There are many different brands of memory
out there, the big names like Crucial, Corsair, Kingston, OCZ,
and Mushkin, and many other smaller names. One of these
is Concord Idea Corp, a Canadian company that 'designs, manufactures
and distributes PC Memory', your standard memory manufacturer
for the last 19 years.
There are many different speeds of DDR2 memory
from the PC2-4200 to current high-end PC2-8000. Latency
has also become a valuable selling point as the JEDEC standards
for DDR2 only require 4-4-4 or 5-5-5 timings on the memory.
So lets see what Concord
Idea (SyncMax) has sent us.
Concord Idea was nice enough to send us their
dual channel kit of the 533Express RAM. The name of the
memory is pretty evident to the actual rated speed of the memory,
in this case 533MHz or PC2-4200. What does this RAM look
like?
Overclocking DDR2 RAM can have quite a few
variables, especially when you factor in the ability to adjust
the timings as well as the actual clock speed that the RAM is
running at, all without overclocking your CPU. Results as
always will vary but how did this memory do when we gave it 2.2v
to play with?
The results are pretty good considering the
ratings of this RAM, you can either look at it as a 200MHz increase
with the same timings, or a 70MHz increase while tightening the
timings of the RAM. Compared to our other test subject,
the Crucial Ballistix PC2-5300 1GB kit, it gives you a 40MHz faster
memory clock. A rather impressive overclock comparatively,
but it still can't quite reach 800MHz at any point. However
as a note the motherboard used for this testing seems to have
issues with trying to set DDR2 memory anywhere near the 800MHz
mark.
| CPU: |
Intel Pentium D 820 |
| CPU
Clock Speed: |
2.8GHz and 3.12GHz |
| Motherboard: |
Asus P5LD2-Deluxe BIOS 0603 |
| Memory: |
SyncMax 533Express |
Crucial Ballistix PC2-5300 (2*512MB) |
| Memory
Timings: |
3-3-3-12-1 |
3-3-3-12-1 |
5-5-5-15-1 |
| Memory
Speed: |
667MHz (QDR) |
| Hard
Drives: |
40GB
Maxtor 34098H4; 4*250GB WD SATA HD (RAID5)
|
| Video
Card: |
Gigabyte
GV-RX18L256V-B (ATi x1800XL) |
| Operating
System: |
Windows
XP Pro SP2 Direct X 9c |
| Drivers: |
Catalyst
6.4 |
| Cooler: |
Zalman
CNPS7700-AlCu (Full Speed) |
| Other
Fans: |
92mm Fan |
80mm Fan |
| Case: |
Generic Case |
| Power
Supply: |
Ultra
X-Finity 500w (Titanium) |
| Software: |
Sciencemark
2.0 Release: 21032005 |
XMpeg
5.03 |
|
UT
2004 3369 (Self made demo) |
DivX
5.21 |
|
7-Zip
4.42 (05-14-2006) |
TMPGEnc
2.59.47.155 |
All testing was run at 1600*1200
@ 32bpp and run a minimum of three different runs. The
Unreal Tournament test was run at 1024*768 with the lowest detail
settings, and no sound card present in the system. The
reason for using DivX 5.21 still is the newer versions do not
encode properly. The 7-Zip test was two different tests,
first was the benchmark mode, in which we made use of the 64MB
dictionary size as well as the multithread option. The
other 7-Zip test was a actual compression test, in which we
used the 7z format, ultra compression, LZMA compression method,
64MB dictionary, and also multi-threaded. Sciencemark
2 was our synthetic memory bandwidth test for which we used
the Membench test. XMpeg and TMPGEnc used the test settings
that I have used previously. So lets get on with our test
results.
Synthetic Tests
First lets get a look at the
'synthetic' tests we used, first is the memory bandwidth test
from Sciencemark, which will show the largest difference between
the RAM we tested.
| |
Crucial PC2-5300 |
SyncMax 533Express |
| Timings: |
5-5-5-15 |
3-3-3-12 |
3-3-3-12 |
3-3-3-12 |
| Memory Clock Speed (MHz): |
667 |
667 |
667 |
744 |
|
Results (MB/s):
|
4486.21 |
4505.95 |
4501.94 |
4906.80 |
What do these results show?
In the way of bandwidth timings don't make much difference,
but latency is reduced. The additional 70MHz in memory
speed show up in the form of a 400MB/s increase in bandwidth
or about a 9% increase in bandwidth for a 11% difference in
clock speed. Does our other synthetic test show the
same results?
7-Zip is an opensource compression
program that has the ability to work in a multi-threaded mode,
which makes sense for most of today's dualcore CPU's, but does
it stress RAM as well, as our test takes about 700MB of system
memory during the test.
| |
Crucial PC2-5300 |
SyncMax 533Express |
| Timings: |
5-5-5-15 |
3-3-3-12 |
3-3-3-12 |
3-3-3-12 |
| Memory Clock Speed (MHz): |
667 |
667 |
667 |
744 |
|
Results (MIPS) Compress:
|
2433 |
2495 |
2505 |
2735 |
|
Results (MIPS) Decompress:
|
1227 |
1230 |
1230 |
1369 |
|
Results (KB/s) Compress:
|
1397 |
1433 |
1439 |
1571 |
|
Results (KB/s) Decompress:
|
12167 |
12232 |
12200 |
13573 |
This results table contains
a bit more information than the last set of results.
All the numbers follow the higher is better format.
Most of the difference occur in the compression section, as
that is where the most memory is used (700MB compared to <100MB).
