Introduction

AMD Ryzen 5000 is here, powered by the new "Zen 3" architecture! We bring you this review of the 12-core Ryzen 9 5900X processor. But wait, where's Ryzen 4000, you ask? AMD cluttered its 4000-series processor model numbers with APU parts based on the "Renoir" silicon powered by "Zen 2" and felt "Zen 3" deserved a new number series. The new "Zen 3" architecture promises an impressive 19 percent IPC uplift over "Zen 2," which had similar double-digit gains over its predecessor.

For AMD to repeat such huge generational IPC improvements with each successive generation is stunning, given its main rival Intel has been using the same "Skylake" architecture for many years, paired only with higher core/thread counts and better boost algorithms. Still, Intel has managed to hold on to one thing that eluded AMD even after its return to competitiveness—gaming performance leadership. AMD claims that the new Ryzen 9 5900X beats the Core i9-10900K at gaming owing to its single-threaded performance gains and should naturally beat it at productivity by virtue of its higher core count.



The Ryzen 9 5900X is a 12-core/24-thread processor AMD is pricing at $549, about the same as the current Core i9-10900K street price (it originally launched at around $500). Ryzen 9 5900X's CPU cores are built on the same 7 nm silicon fabrication process as the Ryzen 3000 "Zen 2" processor, but with several refinements to the microarchitecture. The biggest change with "Zen 3" has to be the company doing away with the 4-core CCX and unifying all cores of the CPU chiplet into a single 8-core CCX. Even within the CPU core, AMD has worked to reduce latencies, improved branch-prediction, optimized the execution engine, fattened the front-end and load/store units, and deployed faster caches, which has a direct impact on IPC, or single-thread performance. IPC is the single biggest contributor to gaming performance, and the 19% claimed IPC gain over "Zen 2" should mean AMD has taken the gaming crown since the "Zen 2" architecture wasn't too far behind "Comet Lake" at gaming to begin with.

Unlike the competition, AMD isn't launching a new chipset with Zen 3. Existing motherboards based on AMD 500-series or 400-series chipsets will work with the new Ryzen 5000 Series processors after a BIOS update. AMD also ensured that the increased IPC on the same 7 nm process doesn't translate into a higher power draw, with the Ryzen 9 5900X shipping with the same 105 W TDP as its predecessor, the 3900X. In this review, we put AMD's biggest claim for "Zen 3"—gaming performance leadership—to the test, along with our vast suite of CPU benchmarks to investigate whether AMD has wholly and comprehensively defeated Intel.

Unboxing and Photography

The Ryzen 9 5900X comes in a fairly large paperboard box. The front face features a brushed metal appearance (compared to the carbon fiber appearance of the Ryzen 3000 series). There are enough pointers to let you know you're buying a Ryzen 5000-series part. A small cutout on the side shows the actual processor inside the package.


The processor looks like any conventional AMD CPU with a large IHS dominating the top and a 1,331-pin micro-PGA for the bottom. The "Zen 3" CCD chiplet is made in Taiwan and the I/O die in the US, and the two are put together at a facility in China.


The retail Ryzen 9 5900X box does not include a cooler. Luckily, it can be paired with a fairly big selection of AM4-compatible coolers that have been released since 2017. Just make sure the cooler can handle thermal loads of 105 W.

Architecture

The Zen 3 Microarchitecture

Since its 2017 debut, AMD has delivered a new iteration of its groundbreaking "Zen" CPU microarchitecture each year, each with IPC improvements. The new "Zen 3" microarchitecture, as we mentioned earlier, claims to offer a massive 19 percent IPC uplift over its predecessor "Zen 2." This is accomplished through improvements at both the micro and macro level. We already detailed the macro (beyond the core) changes above. In this section, we talk about what's new inside each core. AMD talks about updates to practically all key core components, including its front-end, fetch/decode, the integer and floating-point components, load-store, and dedicated caches.


Modern processors execute multiple instructions in parallel to improve performance. Computer programs consist of huge amounts of "if ... then ... else" instructions, which slow down the processor because it has to evaluate the condition before picking a branch to execute. In order to overcome this limitation, the branch predictor was invented, a piece of circuitry that takes a guess on what's the more likely outcome of the condition check and just speculatively executes that branch's instructions. Of course, there's a chance that the prediction is wrong, in which case a performance penalty is incurred from undoing the executions that were already executed. With "Zen 3", AMD uses an improved TAGE branch predictor, which is more accurate and recovers faster from mispredictions. They also changed the design to be "bubble free," which avoids inserting "wait for result" instructions in the instruction stream whenever a branch is encountered.

AMD generally increased ops/cycle; the front-end now switches between the op and instruction caches faster. The 32 KB L1 instruction cache has been tweaked to offer better utilization due to efficient tagging and pre-fetching. Streamlining was done to the Op cache. Improvements to the branch predictor and front-end add up to nearly a quarter of the overall 19% generational IPC uplift.


