two klevv cras x rgb ddr4-3600 stick in the cross formation obligatory for every memory review

Let me tell you a story. Back in November, I was working on a review of COLORFUL’s CVN Guardian DDR4. The lighting looked great, but I couldn’t get ASRock’s RGB control software working on my X570 Taichi. I’ve now figured out how to get memory RGB control working on an X570 Taichi, but at the time I needed another kit for a sanity check. The KLEVV CRAS X RGB DDR4-3600 is what I picked up, and now it’s getting a review.

There are a couple of reasons KLEVV caught my eye. Firstly, the price is good – £85.99 on Amazon UK is nice and affordable. Secondly, KLEVV are a brand of ESSENCORE, themselves a division of memory chipmaker SK Hynix. Hynix’s DJR chips have been behind some impressive overclocking results, and CJR is also no slouch. This kit’s rated 18-22-22-42 timings at DDR4-3600 are on the loose side, but we’re hoping for some good gains when we overclock.

In this review we’ll be comparing KLEVV’s kit to a field including two of its peers. Firstly, a TEAMGROUP 8Pack RIPPED EDITION DDR4-3600 14-15-15 kit from Overclockers UK, which uses Samsung b-die. Secondly, an offering from Micron’s in-house brand – a Crucial Ballistix DDR4-3600 RGB kit. Crucial list the Ballistix RGB DDR4-3600 2x8GB kit for £88.79 on their store, just about the same as this KLEVV kit. Meanwhile, Team’s “RIPPED EDITION” kit is a top-end enthusiast kit that uses high voltage to achieve extra-tight timings and now costs £199.99 at Overclockers UK – following recent significant price hikes.

KLEVV CRAS X RGB DDR4-3600 (2x8GB) Specifications & Features

Model KLEVV CRAS X KD48GU880-36A180X
Kit Configuration 2x8GB
SPD Speed DDR4-2666
SPD Timings 19-19-19-43
XMP Speed DDR4-3600
XMP Timings 18-22-22-42
XMP Voltage 1.35V
XMP #2 Speed n/a
XMP #2 Timings n/a
XMP #2 Voltage n/a
Cooling Aluminium Heatspreader
Special Features Addressable RGB
Memory IC 8Gbit Hyinx
IC Guaranteed? No
Ranks Single Rank
PCB Layout A2
Warranty Lifetime (warranty seal on box)

KLEVV don’t specify the ICs used, but as a division of SK Hynix they can be expected to use Hynix’s offerings.

KLEVV CRAS X RGB DDR4-3600 (2x8GB) Closer Look

Packaging and Visual Inspection

The CRAS X RGB DDR4-3600 2x8GB kit comes in a simple cardboard box with a window through to the RGB diffuser. The box isn’t much bigger than it needs to be, and contains a plastic protective clamshell within which are the sticks themselves. There’s no manual, and no link to online information. Thankfully, the motherboard manual should cover installation instructions.

Irritatingly, there’s a warranty seal on the base of the outer packaging. What on earth is the point of that?!? It can still be opened from the top, but this just feels hostile.

We were critical of Crucial for not covering XMP in their support material. KLEVV’s support material, on the other hand, could be improved by existing at all.

The modules themselves are black with gold branding, and a large crystal-like RGB diffuser. They’re also fairly tall. The heatspreaders feel solid, look classy, and are noticeably thick. There are also plenty of grooves cut into the thick aluminium to help with heat dissipation. There’s certainly a quality feel here.

Peeking under the heatspreader, we can see the memory chips arranged in two tight clusters of four chips each. This is consistent with the JEDEC-standardised “A2” PCB layout typical of modern 8GB sticks.

RGB and Aesthetics

The modules look nice, but the height isn’t ideal in our cramped test build. We haven’t been able to manage away our 24-pin cable, and the tall CRAS X module in the outer slot presses into it a bit. Our test build also doesn’t match the gold lettering on the modules, but that’s fine. They’d look great with a more understated board like an MSI B550 Unify. Worth mentioning is noticeable bright spots on the RGB when examined from an angle – the other two RGB kits we’ve looked at recently were better diffused.

One test we do for RGB memory is to look for PWM flicker, using a very fast shutter speed. PWM is commonly used in LED lighting solutions to control brightness, and works by turning the LEDs on and off imperceptibly fast. If the PWM is too slow, this can lead to a flicker that causes eye strain or may even be visible and irritating.

One way to view PWM flicker is by using a camera. As you increase shutter speed, faint banding will appear. At first there are faint blurry lines, but go fast enough and you see solid black stripes.

