Kingston HyperX Beast DDR3-2133 64GB RAM Kit Review

Kingston is coming at us with an entire lineup of RAM bearing the name HyperX Beast, complete with a heatspreader redesign and impressive specifications, with speeds up to DDR3-2400 and capacities up to 64GB. The word “beast” implies big things, and big is what you’ll get today, in the form of their biggest kit at the fastest speed (at that capacity) – the DDR3-2133, 64 GB beast of a kit, model # KHX21C11T3K8/64X

Specifications & Product Tour

Features / Benefits

• Unique — black heatspreader
• Reliable — 100-percent factory tested
• XMP profiles — factory configured for best maximum performance
• Guaranteed — lifetime warranty, free technical support

Specifications

• Capacities 8 GB, 16 GB, 32 GB and 64 GB in kits of 2, 4 and 8
• Frequency speed 1600 MHz–2400 MHz
• Operating temperature 0 °C to 85 °C
• Storage temperature -55 °C to 100 °C
• Dimensions 132 mm x 41.4 mm
• DDR3 only
• Available in a variety of CAS latencies
• Voltage 1.5–1.65 V

The kit supplied today is model # KHX21C11T3K8/64X. After talking with Kingston, we’ve agreed to bring this to you in testing as a 32 GB kit (which would be model # KHX21C11T3K4/32X) because to be brutally honest, the memory controller on our i7 3960X is less than stellar and pushing past DDR3-2133 is not something it likes to do. Thus, to show whether these can be overclocked, we needed to test on our i7 3770K, which has a stellar IMC. As the Z77 platform maxes out at four sticks, you’ll see how 32 GB performs in a 4 x 8 GB configuration.

When you plug these sticks in and choose to have the system read via XMP, this is what you’ll get, with the exception of the Command Rate, which was manually set at 1T since the RAM is happy there.

Now we’ll take a look at the 64 GB HyperX Beast kit. Never have I seen a set of RAM come in a bigger box. You’d think I had received a massive heatsink or something. Its packaging is definitely befitting its name.

After unloading everything, you can see the 64 GB kit is sent with two HyperX fans (for use in separated X79-based RAM configurations), with the RAM packaged in two-stick sets. They didn’t just throw four two-stick sets together; each stick shares the same model number and specification as a 64 GB kit, it was probably just more convenient to use existing packaging than come up with something new.

The HyperX fan assembly is metal and has two 60 mm fans mounted. The fans are rated for 0.16 A and operate very quietly.

The assembly goes together quickly and easily. I like the metal construction, these won’t vibrate or fall apart. The fans will croak long before the housing ever has any issues.

Because of the configuration on the motherboard used for testing (ASUS Maximus V Extreme), the first graphics slot is too close to the RAM slots to use the fan, so there won’t be any installed photos for you today.

Now on to the main event – HyperX Beast, in the flesh. The new heatsink design is very aggressive looking and befitting the name.

When looking at the kit not installed, it makes you wish they had considered switching to black PCB, but remember you won’t see any PCB when the sticks are installed unless you’re looking very hard for about a 1/4″ x 1/16″ piece just above the RAM tabs on your motherboard. Everything else will either be covered by the motherboard or by the nice looking heatspreaders.

When pulling the kit out, I just kind of let them go where they like, then smacked them and told them to get in line.

64 GB is a LOT of RAM. It’s also fun to take pictures of so much RAM, so there are a few more before we’re done with the photo shoot.

Ok, ok…one more and we’re done.

The HyperX Beast sticks sure look the part. The heastinks are aggressive looking and lightweight. Importantly they are also thin – thin enough to go into tight-fitting mITX configurations, unlike some (*cough* Adata). Heatsinks are mostly for looks now-a-days, with DDR3 RAM not really getting too warm unless you over-volt it, which most IMCs don’t appreciate very much. Since they’re mostly for looks, I’d say Kingston did a good job in that department.

Test System

Here they are installed in our Z77 test bed. As you can see, the green PCB isn’t nearly as obtrusive as it was when they weren’t installed. Two thumbs up for how they look in-system.

Our test system today is an Ivy Bridge based system using ASUS motherboards and several varieties of RAM.

Of course, a kit this large is really designed for use in an X79 platform. While we’re not testing them in that platform due to its weak IMC, it doesn’t hurt to make sure they’re stable at the rated speed and have a look at them installed. They definitely look great when paired with a beautified Rampage IV Extreme-based system. I don’t think I’d even want to put the fans in there; they look stellar without them.

Stability at Rated Speed

The first thing we need to do with RAM is test for stability. Our older, out of date UltraX R.S.T. Pro 3 PCIe couldn’t handle 32 GB (and was promptly and graciously replaced by UltraX), so the first thing tested was stability in-OS with HyperPi. This is a killer version of SuperPi that runs as many instances as your CPU has threads. If there is instability, this will show you pretty quickly.

On the Z77 platform, stability testing was quick and painless, with no failure to speak of.

The X79 platform with all 64 GB installed gave the same quick and stable result.

