All these reviews test the PT800 reference board from VIA, and not an actual shipping product...
Which, if anything, means the retail boards should be at least as fast if not faster than a reference board, being an early
pre-production unit with a less optimized
pre-production BIOS.
ALL Canterwood and Springdale boards competing against the PT800 in these reviews were production boards that had been shipping for months and undoubtedly had better optimized BIOS. For example:
The Intel D865PERL and D875PBZ motherboards tested in The Tech Report review already had several BIOS updates before the date of publication (Aug. 10th, 2003). Intel had released
four BIOS updates for the D865PERL and
seven BIOS updates for the D875PBZ before Aug. 10th,
not counting the initial release version.
The Abit IC7 875P also had
four BIOS updates released before publication of the TweakTown review on July 11, 2003.
The DFI LANPARTY PRO875 used in the Hexus review had
five BIOS updates released before publication on August 29, 2003. The
Epox 4PDA2+ 865PE had
four.
All dual channel boards in question were well optimized production boards. A board that has been in production long enough to see several BIOS updates released is far more likely to be near its maximum performance potential than a
pre-production reference board with a
pre-production BIOS.
Questions?
and they even mention that the board came slightly overclocked from VIA, and they could not change that (which is strange).
Two of the
five reviews measured the PT800 board's FSB to be a shocking .5% higher than spec, one-half of one percent, or 1MHz. 1MHz overclocking clearly would result in a hugely unfair advantage. :roll:
On the other hand, the Legion Hardware review noted:
"...the ASUS P4C800 is running on a 202MHz FSB where as the
PT800 runs the proper 200MHz FSB. The ASUS P4C800 is
one of the fastest Dual-Channel DDR boards available though it is a shame ASUS have made it impossible to test the board at the 200MHz FSB. Anyway at the end of the testing I have increased the FSB of the PT800 for a more accurate comparison."
So I wouldn't take these results as an accurate real world comparison.
Me neither. Final PT800 production boards should perform slightly better against dual channel than these tests reveal.
A better roundup review which tests PT880 (the dual channel successor to PT800) based MSI board shows that it lags the 865/875 by up to 5% in gaming, which means that the single channel PT800 is probably 5-10% slower than the leading Intel chipsets.
lol! A better indication of the PT800's real world performance than
five different PT800 reviews with extremely consistent results is the measured performance of a different chipset? Ooookay!
And btw, the quote from HardOCP saying that dual channel memory does not help performance is laughable,
But not as laughable as your apparent illiteracy. The HardOCP review never stated dual channel memory does not help performance.
as there are many tests which show a significant performance difference between single and dual channel using the same board and memory (which is the right methodology to measure the impact of dual channel).
Its the right methodology to measure the performance impact of single channel vs. dual channel on
the same chipset. This result cannot be transferred to a different chipset, as the unanimous results of these
five different reviews prove beyond any doubt.
I think they came to this wrong conclusion because they compared a manually optimized and overclocked (single channel) PT800 reference board to a below average (dual channel) 875 board, so they tried to explain it by saying the dual channel doesn't matter.
Well don't think too hard, it doesn't seem to be your strong suit.
But the fact is that if they ran the 875 in single channel mode, it would have been slower than the PT800, which is of course untrue as we know that 875 is even faster then PT880 in dual channel mode. As I said, the reason for the skewed results is that the boards they chose are not true representatives of their chipset class.
True, the competing boards tested aren't representative of their chipset class. The boards tested have consistently been among the performance LEADERS in their chipset class.