Fifth is the Abit KR7A-raid with Creative Live:
41. resolution: 1024*768
Min: 27
Max: 44
Avg: 34.55

42 resolution: 1280*1024
Min: 23
Max: 41
Avg: 33.617

43. resolution: 1600*1200
Min: 22
Max: 39
Avg: 30.183

Half life 2 short summary: VIA KT266A kills everyone else!
Long summary: even KT133A manages to outscore nforce1 by 9% at 1024*768 leaving NVidia behind. At the same resolution AMD760 scores a 33% lead over the slowest platform: nforce1; while KT266A lead is 37%!!
By now most of you will think: this guy screwed up something, that kind of results with the same CPU, RAM and GPU should be impossible! Well, that is exactly what I thought. I formatted the drive, reinstalled XP + all the benches and games and tested again. Pulled up the reserve NV7-133R from storage and again all tests run on the second board. The scores remain just as bad. For the third time I used a different hard drive, full install of XP + everything else and full set of results on the primary and secondary NV7-133R. So I decided to stop repeating the tests and that these are indeed the scores obtained on nforce1. That chipset is suffering from "childhood disease", is a bad product and nforce2 should be a matured chipset where NVidia solved the problems with the first.

44. & 45 & 46. The final game to be tested is Doom 3. It is an exceptionally GPU bound game so I disabled AA in order to give 7600GT some breathing room.
First is again nforce1 + sound storm; second is nforce1 and live (the scores are IDENTICAL - this is the test done with the timedemo demo1 from the command line so everything stays exactly the same between every single run, 2 times tested because the first time the results will always be lower):
1024*768
1280*1024
1600*1200

Third is the system with the KT7A-raid. Note that testing Doom 3 from the command line does NOT use sound:
1024
1280
1600

Fourth is the one based on KG7-raid:
1024
1280
1600

And the fifth and final one is the system with the KR7A-raid:
1024+1280+1600

Doom 3: again KT133A and nforce1 are at the bottom of the charts while AMD760 with KT266A are in front with a 10%, respectively 15% lead over the nforce.

47. Only 2 tests remain. The second to last one is Super PI. In this test a shorter time to calculate 1M is actually a better result!
Same order:
1. nforce1: NV7-133R + sound storm
2. nforce1: NV7-133R + creative live
3. KT133A: KT7A-raid + creative live
4.AMD760: KG7-raid + creative live
5. KT266A: KR7A-133R + creative live

VIA KT133A is let down by its low memory bandwidth so the score is substantially lower than the rest. AMD760 and KT266A are in front with equal scores!

48. We reached the final test: Cine bench 2003. We are interested in the CPU score:
1. nforce1: NV7-133R + sound storm
2. nforce1: NV7-133R + creative live
3. KT133A: KT7A-raid + creative live
4.AMD760: KG7-raid + creative live
5. KT266A: KR7A-133R + creative live

All scores are so close that there is no clear winner here.

My first proper gaming build after a long long break from PCs (K6-2 266 and Voodoo 1) was an nForce 2 build. It was awesome. Paired it with a Gainward 6800GT GS. Just an absolute beast.

The main reason DFI Lanparty Ultra B ruled this period (and something that your benchmarks will not reflect) was it's insane over clocking potential when using the right components.

The Ultra B could push 3.3v to the RAM (most boards top out at 2.8v). We paired this with Mobile Barton 2500 CPUs and legendary Winbond BH5 RAM ICs. The result? 3GHz stable on air at silly FSB (250-300MHz) and super tight timings. The raw clock speed combined with massive memory bandwidth was hard to beat at the time.

I was a proud member of the 3GHz stable club on DFI Street.

Also, a 1.8GHz spec CPU doing 3GHz vroooooooooooom!

(I've been trying to find a DFI NF2 ultra B for several years now. I wish I'd never sold it... Never loved the Lanparty nf3 939 I replaced it with...)

Lostdotfish wrote on 2024-06-26, 18:54:

My first proper gaming build after a long long break from PCs (K6-2 266 and Voodoo 1) was an nForce 2 build. It was awesome. Pa […]
Show full quote

My first proper gaming build after a long long break from PCs (K6-2 266 and Voodoo 1) was an nForce 2 build. It was awesome. Paired it with a Gainward 6800GT GS. Just an absolute beast.

The main reason DFI Lanparty Ultra B ruled this period (and something that your benchmarks will not reflect) was it's insane over clocking potential when using the right components.

The Ultra B could push 3.3v to the RAM (most boards top out at 2.8v). We paired this with Mobile Barton 2500 CPUs and legendary Winbond BH5 RAM ICs. The result? 3GHz stable on air at silly FSB (250-300MHz) and super tight timings. The raw clock speed combined with massive memory bandwidth was hard to beat at the time.

I was a proud member of the 3GHz stable club on DFI Street.

Also, a 1.8GHz spec CPU doing 3GHz vroooooooooooom!

My best Socket A build ever was an Epox 8RDA3+ with a mobile Barton cooled with an SLK-900 and using a 9800 Pro for video. It never came close to a number like that, I think the best I ever achieved was maybe 2.5 GHz. I would likely have stuck with my Epox 8K3A+ until I moved to an Athlon 64 system since it was an overclocking beast, but the caps on the board began failing in the summer of 2004 and Newegg had a sale on the 8RDA3+.

