Hi guys,
here the original link to the italian review of this new DDR3 kit from TG.
http://news.tecnocomputer.it/forums/...ad.php?t=13166
For you, here a brief translation, hope you like it
-----------------
Images and specifications
New packaging, in a trasparent plastic box, increased in size:
Inside a plastified guide with some advices for a correct installation.
On the back, the news: a brief summary of great scorse achieved by TeamGroup memories. Even if the scores are not really updated, the basic concept remains: TeamGroup is leader in memory production and shows it every overclock session, everywhere in the world.
Also on DDR3, Teamgroup is faithful to its simple and essential design, still usign those “light” heatsink. At a first sight they might look inefficient, but the manufacturer adopts this solution since DDR1 series, so we think this is a well known and efficient technology.
TeamGroup also changed the look of the modules: finest painting and mat grey, coupled with bronxed tonality give a new style to the new product.
On the old-style label, frequencies, timings and voltages for default conditions.
800 Mhz and timings 7.7.7.21 still let us a bit astonished: we took some time to pass to DDR1 to DDR2 and move from timings like 2.2.2 to 5.5.5
But we are not so scared about this: experience has showed us that the overclock capability of the Xtreem series shoud allow us to reach great results, as compromise of high frequency and low timings. An example from a recent past: Teamgroup PC2-9600 kit, workin at 600 Mhz 5.5.5 at default, have in certain case reached almost DDR2-1400 5.5.5 (playing with frequencies) or DDR2-1200 4.4.4 (playing with timings).
Test configuration
Mainboard: Asus P5K3-DELUXE / Asus P5K-E
CPU: E6700 / E6800
PSU: Enermax Galaxy 850W DXX
VGA: Club3d Radeon 2900XT (Crossfire Setup)
RAM: 2x1gb TeamGroup DDR2-1200 5.5.5 / 2x1Gb TeamGroup DDR3-1600 7.7.7.21
DDR3 Max Frequency
In this first part, we tried to study and understand this new DDR3, trying to find out the limits with different combination of frequency and timings.
So we have:
Played with straps and memory divider, varying the FSB (and so the CPU frequency, always below 3500 Mhz).
Tried to optimized timings, lowering the frequencies in order to stay close to the actual best frequencies for DDR2 product (near to 700 Mhz DDR2-1400 but commonly 600 Mhz DDR2-1200).
Increased the frequency relaxing the timings, moving beyond the default value (800 Mhz, DDR3-1600).
In this analysis, voltage has beed set to 2.05 V, already a +10% overvolt, keeping the modules constantly cooled by a 120mm fan.
For the maximum frequency we based our investigation on pi1M test.
DDR3 coupled with Intel P35 chipset is actually the best choice. It’s quite natural and easy to move between the memory dividers, and finally we can use also 5:8, 3:5 and 1:2 divider.
In the middle of the Y scale, we find the 955 Mhz with default timings (7.7.7). It’s +20% overclock, with a +10% overvolt.
Teamgroup modules confirm their typical behaviour, moving easily both to high frequencies, and sharp timings (7.6.6.).
We also forced the to work as a DDR2 kit, pushing them to run at 630 Mhz and 5.5.5 timings.
On the other side, realaxing the timings, we didn’t increase the frequencies. Anyway other dual channel kits we have tested have been able to reach higher values as the following screeshots show: real a high value for pi1M benching and the validation at 1000 Mhz (DDR3-2000) barrier.
One note for all the tests: these frequencies have been reached only working on subtimings, using Memset. Probably the Asus P5K3 mainboard was not capable to set correctly this parameters and limited the maximum achievable frequency.
Comparison tests @ 3600 Mhz
Once we found out the limits for the chosen timings set, we decided to make a brief comparison with the “old“ DDR2 modules.
CPU frequency has been kept constant and we played on FSB and memory divider to test some configuration
Basically there is no overlap between DDR2 and DDR3: DDR2 stop working at about 600 Mhz, even if in this review we played with a “particolar” DDR2 kit. This is a underclocked condition for DDR3, which runs at 800 Mhz and higher. We forced them to run as a DDR2 module in order to evaluate their behaviour.
Here the comparisons based on different sintetic benchmarks:
Looking at bandwidth we see that DDR3 are completely different from DDR2. Even if forced to run at same frequency and timings, bandwidth is quite low and only pushing on Mhz and lowering the timings we hava a consistent benefit, being aligned at about 800 Mhz
Also looking at the latency we can see that is necessary to move towards high frequencies and low timings to have a real gain on actual DDR2
On Pi32M we find again the same scenario: The extra time required to complete the test in underclocked conditions is really heavy and brings us to exclude this configuration from further tests. Again, working on frequency and timings we can fight the DDR2 product on this kind of test. And for sure, in order to reach these conditions, we need particular memory modules just like this one we are testing now.
On the top of the chart the best configuration we found: enabling the command rate = 1T, we reach the best performance of our comparison. Finally it seems that this 1T is getting familiar also for an Intel Chipset.
3dmark2005 doesn’t show big improvements. The basic knowledge is the same. Just look at the delta points we gain in the last transition from 9.9.9 to 8.8.8
Also we have to notice that we didn’t use the same settings we found as “best” during our Pi32M session. Graphic test requires more relaxed timings and, mostly a different voltage, nearer to the default one (1.9V) to obtain the sufficient stability to run this test.
Again 1T command rate configuration shows the boost on the top of the chart.
3dmark2001 is aligned with the previous tests, but absolutely more sensitive to the change DDR2-DDR3. In this test the benefits of DDR3 is evident and it grows higher and higher.
The 1T set is still the best in this comparison with a significant +900 points.
A small contest
Gandalfone, co-author of the present review, played a bit, trying to fine tune his best pi32M. It’s a challenge that requires optimization both on hardware and software side. Screenshots are clear and using DDR3 he heavily gain 6 seconds against DDR2 setup.
The Best DDR2-14'20"703
Tuning the DDR3-14'16"110
Fine Tuning the DDR3..1T-14'14"390
Just note the timings, the frequencies and the command rate value in the last screen !!
Conclusions
As anticipated in our intro, measurements has just confirmed that the transition to DDR3 is just began.
For suredevelopment will bring to customers products faster day by day. This is just the way it goes, as we already lived in the past transition of DDR technology
As every new product, it’s necessary to study and understand how to use it. We saw that just increasing the frequency is not sufficient. It’s mandatory to work on timings and find out the best compromise test by test.
Test teached us that we can’t stop at the first glance: Understanding is the base and by this way we have been able to reach and mostly pass the actual DDR2, taking into account that DDR2 have enough years of development behind.
DDR3 just came into the scene and for sure this kit is the best in class for this new standard today.
Reply With Quote
i knew teamgroup will not disappoint. 
Originally Posted by jbartlett323
Bookmarks