• CONTEST ALERT - Experience the power of DDR5 memory with Kingston Click for details

Kingston FURY DDR5 Review

SaiyanGoku

kamehameha!!
For the contest, all of the reviewers were given 2 sticks of 16GB Kingston FURY Beast DDR5 RAM clocked at 5200 MT/s.
Of course, since you can't do anything with just RAM, we got the following parts for assembling a config:
CPU: Intel Core i9-12900K
Motherboard: MSI MAG Z690 Tomahawk WiFi (DDR5 variant)
Storage: Kingston FURY Renegade 2TB PCIe Gen4 x4 SSD
GPU: MSI GeForce RTX 3060 Ventus 2X OC
PSU: MSI A750GF 80+ Gold Modular Power Supply
Intel doesn't includes a cooler with "K" series CPUs so a Deepcool Assassin III 140mm Dual Tower Cooler was arranged as well with LGA 1700 mounting brackets.

qBCMKOl.jpg


Personal peripherals used: Acer VG240YS monitor, SteelSeries Rival 110 mouse, Xbox One Controller, Corsair HS50 Pro, Epomaker S68 Keyboard, AmazonBasics mousepad.
guz4FtD.jpg

There are links to each part from their respective manufacturer's website so that you can go over the official specifications and where to buy them from.

The star of the show is cladded in metal heatsinks (which I assume is Aluminium) and they really are necessary to prevent over heating. These are somewhat low profile and do not have RGB (so, no +10% FPS) which is a good thing from my point of view. Consumer DDR5 RAM, unlike DDR4, has it's own Power Management integrated circuits (PMIC) and Voltage Regulation Modules (VRM) on the DIMM. Earlier, both of these were handled by the motherboard. This is the reason why there isn't a single motherboard which supports both DDR4 and DDR5 DIMMS simultaneously, even when the CPU can support it, at least from a major manufacturer. You can get away with using bare DDR4 DIMMs but I wouldn't risk doing the same with DDR5 DIMMs. It would be even better if your case has cool air flowing directly over the modules. I never thought that RAM would get as fast as or even faster than the CPU itself since i9-12900k is rated at 5.2GHz.
gAVTDAF.jpg


The Setup:
For my open bench setup, I referred the motherboard manual first. RAM went to the designated slots for dual channel operation. I put the SSD in PCIe NVMe Slot no 1 which is directly connected to the CPU. Then, I connected the GPU to top x16 slot, put the CPU and cooler such that the fans would have been pushing air towards the top of the case. For the entirety of the tests, I was touching the ram heat spreader. While it got a bit warm, it never became too hot to touch. My ambient temperatures were around 40° C and only ceiling fan was powered on. I limited the CPU power limits to PL1=200W and PL2=241W using Intel XTU since the default power limits were set to 4096W (YIKES!!)

ec5LiJv.jpg



I have mixed usage and any comparisons to DDR4 performance are based on my current laptop (mentioned in signature).

Synthetic Benchmarks:
Starting off with read/write performance, I created a 16GB dynamic RamDisk using ImDisk. Then CrystalDiskMark v8.0.4 was used to run these benchmarks.
PCIe Gen3 x4 SSD 5 passes
KCbrKgS.png


RAM Disk DDR4 2666 MT/s 5 passes
SMurioO.png


RAM Disk Stock DDR5 4800 MT/s 5 passes
xOB8GHk.png


RAM Disk XMP1 DDR5 5200 MT/s 5 passes
s78r1Pw.png


PCIe Gen4 x4 SSD 5 passes
9nml7x1.png


Overall
L6nLLLl.png


Benchmark Conclusion: DDR4 2666 is already faster than a PCIe Gen3 x4 SSD by a minimum of 91% and as high as 972%. DDR5 just makes the latter look even worse. DDR5 is atleast 37% faster than DDR4 and 43% faster than the same SSD. PCIe Gen4 SSDs sit at the top for Read/Writes currently but faster RAM in future should eventually surpass them as well. The Random 4KiB read/write values of DDR5 RamDisk are higher than Sequential read/write speeds of SATA SSDs (which is usually used for advertisement purposes). I never thought SSD would be a bottleneck for RAM. Potential buyers should not buy SATA SSDs for a DDR5 build.
KtTSuiE.png



