Chaintech 7VJL (VIA KT333) Page 4
| Date: 01/10/2002 Author: Hark |
Chaintech 7VJL (VIA KT333) |
Overclocking
and Stability
The board was working
reliably during testing with different "hard" applications. A qualitative
3-channel power transformer helped to achieve this. It uses six condensers of
2200uF each and three condensers of 1500uF.
After I have received
the board, I noted a very broad scope of overclocking capabilities. The board
has an ability to change FSB clock in the range from 100 up to 200 MHz with
1 MHz step increment. A user can also change CPU clock multiplier from 5 to
12 with a step increment of 0.5. Voltages on all necessary components can be
varied.
Firstly, I'll tell
about CPU voltage. Vcore is available in a range of 1.1V - 2.0V. The step increment
equals 0.025V inside 1.1V - 1.85V interval and 0.05V inside 1.85V - 2.0V.
It is implied that
voltage increase over 1.85-1.9V is possible with good cooling equipment, i.e.
a cooler with copper heatsink.
Secondly, memory
voltage can be set to 2.65V, 2.8V, 2.95V or "Default" value.
And finally, a
user can increase AGP bus voltage and I/O voltage (Vagp and Vio respectively).
The possible values for Vagp are "Default", 1.65V, 1.8V and 1.95V.
For Vio these are "Default", 2.65V, 2.8V and 2.95V.
So, overclocking "stock-in-trade" is very broad. Let's see how it
works in practice.
Maximal FSB clock
was fixed at 159MHz. At this I used both of my processors (Duron 700 and Athlon
XP 1600+), which can work well at 166MHz and higher. Even when I increased the
voltage up to 2.0V, my Athlon XP 1600+ could not reach 166MHz. However Duron
700 could work at 1162MHz (7x166) at the voltage of 2.0V and even passed entire
3dmark (on Epoxx 8K3A it could work so at 1.9V). But Serious Sam SE shooter
in minimal resolution (640x480) used to hang or terminate with an error. That
is why I can't qualify this mode as a stable one. In overclocking test I used
Samsung DDR440 (PC3200, CL3) memory, which worked well in all possible modes
(also as PC2700 with CL2 - 2-5-2, 4Way, 1T timings.
Once again I return
to 166MHz FSB clock. All my attempts to increase FSB clock were made with JP31
jumper set to OFF. As the result, AGP and PCI frequencies differed from standard
values (sometimes they used to reach 86/43). My attempt to set JP31 to ON, i.e.
to set base frequency to 166MHz and to use 1/5 divisor (PCI/CPU), lead to boot
failure. Certainly I used processors, which do not imply such FSB clock and
it is possible that new Athlons XP with FSB clock of 166MHz will work well.
But I doubt in it.
I have tried to
used Chaintech Overclocking utility during testing, but very soon I stopped
doing it. Firstly, the utility used to overestimate CPU and FSB clocks and,
secondly, it contained small (and sometimes big) bugs. Well, this is a beta
version now and I think that Chaintech programmers will soon fix all the bugs.
Here are some first conclusions. Despite rather powerful overclocking tools,
the results of overclocking proved to be very weak.
