Oh DDR3… it was real, it was fun, but is hasn’t been really fun. DDR3 has came, saw, kicked some rear, and is now starting to see its way out. While Intel’s mainstream platform and all of AMD still use it, the release of DDR4 spells the eventual end to our close friend. Its time to learn something new folks. DDR4 brings lower voltage requirements, higher densities, and higher speeds. However, the trade off is higher latencies for now. While that might sound like a negative, in the end it doesn’t matter terribly much. In time, just like in the DDR3 life-cycle, we will see speeds increase and the timings go down.
My first dip into the DDR4 arena is reviewing a kit from Kingston. We all should be familiar with their Hyper X Predator line, and that is what we have in our hands. This kit from Kingston comes in a 4×4 GB flavor (since X99 is a quad channel platform) set to 3000 MHz. It has a CL rating of 15 and a base voltage of 1.35 V. It is time to show off the sticks in a photo shoot, put them through my testing suite, and end up with some overclocking!
Meet the Ram – Specifications, Features, and Product Tour
Kingston sent their HX430C15PB2K4/16 kit. When breaking this ‘code’ down, we can discern the following. The “HX” stands for Hyper X, the “4” for DDR4, the “30” for 3000 MHz, the “C15” is the CL rating, “PB” is Predator Black, and the “K4/16” is for a kit of 4 totalling 16GB. The ram I have in hand was still warm from the factory when it landed. So much so in fact that it still had older specifications burned into the JEDEC. Shortly after their arrival, Kingston mentioned to be sure to lower the voltage to 1.35 V instead of the default 1.5 V it boots up to as that is what the retail voltage will be. This was not a problem at all. Rock solid there too. I mentioned all this above for specification reasons of course, but wanted to note that what you see in my pictures will not match what I have described and is on their website. But moving forward, what I have above is the part number. Please do not be too confused as you will not see these differences in the wild. We just happened to get the kit before they made the changes.
Moving on, below is a list of specifications and Features from the Kingston website.
|Row Cycle Time (tRCmin)||46.5ns(min.)|
Refresh to Active/Refresh
Command Time (tRFCmin)
|Row Active Time (tRASmin)||33ns(min.)|
|Maximum Operating Power||TBD W*|
|UL Rating||94 V – 0|
|Operating Temperature||0c to 85C|
|Storage Temperature||-55C to 100C|
|Module Height (w/spreader)||2.166″ (55mm)|
|XMP Timing Parameters|
JEDEC: DDR4 2133 CL 15-15-15 @ 1.2v
XMP #1: DDR4 3000 CL 15-16-16 @ 1.35v
XMP #2: DDR4 2666 CL 14-14-14 @ 1.35v
- Quad channel available
- Capacity 16GB (4×4 GB kits)
- Speeds up to 3000MHz
- 1.2 & 1.35 voltages for stable overclocking
- Intel XMP-ready profiles optimized for Intel X99 series motherboards
- Exceptional clock and latency timing specifications to enhance overall system performance
- Heat sink design achieves effective maintenance of speed while prolonging memory life-cycle
- 100-percent factory tested
Below are pictures of CPU-Z and what its Memory and SPD tabs have to show for the kit at its rated speed. Again, the voltage listed is different as the updated retail version runs off 1.35 V in XMP mode.
We normally start off our product tour by showing you the retail packaging. Due to this kit being hot off the presses from Kingston, we simply have the cardboard box and a label. Note the part number on the label is not correct. Retail versions will see part number HX430C15PB2K4/16.
After opening the box, we see the kit comes in a form fitting clam-shell style plastic packaging to protect it from movement and damage. Again, retail versions I am sure will have some marketing on it to pretty it up.
Our first look at a stick in the set shows the black Hyper X Predator heatsink that covers the DIMM. I like the look of these, especially in black, but you have to make room for these as they are tall coming in at 55 mm. Be sure you have the clearance to use these! On that note, I am still not sure why memory makers are using these tall heat spreaders. I put 1.5 V through them and even after running Hyper Pi 32M back to back several times, they were barely warm to the touch. I have seen much smaller heatsinks do the same great job and would like to see these shrink as well. If for no other reason, so it doesn’t alienate potential buyers due to the ever growing CPU heatsinks available on the market today.
