Test System
Test System |
---|
Processor: | Intel Core i9-12900K Locked P-core 5.0 GHz, E-cores 3.8 GHz
AMD Ryzen 7 5800X @ 1.325 V 4.6 GHz OC All-Core |
---|
Motherboards: | Gigabyte Z690I Aorus Ultra Plus ASUS TUF Z690 Gaming
MSI MEG B550 UNIFY-X |
---|
Cooling: | Custom loop: EKWB Quantum Velocity2, D5 pump, 380/480 mm radiator |
---|
Harddisk: | 1x Neo Forza NFS01 480GB SATA 6 Gb/s SSD (OS) 1x Neo Forza eSPORTS 1TB NVMe M.2 (data) |
---|
Graphics Card: | NVIDIA RTX 3080 Ti Founder Edition |
---|
Power Supply: | Corsair HX1200 Platinum 1200 W |
---|
Case: | Primochill Praxis Wetbench |
---|
Software: | Windows 11 Pro 64-bit (Intel) Windows 10 Pro 64-bit (AMD) NVIDIA GeForce 512.95 WHQL |
---|
Comparison Memory Kits
Memory Comparison |
---|
SKU | Brand | Name | Config | Speed | Timings | tFAW | tRFC | Voltage |
---|
F4-3200C14D-16GVR | G.Skill | Ripjaws V | 2x 8 GB | 3200 MT/s | 14-14-14-34-48 | 39 | 655 | 1.35 V |
---|
BL.9BWWR.300 | Acer | Predator Vesta RGB V | 2x 16 GB | 3600 MT/s | 14-15-15-35-50 | 44 | 631 | 1.45 V |
---|
F4-3600C14D-16GTZNB | G.Skill | Trident Z Neo | 2x 8 GB | 3600 MT/s | 14-15-15-35-50 | 44 | 631 | 1.45 V |
---|
F4-3600C14D-32GTRS | G.Skill | Trident Z Royal | 2x 16 GB | 3600 MT/s | 14-15-15-35-50 | 44 | 631 | 1.45 V |
---|
54N61AA#UUF | HP | V10 RGB | 2x 8 GB | 3600 MT/s | 14-15-15-35-50 | 42 | 631 | 1.45 V |
---|
AX4U36008G17H-DC50R | XPG | Spectrix D50 ROG | 2x 8 GB | 3600 MT/s | 17-21-21-43-64 | 44 | 630 | 1.35 V |
---|
BLM2K8G40C18U4B | Crucial | Ballistix MAX | 2x 8 GB | 4000 MT/s | 18-19-19-39-64 | 42 | 701 | 1.35 V |
---|
NMGD432F82-4000FF20 | Neo Forza | MARS | 2x 32 GB | 4000 MT/s | 19-23-23-46-70 | 43 | 699 | 1.4 V |
---|
F4-5066C20D-16GVK | G.Skill | Ripjaws V | 2x 8 GB | 5066 MT/s | 20-30-30-50-80 | 61 | 889 | 1.6 V |
---|
Initial Setup
Once installed in one of the test systems, we can see the memory RGB lighting. Unfortunately, the camera photos aren't flattering. The RGB lighting isn't as dim in real life, and the edges are darker because the camera caught the memory in the transition state between colors. After taking a number of photos, the best examples are when the LEDs are set to static color. The issue is that those photos are a blurry mess, so these seen above are used instead.
Before moving on to testing, I have some technical notes to go over before proceeding to benchmarks. Getting straight to the point, when it comes to newer generations of processors from Intel and AMD, the best performance is achieved by keeping the memory synchronized in a 1:1 ratio (Gear 1) to the memory controller. By default the memory should operate in the 2:1 Ratio (Gear 2) mode. However, this can be detrimental to performance in some applications.
Seeing as this is dual-rank memory (i.e., two-sided DIMM and two pools of addressable memory), Command Rate 2 (2T) was enabled on Intel LGA 1700 and Gear Down Mode (GDM) on AMD AM4. On the Intel LGA 1700 platform, adjustments to the System Agent and VDDQ_TX voltages were required (at first). Starting with the ASUS Z690 TUF Gaming DDR4 motherboard, it could not operate the memory at DDR4-4000 1:1 Ratio. Switching to the Gigabyte Z690 Ultra Plus DDR4 motherboard, the kit worked without needing to change any voltages (just memory ratio). This Gigabyte motherboard was used for the benchmarks. For AMD, voltage tuning was required to keep a FCLK ratio of 2000 MHz, and therefore retain 1:1 synchronization for optimal performance.
To reiterate, there is nothing wrong with the hardware itself, but it likely requires tuning of voltages, and knowledge more than just enabling X.M.P to get this kit stable if you plan on operating the memory in a 1:1 synchronous ratio to the memory controller. It should be stated that not all CPUs have high enough quality memory controllers. Motherboards also need to be up to a certain standard for the memory traces, otherwise this memory kit will not work for all setups if you are trying to achieve peak performance, i.e. Gear 1.
Supporting Hardware