Page 2: Stability, Testing, Conclusion
Stability:
Let me reiterate how ASE Labs determines stability. I shall do this for every RAM review that is done. At ASE Labs, RAM is only stable if it passes Memtest86+ and a plethora of memory intensive applications for a few hours like mprime and doing continuous compiling of specific programs. Just because the memory POSTs and runs doesn't mean that it is stable. Linux users may find memory errors by having errors in compiling (very common) to errors in tar as well as segfaults. Windows users can expect program faults and BSODs. Memory stability is very important. When we say that a specific specification of RAM is stable, we mean it!
Unfortunately, the RAM was not totally stable at 1.90V at the rated timings and speed of 1800MHz. To get a fully stable system, I had to up the voltage slightly to 1.96V. This might be due to how well the power is handled on the board, but I would expect to always need more voltage at this speed. This type of stability testing yielded these results after hours of testing. You cannot trust that a computer booting up is proof of a stable system.
Overclocking:
Needless to say, since this RAM is already running at 1800MHz at 1.96V, there is little room for more speed to be found in these modules. I could POST at 1900MHz but the modules were not stable even at 2.1V. I couldn't get high enough speed to warrant a speed overclock that was worth it. I only got to 1850MHz before I hit a wall. I once again opted to find out how tight I could make the timings of the RAM. At 2.0V running at 1800MHz, I yielded a result of 7-6-6-18 on timings. This is a fairly respectable decrease in latency, but still the modules are running with more power. I could not set the CAS latency to 6 without errors at any voltage.
Testing:
I want to preface this section by saying how ASE Labs will be doing benchmarks from now on. There will be no synthetic benchmarks at ASE Labs. This means that for fair comparisons, an apples to apples system will need to be made. We will be using tools that anyone will be able to get so anybody can reproduce our results. Current memory tests include compilation of LAME and FFMPEG, encoding in each of those programs, and a round of mprime to round everything out. We will be adding testing as we see fit during future reviews. If you would like to see specific (real world stuff) programs, please send me an email.
As shown from the previous DDR3 review, memory is pretty useless to review with real world apps. The differences in modules are really statistically negligible unless there is a big problem with a set of modules. This is why testing is less important for RAM and more weight should be giving to the stability and overclocking results. Regardless, I've tested these modules against the »Crucial Ballistix PC3-12800 that was just reviewed.
Testing was done with a Core 2 Duo 2.4GHz on an Asus P5E3 Deluxe (review coming soon) for DDR3 and an Asus P5B Deluxe for DDR2. Same specs for both other than the motherboard. The DDR2 was running at DDR2-1066 with timings of 5-5-5-15. Dark gray denotes DDR2. Dark blue is the Crucial Ballistix DDR3. Light blue is the Kingston HyperX DDR3. Both modules are running at 1600MHz at the same 8-8-8-24 timings for a fair comparison. Remember that the Kingston RAM can go much higher for its rated speed, but this wouldn't be a fair test without running at the same speed.
Even as compared to DDR2, DDR3 has a ways to go before gaining enough of an advantage. Still, it is a logical progress towards more speed and lower latencies and hopefully the latencies will catch up soon for DDR3. I believe that the testing has shown that there are negligible difference between modules of the same quality. Kingston makes some nice RAM.
Conclusion:
The KHX14400D3K2 2GB kit retails for about $420 (Newegg) at the time of writing. This is a bit under other offerings in the same speed range. Yes, you are paying for a premium in two places; One for DDR3 and another for 1800MHz. You do gain the ability to really pump your front side bus to extremely high levels with this RAM. Overclockers, this RAM is for you. If you aren't an overclocker, look elsewhere since you need to be to unlock the potential of these modules.
I'd like to thank David from Kingston Technology for making this review possible.
Let me reiterate how ASE Labs determines stability. I shall do this for every RAM review that is done. At ASE Labs, RAM is only stable if it passes Memtest86+ and a plethora of memory intensive applications for a few hours like mprime and doing continuous compiling of specific programs. Just because the memory POSTs and runs doesn't mean that it is stable. Linux users may find memory errors by having errors in compiling (very common) to errors in tar as well as segfaults. Windows users can expect program faults and BSODs. Memory stability is very important. When we say that a specific specification of RAM is stable, we mean it!
Unfortunately, the RAM was not totally stable at 1.90V at the rated timings and speed of 1800MHz. To get a fully stable system, I had to up the voltage slightly to 1.96V. This might be due to how well the power is handled on the board, but I would expect to always need more voltage at this speed. This type of stability testing yielded these results after hours of testing. You cannot trust that a computer booting up is proof of a stable system.
Overclocking:
Needless to say, since this RAM is already running at 1800MHz at 1.96V, there is little room for more speed to be found in these modules. I could POST at 1900MHz but the modules were not stable even at 2.1V. I couldn't get high enough speed to warrant a speed overclock that was worth it. I only got to 1850MHz before I hit a wall. I once again opted to find out how tight I could make the timings of the RAM. At 2.0V running at 1800MHz, I yielded a result of 7-6-6-18 on timings. This is a fairly respectable decrease in latency, but still the modules are running with more power. I could not set the CAS latency to 6 without errors at any voltage.
Testing:
I want to preface this section by saying how ASE Labs will be doing benchmarks from now on. There will be no synthetic benchmarks at ASE Labs. This means that for fair comparisons, an apples to apples system will need to be made. We will be using tools that anyone will be able to get so anybody can reproduce our results. Current memory tests include compilation of LAME and FFMPEG, encoding in each of those programs, and a round of mprime to round everything out. We will be adding testing as we see fit during future reviews. If you would like to see specific (real world stuff) programs, please send me an email.
As shown from the previous DDR3 review, memory is pretty useless to review with real world apps. The differences in modules are really statistically negligible unless there is a big problem with a set of modules. This is why testing is less important for RAM and more weight should be giving to the stability and overclocking results. Regardless, I've tested these modules against the »Crucial Ballistix PC3-12800 that was just reviewed.
Testing was done with a Core 2 Duo 2.4GHz on an Asus P5E3 Deluxe (review coming soon) for DDR3 and an Asus P5B Deluxe for DDR2. Same specs for both other than the motherboard. The DDR2 was running at DDR2-1066 with timings of 5-5-5-15. Dark gray denotes DDR2. Dark blue is the Crucial Ballistix DDR3. Light blue is the Kingston HyperX DDR3. Both modules are running at 1600MHz at the same 8-8-8-24 timings for a fair comparison. Remember that the Kingston RAM can go much higher for its rated speed, but this wouldn't be a fair test without running at the same speed.
Even as compared to DDR2, DDR3 has a ways to go before gaining enough of an advantage. Still, it is a logical progress towards more speed and lower latencies and hopefully the latencies will catch up soon for DDR3. I believe that the testing has shown that there are negligible difference between modules of the same quality. Kingston makes some nice RAM.
Conclusion:
The KHX14400D3K2 2GB kit retails for about $420 (Newegg) at the time of writing. This is a bit under other offerings in the same speed range. Yes, you are paying for a premium in two places; One for DDR3 and another for 1800MHz. You do gain the ability to really pump your front side bus to extremely high levels with this RAM. Overclockers, this RAM is for you. If you aren't an overclocker, look elsewhere since you need to be to unlock the potential of these modules.
I'd like to thank David from Kingston Technology for making this review possible.