Previously reviewed by Trembledust last month, this review is covering more modern Intel systems. Both the P67 family these sticks are designed for as well as some X79 results.
As this is part two of a review we will not continue with our usual methodology of discussing the sticks themselves the company as a whole and the aesthetics of the sticks. This will cover only the performance numbers and a conclusion.
Due to poor bus clocking on the P67 platform, and repeated memory no POST errors, overclock results were included only on the X79 platform. Tests were run at stock and overclocked to 1866 10-10-10. While the sticks were capable of reaching the same speeds as Trembledust’s on the X58 platform with even less voltage (1940 MHz 10-10-10), stability was not 100% and PCMark tests continued to fail. Dropping back to 1866 MHZ allowed all tests to pass and blend prime x64. As such only those results are included.
For strict bandwidth and latency testing both AIDA64 and MaxxMEM2 were used. Results were mixed with both tests performing unusually.
MaxxMEM reported a 150 ns latency on the P67 tests which is completely aberrant. A deficiency in the chip or board used is possible multiple tests kept giving a bizarrely high result. Bandwidth results were pretty much as expected however. Conversely AIDA64 gave a really good latency performance and was a bit better than expected but not unreasonably so.
Both tests on X79 however were very poor. considering the X79 platform is little more than Sandy Bridge with an enhanced IMC for quad channel and PCIE 3.0, this was unexpected. Since the attempts to OC the memory on P67 were unsuccessful this is what we were left with.
Since the 2600K defaults to a higher CPU clock it is expected that P67 results should always result in higher performance than X79 except in multithreaded heavy benchmarks.
No surprise that P67 wins this even at default. Normally relaxed latencies on P67 are overcome by enhanced bandwidth. In the case of the X79 however this is not the case, and despite the OC every test at 1866 MHz results in reduced performance.
PCMark is a whole system test that is generally single threaded.
Once again the enhanced clocks of the 2600K proved to be superior to even OC memory results on the X79. The memories benchmark was a notable exception in which the OCed results performed about even with P67. TV and Movies same a small gain with the improved memory clocks as well.
PCMark 7 is no exception to the rule of CPU clocks beats multithreaded goodness.
Once again we see the 2600K dominating in almost every area especially in the entertainment and lightweight tests. However the 3930K show its performance in the Computation tests with the 1866 MHz memory tests far exceeding all other results.
SuperPi is a long standing tradition for benchers and really highlights both CPU and memory performance.
While the P67 results in better overall times, the results of x79 are more telling. Reduced latencies impede 1M results, but increased bandwidth still improves 32M results. This is exactly how it performs on P67 as well.
Memory is a fickle creature, many companies have followed suit with increasing some latencies in order to get high clock speeds or lower voltages. We have all seen the 7-9-7 9-11-9 and other such bologna over the last couple of years, and Kingston has done it to. It is good to see at least at 1600 that Kingston did not go with 8-9-8 for this ram at 1.6v although it will run that. Because at 9-9-9 1.5v this RAM is suited at 1600 for an 8GB density.
Overclocking these sticks required both increase in voltage and loosening of timings, which was surprising, normally loosening if timings precludes the need for increased voltages. As mentioned before, the sticks were capable of doing what they could on X58 per the previous review, on the newer Intel IMC with even less volts. Increasing latency to 10-10-10 to achieve 1866 MHz should have resulted in no need for a voltage or at least only a tiny bump. Not the full .1 volts it required.
These configuration of the heatspreader is for appearance over performance. There is no real need for a heatsink on these sticks but at least they are standard height so will not interfere with any mounting configuration. The sinks to provide a nice bling factor that we saw on the last Kingston memory review.
Price is a key factor in any purchase, and a high MSRP on an item can be offset by extreme performance. That is not he case with these sticks. At about half the current market price these sticks are a great buy, and would get a Gold award hands down. Even 65% of market value would put them in competition with similar performing sticks. But at $70 these sticks are priced out of consideration.
Overall the Kingston sticks get a TechREACTION bronze award for fair performance, good aesthetics and a terrible price point (which to be fair can change). Their customer service is outstanding however and is a consideration when purchasing new sticks of memory.
Thanks to Kingston for letting us test their sticks let’s hope the do better next time.