To see the solid black stripes with this RAM we had to go up to 1/3000th shutter speed, the same as the COLORFUL CVN Guardian. For comparison, a mouse from a big name in the RGB industry shows the same effect at 1/1000th. A cheap remote control LED light bulb also hits the point where the banding is solid black at 1/1000th. We’d rate the flicker on this memory as very low. It is worth noting that COLORFUL’s offering somehow looked better at 1/2000th, so there may be a slight difference in the nature of the PWM control – this kit is also dimmer. Still, it’s a good result.

SPD and XMP Profiles

The Serial Presence Detect, or SPD, is a small chip on a memory module that reports the stock settings to the motherboard. This lets the motherboard know how to run the memory at stock, and also provides XMP profiles for a one-click overclock. Here, we’re looking at how the SPD is programmed.

The “JEDEC” rows are true stock speeds, named for the organisation that defines memory specifications. When the motherboard is left to set up memory on its own, it will go for the best JEDEC speed that is supported by the CPU and any other memory in the system. In this instance, that’s DDR4-2666 at 19-19-19 timings. The XMP #1 row defines settings that are technically an overclock, but which the stick is tested for. To run at XMP settings, you need to go into the bios and enable XMP.

It looks like Essencore/KLEVV are using chips that are rated for only DDR4-2666 and uprating those. As such, the true stock speed is only DDR4-2666 rather than one of the higher standardised speeds of 2933 or 3200. This is standard practice among enthusiast memory manufacturers. However, while it’s not unusual, we would rather see higher JEDEC speeds. Not all users will know to enable XMP, and others may not be comfortable with it. Higher JEDEC speeds mean everyone can get good performance.

Some sticks may also include a second XMP profile. This would be nice to have here as some older systems may find DDR4-3600 a little aggressive, though the latest platforms should all be totally fine. Overall, the SPD is standard – our grumbles apply to just about every manufacturer.

Performance – Test Setup

Since our first DDR4 review, we’ve tweaked the test setup a little. Firstly, it’s in a case now. This helps us evaluate the aesthetics in a more typical setup, as well as frankly saving some space compared to the open test bench. We’ve also updated the BIOS on our X570 Taichi to version 3.80. We have a new PSU, to free up our EVGA G3. Finally, the untimely passing of our RTX 3070 means we’re now on an RX 6800. Because of the changes, all our benchmarks have been rerun. We’ve also dropped CS:GO, which isn’t really memory sensitive on Zen 3, in favour of F1 2019 and Shadow of the Tomb Raider.

CPU Ryzen 9 5950X @ 4GHz 1.1V
Motherboard ASRock X570 Taichi (BIOS 3.80)
Cooler Arctic Liquid Freezer II 280
Power Supply Seasonic Core Gold GC-650 650W
GPU AMD RX 6800
Storage Adata XPG SX8200 Pro 256GB
Case Fractal Design Meshify C TG
Performance – Overclocking

We’ve updated our overclocking methodology to be a bit more in-depth, with the aim of being consistent for reviews going forwards. All the overclocked results you’ll see follow this new methodology.

First, we lock in the tRC specified in the XMP profile, which is usually tighter than what our X570 Taichi sets on its own. Second, we set tRRD_S 4, tRRD_L 6 and tFAW 20 – if needed we could loosen these, but it hasn’t been needed yet. Third, we push to DDR4-3733 with a synced (1866MHz) FCLK and UCLK – loosening any timings that need loosening to get there. We also try to run 1T with geardown mode off if we can without extra tweaking. Finally, we tighten up the main primary timings – tCL, tRCDRD, tRCDWR and tRP. This isn’t a comprehensive overclock, but comprehensive overclocks take a lot longer.

It’s not clear what ICs are in use here, other than the manufacturer being SK Hynix – but it’s most likely CJR or DJR. DJR can take plenty of voltage, but CJR doesn’t show up in kits rated for high voltage. As such, we’re not comfortable with more than 1.4V on this kit. In the end, however, 1.4V didn’t let us tighten any timings more compared to 1.35V.

This is disappointing, to say the least. We’d expected tCL of 16 and tRCDRD of maybe 19 or 20. Being stuck at tCL 18 is unfortunate. This is going to hurt when it comes to overclocked benchmarks. Running with geardown mode off was frustratingly close to stable, which would have helped, but not quite there. More tweaking could probably stabilise GDM off, but more tweaking could also get our other kits faster.

Performance – AIDA64

Produced by FinalWire Ltd, AIDA64 is a system information and diagnostic suite with an extremely wide range of features. Memory enthusiasts are particularly drawn to the synthetic cache and memory benchmark. AIDA64 latency is a common point of comparison for those looking to get the absolute most out of their systems.

It’s a narrow loss, but a consistent one. This is hardly surprising – KLEVV are hitting the same clock as Crucial’s equivalent kit, but with looser timings.

AIDA64 scores wouldn’t be complete without the screenshots people love to use to compare results – here are all three, for JEDEC, XMP and our manual overclock.

