New memory standards don't come out
all that often, and when they do it takes a while for them to
be adopted. Take the move to DDR2 as an example. Its
been over a year since the release of DDR2 motherboards and CPU's
but still there is quite a few new systems that come with the
standard memory in them.
Crucial has earned a reputation as one of
the top memory companies in the world, simply because they have
solid products and good prices. However they have not always
kept up with the enthusiast communities desire for always faster
products. They have started to do this with their Ballistix
line of memory which has tighter timings and faster clock speeds.
Previously I have only used DDR memory in
my computers, as I haven't needed DDR2, even though I'm running
a LGA775 processor. However I finally got my hands on some
DDR2 memory, so lets see how it does.
Crucial Ballistix PC2-5300 CL3-3-3 Memory
Crucial kindly sent us a 1GB kit (2*512MB)
of their newest 'low latency' DDR2 memory. Before we go
on with the information, lets take a look at the RAM itself to
see what Crucial has done.
There isn't much that comes
with this memory, basically the memory itself, a simple manual,
and an offer for a free software program.
The RAM itself is packed individually meaning
that it might not be an actual matched pair, but two sticks of
RAM selected randomly. It might have been a matched pair
but in this case putting them in a combined static free container
would convey this better. Looking at the actual RAM
we see that it has a copper/gold coloring to the heat spreader
as many other sticks of 'enthusiast' memory have.
The 'free' software offer that is included
is more or less free, minus a delivery fee for the CD. This
really isn't free as you are paying a decent amount in shipping.
Crucial should have included it in the package or offered free
shipping on it like it does with this memory, in the US at least.
Installing the memory is as easy as usual
with either DDR or DDRII memory, line up the notches push the
RAM in until it locks in. For dual channel operation you
will need to take a look at your motherboard manual for which
slots the memory goes in but that is the extent of installation.
Overclocking
This RAM is rated at PC5300 speeds or 667MHz
(166MHz QDR) with timings of 3-3-3-12-1, which are pretty tight
timings for DDRII memory as this is basically what DDRI hits when
you are at PC4200 clock speeds. The memory made it to this
clock speed without a problem at stock voltage, so lets see how
high it can go when we give it 2.2v rather than its stock 1.8v.
We can see that there isn't much more headroom
here but it does make it to 700MHz with these tight timings.
Loosening the timings at all didn't help overclockability in this
case. However since most memory has rather loose timings
in comparison, does this help this memory perform better?
Lets see in our benchmarks.
System Setup
| CPU: |
Intel
PIV 2.8E 775LGA |
| CPU Clock Speed: |
2.8GHz |
2.8&2.94GHz |
| Motherboard: |
ECS 915P-A
|
Asus P5LD2-Deluxe |
| Memory: |
Corsair TwinX PC4000
(2*512MB) |
Crucial
Ballistix PC2-5300 (2*512MB) |
| Memory Timings: |
3-4-4-7-1 |
4-3-3-12-1 |
3-3-3-12-1 |
| Memory Speed: |
400MHz (DDR) |
533MHz (QDR) |
533 & 700MHz
(QDR) |
| Hard Drives (Master)
: |
80GB Western Digital 7200RPM SE 8MB Buffer
|
| Video Card: |
Asus
Extreme AX800XL-2DTV |
| Operating System: |
Windows
XP Pro SP2 Direct X 9c |
| Drivers: |
Catalyst
4.7 |
| Cooler: |
Swiftech
H20-120 REV. 3 Liquid Kit |
| Case: |
CoolerGuys
Windtunnel IV |
| Power Supply: |
RaidMax
400Watt Power Supply |
| Direct X Benchmarks: |
Unreal
Tournament 2004 (CPU timedemo) |
| |
Half
Life 2 - Anandtech Canals Demo |
| Other Benchmarks |
XMPEG 5.03 |
VirtualDub 1.6.4 |
DivX 5.21 |
| |
ScienceMark
2 Build 171102 |
SiSoft Sandra 2005
SR2a |
| |
TMPGEnc
Plus 2.59.47.155 |
Sysmark 2004 |
We used two motherboards here,
the ECS board was used to show the differences (if any) between
the higher frequency but somewhat slower timings of the DDRII
memory and the DDRI memory. The Asus board is the overclocking
board as the ECS gives absolutely no control over memory timings
in the BIOS.
We used two synthetic memory tests in
SiSoft Sandra and ScienceMark 2, these will give us an estimate
of any improvements that the memory should cause. Our video
tests are quite CPU demanding but can have some memory intensive
parts to them as well. Games can be memory hogs and we take
a look at two of them UT2004 and Half Life 2, both of which were
run at 640*480 at low detail to show the best that the memory
can do. Sysmark 2004 is your standard office software benchmark
to see if this will provide any difference in normal office tasks.
So lets take a look at how the DDRII memory fairs.
Synthetic Tests
First lets take a quick look at the synthetic
tests, in the form of memory tests. First up is ScienceMark
2 which provides a memory bench that shows you the theoretical
bandwidth available to the programmers who take time to write
proper code.
| ScienceMark Results |
Bandwidth (MB/s) |
| ECS 915P-A DDRI: |
4010.42 |
| ECS 915P-A DDRII: |
4312.75 |
| Asus P5LD2-Deluxe DDRII: |
4388.34 |
| Asus P5LD2-Deluxe DDRII (700MHz): |
4553.80 |
We can see that with almost
no clock speed increase, except the last result which is run
at 2.94GHz) there is a steady increase in available bandwidth.
Moving to DDRII gives a 7.5% bandwidth boost and overclocking
from 533MHz to 700MHz gives a 3.7% increase in bandwidth which
is nice but nothing to write home about. Does SiSoft show
similar results? Lets see.
| SiSoft Sandra Results |
Integer Bandwidth (MB/s) |
Floating Point Bandwidth (MB/s) |
| ECS 915P-A DDRI: |
4656 |
4658 |
| ECS 915P-A DDRII: |
4871 |
4897 |
| Asus P5LD2-Deluxe DDRII: |
4890 |
4983 |
| Asus P5LD2-Deluxe DDRII (700MHz): |
5149 |
5151 |
This test returns higher bandwidth
numbers, about 13-16% higher than the ScienceMark tests.
In this test the move to DDRII nets a 4.6% increase in bandwidth
while overclocking gives a 5.3% increase in bandwidth.
This is slightly backward from what the other test showed us
but as these are synthetic tests we don't put too much stock
into their results. Let us see if the remaining tests
follow either of these results.
Video Encoding Tests
Our first test involves getting data from
a MPEG-2 video stream and outputting it to a DivX based AVI without
any audio. We used XMpeg 5.03 which has specific instructions
for the various CPU types. So lets see if the RAM change
makes any difference.
What do these results show us? First
we will consider the move from DDRI to DDRII. There is a
1.9% 'difference' in this test, which is just within the margin
of error, and really shows no major difference at all. What
about overclocking the RAM from 533MHz to 700MHz. Here we
see a more significant difference, of about 6.5%. However
when we consider that the clock speed of the CPU is 5% higher
(210 vs. 200MHz) the difference shrinks to a minimal gain or none
at all. We've see that one video encoding program doesn't
need the higher bandwidth memory, without an accompanying FSB
increase, so lets see if VirtualDub shows any improvement.
We see that there is a total of 0.29fps increase
associated with moving to DDRII in this test compared to DDRI.
Overclocking gives a 5.2% increase or a clock speed based increase.
Lets move on to our last video test, that of MPEG-2 encoding using
TMPGEnc.
NEXT
