Stability Testing & Overclocking
Stability Testing & Overclocking
<p style="text-align: justify;">There is no point in me wasting my hot air on explaining our stability testing methodology, I blow enough hot air in the forums, so here is a cut and paste job from previous memory reviews to explain what we consider stable at HWC.</p>
<b>Stability Testing Methodology</b><p style="text-align: justify;">Memory stability, what constitutes stable? What is not considered stable? These questions get hotly debated in enthusiast forums all over the internet like little brush fires on the fringe of an inferno that play havoc with forest fire crews. Everyone has their own opinion about stability, especially when it comes to memory stability. For some, stable means they can do whatever it is on their computer without it crashing, blue-screening, or restarting; whether that means gaming or just surfing the internet. To this user, stable means simply using the computer as they normally would.</p><center><img src="http://images.hardwarecanucks.com/image/3oh6/corsair/tr3x6g1866c7gtf/tr3x6g1866c7gtf_dominator-gt_setup-5.jpg" alt="Corsair Dominator 3x2GB PC3-12800 C8"></center><p style="text-align: justify;">Of course, that is not what we would call stable and we do a lot more thorough testing before labeling a memory frequency/timing stable. One of the toughest types of programs on a system has been found to be distributed computing projects such as Rosetta@Home, Folding@Home, World Community Grid, and more. Running 24/7 crunching for one of these great causes is a sure way to find holes in a system if there is truly some instability, unfortunately it takes a considerable amount of time to use them for stability testing so we use the list of programs below to all but guarantee the system to be
24/7 distributed computing stable:</p>
<p style="text-align: justify;">As we have said, stability testing is a fairly fluid discussion. There are a million ways to properly test an overclock for stability. The above recipe we have found to bake a pretty solid and consistent cake and like to think it is a pretty good method for testing 24/7 overclocks of memory and a system. With memory involved in today's overclocking, the modules themselves aren't the only component being pushed. With the Intel i7 processor housing the memory controller now instead of the chipset, the CPU used in running a kit of memory over 900MHz at 7-8-7 is going to play a key role. This is even more evident when we start to approach the 1000MHz mark. Processor voltages like VTT are going to play a huge role in overclocking and balancing all of these with cooling is going to be tough. At some point, the processor is going to run out of gas with just air cooling before these Corsair Dominator-GTs do. Keep this in mind when looking at the results.</p>
<b>
Specification Stability Testing</b>
<p style="text-align: justify;">After the little rant earlier about lack of XMP support, stability testing at the specified timings and frequency obviously involved a little setting manipulation in the BIOS of the DFI X58-T3eH8 and the EVGA X58 SLI. Both boards were rather simple to get setup and running for an experienced user like myself, but for the average person, there will be a bit of a learning curve. Rumors on the internet were that we would run into one major speed bump with the EVGA X58 SLI as it was reported not to run the 2:14 ratio so we would have to go with the 2:12 ratio and increase the BCLK. This wasn't the case in any way, shape, or form. Like the DFI X58-T3eH8, we simply had to go in and adjust a couple voltages, manipulate the settings to get DDR3-1866 7-8-7-20 @ 1.65v running and we were off to stability testing.
As mentioned the DFI X58-T3eH8 was a breeze to get up and running at spec like the EVGA X58 SLI. A slight bump in voltage here, an adjustment of a timing in the memory section there, and we were up and running. We of course are quite familiar with overclocking and know the boards we are using quite well. For the average user, this might not be the case. So like the EVGA settings, we have posted the DFI settings in the
comment thread for the Corsair Dominator-GT 2x3GB PC3-15000 C7 here in the HWC forums. Both of these boards were virtually painless to get this memory running at spec, but we would still love to see the XMP profiles working on these, and other X58 boards...in case that wasn't obvious enough already.</p>
Click for full size screenshot...
<center><a href="http://images.hardwarecanucks.com/image/3oh6/corsair/tr3x6g1866c7gtf/stable_spec-1.png" target="_blank">
</a> <a href="http://images.hardwarecanucks.com/image/3oh6/corsair/tr3x6g1866c7gtf/stable_spec-2.png" target="_blank">
</a></center><p style="text-align: justify;">Both setups required pretty much identical voltages as far as VTT goes to get the systems stable. The key to high memory overclocks with the Dominator-GT memory, or any DDR3 on the Core i7 platform, is the VTT or sometimes incorrectly referred to as QPI voltage. This voltage setting is what powers all things on the processor that aren't the actual core, like the memory controller. Every processor is going to require a different amount of VTT in order to get this kit stable at DDR3-1866 7-8-7-20 but we found that both of the i7 965EE processors we used today required about the same. The EVGA screenshot is a little misleading as it shows a VTT of 1.400v set in the BIOS, but that actually equates to 1.35v when measured by a DMM. This is about the same VTT that the DFI X58-T3eH8 required as seen in the screenshot. The Everest reported voltages for the DFI shown in the screenshot are pretty much bang on to what is measured with a DMM. The EVGA voltage readings are pretty far off so we opted to show what voltages were set to in the BIOS through E-LEET.
