Test System
In typical TechPowerUp style, let's dive into the performance nuances of various DDR5 memory configurations. Starting with the JEDEC baselines, we have the DDR5-4800 CL40 in a 2x 16 GB single-rank setup at 1.10 V. This is your standard fare, often found in prebuilts or older systems. Alongside it, the JEDEC DDR5-5600 CL46 as 2x 8 GB offers a slight bump in frequency but has fewer memory chips as covered in our
XPG Lancer Blade RGB DDR5-5600 16 GB CL46 Review. However, it still remains within the realm of typical lower-end consumer setups and prebuilts.
The recent appearance of a single 16 GB stick of DDR5-6000 CL30 in combo bundles with motherboards and CPUs is particularly noteworthy in the context of the ongoing RAM pricing challenges, providing a potential cost-saving measure. By using only half the memory interface, or one channel, we can closely examine the performance implications. The results are quite telling, shedding light on how memory bandwidth and overall system responsiveness are affected when operating with a reduced channel configuration. It's an intriguing development that could influence purchasing decisions for those looking to balance performance and budget.
Stepping into the realm of better XMP/EXPO settings, we find the DDR5-6000 CL30 ULL operating at a reasonable 1.35 V. This configuration showcases the potential for tighter timings without significantly increasing voltage, offering a taste of what Ultra Low Latency memory can achieve. For more details, you can check out the
G.SKILL Trident Z5 NeoX RGB 6000 CL30 32 GB Review. Moving on, we have the DDR5-6400 CL32 setups, available in both 2x 16 GB single-rank and 2x 32 GB dual-rank configurations. This comparison will allow us to explore the impact of rank versus frequency, providing valuable insights into how these elements affect overall system performance.
Pushing the boundaries further, the DDR5-7600 CL36 and DDR5-8000 CL38 at 1.45 V step into the enthusiast-performance arena. These configurations necessitate a 2:1 divider for the memory controller, showcasing what happens when you push the limits of current DDR5 technology. Across these configurations, we observe a 25% variance in measured latency from the most economical to the priciest setup. This range provides ample room for enthusiasts to fine-tune their systems, balancing cost against performance to suit their specific needs.
In this article, we do our best to maintain consistency by running every game through the same comprehensive suite on identical hardware. On the synthetic and productivity side, we used AIDA64's memory read, write, copy, and latency tests, along with Cinebench R23 and R24, and y-cruncher 2.5B. For graphics and CPU benchmarks, 3DMark Time Spy Extreme, Steel Nomad, and CPU Profile were included, as well as UL Procyon Photo and Video Editing benchmarks. Our gaming suite was extensive, covering eight popular titles: Palworld, Baldur's Gate 3, Elden Ring, Resident Evil 9, Spider-Man Remastered, Forza Horizon 6, Counter-Strike 2, and Cyberpunk 2077. Each game was benchmarked at resolutions of 1920x1080, 2560x1440, and 3840x2160, all on the High preset using an RTX 4090. We reported results as average frame rates along with 95th percentile, 5th percentile, 1% lows, and 0.1% lows.
The additional tests provide a more comprehensive view of the 9800X3D's performance capabilities. By re-running the full 1080p gaming suite with Precision Boost Overdrive enabled, we can observe how memory scaling measures up against another free performance enhancement available to 9800X3D users. A second test was conducted to minimize GPU influence. The graphics card was effectively removed from the equation. Counter-Strike 2 was tested at 1080p on Low settings, achieving frame rates well over 600 FPS. This allowed for a comparison between the slowest and fastest memory options, complete with full frame time analyses and FPS distribution histograms. This setup arguably presents the most unbiased scenario to evaluate the impact of fast memory. The results are detailed in the following sections.
Memory Kit Specifications / Comparison
Spoiler: AMD (AM5) Comparison Memory Kits
| Memory Comparison |
|---|
| SKU | Brand | Name | Type | Config | Speed | Profile Timings | tRFC2 | tRFCSB | Voltage | Profile | FCLK:UCLK:MCLK |
|---|
| DDR5-4800 | JEDEC | | DDR5 | SR 2× 16 GB | 4800 MT/s | 40-40-40-77-117 | 384 (160 ns) | 312 (130 ns) | 1.1 V | JEDEC | 2000:1:1 |
|---|
| DDR5-5600 | JEDEC | | DDR5 | SR 2× 8 GB | 5600 MT/s | 46-45-45-90-135 | 448 (160 ns) | 364 (130 ns) | 1.1 V | JEDEC | 2000:1:1 |
|---|
| DDR5-6000 | | | DDR5 | SR 2× 16 GB | 6000 MT/s | 30-38-38-32-68 | 402 (134 ns) | 300 (100 ns) | 1.35 V | EXPO ULL | 2000:1:1 |
|---|
| DDR5-6000 | | | DDR5 | SR 1× 16 GB | 6000 MT/s | 30-40-40-76-116 | 480 (160 ns) | 390 (130 ns) | 1.35 V | EXPO | 2000:1:1 |
|---|
| DDR5-6400 | | | DDR5 | SR 2× 16 GB | 6400 MT/s | 32-38-38-78-116 | 512 (160 ns) | 416 (130 ns) | 1.35 V | EXPO | 2000:1:1 |
|---|
| DDR5-6400 | | | DDR5 | SR 2× 32 GB | 6400 MT/s | 32-39-39-102-156 | 512 (160 ns) | 416 (130 ns) | 1.40 V | EXPO | 2000:1:1 |
|---|
| DDR5-7600 | | | DDR5 | SR 2× 16 GB | 7600 MT/s | 36-46-46-86-132 | 608 (160 ns) | 494 (130 ns) | 1.40 V | EXPO | 2000:2:1 |
|---|
| DDR5-8000 | | | DDR5 | SR 2× 16 GB | 8000 MT/s | 38-48-48-128-176 | 640 (160 ns) | 520 (130 ns) | 1.45 V | EXPO | 2000:2:1 |
|---|
Supporting Hardware