38. The next game is one of my favorites: Far cry! 1024*768 max settings including AA this time because geforce 7800gs is more than capable of handling it. This is mostly a CPU bound game therefore so this is a good test to see if one of the platform is limiting the CPU performance. 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.
nforce2:
The benchmark started at 15.01.2021 21:37:44

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) XP 3200+
CPU speed: 2200 MHz
Sound system: : NVIDIA(R) nForce(TM) Audio
VGA Information
Graphics card: NVIDIA GeForce 7800 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0011.7519 (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 = 68,41 FPS (Run 1)
Score = 66,77 FPS (Run 2)
Score = 66,89 FPS (Run 3)
Average score = 67,35 FPS

kt880:
The benchmark started at 06.03.2021 18:33:30

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) XP 3200+
CPU speed: 2200 MHz
Sound system: : Realtek AC97 Audio
VGA Information
Graphics card: NVIDIA GeForce 7800 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0011.7519 (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 = 59,86 FPS (Run 1)
Score = 55,59 FPS (Run 2)
Score = 59,21 FPS (Run 3)
Average score = 58,22 FPS

nforce2 wipes the floor with kt880!! 15.68% is one of the greatest leads NVIDIA has in this comparative.

39. Far cry 1 1280*1024 max settings
nforce2:
The benchmark started at 15.01.2021 21:46:10

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) XP 3200+
CPU speed: 2200 MHz
Sound system: : NVIDIA(R) nForce(TM) Audio
VGA Information
Graphics card: NVIDIA GeForce 7800 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0011.7519 (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 = 68,85 FPS (Run 1)
Score = 65,59 FPS (Run 2)
Score = 66,16 FPS (Run 3)
Average score = 66,86 FPS

kt880:
The benchmark started at 06.03.2021 18:44:48

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) XP 3200+
CPU speed: 2200 MHz
Sound system: : Realtek AC97 Audio
VGA Information
Graphics card: NVIDIA GeForce 7800 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0011.7519 (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 = 57,89 FPS (Run 1)
Score = 57,87 FPS (Run 2)
Score = 60,22 FPS (Run 3)
Average score = 58,65 FPS
Another crushing victory for nforce2 - kt880 is far slower and this is observable during game play! I would not play Far cry on a VIA based system.

40. Far cry 1 - 1600*1200 max settings
nforce2:
The benchmark started at 14.01.2021 18:55:49

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) XP 3200+
CPU speed: 2200 MHz
Sound system: : NVIDIA(R) nForce(TM) Audio
VGA Information
Graphics card: NVIDIA GeForce 7800 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0011.7519 (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 = 57,63 FPS (Run 1)
Score = 56,11 FPS (Run 2)
Score = 58,20 FPS (Run 3)
Average score = 57,31 FPS

KT880:
The benchmark started at 06.03.2021 18:53:38

System Information
Operating system: Microsoft Windows XP
System memory: 2,0 GB
CPU: AMD Athlon(tm) XP 3200+
CPU speed: 2200 MHz
Sound system: : Realtek AC97 Audio
VGA Information
Graphics card: NVIDIA GeForce 7800 GS
Memory: 256.0 MB
Current GPU speed: 375 MHz
Current memory speed: 1200 MHz
Driver version: 6.14.0011.7519 (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 = 59,36 FPS (Run 1)
Score = 55,47 FPS (Run 2)
Score = 56,48 FPS (Run 3)
Average score = 57,10 FPS
At this resolution there is no more any difference between the 2 platforms.

For the final game tests only one remains, but it is the most highly awarded game in the last 25 years: Half life 2! This is one of the best optimized game that requires both a good CPU and a strong GPU if using period correct components - however because I am using a newer GPU than the rest of the platforms all results will show a CPU bottleneck. All settings are at their maximum, including AA. I used Fraps 3.4.7 because HardwareOC bench tool absolutely refused to work. I am sorry for the lack of screenshots - Fraps outputs the results in Excel.
41. Half life 2 1024*768
nforce2
Min: 38
Max: 63
Avg: 48.917

kt880:
Min: 31
Max: 62
Avg: 49.933

42. Half life 2 1280*1024
nforce2:
Min: 38
Max: 63
Avg: 47.750

kt880:
Min: 31
Max: 62
Avg: 48.717

43. Half life 2 1600*1200
nforce2:
Min: 26
Max: 82
Avg: 47.550

kt880:
Min: 28
Max: 62
Avg: 45.700

This is the only game when VIA has the upper hand at lower resolutions, however NVIDIA is back on top at 1600*1200!

