Team Group Xtreem LV DDR3 2600MHz 8GB Memory Kit Review
Andy Ruffell / 12 years ago
Once we had set the XMP profile and changed the timings to reflect what they should be, we managed to boot into Windows to make sure that the appropriate settings had been applied in the BIOS, and we did this by using CPU-Z. Once we had CPU-Z open, we could confirm that the memory was in fact an 8GB kit and was running at 2600MHz with timings of 10-12-12-31-1T. Quite surprising to find this memory operating with a command rate of 1T, but we’re not exactly going to moan about it.
After we had checked our settings, we could fire up AIDA64 so that we could check to see how the memory performed under stock speeds and being rated at 2600MHz with some good timings on our Z77 motherboard, we were expecting to see some strong numbers here.
As we expected, the numbers were fantastic with 21797MB/s read, 18946MB/s write, 23055MB/s copy and a low latency of 32.7ns.
The next thing for us to do was to push the memory as far as we could on stock timings, so we kept everything as it was within the BIOS, but increased the CPU speed of our i7 3770k to 4.5GHz to alleviate any bottleneck from the processors memory controller. Once this was clocked to 4.5GHz, we first tried to increase the memory divider from 2600MHz to 2666MHz but while this booted, it froze when loading Windows. We knew that the only way of getting further would be to utilise the BCLK and slowly and gradually moved this up from 100MHz to 101MHz and managed to boot into Windows and run all of our tests fine. Not being 100% happy with this, we pushed things a little further and found that 101.5MHz was the limit to keep the memory 100% stable.
We ended up at 2640MHz using a BCLK of 101.5MHz while keeping the timings at 10-12-12-31-1T at 1.65V.
Keeping the timings the same as what we had at stock but just increasing the speed allowed us to push the bandwidth of the memory higher than before and that is show in AIDA64. The read, write and copy speeds all increased to 24990MB/s read, 23365MB/s write and 25149MB/s copy and the latency decreased dramatically from 32.7ns to an extremely low 29.9ns.
Now that we’ve seen how far the memory can be pushed with stock timings, it’s time to mix things up a little bit and this saw us going back into the BIOS to set all timings to auto and then start cranking things up again.
We immediately tried the 2666MHz divider and this time round, it worked and managed to boot into Windows and run our stress testing software. Wanting to squeeze even more juice out of these modules, we pushed them to the 2800MHz divider, which we wasn’t that sure about and this was confirmed when the system refused to boot. Going back to the 2666MHz divider, we had to start using the BCLK again and we managed to push this up to 100.6MHz leaving us with a final overclock of 2684MHz with timings of 11-13-13-35-2T at 1.65V.
Once we had confirmed this within CPU-Z, we opened up AIDA64 for the final time and were keen to see how the numbers looked with a higher clock speed, but slacker timings. Quite surprising, the numbers weren’t to far off the results before with 24562MB/s read, 23570MB/s write and 24982MB/s copy and a latency of 30.9ns.
On a side note, we did manage to push the memory up slightly higher using a BCLK of 100.8MHz but the performance on all tests (read, write, copy) started to drop below that of the stock level.
