With our
test system sitting at idle and the same default software load I
tested the OCZ DDR3 Kit at the default XMP settings and at the
recommended JEDEC timings. With the XMP Profile #2 timings set to a
more aggressive 7-6-6-28 timings I expected to be at the end of the
envelope for performance on these modules. With the timings
loosened to the JEDEC #4 timings I was able to creep up to a
respectable 800 MHz and bumped the performance up a notch more. Sadly I was
unable to find a setup that matched the JDEC #5 settings that booted
over 800 MHz without resorting to changing the processor timings despite the recommended 857 FSB.
But lets not let that stop us!
|
The majority of testing will be done using
SiSoft Sandra XII Pro Business SP1. ClubOC has
been using SiSoft Sandra for many years now and it's
pretty much the main benchmarking program we use for
testing memory. The first four tests represent the
SPD settings shown bu CPU-Z with an 800 MHz FSB.
The 5th setting shows the actual JEDEC 5 setting at the
recommended ~450 FSB. After that we took the
timings up a notch to see what would happen..
|
|
Memory Speed / Bus Speed |
Timings |
|
XMS Profile 1 |
8-8-8-28 2T |
|
XMS Profile 2 |
7-6-6-28 1T |
|
JEDEC #4 |
8-9-9-23 2T |
|
JEDEC #5 (~450 MHz) |
9-9-9-26 2T |
Factory SPD Settings
|
Memory Speed / Bus Speed |
Timings |
Int
Buff'd |
Float
Buff'd |
|
1600 MHz/400 MHz XMS |
8-8-8-28 2T |
7324 MB/s |
7324 MB/s |
|
1600 MHz/400 MHz XMS |
7-6-6-28 1T |
7479
MB/s |
7511
MB/s |
|
1600 MHz/400 MHz |
8-9-9-23 2T |
7141
MB/s |
7159
MB/s |
|
1600 MHz/400 MHz |
9-9-9-26 2T |
7028 MB/s |
7099 MBs |
|
1700 MHz/425 MHZ |
9-9-9-26 2T |
7246
MB/s |
7240
MB/s |
|
1800 MHz/450 MHz |
9-9-9-26 2T |
7682
MB/s |
7700
MB/s |
|
1900 MHz/ 475 MHz * |
10-10-10-30 2T |
7956
MB/s |
7958
MB/s |
|
1850MHz/462 MHz |
9-9-9-26 2T |
7818
MB/s |
7811
MB/s |
Here you can see the default XMP
settings as tested by Sandra in the first two rows. The OCZ warranty covers
extended VDIMM voltages up to 1.95V without voiding the warranty
but I found with the XMP selected 1.9V was the highest setting needed
to boost the speed.
Based on the JDEC #4 timings I was able
to boost the DDR3 modules past the recommended 762 MHz setting
and go directly to a stable 800 MHz. Of course your
mileage may vary but the resulting numbers provided the fastest
results tested with the standard 800 MHz FSB.
To get the FSB to clock higher I set
the CPU and memory settings over to manual and adjusted the CPU
ratio. First I set the CPU to 6 x 425 to get the FSB up to
1700 for the JEDEC 5 test. I then bounced the FSB up 25
MHz at a time by dropping the multiplier to 6 x 450 and 6 x 475.
At 475 MHz the memory was only stable at a 2.0 Volt setting.
*Although the standard benchmarking was successful, an attempt
to prove stability failed with SuperPI at 32M. The highest
clock to pass our SuperPI stress test was the last set of clocks
at 462 MHz with a 6 x multiplier.
|
Speed |
Timings |
Read |
Write |
Copy |
Latency |
|
1600 MHz/400 MHz XMS |
8-8-8-28 2T |
8670 MB/s |
6251 MB/s |
7186 MB/s |
61.2 ns |
|
1600 MHz/400 MHz XMS |
7-6-6-28 1T |
9674 MB/s |
6302 MB/s |
7331 MB/s |
53.0 ns |
|
1600 MHz/400 MHz |
8-9-9-23 2T |
8879 MB/s |
6296 MB/s |
7208 MB/s |
59.0 ns |
|
1600 MHz/400 MHz |
9-9-9-26 2T |
8961 MB/s |
6292 MB/s |
7223 MB/s |
59.6ns |
|
1700 MHz/425 MHZ |
9-9-9-26 2T |
8883 MB/s |
6641 MB/s |
7584 MB/s |
61.2 ns |
|
1800 MHz/450 MHz |
9-9-9-26 2T |
9489 MB/s |
7037 MB/s |
8033MB/s |
57.7 ns |
|
1900 MHz/ 475 MHz * |
10-10-10-30 2T |
9551 MB/s |
7409 MB/s |
8370 MB/s |
56.9 ns |
|
1850MHz/462 MHz |
9-9-9-26 2T |
9657 MB/s |
7238 MB/s |
8210 MB/s |
56.1 ns |
The
Everest tests push memory in a bunch of different ways
offering a more complete picture of the DDR3 performance.