It's very interesting seeing SoundStorm making that big of an impact vs the onboard sound on the KT880, even with a kind of overpowered cpu (for the time). I wonder how things would look if you use an Audigy 2 or X-Fi on the KT880...

Using a Creative audigy 2 ZS on both platforms would have negated the biggest advantage NVIDIA has over VIA: Sound Storm! As you will see in the final table, despite winning in most tests the gap between nforce2 and kt880 is not that large! It wins by large margins ONLY when using onboard sound. In fact when using another platform based on Abit NF7-S2G with nforce2 ultra without Sound Storm NVIDIA no longer has the upper hand and both nforce2 and kt880 are really close to each other.

Yes, but at the time there was another factor that played a part in the decision, price! Nvidia was always more expensive, so it was interesting to compare Nvidia with SoundStorm vs Via + dedicated sound card. Maybe the money you saved by going the VIA route could be spent on a better sound card with features like EAX.

This is correct! Back in 2003-2004 nforce2 boards were always more expensive than kt600/kt880 boards.
Now, after 200+ tests, it is time for conclusions:
1. VIA and DUAL Channel memory: there is absolutely no difference between KT600 with single channel memory and KT880 with memory in dual channel. The only difference between the KT600 and KT880 chipsets is DUAL Channel memory on the latter, I did all the above tests on a KT600 platform - motherboard Abit KV7 - and the results are within the margin of error: sometimes KT880 is 1-2% faster than KT600, sometimes is 1-2% slower. Dual channel does not help VIA in any way!
2. NVIDIA and DUAL Channel memory: there is always a 5% difference between single and dual channel memory on nforce2 platform. On the same system - Motherboard Abit AN7 - when using memory in single channel configuration - DIMM 1&2 - you loose in all tests at all resolutions and at all settings around 5%, more at lower resolutions, less at higher ones. Nforce2 with SINGLE channel = KT880. This means that Athlon XP 3200 is capable of saturating the first channel and using the second to provide a little extra performance!
3. Audio: Sound Storm IS the reason nforce2 won! If I have completed the tests without including those which included audio the final result would have looked like KT880 and nforce2 performance is within the margin of error - <5%! There are tests in which Sound storm provides huge gains regardless of resolution such as 3dmark 2003 when there is always an 34-35% advantage in the 24 sounds test and tests where as the resolution increases the gains are lower and lower.
Final conclusion:
Using the best combination of hardware and software that was not possible back then NFORCE2 is the best socket 462 platform. The 2 factors that helped nforce2 win are: Dual Channel memory and Sound Storm!

This is the document where all results are put in a nice little table and final scores are calculated:

Missing from that document are the drivers used:
nforce2: 5.10
geforce: 175.19
KT880: 4.56v
realtek: 4.06

Guess I'll keep my EPoX 8RDA+ then. 😉

Good work on the benchtesting, nice to read.

Eh, supporting v5 more important than all that stuff.

Gimme socket 462 with 3.3v agp and 12v aux atx power and I'll give you my right leg.

KT333 is best chipset as it supports the former and ddr400, not sure if ever put on mobo with aux power though.

these are cool results. I have 2 nforce boards and never bothered with the onboard audio. does it sound good and is it useful tho? i understand this thread isnt for those tests but would be interesting.

dual channel memory performance increase on nforce2 was a novelty i welcomed with open arms. frames are frames!

funny thing about my nforce boards - the Asus A7N8X-X I have doesn't have the 12v aux power plug but the Soltek 75FRN2-L does. The soltek board is pretty as heck, it's sparkly and gold.

kolderman wrote on 2021-03-09, 06:26:

Eh, supporting v5 more important than all that stuff.

Gimme socket 462 with 3.3v agp and 12v aux atx power and I'll give you my right leg.

KT333 is best chipset as it supports the former and ddr400, not sure if ever put on mobo with aux power though.

There is no such board and this is not the purpose of this comparative which is to find the highest performing socket 462 platform.
I also tested an Asus A7N8X-E deluxe in all synthetic benchmarks and that board is always 3-4% behind AN7, not something that you would notice when gaming but enough to put it on equal footing with the KW7. I also done all tests on the Abit NF7-S 2.0 and it scores pretty much there with the AN7 - sometimes it gets even identical points in tests such as 3d mark - so whatever "secret sauce" Abit used in its nforce2 boards is present in both of them!