For real world test, Handbrake was used to transcode videos from x264 (H.264) to x265 (HEVC) with Nvidia NVENC Encoder.
For this test I used a 1.37GB 480p H264 mp4 file with 4 different scenarios.
1. Original on RamDisk, Transcoded to RamDisk
2. Original on RamDisk, Transcoded to SSD
3. Original on SSD, Transcoded to RamDisk
4. Original on SSD, Transcoded to SSD

Eza7ly0.png


By using a RamDisk, you can save good amount of writes to your SSD and prolong its life if yours is a low endurance one. As most people are used to the 4th scenario they can save around 13% of time if they choose to opt for either 1 or 3. Also, there a massive reduction of around 58% transcoding time when comparing DDR4 vs DDR5 RamDisks.

Link to setup pics, benchmark, etc albums:
 
Last edited:
OP
SaiyanGoku

SaiyanGoku

kamehameha!!
GAMING:
Now ends the "non fun" part and "fun" Gaming section begins.

My current laptop has i7 9750H, 115W RTX 2060, 32GB DDR4 RAM at 2666MT/s.
I play mostly single player games with only 2 eSports titles. My gaming test suite comprised of the following games:
  • GTA V
  • Cyberpunk 2077
  • Horizon Zero Dawn
  • Metro Last Light Redux
  • Metro Exodus
  • PUBG PC
  • Rocket League
  • The Last of Us (via RPCS3 version 0.0.21-13399)
  • Red Dead Redemption (via Xenia version Master@9b1fdac98)
9q8h94P.png


The games were configured to a mix of medium-high preset at 1080p with ray tracing and DLSS both set to off.
This how they performed:

Test Bench (fps)Current Laptop (fps)
Cyberpunk 207713270
GTA V180125
Horizon Zero Dawn12080
Metro Exodus10360
Metro Last Light195132
Metro Last Light (RAMDisk)195133
PUBG PC16598
Rocket League165165
The Last of Us (RPCS3)5230
Red Dead Redemption (Xenia)6335

The test bench was 60% better than my current laptop in native PC games. I capped Rocket League to 165 fps because that's the refresh rate of my monitor and going above that hasn't made any difference till today in the same.

I will not be explaining how to setup the emulators since there are better guides available on YouTube. The test bench was 73% better in RPCS3, which is CPU intensive while it was 80% better in Xenia since it prefers GPU more. I do play PS2 games on my laptop but since phones have started to get powerful enough and have decent PS2 emulators, spending time on testing it was a moot point. Same is applicable for other retro consoles.

Just to see if there would be any performance delta, I moved Metro Last Light Redux from the SSD to RAMDisk. Needless to say, the performance was identical. As it was during the synthetic tests, RAM heat spreader wasn't too hot to touch.

VRAM usage in all cases never exceeded 6 GB. This would seem like a rant but 12 GB VRAM for a RTX 3060 is overkill in my opinion. 8GB would've been more than enough and could've saved some cost to the end consumer.
 
Last edited:
OP
SaiyanGoku

SaiyanGoku

kamehameha!!
Final words:
DDR5 RAM works as an incremental upgrade over sub 4000 MT/s DDR4 RAM but there's an early adopter tax, which is the trend across every newer hardware launch. The prices would eventually normalize, just like it happened with DDR4 when DDR3 used to be mainstream since the latter is on its way to being obsolete.

Official links to products provided by Digit and Kingston for the review:
RAM: Kingston FURY Beast DDR5 Memory - 16GB, 32GB 4800MT/s, 5200MT/s, 5600MT/s, 6000MT/s - Kingston Technology
SSD: Kingston FURY Renegade – High-performance PCIe 4.0 NVMe M.2 gaming SSD - Kingston Technology
CPU: Intel Core i9-12900K Processor
Motherboard: MAG Z690 TOMAHAWK WIFI
GPU: MSI GeForce RTX 3060 VENTUS 2X 12G OC | Graphics Card | MSI Global
Power Supply: MPG A750GF | Power Supply | Power Within Reach
CPU Cooler: ASSASSIN III - DeepCool
 
Last edited:
Top Bottom