The execution engine, or combination of the integer and floating-point execution units, is the main math muscle of the CPU core. The "Zen 3" microarchitecture features improvements to both over "Zen 2." Both the INT and FP issue queues (which feed work to the two engines) have been widened and the execution window enlarged. This ensures fewer units are idle in typical programs, which increases overall performance.


AMD worked to minimize latencies at every stage of the INT execution engine, and enlarged its key structures, including the integer scheduler (96 entry vs. 92 on "Zen 2"), physical register file (192 vs. 180 on "Zen 2"), and 10 issues per cycle, up from 7 on "Zen 2." Data picker bandwidth has been significantly increased despite the same number of ALUs. The floating-point engine features the same 256-bit FPUs, but just as with the INT engine, the FP engine has latency and bandwidth improvements across the board, a faster 4-cycle FMAC, and a larger scheduler. The INT and FP improvements contribute around a fifth of the 19% overall IPC uplift.


With the "Zen 3" microarchitecture, AMD addressed many bottlenecks and "intelligence" issues with the Load/Store unit. The biggest has to be bandwidth. The entry store queue has been widened to 64 from 48 on "Zen 2," and the L2 cache DTLB is 2K entries wide. The 32 KB L1 data cache has been made faster, with lower latencies. Memory dependence detection has been improved. Much like the front-end and scheduler, the load/store improvements contribute nearly a quarter of the 19% overall IPC uplift, meaning that by just optimizing the non-execution components of its core, AMD managed to pull off a vast 9% overall IPC uplift.

ISA and Security Changes

Each new microarchitecture heralds support for newer instruction sets and security hardening, and the same is the case with "Zen 3"; however, a notable absentee is AVX-512. Granted, Intel has adopted a less than perfect method of proliferating AVX-512 with certain instructions being exclusive to enterprise-segment microarchitectures and only a handful client-relevant instructions on its "Ice Lake" and "Tiger Lake" architectures, but there's no movement from AMD in this direction.

You still do get 256-bit instructions from within the AVX2 set. Also missing in action is something to rival Intel's DLBoost, which is essentially a software exposure of fixed-function hardware that accelerates matrix multiplication, in effect AI deep-learning neural net building and training. A lot of client applications, particularly image manipulation and video editing, are leveraging edge AI, and some investment from AMD on this would have been nice. That said, "Zen 3" adds two new ISA instructions, MPK (memory protection keys) and AVX2 support for AES/APCLMulQD. AMD has been ahead of Intel with CPU core security vulnerability perception, and with "Zen 3," AMD is introducing CET, or control-flow enforcement, which should provide hardening against ROP-type attacks.

Vermeer Multi-Chip Module

The AMD Ryzen 9 5900X "Zen 3" processor is built on a Socket AM4 multi-chip module package the company refers to as "Vermeer." Since the Ryzen 3000 "Matisse," which was the first desktop processor to implement the 7 nm silicon fabrication process, AMD figured out a way to optimize the utilization of its 7 nm foundry allocation by using two things—building only those components that tangibly benefit from the new node on 7 nm, namely the CPU cores, and moving all other components to a separate die built on older 12 nm process, the cIOD (client IO die). The CPU cores are built on tiny dies with 8 cores each, which AMD refers to as the CCD (CPU core die), and on the older "Zen 2" microarchitecture, the eight cores were split into two groups of four cores, each, called CPU core complexes (CCX). Each of the two CCX on the "Zen 2" CCD had its own 16-megabyte L3 cache shared between the two cores, and communication between cores of different CCX required a round-trip to the cIOD.


With the new "Zen 3" microarchitecture, the biggest high-level change with the CCD is AMD's enlargement of the CCX to now include up to eight cores (essentially taking up the whole CCD). There's now one 8-core CCX per CCD. The biggest dividend of this change has to be improved inter-core latency since the eight cores now share the same L3 cache; the other big dividend has to be cache size. Each core on the CCD now has access to the full 32 MB L3 as a victim cache, so lightly threaded workloads should see a performance uplift. Eight-core Ryzen 7 5000-series models, such as the 5800X, feature a single CCD with all cores enabled. 6-core parts, such as the Ryzen 5 5600X, feature one CCD with any two of the cores disabled (shouldn't matter which ones). The 12-core Ryzen 9 5900X and 16-core Ryzen 9 5950X are parts that have two 8-core CCDs besides the cIOD. The 5900X is carved out by disabling any two cores per CCD, while the 5950X has all cores enabled on both CCDs. We confirmed with AMD that Ryzen 5000 "Vermeer" uses the same exact 12 nm cIOD as the Ryzen 3000 "Matisse" with only a couple of non-physical improvements, such as improved memory clocks and clock domains.