Now we get to the hard part. The UltraX team actually sent two R.S.T. Pro 3’s to us this time, the R.S.T. Pro 3 PCIe card and the R.S.T. Pro USB. Both do the same things, they just (obviously) plug into different slots. Both of these are dedicated pieces of hardware that directly interface with the system to ferret out any possible RAM instability. They work great for doing just that, but the test can take a loooooong time. They test every possible address, and when you’re talking about 32 GB of addresses, there’s a lot of work to do.

Many manufacturers use the R.S.T. Pro tools to test their units before they go out the door. This is an industry accepted, module manufacturer’s tool. They usually only loop the test three times. We’re overclockers though, so the stress we can put on RAM is more than your average Joe Blow stock configuration user. Thus, we run it five times.

I’m happy to report after 17 hours of being pummeled by the R.S.T. Pro, the HyperX Beast came out with zero errors and five tests passed.

Performance

Performance at rated speeds on much of the RAM on the market is always going to be right where you’d expect it to be. To be quite honest, anything north of DDR3-2133 is aimed at benchmarkers and anyone that just wants fast RAM because they can have it.

The biggest differences you’re going to see in testing are in synthetic benchmarking like AIDA64. Aside from that, SuperPi 32M is really the only benchmark you’ll see a very noticeable difference between speeds like DDR3-2133 and DDR3-2666. In “real world” use (for our purposes, rendering & encoding), you’ll see at most a 2% difference, most of the time even less than that.

As such, we’ll just present the results without comment. You can read the graphs as well as I can tell you what they say. All results are graphed using relative performance, with the HyperX Beast results always 100% and the other two kits expressed as how they did relative to the HyperX Beast kit. Actual results are listed next to the percentages.

Synthetic Testing

Side note: “Relatively Expressed ns” means as the time goes down, the percentage goes up, so if one kit comes in at 40 ns (100%) and the other comes in at 45 ns, the 45 ns result will be 88%, meaning the kit whose latency was 40 ns is 12% better than the kit whose latency was 45 ns. If that’s clear as mud and you need further clarification, don’t hesitate to ask in the comments!

Video Conversion and Rendering

Benchmarking

Overclocking

Overclocking is where modules we review really separate themselves from the others. This is where the rubber meets the road for these kits and is, frankly, much more interesting than the more time consuming benchmarking & graphing at stock speeds!

First off, I wanted to see how far the kit could go at its stock timings and voltage (11-12-12-32 / 1.65 V). Remember, this is with four large DIMMs, which is always more difficult to overclock. So, when they easily went to DDR3-2200, SuperPi 32M stable, I was impressed. Things are starting out swimmingly.

Next, I bumped the voltage to 1.75V (not recommended for 24/7 use!), which allowed successful completion of SuperPi 32M at an impressive DDR3-2326.

Then I tried to see how tight the timings could get. Frankly, I expected even less in this department than in the MHz one. Fortunately, the Beasts just kept on giving! At their rated 1.65 V, the HyperX Beast kit, which remember is rated for 11-13-13-32 at 2T (though I ran all tests at 1T), managed to be SuperPi 32M stable at its stock speed with impressively decreased timings of 10-11-10-28.

Last but not least I went back to stock timings and 1.75V but wanted to see how far the MHz would go and complete the (much easier on memory) SuperPi 1M. Turns out, it wasn’t much further than it would complete SuperPi 32M. The final result? DDR3-2400!

Final Thoughts & Conclusion

Kingston has a winner with this HyperX Beast kit. At stock, it is stable and then some. You can’t get better stability testing (as a consumer) than with the UltraX R.S.T. Pro hardware and the HyperX Beast DDR3-2133 kit passed with flying colors, even with the 1T command rate at which we ran all of our tests. Stability, check.

The new Beast heatspreaders also look great. They will match almost any system (what motherboard out there doesn’t have some -or a lot of- black?) and their aggressive styling is sure to make any system look good. Looks, check.

Now, what am I forgetting? Oh, yea, overclocking! This HyperX Beast kit overclocks, well, like a beast. If you want tons of capacity combined with impressive overclocking ability, this kit has it in spades. Overclocking, check.

Of course, price is always a factor. RAM has been getting a bit more expensive lately due to somewhat limited IC supply. Coming off an insanely cheap market for quite a while now, that give some people pause when thinking about extra capacity. It’s out of stock right now, but the 32 GB HyperX Beast DDR3-2133 kit retails for $264.99. If you don’t quite need all that capacity (or need four sticks for quad channel), you can get a 16 GB kit with the same specs for $133.99. Going even smaller, the same ratings and only 8 GB of capacity goes for $73.99. As far as the 64 GB kit, Newegg doesn’t have it listed currently, but some Google results show prices around $545.

All of those prices are actually quite reasonable and right where they should be at their place in the market. This kit is stable as it needs to be, looks good and overclocks like its name implies. If you need a lot of capacity at higher rated speeds, the HyperX Beast DDR3-2133 series should be a serious contender for your computing dollars.

Click the Approved stamp for an explanation of what it means.

– Jeremy Vaughan (hokiealumnus)