Performance – Geekbench3

Geekbench3 is a cross-platform synthetic benchmark, made by Primate Labs. Like AIDA64, this is a synthetic task. Geekbench3 is heavily weighted towards the artificial memory bandwidth test.

Much like AIDA64, Geekbench3 shows a narrow but consistent loss.

Performance – y-cruncher

A “high school project that went too far”, y-cruncher is a highly optimised constant calculator. Written by Alexander J. Yee, the program can compute pi and other mathematical constants to trillions of decimal places. Because the Ryzen 9 5950X is so strong, y-cruncher is extraordinarily memory sensitive on our test system. We’re benchmarking the time to calculate 1 billion digits of pi.

Another narrow loss. This review isn’t going well for the KLEVV CRAS X RGB DDR4-3600 kit. In all fairness, it seems like the frequency is ultimately mattering a lot more than the timings.

Performance – OpenShot

OpenShot is a free, open-source video editor ideal for content creators on a budget. We’re taking a 1080p 50fps source, and exporting a short video with basic transitions at 720p 30fps. The rendering speed, in frames per second, is a way to measure real-world performance in a demanding application.

OpenShot is a little more variable than our other benchmarks. Functionally the Crucial Ballistix, KLEVV CRAS X and TEAMGROUP RIPPED DDR4-3600 kits are all close enough at XMP to be just about equivalent.

Performance – F1 2019

F1 2019 is a well optimised DirectX 12 game that runs well even on an Athlon 3000G, and has a built-in benchmark. Since the GPU side is so well optimised, our high-end setup is memory sensitive even at 1080p Ultra High settings.

It seems like F1 2019 might bump into a bottleneck elsewhere around 280 FPS in our setup, with the 1080p Ultra High settings we’re using. As such our manual overclock pulls even. However, the performance deficit suffered by the KLEVV CRAS X at XMP compared to its DDR4-3600 peers is bigger than elsewhere in this review. Even lower clocked kits that we benched to fill out the review are performing better than the KLEVV CRAS X, thanks to better latency.

Performance – Shadow of the Tomb Raider

More than just a showcase for Nvidia RTX, Shadow of the Tomb Raider has become popular among memory enthusiasts since it shows off the benefits of overclocked RAM. We ran the built-in benchmark at both 800×600 Lowest, which is entirely CPU/memory bound, and 1080p Medium for a more realistic test.

This isn’t a great showing – we’re seeing losses to lower clocked kits, notably our token dual rank DDR4-3000 kit. The loose timings are certainly rearing their heads in games.

The KLEVV CRAS X RGB DDR4-3600 (2x8GB) Review: The Verdict

If you’re a regular reader, you might be able to guess the conclusion here already. When we were reviewing Crucial’s Ballistix RGB DDR4-3600, we compared it directly to the KLEVV CRAS X RGB DDR4-3600. We said then, “The KLEVV kit in the same price range is consistently behind”. This time around, the Crucial kit in the same price range is consistently ahead. Loose timings aren’t a big deal in some of our workloads, but embarrass the KLEVV CRAS X RGB DDR4-3660 2x8GB kit in the gaming section of this review.

What’s hot:

  • Nice heatspreaders
  • Works well with motherboard RGB software
  • Performance could be worse

What’s not:

  • Slower than competing kits around the same price
  • Poor diffuser with noticeable bright spots from RGB LEDs
  • Disappointing overclocking

Overall, the kit works fine. It’s slower than the competition, but could be worse. If you just love the looks, you can go for it. It won’t make your system bad. However, it’s not the kit we’d recommend.

This kit was purchased independently, so we don’t have to thank anyone for providing a sample. If you found the review helpful, feel free to leave a comment below or on the Play3r Facebook page.

Related Reading:

REVIEW OVERVIEW
Design
Build Quality
Value
Performance
Previous articleWired2Fire Sim X-2020 Valkyrie PC Review
Next articleTerraMaster Launch F5-422 5-bay NAS with 10GbE Networking
klevv-cras-x-rgb-ddr4-3600-2x8gb-reviewIt isn't bad, but with loose stock timings and disappointing overclocking, the KLEVV CRAS X RGB DDR4-3600 (2x8GB) isn't the kit we'd recommend.

2 COMMENTS

  1. tested on 11th gen i5-10400 mobo is msi b560m pro vdh-wifi..stuck @2666,wont go higher no matter how i configure in bios. i have corsair 3600 and it works fine on same specs.

  2. Got these Klev to just under 4000 stable 1:1 with the infinity fabric ,stock voltage, tried tightening the timings just gives whea errors ,dram calculator numbers are not stable, 5800x, Asus b500 F gaming, and these work a lot better with Armoury Crate compared to my Crucial Ballistic 3600 which are just a nightmare with Armoury crate.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.