The other item to note when comparing the two setups is that with the DFI, we manually set the memory to 2T, but with the EVGA we left command rate at Auto. This left us with a 1T command rate on the EVGA board that we didn't notice until half way through the stability testing. Needless to say, the memory XMP profile shows a 2T command rate for the specified frequency and timings but our kit had no problem running 1T.</p>
<b>
Stability Overclocking:</b><p style="text-align: justify;">The overclocking section of this Dominator-GT review is going to be a little flat. Simply put, this kit of memory is already so highly clocked at the specified ratings, that they really are pushing the limits of the Intel i7 processor. As we mentioned earlier, the memory controller is now on the processor die instead of the north bridge like Intel processors of past. Some weaker processors are going to have trouble pushing memory much past DDR3-2000, let alone 3x2GB at ultra tight timings like 7-8-7 without excessive voltages.
The balance between stability and VTT/vDIMM voltages becomes a rather fluid one. We are still very early in the Core i7 life cycle and it isn't known how much voltage is too much so we will be keeping things relatively conservative for our 24/7 overclocking results. Since these results are geared towards actual stability in a daily use environment on normal air cooling, we don't want to be feeding the CPU/Memory/Memory controller excessive volts. Our self imposed limits were 1.755v as set in the BIOS for vDIMM, and 1.55v for VTT. Let's look at the stable 24/7 overclocks.
</p><center>
</center><p style="text-align: justify;">Before we discuss what is shown above, why don't we address what is not. We don't have any results listed above for looser timings than the rated 7-8-7. The reason for that is quite simple; the tight timings of 7-8-7 outran our processor with air cooling, or so we suspect. Our system simply can't handle any higher stable clocks than what this kit was willing to offer at the tight timings of 7-8-7. We didn't even have to go up to our self imposed limit of 1.75v for the 7-8-7 clocks. This is assumed to result from a lack of our processors ability to stably clock the unCore frequency much past 4000MHz. At the same time, the memory could simply be out of gas. Further testing with sub-zero cooling and observation of the memories scaling with stability taking a back seat will provide more evidence to what held us back here.
What we were able to maximize, however, were frequencies at 6-7-6. With a rather impressive DDR3-1816 clocking, the Dominator-GTs really showed their versatility. Going into this round of overclocking, we hadn't seen a whole lot of CL6 testing of these modules and the new Elpida Hyper ICs. Needless to say, we were impressed. At one point we thought we were going to be able to reach the rated frequency of DDR3-1866, but 1.75v vDIMM just wasn't enough to get us there. Either way, still a great showing for CL6 from the Corsair Dominator-GTs.
The 7-8-7 clocking speaks for itself. We really must have gotten lucky with our modules as they came very close to meeting the DDR3-2000 7-8-7-20 frequency and timings of the higher binned Dominator-GT kit at the rated 1.65v. Unfortunately we topped our Dominator-GT sample out at a very respectable 1010MHz 7-8-7-20 2T with less than 1.70v. Keep in mind, this is completely 100% stable through our very thorough stability testing suite. We are already hunting for better unCore clocking processors but at the same time, we haven't seen much - if any - stable overclocking of 3x2GB kits at these types of clocks. A lot of guys are benching these clocks and much higher, but not many showing lengthy stability testing. Below are the screen shots of the testing for the listed overclocks.</p>
Click for full size screenshots...
<center><a href="http://images.hardwarecanucks.com/image/3oh6/corsair/tr3x6g1866c7gtf/oc-2.png" target="_blank">
</a> <a href="http://images.hardwarecanucks.com/image/3oh6/corsair/tr3x6g1866c7gtf/oc-3.png" target="_blank">
</a> <a href="http://images.hardwarecanucks.com/image/3oh6/corsair/tr3x6g1866c7gtf/oc-4.png" target="_blank">
</a></center><p style="text-align: justify;">It is now time to turn this rather impressive overclocking display pu on by the Dominator-GTs into some benchmark results. We will be utilizing the results of the above overclocking to come up with an interesting comparison for the benchmark suite. With the versatility in frequency and timings available from the Dominator-GT, there really is no lack of possibilities for us to test, but we think we came up with something that will provide some insight into memory performance for end users.</p>