As you can see here the memory posts some of the fastest
numbers we have seen on DDR3. To my surprise some of
the Everest results beat the numbers posted by top
performing DDR2 4 GB kits from companies like Mushkin.
Despite the impressive numbers the Latency speeds are slower
than most DDR2. When combined with the Higher DDR3
timings these modules have a ways to go to totally replace
DDR2 in the fastest builds.
|
Speed |
Timings |
Score |
|
1600 MHz/400 MHz XMS |
8-8-8-28 2T |
5879 |
|
1600 MHz/400 MHz XMS |
7-6-6-28 1T |
5990 |
|
1600 MHz/400 MHz |
8-9-9-23 2T |
5778 |
|
1600 MHz/400 MHz |
9-9-9-26 2T |
5804 |
In
our PCMark05 Memory tests the
tightened timings resulted in a modest speed increase. With the
JDEC numbers we see an even more dramatic increase
of almost 20% on the nose! It should be readily
apparent by now that XMP is great but manual adjustments
can still provide the best return with these modules.
I tested the DR3 using the factory SPD settings for ease of
comparison to some of our older reviews as Futuremark has
replaced this test suite with the newer Vantage line.
|
Speed |
Timings |
Score |
|
1600 MHz/400 MHz XMS |
8-8-8-28 2T |
3179 |
|
1600 MHz/400 MHz XMS |
7-6-6-28 1T |
3340 |
|
1600 MHz/400 MHz |
8-9-9-23 2T |
3998 |
|
1600 MHz/400 MHz |
9-9-9-26 2T |
4073 |
To get a better picture of how overall
system performance is affected by the memory clocks, I
tested each SPD setup using the new PCMark Vantage. PCMark
Vantage is designed to pus modern builds to the limits by
simulating a broad range of common tasks and conditions. The
test results show a more gradual increase in performance
with each set of clocks but still a marked increase is
visible. This was the only test where my manual
overclock did not take the top score. Lastly, lets test the memory
using the 1M test in SuperPi mod 1.5XS for easy comparison.
|
Speed |
Timings |
Score |
|
1600 MHz/400 MHz XMS |
8-8-8-28 2T |
21s |
|
1600 MHz/400 MHz XMS |
7-6-6-28 1T |
21s |
|
1600 MHz/400 MHz |
8-9-9-23 2T |
21s |
|
1600 MHz/400 MHz |
9-9-9-26 2T |
21s |
|
1700 MHz/425 MHZ |
9-9-9-26 2T |
20s |
|
1800 MHz/450 MHz |
9-9-9-26 2T |
19s |
|
1900 MHz/ 475 MHz * |
10-10-10-30 2T |
18s |
|
1850MHz/462 MHz |
9-9-9-26 2T |
18s |
Super Pi is the result of raw computing at its best.
Bu pushing the components to the fastest computation
possible we can isolate system performance to the core
components. Our XMP profiles provided us with a
identical 21s score showing a solid baseline for overall
performance. These are the numbers you should expect
right out of the box with these modules. With our overclocked timings we were able to drop that score to a
fast 18s time, that's on par with the fastest DDR3 we have
tested here at the club. With newer systems based
around eth Quad Core Intel processors the 1M test is often
not enough to really stress the system So each
overclocked setting was tested wit the 32M test as well.
The last row shows the fastest overclock to pass this stress
test.
Conclusion:
The OCZ DDR3 PC3-12800 Intel XMP Ready Titanium Edition
Review modules have the first XMP profiles we have
tested here at the Club. I was pleasantly surprised by
just how easy it is to set up one of these profiles and get
stable performance right out of the box. As with all
high performance memory, the modules come clocked close to
the maximum speed already, but with some simple clock
timing changes I was still able to take things up another
15% in some tests! As with any overclocking setup your
actual performance will vary but history has proven OCZ
memory is a constant high performer.
DRD3 is still a new technology and the
addition of XMP profiles is even newer making the jump to
DDR3 an expensive prospect. OCZ prices these modules
at a pretty solid $385 as of the date we tested these
modules, making them a nice middle of the pack option when
it comes to price. However our testing shows that with
very little effort you can get the same performance out of
these modules as memory costing as much as $200 more!
Throw in the OCZ lifetime Warranty and you just might have
the right combination to make the jump to DDR3 and the Intel
X38 chipset.
