I
am using the main testbed here in the dungeon to test
this RAM, the specs are as follows:
Chaintech
7NJS nForce2 based motherboard
Athlon XP2700+ 333FSB CPU
400W Q-Power Gold series PSU
GeForce FX5900 graphics card
As
mentioned earlier, I will be comparing the Kingston HyperX
to a pair of 256MB (512MB total) XtremeDDR 400true DIMMS
in two configurations; single channel and dual channel.
I am choosing to run both configurations because performance
can be markedly different especially where bandwidth and
latency are concerned. The second reason for doubling
up on benchies was that not everyone has the capacity
to run dual channel, so I only thought it fair to provide
both sets of results. Since the nForce2 chipset is happiest
in dual channel when the RAM is in sync with the CPU,
I will run the dual channel tests at 166MHz. I will run
the single channel tests at 200MHz for the RAM, and 166
for the CPU because those are the specs for the motherboard.
Then we will overclock the FSB when I have established
basic performance to seperate the men from the boys.
One
point to make is that the XtremeDDR is capable of very
low latency (2-2-2-2 CAS2) because it is rated at 400MHz,
however, it takes at least 3 Volts to do so; I have been
running it at 3.2V for nearly a year with no hiccups at
all. In order to squeeze the extra speed out of it, the
Kingston is rated at 3-4-4-8 CAS3 and can do so on 2.6V.
In order to truly test performance differences, I ran
all tests at the fastest latencies that I could muster.
The other point along these lines is that I am comparing
a 512 dual channel kit to a 1GB dual channel kit. While
it may not have that much difference in Science Mark and
SANDRA bandwidth scores, 3DMark01 and PCMark04 should
run smoother with the extra 512MB. Just something to point
out.
That
brings me to the suite of tools that I will be using to
test with.
SiSoft
SANDRA (memory bandwidth tests)
Science Mark (Membench tests only)
3DMark 2001 SE
PCMark 2004
OK..
lets get into it…
For
starters I will show performance in SANDRA bandwidth tests.
For those of you unfamiliar with SANDRA, here is a brief
description copied from their website:
|
SiSoftware
Sandra is a Windows 32/64 system analyser that includes
benchmarking and testing modules. It works along
the lines of other Windows utilities, however it
tries to go beyond them and show you more of what's
really going on. Giving the user the ability to
draw comparisons at both a high and low-level. You
can get information about the CPU, chipset, video
adapter, ports, printers, sound card, memory, network,
Windows internals, AGP, ODBC Connections, USB2,
Firewire etc.
|
And
the results of the Memory Bandwidth test...
This is interesting… as a dual channel kit at 166MHz (333DDR),
the XtremeDDR pulls ahead, but in a single channel configuration
at 200MHz (400 DDR) the Kingston has an advantage. Also
note the performance gains to be had when running dual
channel versus single channel. The timings may have something
to do with this, or the Kingston just might like single
channel better. I was concerned that I made an error in
the BIOS setings, so I checked everything and reran the
tests twice with roughly the same results.
Here
is a brief description of ScienceMark, which is our next
test.
|
Science
Mark 2.0 is an attempt to put the truth behind benchmarking.
In an attempt to model real world demands and performance,
SM2 is a suite of high-performance benchmarks that
realistically stress system performance without
architectural bias.
|
And
here are the results of MemBench...
Again, we see the difference in running single and dual
channel with the Kingston favoring a single channel configuration.
Now for you Futuremark fans out there, we'll look at PCMark
04 and 3DMark 01 scores on the next page.
