Hirdetés
ProTest
The board we had was "only" revision 1.0, but before we ran any benchmarks we knew we had the right board... there was a tiny SMD resistor on the PCB that was in the "good place". Had it been placed a few millimeters elsewhere, it would have been a whole lot different story. Located right from the South Bridge are the R126/R127 resistors. Out of these R126 is the key - if it is "resistored", then we have a "good" board, if it isn't, then we don't. This behaviour was discovered originally by AnandTech and OCWorkbench, although the whole thing is probably becoming less and less critical as KT266 production increases. (believe it or not, when Marton called me on the phone and told me that all that matters is R126, it took me two seconds to locate the resistor... out of a few hundred... 
I couldn't believe it myself)
Before we move on to the benchmark results that decide it all, take a look at the test configuration:
- MSI K7T266 Pro 1.0
- Duron 600 @ 828 (6*138MHz)
- Coolink COOL141 "Fox Shark" CPU cooler (60*60*15mm, BB, 4k rpm, 27dBA, 82*60*23mm, 177g)
- 2*128MB PC2100 Samsung DDR SDRAM @ CL2, 4-Way, SDRAM 1T "on"
- Voodoo5 5500 @ 166/166MHz
- 17"-os Panasonic PanaSync/Pro P70 @ 85Hz
- Win98
- VIA 4-in-1 4.29
- DirectX 7.0a
- 3dfx drv. 2000.8.22
- UT 4.36
- Sisoft Sandra 2001SE
Next is a screen-shot from Sandra 2001SE, which needn't really be commented. We weren't expecting a lot from the board a priori, but due to the DDR RAMs, Interleaving, low CAS settings and a relatively high FSB, we got result that quite surprised us. The KT266 can be considered as a mature DDR chipset, therefore we don't expect any serious benchmark increases in the future. Which in turn also means that the top KT133A mainboards, like the Iwill KK266 or MSI K7T Turbo are still in the game, supposing you can achieve an FSB high enough. But the K7T266 Pro still has a few tricks up its sleeve and the KK266 is barely cheaper, so the only thing that could decide against MSI is the price of DDR RAM - which is dropping even as we speak (write).
Got some power to unleash.. (Samsung PC2100 DDR-SDRAM) click4big!!
The other DDR combo to leave our lab, the ALi MAGiK 1 based Iwill board (but we could also mention Transcend's TS-ALR4, which will get it's necrology on our site in a few days) was only slightly worse than the K7T266 regarding the FSB, still, the benchmarks clearly show that VIA is the one to vote for. It is also more flexible for mainboard manufacturers to use the KT266 as it can be configured in numerous ways, so all in all the price/performance ratio of this chipset, and thus the MSI K7T266 Pro is very hard to beat.
We have written before that the chipset could go as high as 138MHz (276MHz DDR), which is somewhat low compared to the ability of setting anything up to 164MHz FSB in the BIOS (the PCI is clocked 1/4 of the FSB at above 133MHz). Since the benchmarks were extremely good at even this speed, we will cease criticizing the board now. But what really makes us forget all that wasn't perfect is the fact that our 600MHz Duron (first series, unlocked) ran at 828MHz without even raising the core voltage! To be honest, you only have +0.1V to give for any CPU as it stands now, but this CPU maxes out at 840MHz no matter how high you set the voltage, so running at 828MHz with default settings is simply amazing. To test the highest FSB possible we set our CPU to a 5X multiplier, but 140MHz still wasn't stable, that's how we, or better said the K7T266, ended up with 138MHz.
We skipped the test of the asynchronous memory clock with the Iwill KK266, as the highest frequency we could have achieved was 132 * 1.33 = 175MHz, which was also possible via simply raising the FSB (which is once again amazing). With the K7T266 Pro it was a different situation as the last FSB stable wasn't near as high as 175MHz. We also had some very high quality DDR SDRAM, so the end for both the RAM modules and the comparison between the sync/async operation had to be found out/made. The highest FSB was 114MHz is this case, which in turn resulted in the memory running at 152MHz. Nevertheless, the guess we took at the beginning of this article turned out to be true, as 138MHz sync was much faster than 152MHz async.
Tied runners?
It is worth looking at the results of the Iwill KK266 board. Few weeks ago we couldn't believe our eyes when we saw these numbers, but we are now talking DDR and as a DDR platform, MSI starts off were the Iwill board finished. Impressive, we must add.
A cikk még nem ért véget, kérlek, lapozz!
