OCZ Platinum Edition PC2-4200 EB - Overclocking
(Page 4 of 5 )
After experimenting with voltages, I found that the OCZ sticks had incredible difficulty exceeding 2.2V, an oddity for Fatbody D9s. The Corsair sticks ran happily at 2.65V and the Crucial’s ran best at 2.4V. In each of these cases the sticks were cooled by their default spreaders and a pair of low-speed Antec 80mm Fans.
After fiddling with timings, I found that changing the tRas didn’t help with the overclock at all, and changing the tRC beyond 3 didn’t help to any great extent either.
Below are the maximum clocks reached that were Super Pi 32M stable, at each timing set.
Timings (tCl-tRC-tRP) | OCZ 4200EB (default: 533MHz 3-2-2) | Corsair 5400UL (default: 675MHz 3-3-2) | Crucial 6400 (default: 800MHz 4-4-4) |
3-2-2 | 710MHz | 810MHz | 733MHz |
3-3-2 | 718MHz | 865MHz | 805MHz |
4-3-2 | 880MHz | 1108MHz | 915MHz |
5-3-2 | 885MHz | 1116MHz | 1016MHz |
5-4-2 | 894MHz | 1116MHz | 1038MHz |
5-5-2 | 904MHz | 1116MHz | 1056MHz |
The maximum clock reached on these sticks.
While the results show the OCZ sticks clearly behind the rest, primarily due to their voltage issues, they still performed spectacularly nevertheless – remember I’m comparing them with the best of the best. A stellar 900MHz from PC2-4200 sticks is huge in anyone’s eyes, representing almost a 70% overclock from default clocks.
One stick was capable of taking up to 2.3V with reasonable stability, however this meant running in single channel. This caused a noticeable loss on benchmarks, in particular those bandwidth intensive ones. At 2.3V, I was able to push this stick to 920MHz at relatively tight timings of 5-3-2.
Maximum clock reached by the best performing stick.
Overclocking Performance
I ran my test set up with the CPU at 2.45GHz with a 350MHz FSB (1400MHz effective speed) and used dividers to keep the RAM as close to default as possible for the first test, and as close to its maximum at each respective timing set.
- Overclocked Test Setup
- Core 2 Duo E6300 @ 2.45GHz
- 975X @ 266MHz
- Stock: 1:1 @ 3-2-2-8 (533MHz DDR2)
- 975X @ 350MHz
- Setup 1: 4:3 @ 3-2-2-8 (525MHz DDR2)
- Setup 2: 1:1 @ 3-2-2-1 (700MHz DDR2)
- Setup 3: 4:5 @ 4-3-2-1 (875MHz DDR2)
Test | Stock | Setup 1 | Setup 2 | Setup 3 |
Super Pi 1M (lower is better) | 30.079 sec | 24.656 sec | 23.828 sec | 23.734 sec |
Super Pi 8M (lower is better) | 5m 36.734 sec | 4m 41.547 sec | 4m 27.110 sec | 4m 24.204 sec |
Sandra Bandwidth (Float) (higher is better) | 5119 MB/s | 5629 MB/s | 6155 MB/s | 6607 MB/s |
Sandra Bandwidth (Int) (higher is better) | 5103 MB/s | 5625 MB/s | 6149 MB/s | 6591 MB/s |
Everest Latency (lower is better) | 68.4 ns | 64.1 ns | 54.1 ns | 55.8 ns |
Everest Copy (higher is better) | 4455 MB/s | 4501 MB/s | 5585 MB/s | 6045 MB/s |
Everest Read (higher is better) | 8197 MB/s | 7827 MB/s | 10406 MB/s | 12722 MB/s |
Everest Write (higher is better) | 4710 MB/s | 6099 MB/s | 6100 MB/s | 6100 MB/s |
Quite impressive results, especially when overclocked, with the top benches far exceeding stock results, supporting my hypothesis that this memory is designed for overclockers.
An interesting oddity is the Everest Read speed on Setup 1 (which I confirmed with subsequent benches). Why was it lower than the bench with equally clocked memory and a lower FSB? Most likely it was the divider used in Setup 1 (as opposed to 1:1 in the stock bench) that hurt the score. In addition, the Everest Latency on Setup 3 being weaker than the latency on Setup 2 also shows the effects of using an asynchronous ratio.
Next: Overclocking on the AM2 Platform >>
More Memory Articles
More By Walter Eigner