Changing the timings gives a 3% improvement in the number
of instructions that are able to be done. The two different
memory sticks show no real difference here, as the difference
is less than 0.5%. Overclocking the SyncMax memory gives
it a nice boost, though note the CPU is also overclocked here.
The difference is also 9%, just like that of the Sciencemark
test. Decompressing seems a more CPU bound environment
as the difference is 11%, which is how much the CPU is overclocked.
But these are synthetic tests, what do real applications show
in the way of a performance difference between these two sticks
of RAM?
Application/Game Tests
Lets continue our look at 7-Zip
with a real compression test, that of three folders (and many
subfolders) with a mixture of file sizes totaling 684MB.
Lets see how long the four different memory test setups take
to compress.
| |
Crucial PC2-5300 |
SyncMax 533Express |
| Timings: |
5-5-5-15 |
3-3-3-12 |
3-3-3-12 |
3-3-3-12 |
| Memory Clock Speed (MHz): |
667 |
667 |
667 |
744 |
|
Results (Minutes:Seconds):
|
9:54.78 |
9:31.37 |
9:32.22 |
8:54.57 |
Here the results are easier
to understand. The total difference between the four
setups is one minute, from ten minutes to nine. Changing
the timings gives a 4% boost, lowering the compression time
by 23 seconds. The two different sticks of memory show
less than one second of difference, nothing at all.
Overclocked the compression speeds up by 7%, which while not
as much as the synthetic test is still pretty good as this
test seems pretty memory limited. However we don't just
compress data, so lets take a look at Unreal Tournament 2004
to see the differences when playing games.
| |
Crucial PC2-5300 |
SyncMax 533Express |
| Timings: |
5-5-5-15 |
3-3-3-12 |
3-3-3-12 |
3-3-3-12 |
| Memory Clock Speed (MHz): |
667 |
667 |
667 |
744 |
|
Results (FPS):
|
112.457 |
114.021 |
115.452 |
126.410 |
Moving to tighter timings
doesn't do anything for this game, as the 1.3% increase is
within the margin of error. So too is the difference
between the two different companies memory sticks. Overclocking
the memory and CPU gives the only real increase of 9.5%, or
slightly lower than the overclock the CPU made. So while
this game is slightly memory dependent, as most games are,
the best way to see the very minor difference is by turning
down all the quality options as well as the resolution which
most of those who are worried about the differences don't
want to do. Will any difference show up in our encoding
tests?
First lets look at converting
a video to a MPEG-2 video file, using TMPGEnc. This program
is multi-threaded, which may use slightly more in the way of
memory bandwidth to the two cores but lets see if that is noticable?
| |
Crucial PC2-5300 |
SyncMax 533Express |
| Timings: |
5-5-5-15 |
3-3-3-12 |
3-3-3-12 |
3-3-3-12 |
| Memory Clock Speed (MHz): |
667 |
667 |
667 |
744 |
|
Real Time Taken (M:S):
|
6:22 |
6:23 |
6:22 |
6:15 |
|
CPU Time Taken (M:S):
|
7:35 |
7:32 |
7:34 |
6:49 |
|
Real Time FPS:
|
22.40 |
22.34 |
22.40 |
22.81 |
Total difference between
all four test setups 1.7%, or within the margin of error.
So this test doesn't really show any difference between timings
and an increase in memory clock speed, as the system was less
efficient when the system was overclocked. Lets move
on to converting from MPEG-2 to a DivX AVI file.
| |
Crucial PC2-5300 |
SyncMax 533Express |
| Timings: |
5-5-5-15 |
3-3-3-12 |
3-3-3-12 |
3-3-3-12 |
| Memory Clock Speed (MHz): |
667 |
667 |
667 |
744 |
|
Time Taken (M:S):
|
2:26.44 |
2:22.90 |
2:23.94 |
2:08.72 |
|
FPS:
|
58.42 |
59.87 |
59.43 |
65.95 |
This test seems to be a bit
more memory friendly. Changing the timings gives a 2.4%
increase in the rendering speed of the memory. Changing
memory without any other differences offers no difference
much like the compression test did. Increasing the clock
speed of both CPU and memory offers a 11% increase, much of
which can be attributed to the CPU though the increased bandwidth
does help somewhat.
Conclusion
So we've taken a look at these
two sticks of PC2-4200 memory, what can we say about it?
First packaging, it was fairly
basic with a clear plastic case that held both sticks of memory
as well as a little booklet to let you know what kind of memory
you have. The interesting thing about the actual sticks
of memory is that they had no heatsinks(spreaders) on them at
all, a rarity among 'enthusiast' memory. However this
didn't seem to hinder its overclocking potential as it easily
reached 744MHz at the timings it was rated at for 533MHz operation,
with only a 0.3v boost in voltage.
Performance was average, it
wasn't faster than the Crucial memory, but wasn't slower than
it. This is good as both were running at the same timings.
The difference between JEDEC standard SPD timings and the performance
of the tighter timings was noticeable in some tests, with at
most a 4% increase, but nonexistent in others. The performance
of the memory when overclocked was very nice, with a few differences
in the 9-10% range.
Lastly we will talk about the
price of this memory, which seems to be in the $100 range right
now, though I'm not sure if it is Canadian or US dollars, though
there isn't much difference at this point. The price is
within the range of most PC2-5300 memory, of which there are
some pretty good overclockers. So while this memory does
well and can provide low timings when overclocked, it doesn't
stand out from the large crowd its in. A good dual-channel
kit, and something to definitely consider for those who will
pick up an AMD AM2 chip.
If
you have any comments, be sure to hit us up in our forums.
HOME
