farcry 1 takes ages!

I will try now to answer some possible questions
a. Where is the 1400C? Maybe with that CPU KT133A can score better!
VIA KT133A does score higher with the 1400C but so does KT266A and nforce2; the distance between the 3 chipsets remain exactly the same.
This are the results from 3dmark 2003 default settings; the relative distance between each system stays exactly the same:
Later edited to reflect the correct order, as the wd raptor is present in the third system, the nforce2 one!
ABIT KT7A-RAID
ABIT KR7A-RAID
ABIT AN7

b. another possible question would be: how about testing with less RAM but with aggressive timings? Maybe using a CAS of 2 and Delay DRAM read latch of 0.5ns and can help improve improve some of the scores for the VIA KT133A platform.
While this is true and the scores DO improve using only 2 sticks of 256mb of SDRAM on the Abit KT7A-raid in some tests such as 3dmark 2000&2001 the impact on tests that require more than 512mb of memory to run and therefore require to activate the swap file is simply catastrophic. In farcry 1 - the best possible example because it needs more than 1gb of ram at max settings - you get visible stutters and the final result will be 15 fps!! - less than half of what we get by using a sufficient amount of RAM in order to prevent the game from trashing the disk drive.

The final battle on socket A is reserved for the very first processors released for the socket: with a FSB of 200 MHz they are the lowest performing of all. AMD released Thunderbird with a 200 MHz FSB starting with the 600 MHz model all the way up to 1400 MHz.
I could not find the 1400 MHz model so I bought the 1200 Mhz one: A1200AMS3B, same frequency as its brother with a 266 FSB.

Also with the Thunderbird VIA released the very first chipset for socket A: KT133. Supporting only 200 MHz FSB it is accompanied by the 686A southbridge, which supports ATA66. The Abit KT7-raid motherboard fully supports the 1200 MHz CPU and recognizes it on the spot.

Screenshots with the KT7-raid system; you can clearly identify the motherboard by the southbridge seen in the first 2 screenshots.

The RAM frequency can be set at 100 MHz - in sync with the CPU - or at 133 MHz. The best performance is obtained by running the RAM at 133 MHz on this platform despite being out of sync.
Also note that, again, CPUZ legacy has great difficulty recognizing the motherboard and the RAM and identifies the chipset as KT133A; I did not took the a screenshot with the memory tab because CPUZ said the frequency was 66 Mhz!!! Everest 5.50 is as reliable as ever and displays correctly all the info.