Initial RAM setup verification completed – first phase testing finished.
Initial RAM setup verification completed – first phase testing finished.
After my laptop upgrade from 1Rx16 to 2Rx8, performance improved significantly. I expanded the evaluation to desktops to address a gap left by previous testing. Nearly two years ago, Linus covered similar topics in a video, but those tests didn’t cover many configurations. My goal is to fill that void. I already possess various 1Rx8 and 2Rx8 RAM units, with some 1Rx16 in order.
Core components: CPU at 11700k, GPU either 1080 Ti or 2070. I’m favoring the 1080Ti for its better raster performance, not because it’s a RT benchmark. The main configurations will be: 1Rx16, 1Rx8, 2Rx8. I’ll also run tests with 1DPC and 2DPC; dual-channel setups are optional unless they’re essential. If you don’t use dual channels, performance likely isn’t a priority. Hybrid testing includes unbalanced channels or mismatched 2DPC types—like my previous machine with a fixed SR module and limited access.
I’m aware that the smallest module size is 4GB, which in two-channel setups leaves only 8GB RAM. This shouldn’t impact computation but might affect gaming benchmarks. I’m open to suggestions on additional titles, but I’ll need built-in or demo versions. I’ll stick to games with official benchmarks.
3DMark isn’t the newest version, so it may not be relevant. CPU Profile isn’t widely adopted. Compute Y-cruncher results (25m and 1b) are useful; I’ll use them once I start testing, aiming for realistic numbers. I won’t just max settings or low settings—I’ll balance both.
I’m considering running games like Call of Duty: Modern Warfare II, Final Fantasy XIV, Endwalker, Benchmark GTAV, Shadow of the Tomb Raider, and The Expanse. I’ll decide on graphical quality later based on real results. I won’t sacrifice performance for ultra-low settings, nor let GPU limits block me from using high-quality benchmarks.
For more insights, I’ll check CPU profiles, compute Y-cruncher, Prime95 throughput, 1b tests, Cinebench, and Blender. I’m also curious about synthetic Aida64 memory operations and rendering tools. This is still a bit outdated, but I’ll keep testing what’s available.
I’ll run a brief test on the new RAM to confirm it’s functioning properly before you invest too much time. I don’t have a benchmark included unless specified, so this would be outside this check. The spec changed from 2x 8GB 1Rx16 to 2x 16GB 2Rx8. More details here.
It seems you're asking about using a bootable drive. Would you like tips on how to properly set it up?
Blender includes its own benchmarks for CPU and GPU performance. It appears to run three tests in sequence, with the middle one requiring about 8GB, possibly causing issues on a 6GB Arc GPU. This could indicate limitations with 8GB RAM setups. Performance seems to degrade noticeably when using more RAM on the 11700k. Scores from 3200 and 2133 are close, averaging across three trials. I’ll look into it later.
This setup seems a bit unusual. The 1Rx16 RAM has arrived with four chips on one side and none on the other. I'm trying to figure out which settings to apply. It claims a 2400 rating, but my current RAM only supports up to 2133. I might have to adjust it to match. My focus is more on how the configuration influences performance rather than clock speeds. To verify functionality, I installed all four modules and ran a Blender benchmark. The results showed a slight slowdown of about 1.1% compared to using four 2Rx8 2166 chips. I suspect this difference becomes more noticeable with two modules instead of four. I also ran memtest and explored the included performance tests, but interpreting the data is tricky. Comparing this setup against others could help clarify the differences. Overall, I’m close to having everything ready; just need to finalize the software before moving forward.
I completed several evaluations. I’ll plot the outcomes later for clearer insight. For now, here’s a detailed summary. The primary comparison looks at four different RAM setups: two 1Rx16 with 16GB (1DPC), two 1Rx8 with 8GB, two 2Rx8 with 8GB, and four 2Rx8 with 8GB. I’ll refer to the reference condition as the baseline. The results should be clearer once plotted.
Most testing centered on benchmarking games at 1080p resolution. FFXIV Endwalker showed the best performance under maximum settings, outperforming the reference by about 4%. The 2x 1Rx8 and 2x 2Rx8 configurations matched or exceeded the reference at roughly 3% above baseline. The 2x 1Rx16 configuration lagged slightly behind, but still within acceptable limits.
For gaming scenarios like GTA V and COD MWII, the improvements were more pronounced. The 4x 1Rx8 setup achieved a 35% boost in performance, while the 4x 1Rx16 configuration improved by around 25%. Prime95 and Blender also indicated similar trends, with best results from 1Rx8 or 2Rx8 modules.
In terms of compute tasks, Y-cruncher and Prime95 showed consistent gains—35% and 40% improvements respectively. The tRFC values for the 1Rx8 and 2Rx8 modules were notably higher than those for the 1Rx16, suggesting a slight edge in speed.
Regarding software benchmarks like Cinebench R23 and SOTTR, the 4x 1Rx8 and 2x 2Rx8 configurations led the pack, with improvements ranging from 25% to over 50%. The results were generally stable across different systems, though some variations appeared due to RAM configuration.
Blender 3.4.0 ran three sub-tests, highlighting that while 8GB of RAM was sufficient for testing, any shortfall could affect outcomes. Overall, the data suggests that higher module counts and better memory alignment yield more consistent performance gains.
In summary, the best configurations were a mix of 4x 1Rx8 and 2x 2Rx8 modules, with clear advantages in both gaming and compute scenarios. RAM quality and quantity played a crucial role, though the impact varied depending on the specific test conditions.
I've prepared the charts for those who prefer a concise view. The numbers refer to typical setups with 2x 1Rx8 modules, usually seen in lower capacity configurations. Results are adjusted so better values stand out. Generally, higher performance is preferable. If you're focusing on time-sensitive aspects like latency, higher scores mean better results. I flipped the data to make comparisons easier at a glance. The tested system ran at 11.7k + 10.8 Ti. Games were benchmarked at 1080p RAM, using the JEDEC 2133C15 FFXIV Endwalker standard for both average frames per second and load times. For GTAV, I selected settings based on V-sync status; the Aida64 version is outdated and may trigger warnings during latency tests. Blender ran three sub-tests, with two highlighted here. The "junkshop" test required about 8GB of RAM, but since my smaller units totaled only 4GB, using two left me with just 8GB—performance dropped due to insufficient memory. Those outcomes aren't included.