This is essentially a stripped-down server board minus two memory slots and a SCSI controller.
You want to see a picture and read specs? Go HERE.
Like other “AMD” chipsets, this is actually a hybrid; it uses an AMD 761 northbridge and a Via 686B southbridge.
As usual, MSI has some humongous tin soldier capacitors guarding the CPU socket. With the socket A being north, and the PCI slots
being south, you have about a half-inch clearance south of socket A for your heatsink/fan before you’d bang into a tall tin soldier, and about 7/8″ from a shorter one to your east.
Five PCI slots. More importantly for some, only two DDR slots on the K7Master: biggest module it can handle is 256Mb. The K7Master-S has four DDR slots and a SCSI controller for another $100.
Passive cooling on the Northbridge, down the road, you might want to put a fan on. Three fan sockets.
Floppy and HD connectors are placed unconventionally; they’re placed in positions parallel to the PCI slots rather than the typical parallel to the memory slots. Maybe a bit more annoying to use.
There’s a relatively minimal number of jumpers on the board, usually for those items you’d find on any motherboard. The overclocking functions are found in the BIOS.
The CMOS jumper is located just to the west of PCI slot #5. Not convenient, not a nightmare. However, you’ll probably not have to get to it too often for two reasons.
First, the board isn’t too fussy about settings. The only time I had to clear CMOS was when I didn’t give the CPU quite enough voltage.
Second, MSI has a handy new feature meant for overclockers. It’s sort of a CMOS jumper that just resets your multiplier/FSB. It’s located right by the socket A. Just take it off for a few, then put it back on. A bit easier for at least the fumble-fingered than the CMOS Kama Sutra moving from position 1-2 to 2-3.
For the future, it has a 12V connector for its AGP Pro slot. Its voltage controller goes down to 1.1V; so future Palomino owners should not worry. However, its hardware system clock only goes to 150Mhz/300Mhz; the Via system clocks appear to be able to go higher.
You will use BIOS 1.12. If you won’t use a beta BIOS, then don’t buy this board yet. Under version 1.00, you can’t save multiplier/FSB settings that will survive a cold boot. The multiplier and FSB adjustments work and save fine with BIOS 1.12. You can get that here.
In general, the BIOS is ugly, but it works. This is pretty important because MSI doesn’t seem to be in any rush to come out with new official BIOSes for it, unlike its Via cousin, where we’re getting one every week.
If you change your BIOS settings (as in any change), the CPU multiplier generally goes from Default up to 12.5. This is not what people with high-end 200Mhz FSB CPUs want to hear, but you can take care of that by clearing CMOS. See Tako-Chu’s comments posted May 03, 2001 09:30 AM for more.
There’s a few cosmetic quirks. When you’re running at 150Mhz, the BIOS shows 149Mhz. WCPUID shows it’s running at 150Mhz.
The CPU voltage settings aren’t particular accurate; set the voltage, and the voltage monitor gives you .1V more. Not an unusual problem, but .10 above is on the high side.
As of the moment, there is no I/O voltage setting. For the moment, this doesn’t present a practical problem, since the 2.5V reading is actually 2.69V, which suits at least Crucial DDR fine,
IRQ sharing theoretically occurs as follows:
AGP and PCI2 shared
PCI1 and PCI5 shared
PCI3 and AC97 shared (also SCSI1 if you buy the full-blown Master-S)
PCI4 and USB shared (also SCSI2 if you buy the full-blown Master-S)
Does this really happen? No. ACPI continues the bad habit also found in Via boards of clumping everything into IRQ11 in Win2K. Go here under the “Uninstalling ACPI” sections to find out how to get rid of it.
I think it’s safe to say it’s not very unstable, but I’m getting quirks.
The most common one is a refusal to boot into Win2K, stopping just short of the Desktop. Rebooting solves the problem, but it really shouldn’t do that. I’ve had a few other sporadic problems like that.
On the other hand, it did run Prime95 at 9.5X150 for fifteen hours with no problem.
The processor itself was capable of running Prime95 OK at 1430Mhz on an IWill KK266, so I don’t think the board is limiting it. The chip had no problem getting into Windows or running Sandra benchmarks at 1500Mhz, which was not the case with the same chip and the KK266. It also was able to run Prime95 before failing at 1500Mhz a bit longer than the KK.
There’s certainly alternate possibilities that something else besides the mobo could be causing these small difficulties, some use will give me a better idea how often they happen and how serious they are.
I’ve generally found that problems tend to emerge after some use, so it’s too early to call it wonderful yet. Promising, yes, but for some strange reason, my daily routine manages to crash most systems better than most benchmarks. Most systems, not all.
We shall see.
We’re going to be comparing this to the A7V133 next week, running at a variety of FSBs and memory settings.
There will be two purposes to the review. First, of course, to tell you it does, and does compared to a KT133 board.
Secondly, though, we’re also out to show how much (or little) memory and FSB settings have on overall performance. We’ll also show what a variety of tweaks do (or don’t do).
As time goes on, when new competition shows up, we’ll try to show in the same fashion how they all do against each other under a variety of conditions.
Hopefully, this will give both you and us a deeper understanding of what works and what doesn’t with this equipment.