Kingston has long been in the memory and flash storage business; and like most other organizations, has kept up with market demand and the wants/needs of enthusiasts. The release of Ivy Bridge CPUs brings an even more robust memory controller on the chips and subsequently, even faster potential memory speeds. To that end, Kingston has brought to market a set of 2666 MHz rated chips within their Predator series of memory. It’s time to test these speedsters and see how they do.
Meet the Ram – Features, Specifications, and Product Tour
The kit we have here today is KHX26C11T2K2/8X. Or in other words, it’s a set of 2x4GB DDR3 2666 CL11-13-13-32 @ 1.65v sticks. Below are the details in a nice table for easy viewing:
|Multi-Channel Kit||Dual Channel kit|
|Tested Speed||DDR3-2666 MHz (PC3-21300)|
|Tested Latency||11-13-13-32 1N|
|SPD Speed||1600 MHz|
|Features||Intel XMP 1.3 (Extreme Momery Profile) Ready|
- Capacities 4GB–32GB (with 4GB, 8GB, 16GB and 32GB kits)
- Speeds up to 2666 MHz
- 1.5V & 1.65 V operating voltages enable stable overclocking
- Intel XMP Ready; optimized performance settings handpicked and tested by Kingston engineers
- Dual Channel kit tailored for P55, H67, P67, Z68, H61 (AG), and Z77 Intel chipsets; as well as A75, A87, A88, A89, A78, and E35 (Fusion) AMD chipsets
- Exceptional clock and latency timing specifications to enhance overall system performance
- Heat sink design achieves effective maintenance of speed while prolonging the memory life cycle
- 100% factory tested
First, we get to take a look at the retail packaging these sticks arrived in. Kingston has opted to place their sticks in a form fitting plastic package. The package holds the modules securely, so there is no room for jostling around while in transit to the stores or your home. You can see it has the HyperX labeling in red tape. The sticker displays the full model number, major timings, and other information.
Our first full shot of the RAM shows the typical blue themed heatspreaders (well, unless you have seen the HyperX “Blu” series red sticks or the ‘Lovo’ sticks). The HyperX Predator series has a large “X” in the middle of the spreader along with their namesake. One thing that is pretty obvious in these first two pictures is the size of these modules. With the heatspreaders, they are 2.12″ tall. So, be sure you have ample room to fit these into your system, as they could pose a clearance problem with some heatsink and fan combinations.
Below are some alternate angle shots of the sticks. Gotta love the creativity and quality of some of these photos. A special thanks to Hokiealumnus for taking these shots before moving the sticks this direction!
Overall, I have to admit I like the look of this set of RAM. It’s not full of bling; and has effective branding by way of the name and model on the heatspreaders.
- i7 3770K (Overclockers Approved!)
- Asrock OC Formula
- Kingston HyperX Predator 2x4GB DDR3 2666 MHz
- Galaxy GTX660 GC 2GB (Overclockers Approved!)
- Seasonic 1KW
- Windows 7 64bit
Setting the XMP profile and booting to the desktop resulted in the screenshots below. Everything looked good with the voltage (1.65 V) and the timings, outside of the command rate. You can see it is set to 2T, but the SPD table lists it as 1T. Not a huge deal; just a quick change in the BIOS and go. But, that’s how this motherboard set it up with the XMP profile enabled, and the same thing happened with a different board as well.
Below you will see the results of our usual battery of tests. We have items from your basic AIDA64 Memory tests showing Latency and bandwidth for Read/Write/Copy, a benchmarker’s delight with Wprime 32M/1024M and Super Pi 1M/32M. Rounding out the performance results are more ‘real world’ tests like compression using 7Zip, video conversion in x264, rendering in PoV Ray 3.7, and CPU computation in Cinebench R10/R11.5. Although we tested several other memory kits, I chose to only use like or similar speed models to test against. In this case, the Kingston HyperX was measured against TridentX 2666 MHz and TridentX 2400 MHz kits from past reviews.
Moving on with the show, we’ll jump into AIDA64 and its memory benchmark. The latency on the Kingston HyperX Predator was right in line with the 2400 MHz kit, but slipped a bit when compared to the other 2666 kit with similar timings. Latency is a fickle thing to measure and varies from run to run and system to system. Results can easily be affected by what is loaded on the system or a simple call to the RAM by the CPU. It’s really not something one would notice, and perhaps a fresh install of windows would reduce that difference…perhaps not.
Sliding down to the Copy, Write, and Read speeds achieved, the kit falls right into place matching its brother from another mother 2666 MHz sticks across the board and taking a slight lead in Copy speeds. This lead can also be attributed to what is happening in the system at the time of the benchmark. So, I will call this a draw for all intents and purposes with the 2666 MHz, and a slight victory over the 2400 MHz sticks (not that it would be noticeable in the real world).
