Will Not Be Reviewing MSI Canterwood For Now
I have returned my MSI 875P FISR2R for a refund and will NOT be completing a review of the MSI Canterwood at AsusBoards. There are just too many things wrong with this board right now when used with a 533FSB Processor. The specs and features are really stand-out, but the execution leaves a lot to be desired.
First and foremost, this board is advertised as an overclockers dream board, and yet the only BIOS that works to OC a 533FSB chip is the oldest Revision 1.0. One revision failing is a bug (shipping 1.1), 2 revisions failing is just plain sloppy (current 1.2). This is an emergency problem that can kill this board in the marketplace since there are virtually NO 800FSB processors available at this point.
2) The BIOS features that are there - like memory multipliers - seem to work intermittently. This makes "getting by" until a BIOS upgrade even more of a problem.
3) The board is exceptionally picky about memory with 1.0 BIOS. Corsair 3200LL performs extremely well, but I could never find timings that work with stability with my OCZ 3500EL. With BIOS 1.2 my OCZ DID work fine at rational timings, but the 533 overclocking does not work. Perhaps a later BIOS will bring all this together into a working board.
4) Reported temperatures are alarmingly high and inaccurate. This needs to be quickly fixed or there will be many concerned endusers.
5) VERY IMPORTANT to me. While this board exhibits very stable voltages and excellent performance at stock speeds and voltages, the voltage regulation on the board I tested was poor at OC speeds and voltages on a known good 520W PS with a 28A rail on 12v and 52A rail on 5V. Voltage drift at 1.7V setting at 190FSB was between 1.55 and 1.69V. I have seen this kind of variation in voltage before (on the poorer Asus boards) and it translated into instability in overclocking. Perhaps it is just a bad board, but I am concerned. The board was unpredictable in overclocking. Every time I thought I had the settings down pat, they often wouldn't work on the next boot. The better overclocking boards are predictable, but in it's present state the MSI is not a predictable overclocker. The Abit IC7, with 4-phase power, does not exhibit this kind of voltage variation in overclocking, and it IS very predictable.
6) The last issue is my problem. I suspect this board is designed to perform best with 800FSB CPU's, and 533FSB operation is just an add-on. I don't have an 800FSB CPU to test with it. I do not believe it is fair to test this board without one. Abit is the only Canterwood so far to make an effort to give some options to 533FSB users with their 'NB Strap' option. This allows you to 'trick' the system into thinking you have an 800 Processor for instance so you get more memory ratios. Without that kind of 'trick' you are stuck with 266 at DDR400, and 266, 333, and 354 (on Abit and MSI) at 533, with a full selection of memory ratios only at 800. The MSI does not yet have any kind of setting to make all memory ratios available at lower CPU FSB's.
I found this MSI to be a great perfomer at stock speeds - and it is even reported to beat the Intel Canterwood at stock according to reviews. However, it is advertised as an overclockers board, and in my honest opinion it falls far short in that very area RIGHT NOW. My other concern is MSI seems to do stock boards fine, but they don't have a reputation for doing a good job with "enthusiast" boards. This definitely lowers my confidence that this board will be turned into a swan.
If the board grows up it should be outstanding, and I will take another look when it matures.