Here’s a letter I just got, the author raises some good points:
First of all, I would like to thank you for your excellent review of the
Iwill KK266. Due in large part to your review, I purchased an Iwill KK266.
Unlike some of the users on support.iwill.net, mine seems to running with
nary a hitch (well, my reset button has turned into an ‘off’ switch, but
that’s very minor). I agree about your emphasis on research and that the
few problems the board has seem to be common to all KT boards.
However, with the extreme overclocking capability of the KK266, and probably
all the KT133A boards, the AMD overclocker is now exposed to a new area of
learning and experimentation previously the exclusive domain of Intel
overclockers. I’m speaking of high FSB overclocking.
In the past, AMD overclockers had, frankly, a relatively easy and lazy life
compared to Intel overclockers. Oh, sure, we had to worry about
multipliers, voltage, and heat. But, really, we didn’t have to worry too
much about peripheral compatibility. Overclocking via the multiplier is
nice that way.
Now, in an effort to have the best combination of high CPU speed and high
FSB, we enter a much wider domain for potential improvements and problems.
As an example, I will be upgrading from my Mushkin Rev. 2 RAM. My Mushkin
will do (stably. . . it will go much higher unstably), 2-2-2 at 145 MHz and
3-3-3 at 150 MHz. Unfortunately, the potential of the Iwill is seducing me,
so I’ve recently purchased some new Kingmax PC150 to see what happens as I
approach 160+ MHz FSB.
According to Iwill technical support, the KK266 uses a divider of 1/4 FSB
for PCI and 1/2 FSB for AGP. This will give me a 40 MHz PCI bus and 80 MHz
AGP bus at 160.
Now, things I’ve noticed when Windows was running (barely) at 157 MHz,
My Linksys NIC seems not to like it. If I run the Sandra 2001 Burn In test,
which includes a network test, the NIC testing bombs sufficiently to hard
lock the whole machine.
Ed. note: Network cards are more than usual susceptible to failing when overclocked to any
significant degree. For people using broadband, this is going to be one of the prime suspects
if your machine doesn’t overclock much.
I *might* have been seeing the first hints of HDD data corruption. Either I
accidentally deleted some of Windows’ .WAV files (such as the recycle bin
sound), or data corruption did. Either way, they seemed to disappear. This
is enough to make me suspicious and scared.
Ed. note: It should.
I suspect that, at 40 MHz FSB, UDMA may stop working all together.
If strange things start to happen in IDE world, will moving to SCSI solve
the problems? Or does SCSI just create a whole new set of problems?
Ed. note: SCSI controllers can be very sensitive to overclocking. In the past,
more than a few wouldn’t overclock at all.
A few times when playing Quake III at 157 MHz FSB, the machine rebooted of
its own accord. This doesn’t happen at lower FSBs. I don’t know if this
caused by a limitation of the RAM, AGP bus, or something else. We’ll see if
it happens with new RAM.
In all my explorations of higher FSB, I have yet to increase the motherboard
IO voltage. I’m not even sure when it’s appropriate.
Ed. note: Increasing I/O voltage looks to be almost mandatory for scaling the heights. Wouldn’t increase more than 10%, though.
All in all, I think the KT133A chipset opens up a great opportunity for a
new article on the implications of FSB overclocking. Possibly multiple
articles, now that I think of it. Given the experience that I’m sure all
the Intel overclockers have in this area, I think this would be a great
cross-over articles for those on both sides of the AMDIntel fence…a sort
of detente, if you will.
Ed. note: We absolutely agree.
Well, in conclusion, I’d like to thank you for all your excellent efforts.
I find overclockers.com to be one of the most sensible, scientific, and
objective hardware sites on the net. Thank you for keeping your standards
Ed. note: Thank you!