Date: April 5th, 2003
Article by: Nathan Glentworth
(Owner, Head Editor & Hardware Reviewer)
Product was donated by: OCZ Technology
<--Shop for this item
Throughout the ram testing, I used exactly the same components with the
only variance being the ram itself. The ram I am putting the OCZ memory
against a 512meg sample of Crucial PC2700 DDR333 and a 512Meg sample of
Mushkin PC3200 DDR400.
The testbed is comprised of:
Intel P4 2.4Ghz with 533Mhz FSB
Via P4PB Ultra Motherboard with Factory Bios
Ati All-in-Wonder 9700pro
Memory set at DDR400 unless testing the DDR333 from Crucial
512meg sample of OCZ PC3500 DDR433 or 512meg sample of Mushkin PC3200
DDR400 or 256Meg sample of Crucial PC2700 DDR333
Windows 2000pro SP3
Via Hyperion 4.46 Drivers
Testing procedures will entail benchmarking the memory at "By SPD" setting, which is governed by the rams onboard settings and the other
test will be to override the BIOS to CAS2 memory latency settings. The
OCZ will also be tweaked further with tighter memory timings which I will
state during the specific benchmarking. In order to test the memory at
maximum stability, the memory voltage was bumped all the way up to 2.8
volts to eliminate any power problems.
In this review, I will not be overclocking this memory and will be keeping
at stock values other than the memory latency settings and the settings
controlled by the rams onboard EPROM ram timing controller. You may ask
"why are you not overclocking?". Well, overclocking today is
dependent on many parameters such as cooling, CPU used, quality of power
supply, motherboard and motherboard chipset, PCI devices, bios etc...
I could go on and on. Having someone buying a product basing their purchase
on my findings and not being able to replicate the performance numbers
at home really makes me feel guilty and responsible. Nevertheless, when
you compare the overclocking population to the mainstream "non-overclocking"
population, the mainstream represents the majority of computer owners
in the world today. Don't get me wrong, I am an overclocking junkie, but
I can't be responsible for you "the reader" not being able to
duplicate the performance at home. With that out of the way, lets put
the modules to the test....
For benchmarking, I used two programs. One is SiSoft SANDRA 2002 Professional
"Memory Bandwidth" and the second is Madonion.com's PCMARK 2002
(Free Edition) focusing on the "Memory Score". Further on, I
will be benchmarking the OCZ exclusively to show you how tightening up
your memory settings will help your scores in 3dMark2001 and 2003.
SiSoft SANDRA Memory Bandwidth Module (Settings: By SPD)
Starting off, I benchmarked the OCZ memory using it's "By SPD"
memory timing. Everything ran smoothly and no instability was noted. Some
people will be thinking "It's timings are set to stock values, of
course it is stable". But, you have to remember that finding a stick
of DDR400 ram that is stable even at stock speeds can be a chore and only
a few are capable of it without ANY instability. Next off is the Mushkin
As you can see, the Mushkin pulled off a better result than the OCZ at
stock values. The stock CAS latency of the Mushkin is 2.5 when the OCZ's
is set at CAS2. Now this looks like a win for Mushkin even when the OCZ
is stock supposed to be faster due to its faster stock CAS2 latency setting.
BUT, even at stock settings, the Mushkin memory was having some stability
problems and some minor hiccups during normal computer use. Even though
it has a higher bandwidth, the Mushkin in my mind was beaten by the OCZ.
Stability is paramount and if a component seems to have a higher bandwidth,
it should utilize that increase in bandwidth without any instability.
And for comparative purposes, I threw in a benchmark of the Crucial DDR333.