To tickle the bencher’s fancy, we will look at the results of Super Pi and WPrime. There isn’t too much to see here to be quite honest; the only benchmark that matters significantly with memory is Super Pi 32M. In this case, the Kingston sticks performed right there with its counterpart in most tests, matching its peers. In Super Pi 32M we saw a three second difference with this kit, which is fairly significant between kits that have matching timings and speed (outside of the 2T timing on the Trident kit). I ran this bench over and over to make sure it wasn’t a ‘cherry picked run’, but I was consistently around the 500 second (8m:33s) range. Perhaps it is a system difference (board/drivers etc), or maybe it’s in tighter sub-timings. Regardless, that is the time posted. Nothing to see really in the rest of these results; but in 1024M Wprime, the Kingston kit did run a hair slower for whatever reason.
Last, we have something a little more tangible than the bandwidth and Super Pi/Wprime benchmarks. Sadly, there isn’t much to see here either in regards to speed differences. For the most part, the tests were within 1% of each other. That said, the HyperX Predator held its own against the other 2666 MHz sticks, so all is well in this testing too. Perhaps when pitted against slower kits we would see more significant differences.
The first thing I tried to do was lower the timings at a slightly raised voltage of 1.725 (same as hokie’s random voltage) to see how tight we could get these things. At these voltages, I was able to drop the CL to 10, but none of the other main timings really wanted to budge. I was able to run a stable Super Pi 32M at 10-13-13-32 at just above its stock frequency at 1339 MHz.
Next, I pushed them at that same voltage (and more in the end) to see what the highest speed I could achieve, and still be Super Pi 32M stable. Sadly, I wasn’t able to go up much at all and managed a 1361 MHz speed (2722 MHz DDR) with stock timings and the voltage upped to 1.8v (you may not want to try that at home!). Even with raising the timings, all of them, she wouldn’t budge past this point.
That isn’t a great result, so I buckled down to see what the deal was. I was trying to determine if it was the RAM or my board/IMC that could be the problem. So, I switched boards to the MSI Z77 MPower, and threw the CPU under an old Single Stage cooler for this testing (approximately -40 °C). I left everything at stock speeds, put up my IO/SA voltages, and crossed my fingers I would see better results. I did, but not by much. I managed to boot and validate at 1400 MHz (2800 DDR). I could pass Super Pi 1M with no problems. However, 32M wouldn’t go through no matter what voltage, timing, slot, or stick I used. Sadly, I think that is the end of the road for my 3770K’s IMC. Not too bad in that respect, but I have to imagine these sticks have a lot more in them than what I can post. Below is a picture of the validation.
It’s pretty tough to wrap things up with memory; about the only thing they are supposed to do is work at their stock speeds. A lot of our readers here benchmark though, so we expect to see some of that too. As you saw earlier, these worked well getting pounded on the RST hardware; so they are good there. Regarding overclocking, I was able to tighten the CL rating down one notch to CL10, overclock to a ‘stable’ 1361 MHz (DDR3 2722) with stock timings, and validate at 1400 MHz (DDR3 2800) with my Single Stage cooler. Again, this isn’t huge overclocking headroom; but I think after testing, this is due to the IMC just banging outside off its limits. However, I looked around some other reviews on the web and their limits were strikingly similar, topping out at 1406 MHz. I personally can’t hold it against these sticks until I see them hangup at the same clocks on a system known to run memory at 2800 MHz+. Not all IMC’s are created equal. With stock speeds at 2666 MHz and few real world results showing differences, I’m not going to hold it against the memory kit from Kingston.
The pricing on this kit comes in at $172.99 at TigerDirect (I was unable to find these on Newegg at the time of publication). Other similar capacity/speed/timing kits (fan included) are coming in at $169 at Newegg.com. This kit does not have a fan, but is a fan needed? Nope. Not even at 1.725v did they get warm on an open air test bench.
About the only drawback I can think of with this kit is the tall heatspreaders. Just be sure you have the clearance with any heatsink and fan combinations you are using, and all will be well.
Overall, Kingston has brought to the table a good looking and fast out-of-the-box memory kit in their HyperX Predator 2666 MHz 8GB (2x4GB) kit. It works great out of the box, one can tighten the CL timing down with a bit more voltage, and it has a bit of headroom to overclock beyond 2666 MHz — just make sure your CPU is up to the task. The Kingston HyperX Predator 2x4GB 2666 MHz kit is…
|Click the Approved stamp for an explanation of what it means|
– Joe Shields (Earthdog)