Here you can see the
aggressive default clocks offer no wiggle room even with the
voltage set to 2.1 V. My only success at 1T offers a small
but noticeable boost in performance. By pushing the
voltage up again I might have obtained better stability at the
215 FSB level but 2.1 V is already pretty high for a factory
setting. Pushing voltages so high in DDR2 is one of the
ways you can do permanent damage to your system.
With the processor/FSB combination
acting as a virtual brick wall, I turned to loosening the
timings up and dropping the multiplier to 12x to see if I could
coax any extra speed out of these modules with a bit more
flexibility. The last set of numbers, marked with the 12x,
show the results of my tinkering. By dropping the
multiplier I was able to get the Mushkin Kit to an aggressive
265 MHz or 1060MHz speed. That's a jump to PC2-8500!
|
Speed |
Timings |
Read |
Write |
Copy |
Latency |
|
800Mhz/400(200MHz) |
4-4-4-12 2T |
8684 MB/s |
8053 MB/s |
8057 MB/s |
45.8 ns |
|
800Mhz/400(200MHz) |
4-4-4-12 1T |
8916 MB/s |
8051 MB/s |
8541 MB/s |
47.2 ns |
|
856MHz/428 (215MHz) |
4-4-4-12 2T |
9272 MB/s |
8594 MB/s |
8645 MB/s |
43.9 ns |
|
832/416Mhz (208MHz) |
4-4-4-12 2T |
9003 MB/s |
8331 MB/s |
8363 MB/s |
44.4 ns |
|
892/441Mhz (225MHz) |
5-5-5-18 2T (12x) |
9172 MB/s |
8713 MB/s |
8487 MB/s |
45 ns |
|
914/507Mhz (255MHz) |
5-5-5-18 2T (12x) |
9969 MB/s |
8659 MB/s |
9245 MB/s |
41.2 ns |
|
1045/523Mhz (260MHz) |
5-5-5-18 2T (12x) |
9939 MB/s |
8096 MB/s |
9568 MB/s |
39.9 ns |
|
1060/530Mhz (265MHz) |
5-5-5-18 2T (12x) |
10095 MB/s |
9047 MB/s |
9722 MB/s |
39.3 ns |
The
Everest tests push memory in a bunch of different ways
offering a more complete picture of the DDR2 performance.
As you can see here the memory posts some impressive numbers
for a 4 GB kit. Even at the highest stable clock
speeds the numbers are a nice increase over stock.
Despite the solid numbers the Latency speeds are slower than
most of the DDR2 we have tested at these speeds. The
12x scores show a small but measurable speed boost with no
loss of stability as well. Next up is the PCmark05's memory test.
|
Speed |
Timings |
Score |
|
800Mhz/400(200MHz) |
4-4-4-12 2T |
5191 |
|
800Mhz/400(200MHz) |
4-4-4-12 1T |
5150 |
|
856MHz/428 (215MHz) |
4-4-4-12 2T |
5579 |
|
832/416Mhz (208MHz) |
4-4-4-12 2T |
5651 |
|
892/441Mhz (225MHz) |
5-5-5-18 2T (12x) |
5610 |
|
914/507Mhz (255MHz) |
5-5-5-18 2T (12x) |
5740 |
|
1045/523Mhz (260MHz) |
5-5-5-18 2T (12x) |
5850 |
|
1060/530Mhz (265MHz) |
5-5-5-18 2T (12x) |
5899 |
As shown by the above chart, the increased FSB
speed provided a dramatic improvement in overall performance
with the Mushkin kit modules. With the 208 MHz clock
speeds alone, I clocked an almost 10% increase in the
results. My first attempt at the 12x multiplier was a bit of
a disappointment but the scores soon began to move again. Lastly, lets test the memory
using the 1M test in SuperPi mod 1.5XS for easy comparison.
|
Speed |
Timings |
Score |
|
800Mhz/400(200MHz) |
4-4-4-12 2T |
30s |
|
800Mhz/400(200MHz) |
4-4-4-12 1T |
31s |
|
832/416Mhz (208MHz) |
4-4-4-12 2T |
29s |
|
856MHz/428 (215MHz) |
4-4-4-12 2T |
Fail |
|
892/441Mhz (225MHz) |
5-5-5-18 2T (12x) |
29s |
|
914/507Mhz (255MHz) |
5-5-5-18 2T (12x) |
28s |
|
1045/523Mhz (260MHz) |
5-5-5-18 2T (12x) |
27s |
|
1060/530Mhz (265MHz) |
5-5-5-18 2T (12x) |
26s |
Here is where we illustrate
the failure point. With any set of tests the product
is only as fast as the last settings that passed every test,
so consider the 208 MHz overclock the fastest stable speed
this time out with the recommended settings. Here at
the club that is just not good enough! With only a
slight jump to 2.2V we were able to keep the 1060 MHz speed
stable through the testing and peaked at a nice 6 second 1M
Super Pi score. That's pretty good for a 2 GB
kit but for a 4 GB kit these are great numbers.
Conclusion
The Mushkin XP2-6400 4 GB Kit modules have some pretty
aggressive timings right from the factory so I was not
surprised to see so little headroom for additional overclocking
by tightening the timings down. Still you have to admit a 4 GB kit with
4-4-4-12 timings is pretty good already. Our testing is further limited by
the ABIT BIOS peak of 800 MHZ for the memory speed
forcing us to use a more basic FSB overclocking method to
push the limits of the whole system together. I was able to
improve the performance
by about 10% and
keep things stable for the duration of testing and a couple
of days of gaming at the 4-4-4-12 settings but with some
simple system overclocking the Muskin kit reached a terrific
PC2-8500 level speed with no instability.
Currently the 4 GB kit is
out of stock on both the Mushkin website and all of the
suggested retailers. But a single 2 GB stick can be
had for about $94 making the estimated street price for this
kit under $200. With so little competition in the
performance 4 GB kit market, Mushkin stands out for making
some attractive DDR2 that performs great right out of the
box.
