|

Judging from what I've found online, Computocx has
recently upgraded the heatsink on the CEC line of memory to the same
type they use on their ultra high-end Diablo models. This was
largely in response to the many suggestions from European tech-site
that had many favorable things to say about Computocx's DDR2-800
memories. They just wanted a little more in the aesthetics
department.
.jpg)
Compustocx answer is blend of style and substance. A
very large and heavy-duty aluminum heatsink which almost completely
encases the memory module. Heat transfer is aided by the large slots
machined into the top of the heatsink.
.jpg)
Considering the placement of most motherboard memory
slots, this is a good thing since there isn't a lot of airflow in
that particular spot in a standard ATX system. However, with the
development of newer heatsinks by companies such as Cooler Master, a
lot of forethought should go into what motherboard, CPU cooler, and
memory combination you intent to use. Memory with a heatsink as
large as this won't fit with their new V10 style CPU cooler.
Test System
| CPU: |
Intel
Core2Duo E8400 |
| CPU Cooling: |
Thermalright Ultra-120 "TRUE Black" |
| Motherboard: |
Gigabyte EP45-DQ6 |
| Hard Drive:
|
2x Hitachi 7K1000 750GB SATA HD
ICH10-R RAID-0 |
| Optical Storage: |
Samsung
DL-DVD +/- R |
| Graphics GPU: |
Sapphire
HD4850 Toxic |
| Power
Supply: |
Tuniq Ensemble 1200watt |
| Enclosure: |
Ultra m998 |
| Operating System: |
Windows 7 32-bit
beta-1 "Build:7000" |
After getting the modules acquainted into the test
platform, it's time to run a few tests. As usual, a few BIOS
settings had to be manipulated to get the system to POST and boot
with the CEC modules running at 1066 Mhz. Along the way in testing,
we'll have to shift the CPU multiplier and the memory dividers to
ensure any instability is caused by the memory and not the CPU.

SiSoftware's Sandra will be used to gain synthetic
benchmark data from the system's Integer and Floating point
bandwidth. We'll also use this section to relay what failed at boot
time and what crashed during benchmarking and stability testing.
3600 Mhz CPU Speed
Memory
Bus Speed
(MHz) |
vDimm
(volts DC) |
CAS Timings
(CAS-tRCD-tRP-tRAS) |
Integer
(MB/s) |
Floating Point
(MB/s) |
| 533 |
1.86v |
5-5-5-16 |
8068 |
8080 |
| 533 |
2.00v |
4-4-4-12 |
FAIL |
FAIL |
| 533 |
2.20v |
5-4-4-14 |
FAIL |
FAIL |
| 533 |
1.86v |
5-5-4-14 |
8116 |
8133 |
3717 Mhz CPU Speed
Memory
Bus Speed
(MHz) |
vDimm
(volts DC) |
CAS Timings
(CAS-tRCD-tRP-tRAS) |
Integer
(MB/s) |
Floating Point
(MB/s) |
| 550 |
1.86v |
5-5-5-16 |
8201 |
8311 |
| 550 |
1.90v |
5-5-4-14 |
8238 |
8347 |
3638 Mhz CPU Speed
Memory
Bus Speed
(MHz) |
vDimm
(volts DC) |
CAS Timings
(CAS-tRCD-tRP-tRAS) |
Integer
(MB/s) |
Floating Point
(MB/s) |
| 570 |
2.1v |
5-5-5-16 |
FAIL |
FAIL |
| 570 |
2.2v |
5-5-5-16 |
8387 |
8461 |
3638 Mhz CPU Speed
Memory
Bus Speed
(MHz) |
vDimm
(volts DC) |
CAS Timings
(CAS-tRCD-tRP-tRAS) |
Integer
(MB/s) |
Floating Point
(MB/s) |
| 573 |
2.4v |
5-5-5-16 |
FAIL |
FAIL |
The end of the line came when I tried to ramp up the
memory speed to anything beyond 570Mhz. No amount of north bridge
voltage or vDIMM would stabilize the system enough to complete a
benchmark routine, let alone a Prime run.
|