|
Processor |
Intel Pentium-4 640
800Mhz FSB / 3.2 GHz Clock
2MB L2 Cache |
|
Motherboard |
GigaByte GA-8N-SLI Roayl |
As
we've just seen in the GigaByte motherboard review, the nForce4 chipset
allows for some serious customization on the topic of RAM front side bus
speed. With most chipsets, you have three settings to choose from at a fixed
ratios. As you move the Front Side bus of the processor, you take your RAM
along with it, and you'll have to pick between the available ratios. On most
Intel based chipsets, those ram dividers are pretty far apart, so you can't
truly maximize the systems performance once you peak out the bus speed of
your memory. With the nForce4 chipset, you have a much higher degree of
fliexibility, because it allows you to enter in the front side bus speed you
want. At that point the BIOS will make the proper ratio adjustment, but the
final bus speed is very close to the value you entered. We'll see more of
this once we begin testing.
.jpg)
200 MHz Processor Bus Speed Test
| Memory Bus Speed |
Latency Timings |
Voltage |
Sandra ALU |
Sandra FPU |
Science Mark 2.0 |
| 533 MHz |
3-3-3-6 |
1.9 v |
4910 |
4902 |
4548 MB/s |
| 600 MHz |
4-4-4-8 |
1.8 v |
4900 |
4907 |
4348 MB/s |
| 667 MHz |
4-4-4-8 |
1.8 v |
4859 |
4854 |
4395 MB/s |
| 685 MHz |
4-4-4-8 |
1.9 v |
4907 |
4912 |
4568 MB/s |
| 700 MHz |
4-4-4-8 |
2.0 v
|
4915 |
4910 |
4568 MB/s |
| 710 MHz |
5-5-5-15 |
2.1 v |
4882 |
4888 |
4548 MB/s |
Setting
the CPU to its default Front Side Bus speed of 200 doesn't let the 2MB of L2
cache have much effect on how the test goes. These tests look very similar
to the scores that were posted with a Pentium 4 -560 with only 1MB of L2
cache. By increasing the Vdimm I was able to achieve both lower latencies at
the lower bus speeds as well as an overall higher bus speed. Much higher
than the rated 667MHz. Both SiSoftware Sandra and Science Mark 2.0 were used
to benchmark memory progress.
233 MHz Processor Bus Speed Test
| Memory Bus Speed |
Latency Timings |
Voltage |
Sandra ALU |
Sandra FPU |
Science Mark 2.0 |
| 531 MHz |
3-3-3-6 |
1.9 v |
5627 |
5629 |
5233 MB/s |
| 598 MHz |
4-4-4-8 |
1.8 v |
5650 |
5641 |
5221 MB/s |
| 664 MHz |
4-4-4-8 |
1.8 v |
5702 |
5693 |
5286 MB/s |
| 675 MHz |
4-4-4-8 |
1.9 v |
5675 |
5672 |
5272 MB/s |
| 699 MHz |
4-4-4-8 |
2.0 v
|
5715 |
5708 |
5307 MB/s |
Here we
see the effects of the memory bus at different speeds after we increase the
CPU front side bus speed to 233, which has the CPU operating at 3.7GHz. We
can also see how the nForce4 is trying to accommodate the request made in
the BIOS to set the memory bus speed. While these speeds are not set
dead-on, thanks to the overclocking of the CPU Front Side Bus, there are
plenty of speed values to choose from since we don't have to be content with
whatever the dividers drag the memory bus speed along to. The setting in the
top tests correlates to the setting attempted in the bottom chart. There is
no test completed after 700Mhz because the tests failed beyond the 699Mhz
speed mark. At 700 MHz we have overcome the higher latency required to run
at this bus speed as all tests have pulled away from lower speed settings.
Conclusion.........
Gold DDR2 represents the third rung in OCZ's
performance series memory. The current retail price for a 1GB Dual-Channel
kit (2x 512MB) is very reasonable. Especially once you consider that the
modules have a lifetime warranty and their potential to perform well beyond
rated specifications. The test results achieved here should be pretty
typical from system to system. I tested this ram very briefly on an Intel
925 chipset equipped motherboard. With a 3.4GHz Pentium-4 540 overclocked
with a 222 Front Side bus, the RAM was clocked at 667 with 3-4-4-8 timings.
The end result with Sandra was 5077/5073. Not bad considering the 1MB cache
of the 540 series processor. As
DDR2 continues to mature, I fully expect to see OCZ Technology leading the
way in performance series modules. They have always been one of the
overclockers best friends.
.jpg)
