DDR2 Memory Roundup Autumn 2007

Memory by thorgal @ 2007-12-14

As two different platforms do not treat memory in the same way, let´s swap motherboards and user a new test system. In this second part of our DDR2 roundup we continue our stress test with a DFI 680i based board. Find out how different nVidia´s 680i chipset clocks your ram in the following review, and see if the latest memory kits can threaten our previous champions

  • prev
  • next

Cas 4 / 2T performance results

Cas 4/2T, 400FSB unlinked testing

As we have not yet tested how far our modules will overclock, we decided to fix the CPU speed and let the memory speed vary. Our 680i board provides us with an almost unlinked memory control allowing us to test the memory by itself, so here we go. For the front side bus and CPU speed, we decided to stick with the same settings as our previous cas 3 tests, giving us ample bandwidth for good memory performance. Do not expect similar results to the synced memory/front side bus tests though, as the DFI board clearly prefers the synced setting.

To see how far the memory would go, we chose a cas 4 setting this time, and a command rate of 2T. In this case we expect the differences between the memory modules to be much bigger, so let's start the testing.

To determine our maximum stable memory setting, we first had to start overclocking the different memory kits of course. The results you can find in the thumbnails below.


Madshrimps (c) Madshrimps (c) Madshrimps (c)
Corsair PC8888C4F, Corsair PC6400C3F, Team Xtreem PC6400C3



Madshrimps (c) Madshrimps (c) Madshrimps (c) Madshrimps (c)
G.Skill PC8500C5, OCZ PC9200 Flex, OCZ PC8500 Reaper (C5), OCZ PC6400 Flex



The results - for this test * - are as follows:

1. Corsair Dominator PC8888F : 600 Mhz (1200 Mhz DDR)
2. Corsair Dominator PC6400F : 571 Mhz (1143 Mhz DDR)
2. Team Group Xtreem PC6400C3 : 571 Mhz (1143 Mhz DDR)
4. G.Skill PC8500C5 : 550 Mhz (1100 Mhz DDR)
5. OCZ PC9200 Flex XLC : 533 Mhz (1066 Mhz DDR)
5. OCZ PC8500 Reaper HTC : 533 Mhz (1066 Mhz DDR)
7. OCZ PC6400 Flex XLC : 510 Mhz (1020 Mhz DDR)

As you can see, OCZ finishes at the bottom this time and Corsair at the top. Let's find out if this is repeated in the benchmark tests.

* "For this test" means that some of the modules were stable a few Mhz higher, but there were no higher dividers available on 680i at 400 FSB, forcing us to test them a little lower. The screenshots give you the real maxima; on the next page we give you the maximum performance graphs.




For our benchmark results, first the overview table once more (click on it to open in a new window) :

Madshrimps (c)



And the details in the following graphs, starting with the bandwidth results :

Madshrimps (c)


Logic is almost followed here, giving the higher overclocked modules advantage over the others. However, OCZ manages to steal a few places with their PC9200 part, in favor of the Dominator PC6400 part which loses some places.

On to the latency:

Madshrimps (c)


Rather strange results here, not following the OC list at all. Reason for this probably is the internal chipset latencies that are different for different dividers. We described this in our test methodology a few pages ago, and our theory is acknowledged. This makes benchmarking and tuning a 680i system all the more difficult as you can get unexpected results while overclocking your memory to the maximum.

Let's look at SuperPi results once more:

Madshrimps (c)


Seems like we get a mixed ranking from the latency and bandwidth charts. The OCZ modules pay the price here for not overclocking as far on this cas setting. The differences are very small though.

Cinebench again :

Madshrimps (c)


Although it only gets an average OC, the G.Skill modules come out on top. It's the second good score for them in a row, as the SuperPi score was not bad either. Our top overclock modules finish last here, a strange result (but we did repeat the test more than once to see if the results were accurate).


3D-Mark 2006 for the last time, starting with the CPU :

Madshrimps (c)


A very nice score for both Dominator parts, and a rather disappointing score for the OCZ PC9200. Again, differences are small.

And the 3D06 total results :

Madshrimps (c)


With only 10 Marks between the first and the last contestant, we can only speak of very minor differences indeed. Team Group does well once more; the rather low CPU score of the OCZ PC9200 also reflects itself in the total score.

Next is PC Mark 2005 :

Madshrimps (c)


Mixed results here, with a large pack of modules within a few PC Marks. The modules that stand out however are the PC6400 Flex modules, this time in a negative fashion. Again we repeated the tests at the end with these modules again, with no better results.

And finally, our game test:

Madshrimps (c)


Only small differences once more, but the PC6400 Flex module does come out as last again.


Let's see how our modules overclock on different cas settings next >>>
  • prev
  • next
Comment from Massman @ 2007/12/15
Excellent write-up, Thorgal
Comment from thorgal @ 2007/12/15
Quote:
Originally Posted by Massman View Post
Excellent write-up, Thorgal
Thanks Massie !

Better late than never I guess, but doing a roundup like this always takes more time than expected.
Comment from Oscar @ 2007/12/15
Excellent review, thorgal ! I especially like the "Value for Money" section at the end. It's what a lot of reviews at other websites are lacking.

I was wondering whether MadShrimps would consider including RAM from Patriot Memory for review in the future? I have heard great things about them from time to time.
Comment from maher @ 2007/12/16
excellent job but that memorys are out my range so I just can look and hope.............
Comment from thorgal @ 2007/12/16
Quote:
Originally Posted by maher View Post
excellent job but that memorys are out my range so I just can look and hope.............
Stay tuned then, because we just might have a give-away coming up that might interest you
Comment from maher @ 2007/12/17
can't wait

 

reply