Archive 1997–2024

Athlon 64 Memory: Rewriting the Rules


Test Results: G. Skill TCCD

To be considered stable for test purposes, Quake3 benchmark, UT2003 Demo, Super PI, Aquamark 3, and Comanche 4 had to complete without incident. Any of these, and in particular Super PI, will crash a less-than stable memory configuration. We have also included results for RCW-ET using the Radar benchmark.

G. Skill TCCD - 2 x 512Mb Double-Bank
CPU Ratio at 2.4GHz Memory Speed Memory Timings
& Voltage
Quake3
fps
Sandra UNBuffered Sandra Standard
Buffered
Super PI 2M places
(time in sec)
Wolfenstein - Radar - Enemy Territory fps
12x200 400 DDR 2-2-2-10
2.6V 1T
512.7 INT 2636
FLT 2767
INT 6117
FLT 6046
81 110.4
11x218 438 DDR 2-3-2-10
2.8V 1T
513.4 INT 2791
FLT 2928
INT 6486
FLT 6415
80 110.7
10x240 480 DDR 2.5-3-3-10
2.85V 1T
520.4 INT 2794
FLT 3035
INT 6707
FLT 6609
80 111.5
9x267 533 DDR 2.5-4-3-10
2.85V 1T
525.5 INT 3032
FLT 3226
INT 6956
FLT 6875
79 112.5
8x292(2.34GHz) Highest Mem Speed
584 DDR
3-4-4-10
2.85V 2T
503.9 INT 2779
FLT 2969
INT 6595
FLT 6514
81 108.2
9x285(2.57GHz) HIGHEST
Performance
570 DDR
2.5-4-3-10
2.85V 1T
557.8 INT 3321
FLT 3429
INT 7408
FLT 7287
74 119.8

When we first tested memory based on the Samsung TCCD chips, none of the modules performed as well on the Athlon 64 as they did on the Intel test bed. At that time, TCCD topped out at a bit over 500 on Intel and about 466 on the Athlon 64. The lone exception to this was the newest OCZ PC3200 Platinum Rev. 2, which reached much higher on the Intel and to at least DDR500 on the A64 platform.

It is now clear with the G. Skill and other later TCCD memory in the roundup that the OCZ performance was not a fluke. All of the recent TCCD is now doing even better on the A64 than they do on Intel.

In the course of testing for this review, we also stumbled upon another mild surprise. We had been using a major brand 465 watt Power Supply in one of the A64 test beds with the FX53 and an nVidia 6800 Ultra video card. We were experiencing serious problems with video corruption and we were also seeing lower overclocks on memory in the 465 watt machine. We switched to the 520 watt OCZ PowerStream power supply in the problem test bed and found that video corruption disappeared and memory overclocks reached higher levels. Out of curiosity, we switched to the ATI 9800 PRO in the same test bed. We did not experience video corruption with the ATI and the 465W power supply, but memory overclocks were still lower with the 465 than what we could achieve with the PowerStream 520W.

Based on our experience, we would recommend a quality 500 watt+ power supply if you plan to seriously overclock memory on the Socket 939 platform. The difference in performance is substantial, with higher overclocks possible with the larger quality power supply. For that reason, we have updated the specifications of our A64 memory test bed to the 520W OCZ PowerStream.

The G. Skill reached DDR582 at 2T and achieved the best performance at 1T at DDR570. Both of these overclocks are substantially higher than what we could achieve on the Intel test bed with the same memory.

Originally published at https://www.anandtech.com/show/1494 (Wayback Machine).