Outside of that, the only differences you may notice is DDR4’s move to 288 pins as opposed to the 240 found on DDR3.
Again, a great looking set of sticks to me…
I put all of the sticks together in some odd nerdy guy’s form of art in the gallery below…
Below you will see the results of our usual battery of tests. We have benchmarks 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, and rounding out the performance results are more ‘real world’ benchmarks like compression using x264 (video conversion), and CPU computation/rendering in Cinebench R11.5/R15. Since this is our first crack at at DDR4 we do not have anything to compare to in an apples to apples environment, so we will simply see the results. As we get more kits to test, we can compare them against each other in a graph. Please check out Lvcoyote’s reviews of DDR4 to see some of what else we have reviewed on the DDR4 front!
Before we start though, I have listed out the system that will be used moving forward for memory testing.
- i7 5820K
- MSI X99S Gaming 9 AC
- Kingston HyperX Predator 4x4GB DDR3 3000 MHz @ 1,35v
- AMD R9 260x
- Seasonic 1KW
- Windows 7 64bit
Super Pi 1M/32M
As I mentioned towards the beginning of this article, we see the blazing speeds of all DDR4 kits, but we also see a big increase in timings to compensate for the bandwidth and low voltage that DDR brings to the table. However it is a function of DDR4 itself from my limited exposure of it. Our very own Memory guru, Woomack, has looked at a couple of kits and has seemingly run into the same thing.
I have an email out to Kingston and we are working together to see if the results I have are in the ballpark or if there is some setting I may have missed to allow for better bandwidth on this specific kit. I will report back with those results. For now though, we can see a lot more bandwidth available on this platform with a lot lower voltage as well.
EDITORS NOTE: After working with Kingston they are not able to reproduce the problem on the same board, so it appears my system and I have some soul searching to do. Though honestly, I have no idea what to even look for as I have done everything I can think of from a fresh OS, to different BIOS to different boards.
This will be a fairly sad section in my world and is of no fault of the sticks. For some reason, I was not able to raise the BCLK past 125, period. I could get to 166 MHz and under 3000 MHz but could not hit the 3334 MHz which would not boot no matter what I did. This made overclocking difficult at best. I have reached out to MSI to see what is going on (and taking a healthy look in the mirror this being a new platform).
That out of the way, we can tighten the timings so that is what I have to show. I was able to drop down to 14-15-15-37 from the stock 15-16-16-39 using stock voltage of 1.35 V and be 32M Hyper Pi stable. I raised the voltage to 1.5 V and kept on tightening the screws. I was able to get 13-14-14-37 and complete the AIDA64 Memory tests, and that was about it as far as stable goes. So, there is room to tighten this kit down, and surely there is MHz headroom too. But as I said, I am likely lacking somewhere on pushing this platform up. I will update this later when I hear back.
Its still pretty early in the DDR4 era, so things on the speed and timings front will likely change as time goes on. Still, we have a wicked fast speed already at 3000 MHz out of the box. Remember that all consumers will see the 1.35 V version out of the box and will not have to manually adjust things as I have done. As far as performance goes, things were a bit off but after working with Kingston, they are unable to reproduce it on different brand boards than mine, and then on my board. Others have reviewed this kit and saw better values on different boards so the concern seems to lay somewhere in my system at the moment. I will update the conversation thread with this information as I get it, but at this time assume the issue is internal and not that of Kingston from the evidence we have.
Timings tightened down fairly well and overclocking just hit a wall for me personally. I cannot fault Kingston for this lack of headroom at this moment. But, like with everything in overclocking, what one CPU/GPU/Memory Kit may do, the next one may not.
We know from being new, DDR4 demands a premium on its price. I haven’t seen these pop up at Newegg yet, but at Provantage I am seeing them priced at $334.87 with a list price of $373. When comparing this price to other kits of the same speed/voltage and similar timings at Newegg.com, it is priced on the cheap… big time. Newegg’s cheapest comes in at $499 with slightly tighter timings (15-15-15-xx). I cannot speak well to its overclocking abilities, however the price is surely right being the cheapest available at this moment against similarly spec’d kits. It was rock solid stable at stock timings and the stock voltage of 1.35 V, and that’s all most want to ask for. She’ll tighten down a bit too. This kit is Overclockers.com approved!
– Joe Shields (Earthdog)