AMD Ryzen 7 5800X3D review - Introduction

Ryzen 7 5800X3D processor review
AMD's fastest $449 USD processor is dedicated to PC Gaming

It's time for another ZEN3 review, this round is something extra especially for PC gamers. It's the much-discussed Ryzen 7 5800X3D. The CPU, which is on many people's radar, features eight cores and sixteen threads but is coupled with a large 96MB L3 cache, which would boost performance for games, regaining AMD's advantage over Intel's Alder Lake processors such as the 12900K. One of the new Ryzen chip's primary advantages is the addition of 3D V-cache memory. For example, the nice Intel Core i7-12700K has 25MB of L3 cache, the new Ryzen CPU now offers 96MB L3 cache (!). This will enable the Ryzen 7 5800X3D to perform substantially better in gaming. According to AMD's internal testing, the new Ryzen 7 5800X3D appears to run at the same level as the Intel Core i9-12900K, so that's going to be our baseline for this review.

Be warned though, the AMD Ryzen 7 5800X3D processor features eight cores yet a lower maximum clock speed of 4.5GHz. This is somewhat sluggisher than the standard Ryzen 7 5800X, which runs at a maximum speed of 4.7GHz. Along with its 32MB of standard cache, the new model has 64MB of 3D V-Cache. Cache memory is placed on top of the chip in this configuration. At 105W, the TDP remains the same. 

ZEN3 series processors

ZEN, of course, is the codename behind the processor architecture. ZEN3 promised an updated architecture with increased IPC (raw core for clock performance) combined with the now-familiar chiplet designs that offer better yields. AMD has been continuously pushing the limits of the chip-fabrication foundries, and sure, 7nm production has been a sweet spot for AMD. They produce good yields ever since the initial launch. Yields are good because of that chiplet design; see when you fab monolithic and only get 30 chips from a wafer with a 60% yield, you end up with 18 working die's. When you use chiplets (multiple chips per package), you can fab perhaps 200 chips per wafer, and with the same yield ratio, you all of a sudden have 120 working die's. And therein is part of the secret sauce to be found in AMD's recent successes. 

Gaming king?

A lot has been said and spoken about Ryzen 5000, or the artists are previously known as ZEN3; AMD single and multi-threaded performance have been great overall but have been struggling a bit more with high FPS and CPU bound games. That last bit is not solely due to the processor, as the gaming industry has been on an Intel intravenous drip for a decade and optimized their procs the best. This thesis is wider than that as AMD has had architectural disadvantages in its processor design, the cluster design, I should say. AMD's ZEN and ZEN2 processor dies each holds 8-cores. However, within that real estate, you'll have learned that these eight cores are clustered in two groups of four. In there are some latency issues to be found as the two 4-core partitions communicate with more latency; it's part of the root cause of that gaming differential. The solution to that last hurdle is often twofold; the first is to brute force your way out of there with increased IPC (your raw per clock cycle performance). You can do that by making more efficient buffers, and caches, basically creating stronger performing cores. Of course, a second methodology is something we see at Intel, raise that clock frequency as high as possible. Overall, Intel is King of High turbo clock frequencies but a notch weaker in IPC. On the other hand, AMD has been powerful on IPC but less so in that absolute peak clock frequency. Now it isn't that AMD processors perform badly in games, as that isn't the truth whatsoever, but that lacking 5% or 10% game performance in the most extreme situations (GeForce RTX 3090 / CPU bound games / high FPS games), no it's the fact that they could not pass Intel in that segment. That has been haunting them and, in the end, is something that the more enthusiast end-user would stare himself blind at. It also created a reputation, a stigma of some sort, and no matter how good the processor series and its infrastructure have been, the first thing that people ask me is this 'is it faster than Intel in games'. Thanks to an IPC increase and a fundamental change in the architectural move from two four-core clusters within a die towards one fully unified 8-core cluster. Then came Alder Lake, these Intel processors do brilliant stuff for game performance. So AMD now releases an in-between, an extra chip inside the package holding more L3 cache (64MB), added towards the existing L3 cache - bringing back performance levels towards the Core i9 12900K even (at least that's the claim).

No overclocking?

Overclocking is not supported on the Ryzen 7 5800X3D. According to AMD, they are still fine-tuning the packaging for their V-cache CPUs. Due to the fact that the Ryzen 7 5800X3D is the first of its sort, they chose to voltage lock it at 1.35V. However, fabric and memory overclocking remain enabled, allowing enthusiasts to modify some parameters. We hope this is not the fate of future V-cache CPUs and AMD is able to configure these new processors properly. In short, multipliers are locked, but also PBO is not working. That with lower clocks will for sure have an effect on application performance. However ... this is a gaming processor and in that segment that's where the extra L3 cache will bring back the performance loss.