After extensive testing I reached the point when I can state some other things not shown in the screenshots and data already published. I am sure there are people who made far more tests and have much more experience with socket 462 so I welcome any suggestions.
Final purpose of this comparative is to find the highest performance socket 462 platform using the best combination of hardware and software that was not available back then.
1. Isn't Windows 98SE better suited for socket 462? No, not at this level of performance; I tested the winner - AN7 same setup but with 2*256mb of RAM - and obtained 3dmark & PC mark results. Nforce2 has around 5% lower performance under 98SE so XP is better suited for the ultimate socket 462 system.
2. Isn't a period correct graphic card - such as geforce 5950 ultra/radeon 9800xt better suited for socket 462? I do not own a 5950 ultra, but with a 9800xt there is a serious GPU bottleneck at 1600*1200 resolution in absolutely all tests. Even with a geforce 7800gs you can see that at extreme settings some tests already are reaching the point where the CPU waits for the video card.
3. Does a discrete sound card helps nforce2/kt880? Yes, it does, especially VIA. KT880 + Creative audigy2 zs = nforce2 ultra + Sound storm. Creative does not seem to help nforce2, there is a very small advantage of around 3% which for me is within the margin of error and using a 2.1 altec lansing speaker setup i do not hear "better" sound with Creative.
4. You tested too few games, why did you not included game X? Guilty as charged! I will try to include more games based on your suggestions. I did tried Need for speed underground 2 but it was absolutely impossible to obtain reliable results!! I spent 1 week doing 100 runs on the same track at 1024/1280/1600 resolutions and the frame rate was all over the place; I could not get 2 runs with comparable FPS so I gave up. I did tested C&C Generals but the results were meaningless as they were identical!
5. You did not included transfer rates for the disk drives and other HDD benchmarks such as HDtach - maybe VIA KT880 with integrated SATA controller is faster the Silicon image 3112 found on nforce2! I do not have 2 identical period correct hard drives so the tests would have been apples to oranges. Even so I doubt that even the fastest drive of the era - with the exception of the original Raptor - would make a difference on this 2 platforms.
6. You did not included a single networking test - how do you know which of the 2 platforms performs best on the internet? On April 2019 I took all my socket 462 systems offline - I consider that the lack of SSE2 instructions, lack of GPU acceleration and general performance is no longer suited for the modern internet. Socket 462 is about retro gaming today and will remain so.

nd22 wrote on 2021-03-18, 11:48:

i do not hear "better" sound with Creative.

Quality wise they are probably the same, but you have the bonus of EAX support, which for some people may be the deciding factor.

Reason for the dual channel is IGP performance will be increased. Everything else remains the same due to socket A data path is bottle necked.

Cheers,

Today I finally got the last piece of the Socket 462 dream machine - manufactured much later and a later revision but originally available in 2003:

The NVAPU uses Sensaura libraries and they do support EAX 2.0. One issue with motherboard audio, including NVidia's, is the analog output is often somewhat noisy. I think the only real advantage for NVidia over the Audigy 2 is the hardware Dolby Digital Live.

pentiumspeed wrote on 2021-03-18, 23:35:

Reason for the dual channel is IGP performance will be increased. Everything else remains the same due to socket A data path is bottle necked.

There are some benefits with a discrete GPU for the CPU and probably AGP transfers.
http://ixbtlabs.com/articles2/nforce2-1vs2cha … nels/index.html

SWAAYE is right! Dual channel memory DOES help NFORCE2 but DOES NOT help KT880! NFORCE2 in single channel = KT880 even with the huge advantage in the sound tests!
I did HDtune, PCMark 2002&2004&2005 tests with the Raptor and to my dismay the performance is not earth shattering; average transfer rate does not exceed 80 MB/s! I have done over 50 HDtune and HDtach tests and the best transfer rate is 75 MB/s; I installed old versions of HDtune and HD tach and it seems it makes no difference! Granted the original drive has 50 MB/s transfer rate but I sincerely expected at least 100 MB/s. It seems that first and second generation Raptors do not exceed PATA133 in average rates.

nd22 wrote on 2021-04-04, 11:47:

SWAAYE is right! Dual channel memory DOES help NFORCE2 but DOES NOT help KT880! NFORCE2 in single channel = KT880 even with the huge advantage in the sound tests!
I did HDtune, PCMark 2002&2004&2005 tests with the Raptor and to my dismay the performance is not earth shattering; average transfer rate does not exceed 80 MB/s! I have done over 50 HDtune and HDtach tests and the best transfer rate is 75 MB/s; I installed old versions of HDtune and HD tach and it seems it makes no difference! Granted the original drive has 50 MB/s transfer rate but I sincerely expected at least 100 MB/s. It seems that first and second generation Raptors do not exceed PATA133 in average rates.

Yup, sequential transfer rates were not the Raptor's main advantage.
The 2.5" platter means you get less area per revolution.
Access times on the other hand!
Funny enough I never had any compatibility problems with my 36GB drives back in the day.
But now I have several main boards they wont play nice with.
The 74GB ones are way better.