Zen 3
AMD microarchitecture released in 2020 with unified core complexes.
Zen 3 is a central processing unit design by AMD that launched in November 2020. It improved performance per clock cycle by nineteen percent over its predecessor through a unified cache structure and refined execution units.
- Delivers 19% higher instructions per cycle than Zen 2
- Uses unified 32 MB L3 cache for all 8 cores in a complex
- First PC product to use 3D vertically stacked L3 cache
- Built on TSMC 7 nm process for compute dies
What it is
Zen 3 is a microarchitecture for central processing units developed by AMD. It arrived in late 2020 as the successor to Zen 2. The design focuses on higher instructions per cycle and better communication between cores. It powers desktop Ryzen 5000 processors and Epyc server chips. This architecture uses a chiplet approach where compute dies are separate from input output components. It represents a significant redesign of the core structure compared to earlier versions.
How it came about
AMD released this architecture on November 5, 2020. It followed Zen 2 and aimed to fix latency issues caused by split cache pools. The team redesigned the core complex to include eight cores sharing one large cache block. This change allowed faster data exchange without routing through external links. Later updates included a three dimensional cache stacking technique for gaming performance. A planned desktop refresh called Warhol was cancelled due to supply constraints.
Inside the chip
The core improvement involves merging two four core groups into a single eight core complex. All eight cores share a thirty two megabyte level three cache pool directly. This setup reduces latency for inter core communication but slightly increases cache access time. Branch prediction buffers were doubled in size to improve instruction flow. Integer and floating point units received wider dispatch paths and larger schedulers. New instructions for encryption and carry less multiplication were added to the set.
How it was made
Manufacturing relies on TSMC seven nanometer process technology for the compute dies. Input output dies use GlobalFoundries fourteen nanometer nodes for servers and twelve nanometer for desktops. A single compute die contains roughly four billion transistors. The packaging supports various sockets including AM4 for consumers and SP3 for enterprise. Some variants use hybrid bonding to stack additional cache layers on top of the main silicon.
Where it went
Desktop computers use Vermeer and Cézanne chips for general computing and gaming. Workstations employ Chagall processors for heavy multitasking with many PCIe lanes. Servers run Milan chips which support up to sixty four cores per socket. Mobile devices utilize Cézanne and Barceló APUs with integrated graphics. The architecture supports DDR4 memory across most platforms and PCIe generation four interfaces.
Versions and family
Zen 3+ is a refined version focusing on power efficiency for laptops. It uses a six nanometer process and adds support for DDR5 memory. The 5800X3D desktop chip features stacked vertical cache for gaming boosts. Server variants include Milan X models with extra cache layers. Mobile APUs like Rembrandt integrate RDNA two graphics cores and USB four support.
Why it matters
This design delivered a nineteen percent average gain in instructions per cycle. It allowed AMD to compete strongly in both consumer and server markets. The unified cache structure became a standard for subsequent designs. Smart Access Memory enabled direct CPU access to video card memory. The cancellation of the Warhol refresh shifted focus to the next major architectural leap.




Questions people ask about Zen 3
How does Zen 3 improve cache performance?
It combines two four-core groups into one eight-core complex. All cores share a single 32 MB L3 cache pool. This allows direct communication without using the external Infinity Fabric link.
What manufacturing process is used for Zen 3?
The compute dies are made using TSMC 7 nm technology. The input output dies use GlobalFoundries 14 nm for servers and 12 nm for desktop processors.
What is the difference between Zen 3 and Zen 3+?
Zen 3+ focuses on power efficiency improvements rather than raw speed. It uses a 6 nm process and adds support for DDR5 memory and AV1 decoding in mobile chips.
Which sockets support Zen 3 processors?
Desktop chips use Socket AM4. Workstations use sWRX8. Servers use Socket SP3. Mobile APUs use Socket FP6 or FP7 depending on the specific model generation.
Good reading and sources
- AMD Dishes on Zen 3 and Zen 4 Architecture, Milan and Genoa RoadmapTom's Hardware, Article
- Checking your browserAnandTech, Article
- GeForce RTX 30 Series Performance Accelerates With Resizable BAR Support | GeForce NewsNvidia, Article
The facts on this page come from the Wikipedia article "Zen 3", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryChip rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.
Like Zen 3
All CPUs and microprocessorsMore from AMD
All of them
🇺🇸 United StatesApple A14 Bionic
Apple's 5 nm phone and tablet chip with six CPU cores and a 16-core Neural EngineSoC · Arm · 2020
🇺🇸 United StatesApple M1
ARM-based system-on-a-chip for Macs and iPads, designed by Apple and manufactured by TSMC.SoC · Arm · 2020
🇺🇸 United StatesRadeon RX 6000 series
AMD RDNA 2 graphics cards with ray tracingGPU · GPU architecture · 2020