Articles: Memory
 

Bookmark and Share

(24) 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 ]

Overclocking

We test modern memory modules, especially overclocker-friendly ones, only in the dual-channel mode since all top-end platforms today support dual-channel memory access. The overclocking potential is somewhat reduced in this mode, so people who use Athlon 64 for Socket 754 or Dothan processors and older platforms like Intel’s 845PE/848P can add at least 10 megahertz to the numbers from the tests, especially at relaxed timings.

Unlike with graphics cards, memory modules of one type generally differ little in their overclockability, so the results quite precisely indicate the overclocking potential of the modules.

We tested the memory in four different modes: with minimal (2-5-2-2-1T), average (2-6-3-3-1T), maximum (2.5-8-3-3-1T) and default (2.5-8-4-4-2T) timings. We performed our tests at 3.0B memory voltage (this is the limit of the employed mainboard in its unmodded state). All the performance-related BIOS settings were set to their highest-performance values.

First let’s take a look at the SPD information from both memory kits:

Both kits have competently written parameters. The lack of extra-low timings at DDR400 frequency guarantees compatibility with a majority of mainboards (sometimes you may run into trouble with low-latency modules that have something like 2-5-2-2 written into their SPD). The second column on the screenshot contains parameters of DDR566 (PC4400) mode with most sparing timings.

The only strange thing is that the SPD chip of the F1-4800DSU2-1GBFR kit contains the PC4400 parameters, including the module type. On the other hand, we have seen even more ingeniously written SPD which doesn’t affect the performance of the module in real situations. And again, keeping the default timings high ensures better compatibility with some mainboards.

Note that the CAS Latency parameter is set to 2.5, as in the specification. The fact is Samsung’s TCCD chips do not like CAS = 3! Setting this parameter to 3 should theoretically lead to better overclockability than with 2.5 CAS at the expense of a certain performance hit at the same frequency. But this is only theoretically true. In practice, overclockability of TCCD chips worsens in this case, sometimes quite much…

Let’s first check the senior model.

The declared DDR600 is a very serious frequency, rarely achievable by DDR modules without your taking special measures. And still the F1-4800DSU2-1GBFR kit really worked at this frequency with the specified timings. The voltage was 2.7V, i.e. the lowest specified (2.7-2.9V). This is impressive already. From this starting point we enthusiastically got to squeezing the juices from G.Skill’s top product.

Our further experiments (at 3.0V) produced the following results:

Our tests prove that these modules easily reach an astonishing frequency of DDR650 (or PC5200) at a modest voltage of 3.0V and with 2.5-8-4-4-2T timings. That’s an outstanding performance even for TCCD-based chips!

 
Pages: [ 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 ]

Discussion

Comments currently: 24
Discussion started: 08/20/05 09:58:44 PM
Latest comment: 08/25/06 12:43:21 PM

View comments

Add your Comment