6. ... and the final and most demanding settings: 1600*1200 maxed out with sound:
A. NFORCE4
B. K8T890

Aquamark3 does not have a clear winner, the very small advantage the nforce4 has over the competition is tiny! PCI-express riser is functioning just fine and the system with the AX8 board never locked up, not even once.

7. 3dmark is next and the version that must be present in every single comparative is, of course, the 2001 one! It is the only DirectX 8 test and both platforms are overpowered for this benchmark!
Default settings first:
A. NFORCE4
B. K8T890

8. nforce4 starts good with a 3% lead!
1280*1024 at max details is next:
A. NFORCE4
B. K8T890

9. Evan at 1600*1200 fully maxed out 3dmark 2001 is very easy on both systems:
A. NFORCE4
B. K8T890

NForce4 is again in front without breaking the 5% limit of error!
Before going further i must underline the fact that period correct reviews used graphic cards that are weak for what is coming: radeon x850xt pe or geforce 6800 ultra are the best available in 2004 -early 2005- however for DirectX 9 games and benchmarks they perform miserably!

10. 3DMARK 2003 is the first benchmark from FUTUREMARK that supports DIRECTX 9 - only in the nature test.
Default settings first:
A. NFORCE4
B. K8T890

For the first time we got a clear winner! nforce4 manages to overcome its competitor, K8T890, by a 10% lead!
11. 1280*1024 at maximum details is next!
A. NFORCE4
B. K8T890

And again Nvidia manages to win!
12. 1600*1200 is the final resolution; again maximum details - that means AA set at 8X. No period correct video card can ever hope to cope with these settings!
A. NFORCE4
B. K8T890

Nvidia scores its first victory in 3dmark 2003; across all resolutions and settings nforce4 wins decisively!
13 & 14. 3dmark 2003 is the only version that includes 3 sound tests; the AC97 codec included in the nforce4/K8T890 chipsets can only do the first and second test - no sound and 24 sounds; the third test - 60 sounds - is only available if you got a hardware accelerated solution such as creative audigy!
Unlike the 3d scores VIA manages to win the sound benchmarks by a tiny 0.3% advantage over the nforce4, nothing to write home about though!

15. 3dmark 2005 is next and, just like with the 2003 version, GTX 285 is bottlenecked by the Athlon 64 X2 4800 across all resolutions and settings!
1024*768 default settings first:
A. NFORCE4
B. K8T890

And again nforce4 wins! This contradicts all period correct reviews and we are not even half way through!
16. 3dmark 2005 1280*1024 max details
A. NFORCE4
B. K8T890

17. 3dmark 2005 1600*1200 fully maxed out including AA 8X:
A. NFORCE4
B. K8T890

One more time nforce4 distances itself from K8T890 with a 6-7% lead across all 3 resolutions! Things are not looking good for VIA and again my results contradicts period correct reviews that showed NVIDIA and VIA neck in neck!

18. For the first time 3dmark 2006 makes an appearance! An enhanced version of the 2005 one it is very tough on any DirectX 9 video card. Not a problem for GTX 285 which is again CPU-limited even at the highest resolution and settings!
Also 1024*768 is no longer the default resolution but 1280*1024 so we have 2 results for 1280*1024: one with default settings that are somewhat easier on your GPU and one with all the bells and whistles turned on.
1280*1024 default settings , no AA, low quality texture filtering:
A. NFORCE4
B. K8T890

Nforce4 has a 4% advantage, not decisive but still in the lead.
19. 1280*1024 max details: AF 16X, AA 8X, full floating point precision in shader calculations:
A. NFORCE4
B. K8T890

Nforce4 still in the lead with the same 4% advantage!
20. 1600*1200 max details
A. NFORCE4
B. K8T890

At the highest resolution nforce4 leaves no more doubt: 8% better. A pattern emerges after synthetic 3d tests: the tougher the settings the higher the advantage that nforce4 has over K8T890 - that means that most likely PCI-express implementation in the NVIDIA's chipset should be better than in the VIA's one! We shall see...

My Turion 64 performs about as well as K8T890 in 3d mark 2005 which represents games up to 2006. 3d mark 2006 is the first benchmark where double cores show significant 30% advantage (~5900 vs 7985 score). The s939 system used still has room for improvement (CR 1T, even tighter timings with 2 sticks at DDR400 or CR 1T at DDR533 ECC) while s754 is at the end of its capabilities. 3d mark 2006 represents best games from 2007-2008 and is very CPU demanding. Even fast AM2 systems struggle with it. It needs AM2+.

AFAIK only 3dmark 2006 is multi-threaded, all previous versions work on 1 thread only.
You are absolutely correct about the command rate - 2T carries a significant performance penalty - however I must stress the fact that it is crucial to run the memory in sync with the CPU so DDR-400 is the best option for a standard clocked Athlon 64 X2 and timings influence is 0(zero)! I got some results with 3-3-3-8 and there is no difference between them and the ones obtained at 2-3-2-6 with high end Corsair modules!