I also had 2 socket 462 systems back in the day. The first system was built around a Lucky star board that was unstable as hell! The second time I used an Abit KR7A that was running rock solid (and still runs today!). I remember the period from 1999 to 2004 where everything was advancing so fast that practically your system was obsolete in 2 years!

Before getting to the final results a few words about each system:
1. Abit KT7A – raid: although runs XP and all the benchmarks just fine I must mention a few things, both good and bad. Good ones: not even once the 1.5gb RAM limit was exceeded; took the geforce 7600GT without any problems; no crashes, no BSOD, full stability throughout all the tests. Now the bad ones: XP start up and shut down are really slow, I did not timed them but it is noticeable; games loading times are atrocious, especially Far cry 1 and Doom 3 – note that using a WD raptor over an adapter did not improved the transfer rates! It is the only system where XP feels slow! My personal opinion is that VIA KT133A is better suited for an earlier era. Abit KT7A – raid with a Thunderbird and 512 mb of RAM would make for a killer Windows 98 rig! This is strictly my personal opinion, I did not test it in W98 to see how it performs but I feel this would be a better use for such a system.

2. Abit NV7 – 133raid: when I started the tests I expected from nforce nothing but a clear victory. Imagine my disappointment when the results were so poor. I simply could not believe it and I repeated the tests 5 times before finally concluding that nforce1 is a poor choice for gaming. In addition memory compatibility is awful. The good thing is that it has the 12V power connector. Also XP runs very well on this system, even games are loading fast.

3. Abit KG7 – raid: it was really, really difficult to get this motherboard going with the 7600GT but I am glad I discovered the correct settings to make it work! XP feels snappy and all games are loading pretty fast. A chipset released a whole year before VIA finally put its act together and came up with a worthy DDR chipset is nothing but amazing. Memory compatibility is also very good.

4. Abit KR7A – 133raid: the board simply worked from the get go; memory and video card compatibility is very good; it never disappointed. Excellent board!

Now the standings:
Nforce1 with integrated audio represents the baseline and is awarded a 100% score and the other 4 platforms are a percentage higher or lower.
1. VIA KT266A: 113.57%
2. AMD760: 109.18%
3. nForce1 + Creative live: 101.87%
4. nForce1 + Sound storm APU: 100%
5. VIA KT133A: 96.31%

Despite being the slowest of all I am not considering Abit KT7A/KT133A a bad board/chipset. It is simply better suited for Windows 98 and games of the late 90’s than the ones from the early 2000’s. When used as it is meant to be I think this motherboard would be great! So I do not consider Abit KT7A – raid the looser in this tests. Also Athlon XP is seriously bottlenecked by the SDRAM and really needs DDR to shine.
The real dissapointment and what I consider to be the actual looser is the Abit NV7 – 133raid. nForce1 is a weak chipset, a lot of performance is lost when using a board with the first Nvidia chipset. Also worth to mention is the fact that sound storm is equal to the Creative live.
Abit KG7 – raid with AMD760 chipset is miles ahead of nforce1 as it demonstrates in all the tests. A good match for the Athlon XP it shows that DDR has a lot of impact on performance. A personal favorite, I would not hesitate to build a system around the KG7 – raid – the best AMD760 board worldwide.
Finally we got to the winner of the 266MHZ battle: Abit KR7A – 133RAID with VIA KT266A chipset. It crushed the competition in all synthetic 3d marks and real games tests. Performance wise was always in front. I know that there are a lot of KT266A boards out there and I did not test any other manufacturer but I still believe this board is among the best.

Lostdotfish wrote on 2024-06-26, 18:54:

My first proper gaming build after a long long break from PCs (K6-2 266 and Voodoo 1) was an nForce 2 build. It was awesome. Pa […]
Show full quote

My first proper gaming build after a long long break from PCs (K6-2 266 and Voodoo 1) was an nForce 2 build. It was awesome. Paired it with a Gainward 6800GT GS. Just an absolute beast.

The main reason DFI Lanparty Ultra B ruled this period (and something that your benchmarks will not reflect) was it's insane over clocking potential when using the right components.

The Ultra B could push 3.3v to the RAM (most boards top out at 2.8v). We paired this with Mobile Barton 2500 CPUs and legendary Winbond BH5 RAM ICs. The result? 3GHz stable on air at silly FSB (250-300MHz) and super tight timings. The raw clock speed combined with massive memory bandwidth was hard to beat at the time.

I was a proud member of the 3GHz stable club on DFI Street.

Also, a 1.8GHz spec CPU doing 3GHz vroooooooooooom!

(I've been trying to find a DFI NF2 ultra B for several years now. I wish I'd never sold it... Never loved the Lanparty nf3 939 I replaced it with...)

DFI was something else (alongside ABIT). However price is a serious issue - both than and now.
About overclocking: because I want my hardware to last as long as possible - my entire life if possible and according to statistics at least it would be a few more years I do not overclock any retro parts. Most of them are over 20 years old and I don't want to stress them.