4GB Gets Cheap: 9 Dual-Channel Kits Compared
Latency And Overclocking Results
While previous DDR2 tests were often stopped at CAS 5, one of the module sets in today’s competition was rated at super-relaxed latencies of CAS 7. With high hopes that several of these memory kits might reach a 1,200 MHz data rate, we tried latencies ranging from CAS 3 to CAS 8 before determining the “Lowest Latency” for data rates of 800, 1,066, and 1,200 MHz.
Many users fear added voltage may cause premature parts failure, so we started with the DDR2-standard 1.80 V setting and worked our way up. Modules are sorted by best latency, starting with the highest achievable speed.
| DDR2-1200 | DDR2-1066 | DDR2-800 | |
| G.Skill F2-8000CL5D | Failed | 5-5-4-14 | 4-4-3-10 |
| Crucial Ballistix PC2-6400 | Failed | 5-5-5-14 | 4-4-4-10 |
| Patriot Viper 6400LLK | Failed | 6-6-5-6 | 4-4-4-5 |
| Kingston KHX8500AD2 | Failed | 6-5-6-15 | 5-4-4-12 |
| PNY MD4096KD2-800-X4 | Failed | Failed | 4-4-3-10 |
| Super Talent T800UX4GC5 | Failed | Failed | 4-5-3-5 |
| Corsair XMS2-6400C5 | Failed | Failed | 5-4-4-10 |
| Mushkin 996557 | Failed | Failed | 5-4-4-10 |
| Wintec AMPX 6400C5 | Failed | Failed | 5-4-4-10 |
While none of our module sets were capable of reaching DDR2-1200 at standard voltage, G.Skill’s DDR2-1000 and Crucial’s Ballistix DDR2-800 both reached DDR2-1066 at CAS 5. Kingston’s DDR2-1066 didn’t require a voltage increase to reach its rated speed, but it suffered from longer CAS 6 latency requirements.
At the basic DDR2-800 setting, Patriot’s low-latency kit stands out for having exceptional tRAS.
| DDR2-1200 | DDR2-1066 | DDR2-800 | |
| Crucial Ballistix PC2-6400 | 6-5-6-15 | 5-4-5-14 | 4-3-4-10 |
| G.Skill F2-8000CL5D | Failed | 5-5-4-14 | 4-4-3-10 |
| Patriot Viper 6400LLK | Failed | 5-5-5-6 | 4-4-4-4 |
| Kingston KHX8500AD2 | Failed | 6-5-6-15 | 5-4-4-12 |
| PNY MD4096KD2-800-X4 | Failed | Failed | 4-4-3-9 |
| Corsair XMS2-6400C5 | Failed | Failed | 4-4-4-10 |
| Super Talent T800UX4GC5 | Failed | Failed | 4-5-3-5 |
| Mushkin 996557 | Failed | Failed | 5-4-4-10 |
| Wintec AMPX 6400C5 | Failed | Failed | 5-4-4-10 |
A modest voltage increase to 2.00 V returned amazing results from Crucial Ballistix, with a 50% overclock achieved ! The stigma of using CAS 6 timings goes away at DDR2-1200, since this 50% increase in latency cycles (compared to rated timings) is accompanied by an equally significant decrease in cycle time (compared to rated speed).
Patriot’s low-latency kit again offers impressively-low tRAS, with the added voltage bringing it down to four cycles.
| DDR2-1200 | DDR2-1066 | DDR2-800 | |
| Crucial Ballistix PC2-6400 | 6-5-6-15 | 5-4-5-14 | 4-3-4-10 |
| Patriot Viper 6400LLK | 7-7-7-11 | 5-5-5-6 | 4-4-4-4 |
| G.Skill F2-8000CL5D | Failed | 5-5-4-14 | 4-4-3-10 |
| Kingston KHX8500AD2 | Failed | 6-5-6-15 | 5-4-4-12 |
| PNY MD4096KD2-800-X4 | Failed | Failed | 4-4-3-8 |
| Corsair XMS2-6400C5 | Failed | Failed | 4-4-4-10 |
| Super Talent T800UX4GC5 | Failed | Failed | 4-5-3-5 |
| Mushkin 996557 | Failed | Failed | 5-4-4-10 |
| Wintec AMPX 6400C5 | Failed | Failed | 5-4-4-10 |
Crucial’s Ballistix DDR2-800 continues its lead at 2.20 V, but doesn’t show any gains compared to its 2.00 V results. Patriot breaks through the DDR2-1200 ceiling as well, but at lengthier CAS 7 timings.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
None of the remaining modules showed notable performance gains in the switch from 2.00 V to 2.20 V, but we still used 2.20 V as the limit when we moved on to seek the maximum clock speed of each module set. Hoping that several modules would at least come close to a 1,200 MHz data rate at reasonable timings, we chose latency limits of 6-6-6-18.
Crucial Ballistix PC2-6400 isn’t the cheapest memory around, but its overclocking potential is nothing short of stunning. Only Patriot Viper PVS24G6400LLK was able to come close.
Some of our samples were rated at far higher speeds than the DDR2-800 of Crucial’s mid-priced Ballistix, but it’s obvious that “overclock-validated” parts are not always the best choice for overclockers. A comparison of rated-speed to achieved speed adds a little more perspective.
The two “high-speed” memory kits were among the worst overclockers, percentage wise. Anyone who is willing to pay extra for overclocking-validation will be especially disappointed in the chart above, as its shows that the only reliable way to estimate how far a memory set will go is to read reviews. Since manufacturers often alter components without telling anyone, user reviews are a great way to stay updated once an official review no longer appears relevant.
While overclocking and low latencies are nice features, the focus of today’s comparison is performance value, also known as “bang for the buck." Our next page will show the value of each kit, from several perspectives.