High-End DDR3 Memory on the Hook


The test results were conducted at two different speed levels: 400 MHz base clock times 6 (2.4 GHz core clock speed) for the DDR3-1600 memory, and 450 MHz times 6 (2.7 GHz core clock speed) for the DDR3-1800 products. This would give our test CPU sufficient clock speed margin to go above FSB500/DDR3-2000 if required.

We stuck to the memory timings specified by the memory vendor and increased the FSB/memory clock in increments of 25 MHz. As this automatically increases the processor clock speed as well, the benchmark results don't allow a direct benchmark comparison of the individual memory products, as the results reflect the combined performance increase of CPU/FSB/memory. However, they show real-life system performance that you'll achieve when you install the memories in our reference system and try to squeeze out as much memory clock speed as possible.

To give you an idea of how to assess performance, we also added a run at 450 MHz FSB clock and a 2:3 memory ratio, setting the memory to DDR3-1352 speed (this is as close to DDR3-1333 as we could get). We paid attention to only select benchmarks that actually show impact on varying memory performance. As you could see there is a difference, but is it worth at least twice the memory cost? Since all other benchmarks hardly take notice of the changes, we clearly have to say "no". Spending a few hundred bucks extra on the processor instead of the memory will give you a more noticeable performance boost. The only exception would be the very highest-end enthusiast PCs, where you need hardcore RAM to support the RAM speeds required at ridiculous FSB speeds. In this case, overclocking memory is required to overcome the multiplier limitations of a memory controller, but it's not a performance tweaking instrument by itself.

We found that all DDR3-1600 DIMMs could be overclocked to at least DDR3-1800 speed with only a little more memory voltage. One product, Crucial's Ballistix DDR3-1600, even reached maximum speeds of DDR3-1950, which is as fast as the high-end DDR3-1800 DIMMs by Chaintech and Kingston could go. Although we were able to boot some of the DIMMs at DDR3-1950 or DDR3-2000, it always required us to considerably relax the memory timings or to live with insufficient reliability. All results we tested were confirmed with two runs of Memtest86+ 1.7.

Although DDR3-1600 and 1800 memory speeds are way ahead of what the mainstream will be in the following months, we have several reasons that speak against purchasing these premium memory products today. On the one hand, these products are still based on first generation DDR3 memory formed from 512 Mbit or 1 Gbit chips. That's not bad, but as I described some pages above, only higher density DDR3 (2 Gbit) silicon will make the new technology affordable - and it will probably perform even better as well. Sub-70-nm manufacturing processes will output 2 Gbit ICs that allow memory vendors to build 2 GB DIMMs at a more attractive cost. On the other hand, we can't tell if and how many of the review samples were hand-picked by the memory vendors in order to achieve ideal benchmark results. Don't be surprised if your memory doesn't achieve the overclocking levels you may expect.

Once again, if you're looking for memory to support your overclocking ambitions at maximum FSB speeds, where you simply cannot set the memory to slower clock speeds due to insufficient memory ratio settings, these products are exactly what you need. If you're looking for more performance at moderate overclocking speeds we recommend staying with DDR3-1333, as the latest products will perform well at a considerably lower cost.

Join our discussion on this article!

Create a new thread in the Reviews comments forum about this subject
This thread is closed for comments
    Your comment
  • ilysaml
    i'm stuck with 1333 MHz and it's really pretty cool
  • Anonymous
    Can you specify what brand and/or model number that ddr3-1352 RAM is? It flattened the 1600 with identical timings, and I'd like to know why.