AMD Ryzen 7 5800X3D: Best for gaming? In practice, rarely

AMD Ryzen 7 5800X3D in detail

AMD’s first processor with 3D V-cache is a rather controversial piece of hardware. Sure, it may be the ultimate in gaming performance, but the desired effect is more “on paper” than practical, and when it does come, it’s in very rare cases. So that you don’t end up disappointed with a virtually single-purpose processor that may not even excel at gaming, we’ve broken it all down in detailed tests.

AMD Ryzen 7 5800X3D in detail

A full stop for AMD Zen 3, which is supposed to change the view of what is really important in “gaming” processors. The Ryzen 7 5800X3D has a large L3 cache instead of high frequencies, which is supposed to be more efficient for games. The 3D V-Cache is made up of a chip soldered directly on top of the CPU chiplets. Architecturally, it’s solved in such a way that its capacity for software is transparently added to the capacity of the original L3 cache with 32 MB. Thus, the CCX octa-core block suddenly has an L3 cache capacity of up to 96 MB. This is a contiguous block and any core/thread can use the entire capacity. The toll of such an expansion is a few cycles of increased latency.

A large, or perhaps even huge, L3 cache is useful for programs with a working data set that exceeds 32 MB (the normal L3 cache capacity), but is ideally below 96 MB (the 3D V-cache capacity). If such a program spends too much of its runtime waiting for data from RAM, then there can be a massive speedup with 3D V-cache. This is because data from it is served directly and with significantly shorter latency.

One thing is better (bigger, faster) cache and another thing is lower frequencies that had to be “sacrificed”. The Ryzen 7 5800X3D is 250 MHz slower in terms of gaming boost compared to the R7 5800X (with “only” 32 MB L3 cache) and single-core boost speed is reduced by up to 350 MHz. The reason for the lower frequencies of the 5800X3D is reportedly that the chiplet connection technology used with the 3D V-cache cannot handle the high supply voltages that can be used in the rest of the CPU chiplet (it is possible that this limitation will be relaxed in future generations). For this reason, AMD doesn’t even allow overclocking on these processors.

The benefits of 3D V-cache are mainly discussed in the context of games, where the benefits should be significantly higher than in the desktop application environment. This is because games access a large amount of data in RAM, and their code therefore spends a significant amount of time idle waiting many cycles for data from DRAM to reach the CPU (its caches and registers).

A high-capacity cache reduces the amount of these delays and results in an overall speedup of the program as some of the downtime is eliminated. The rate of game speedup naturally depends on how large a percentage of the most actively used data can fit in the cache. If a large cache works, the performance increase can be so significant that it wipes out the deficit from lower clock speeds.

3D V-Cache… It’ss not a completely new invention

Hardware enthusiasts are familiar with the large cache phenomenon from Intel “Broadwell-C” processors (Core i5-5675C and Core i7-5775C). These were quad-core processors with Broadwell architecture that came to the market with some delay in 2015. Thus, they had to compete with architecturally newer Skylake (Core i7-6700K, etc.) for a few months.

Compared to Skylake processors, Broadwell-C processors not only had worse IPC (performance per clock), but also lower clock speeds. For example, the Core i7-5775C had clock speeds of only 3.3–3.7 GHz. Nevertheless, the specialty of Broadwell-C processors was 64 MB L4 cache (this technology is sometimes referred to as Crystalwell). It consisted of eDRAM and was originally intended for the integrated graphics core. It was quite helpful for this (it was one of the most powerful iGPUs), but this L4 cache was also available for the CPU.

Back then, gamers quickly discovered that in select games, the large L4 cache of Broadwell-C processors helped a lot (now we’re referring to CPU performance when using a graphics card, not iGPU performance), and the benefit was so great that these processors were beating even newer Skylakes in some cases.

Similar to Broadwell-C, the high-end Broadwell-E and Broadwell-EP Xeons also worked in this way. These also had a very large ringbus-based cache, which may have had a positive effect. As with Broadell-C, the large L3 cache in the Ryzen 7 5800X3D is capable of increasing performance beyond what can be expected from cores of its architecture (Zen 3) and its clock speeds.

* The processor is tested with AGESA 1.2.0.7. Compared to AGESA 1.2.0.6c, it achieves slightly lower raw performance (up to 3 %), but this is the final microcode with which this processor will be mostly used. The results are also more relevant for later comparison with Ryzen 7000, with which AMD will follow up AGESA 1.2.0.7.
** After official discount by AMD. The suggested price at launch (strickenthrough value) was higher.

• Contents
• AMD Ryzen 7 5800X3D in detail
• Methodology: performance tests
• Methodology: how we measure power draw
• Methodology: temperature and clock speed tests
• Test setup
• 3DMark
• Assassin’s Creed: Valhalla
• Borderlands 3
• Counter-Strike: GO
• Cyberpunk 2077
• DOOM Eternal
• F1 2020
• Metro Exodus
• Microsoft Flight Simulator
• Shadow of the Tomb Raider
• Total War Saga: Troy
• Overall gaming performance
• Gaming performance per euro
• PCMark and Geekbench
• Web performance
• 3D rendering: Cinebench, Blender, ...
• Video 1/2: Adobe Premiere Pro
• Video 2/2: DaVinci Resolve Studio
• Graphics effects: Adobe After Effects
• Video encoding
• Audio encoding
• Broadcasting (OBS and Xsplit)
• Photos 1/2: Adobe Photoshop and Lightroom
• Photos 2/2: Affinity Photo, Topaz Labs AI Apps, ZPS X, ...
• (De)compression
• (De)cryption
• Numerical computing
• Simulations
• Memory and cache tests
• Processor power draw curve
• Average processor power draw
• Performance per watt
• Achieved CPU clock speed
• CPU temperature
• Conclusion