Neoseeker : Articles : Memory : DDR : Corsair TWINX2048-4000PT
Hardware Newsletter:
Email:

News Headlines
New Articles

Compare Prices

Motherboards
Abit
ASUS
Gigabyte
MSI
eVGA
Intel
Tyan
More...

Processors
AMD
Intel
More...

Memory
DDR
DDR2
DDR3
More...

Video Cards
ATI
eVGA
XFX
BFG
Sapphire
More...

search for lowest prices

send article   hardware newsletter   article comments (1)
Corsair TWINX2048-4000PT - PAGE 6
William Henning - Tuesday, November 1st, 2005

Doom 3

Doom 3, in low resolution with fancy effects turned off makes a pretty good processor and memory benchmark as it is quite sensitive to processor speed and even memory bandwidth (as long as the resolution and graphical feature set is kept to the minimum).

Take a peek here... the XMS4000 dominates this chart, and is a whopping 35fps faster than at stock settings; granted the processor is also overclocked like crazy... but it is still a very impressive gain.

Tweaking Memory

I was very pleasantly surprised by the XMS 4000.

It was rated for 3-4-4-8-1T at 250MHz on the Corsair web site.

But it significantly outperformed its rated specifications, as it was stable at these additional settings:

  • 3-4-4-8-1T at 265MHz (DDR 530)
  • 3-4-4-8-2T at 275MHz (DDR 550)
  • 3-4-4-8-2T at 285MHz (DDR 570) might have been stable, it was probably the CPU that was unstable

I used MemTest to check for stability at various timing settings; then tried the benchmarks (Sandra, RightMark, WinRAR and Doom3) to confirm stability. 285MHz was not totally stable, but it was probably due to the processor not wanting to run at 3.025GHz - after all, it was only rated for 2.2GHz!

The processor was run at a 1.55V core voltage, the memory at 2.9V and the LDT multiplier was set to 4 for 265MHz and 275MHz, but to 3 for 285MHz.

Conclusion

The Corsair XMS 4000 PC4000 memory performed very well, quite frankly, well above my expectations.

As you can see from the benchmark results, running the memory at a high data rate can make quite a difference for applications that are memory bandwidth limited; however it is also important to keep the memory latencies under control, as sometimes lower latencies at a lower memory bus speed will outperform higher bus speed memory with higher latencies.

The boundary lies in how the memory is used - if it is mostly reading large chunks of contiguous data, the higher memory speed will dominate; if it is writing a lot or reading in a non-sequential manner, the latencies will tend to matter more than the raw bandwidth.

The XMS 4000 tries to strike a nice balance between high memory speed and low bandwidth, and manages to do so quite nicely for most circumstances. Of course I wish it was able to run say at 2.5-3-3-7-1T, but that is not realistic at this time, as very few memory chips would bin at high enough speeds to allow that.

Recommended.

Recommended

What's Next?

Article Index

1.Introduction
2.Benchmark Setup and Benchmarks
3.Sandra Results
4.RightMark Memory Bandwidth
5.RightMark Read Bandwidth
6.Doom 3, tweaking and Conclusion

Submit our article to: diggDigg this! de.le.ciousdel.icio.us

Get updates when we publish new articles
Email Address:
(0.1551/d/ascension)