What bus is the most important to get the fastest?

With a bit of testing yesterday I think I know the answer but I figured I would get the experts opinions. I have a 3200 Winchester on a Neo2 Platinum. I currently am running it at 9x278, HTTx4, ram 1:1. I can get a higher fsb by lowering the HTT to 3x, but performance seems to take a hit. So I am guessing that the HTT speed does matter in the equasion. I have tried a ram divider but my Hyperam has never played well with them so I try to stay at 1:1. 285 is about the limit at 1:1. A 133 limit will work and I have done a 9x300, but it was not quite stable. I really need to tear it down and check the N/B cooler, which I forgot to do when I put things together. BUt for now it seems to me that 278 with 4x HTT and 1:1 ram speed is my sweet-spot. Is my logic flawed?

could u tell me what board your using because i dont seem to take any hit when i lower the HTT speed to 3x from 4x HTT

He states in his thread its the Neo2 Platinum.

But With him saying he is getting a performance hit, I have tried it all, and I dont see any performance hit going from 3x to 4x.

oh yeah sorry couldnt of read it right

I did get lower 3DMark 2001 scores when going to 3x. Not much lower, 300-400 points, but they were lower. So what you are saying is that the HTT speed means nothing? Why not run it at 1x then.

i always wondered this too, if HTT makes a diff, but i was thinking if Ram runs at DDR400, dual channel mode makes it like 800, right? so if you have HTT 1000 or 800, you wont see any difference. but in Pntgrd's setup, he's running ram at 1:1 @ 278HT, which makes it DDR556, dual channel mode = 1112mhz? so lowering his HTT to 3x would bring it down to 834mhz while ram is at 1112, could that be why he's getting a performance hit?

dont take this as fact. i'm just rambling on some thoughts, i'm not exactly sure how HTT and Ram speed interact with each other. can anyone clarify?

You make a good point, all of which are correct. But take my setup for example, I run 289mhz (DDR578)1:1 with 3x HTT. if I go to 4x HTT I can really unstable, its just too high considering I have a multi of 9x. I ran all the benchmarks including Sisoft and didnt notice any real world performance lose at all. Sure if your a hardcore benchmarker, then 300-400 points will hurt.

Something has come to me which might explain the performance hit. What is the percentage of error in the bench? 300 to 400 3DMarks is actually less than 2% when talking about a score in the high 27,000's. I never thought about that until after starting this discussion. Maybe I don't have the hit I thinnk I do.

Way bigger than 300-400 marks where you're at. The HyperTransport bus is way too wide for you to come even close to saturating at 3x. You just had a bad run. 9x285 is definitely the best option.

i kinda wish they had a 3.5 HTT multiplier on the MSI boards. It would make it easier to fine tune your overclock without sacrificing your HTT speed.

Thanks for all the imput, guys. I am going to make another run at Max o/c and just ordered a new chipset cooler and 3-blade Delta to make sure everything is optimized as far as hardware is concerned. Hope Sidewinder gets it here by Saturday. Like I stated, I had somehow forgot to even check the chipset cooler when I built the A64 system and the Neo2 is not known for having an efficient solution. I plan to resink all PLL's and mosfets, m,ost are done now but I will redo to make sure. Also replacing the Panaflo on an XP90 with the Delta to maximize cooling. I will play around with some suggestions to get the memory dividers to work with my Adata chips. I have booted and loaded XP at 2.7 (9x300) so maybe by getting everything tweaked I can get that stable.