AMD’s next major CPU architecture, AMD Zen 7 Grimlock, is already stirring excitement across the tech world. Following years of innovation from Zen 1 through Zen 6, this generation looks to deliver a massive leap in both performance and efficiency, setting the stage for AMD’s 2028 lineup.
Below, we break down everything known so far about Zen 7 Grimlock, from architecture and manufacturing to performance expectations, laptop APUs, and server-grade designs.
AMD Zen 7 (Grimlock) release timeline and versions
The AMD Zen 7 Grimlock family is expected to roll out through 2028, following AMD’s traditional two-year cadence.
- Server (EPYC – Steamboat): Launch planned for early to mid-2028.
- Laptop & Desktop (Grimlock Point, Halo & Ridge): Arriving by late 2028, with desktop models likely landing first thanks to groundwork laid by Zen 6.
For context, Zen 6 (EPYC Venice) will debut in 2026. That timeline makes a 2028 Zen 7 launch “totally believable,” especially since Zen 6 desktop chips might slip into early 2027.
Architecture and manufacturing details
At the heart of AMD Zen 7 Grimlock lies the Prometheus classic core, a fundamental redesign rather than a simple refresh.
- Process Node: Most Zen 7 variants (EPYC, Laptop, Desktop) use TSMC’s A14 node, with potential A16 or 2 nm nodes for budget models.
- Cache Structure: Each core features 2 MB of L2 and 4 MB of L3 cache.
- AI Acceleration: Built-in FP512 half-A AI acceleration is powerful enough that some questioned whether an NPU would even be necessary.
- Instruction Set: Includes new ISA instructions for quantization format conversions and optimized parallel data preparation with MI accelerators, and delivers 4× improvement in FP8 and 2× improvement in Intake per cycle.
This means Zen 7 is a true architectural upgrade, not just a “Zen 6 Plus.”
Performance expectations (Zen 6 → Zen 7)
Zen 7’s projected uplift is enormous, maybe one of AMD’s biggest generational jumps ever.
| Metric | Uplift (Zen 6 → Zen 7) | Details |
|---|---|---|
| IPC | +15–25% | New cache design adds ~8% IPC alone; SPEC Int 2017 tests show 15–25% gains. |
| Single-Threading | +15–30% | Improved ISA and cache synergy. |
| Multi-Threading | +50–67% | Driven by higher IPC and ~33% more cores. |
Compared to Zen 5
- Up to 50% faster single-threaded and 150%+ faster multi-threaded.
- Potentially doubles single-thread and triples multi-thread performance over Zen 5.
- The overall leap from Zen 4 to Zen 7 rivals the historic Zen 1 → Zen 3 jump.
3D V-Cache (X3D) performance
AMD will continue leveraging 3D V-Cache technology for extra performance headroom.
- Performance Gap: ~20% difference between standard Zen 7 and Zen 7 X3D in non-gaming apps.
- Broader Benefits: Now boosts multi-thread and non-gaming tasks, not just games.
- Manufacturing: V-Cache dies built on TSMC N4P.
- Design Limit: Only a single layer of V-Cache is planned for the foreseeable future (no two-high stack).
Laptop APUs: Grimlock Point & Grimlock Halo
Zen 7’s laptop lineup, codenamed Grimlock Point and Grimlock Halo, introduces huge efficiency and core-count improvements.
Grimlock Halo (Flagship APU)
- Core Config: 8 Zen 7 classic + 12 Zen 7C cores = 20 cores (base).
- Chiplet Expansion: Supports two optional 8-core Silver King CCDs → up to 36 cores.
- Efficiency: Compact design, more cores than Strix Halo (16 cores today).
Grimlock Point (Mainstream)
- Base Config: 4 Zen 7 classic + 8 Zen 7C cores = 12 cores (base).
- Expansion: Adds one 8-core Silver King chiplet → 20 non-LP cores.
- Efficiency Uplift: Up to 2× more efficient than Zen 5 APUs.
- +13–17% performance per core @ 22 W/core.
- +30–36% uplift @ 3 W/core (low power).
- Product Stack: Includes Grimlock 1 – 4 models (only Grimlock 1 supports chiplet add-on).
- Graphics: No confirmed info yet on integrated GPU performance.
Desktop CPU: Grimlock Ridge
The desktop Zen 7 Grimlock Ridge line will continue using the AM5 platform, with a refined I/O die borrowed from Zen 6.
- I/O Die: Likely 6 nm for cost efficiency (4 nm and 3 nm tested but not selected).
- Flagship Chiplet – Silverton:
- Size: 98 mm² (AMD’s largest desktop chiplet yet).
- Process: 1.4 nm.
- Cores: 16 Zen 7 cores.
- Cache: 32 MB L2 + 64 MB L3 on-die.
- V-Cache: Adds 10 MB per core → 160 MB extra L3 = 224 MB L3 per CCD.
- Clock Target: ~7 GHz.
- Max Desktop Config: Two 16-core chiplets → 32 cores / 448 MB L3 cache (with dual V-Cache).
- Budget Chiplet – Silver King: 56 mm² (8 cores, no V-Cache), shared with laptop APUs.
AMD Zen 7 pricing and value
While Zen 7’s cutting-edge silicon will raise prices, performance gains could easily justify the jump.
- Flagship: ~$999+, 32-core variant with 448 MB L3.
- Mid-Range: 16-core V-Cache model (~$699) ≈ 50% faster than Zen 5 at the same price.
- Budget: 8-core Silver King (~$299) expected to outperform 12-core Zen 5 X3D chips in gaming and multithread workloads.
Even if flagships cost ~50% more, the performance-per-dollar could double, making Zen 7 Grimlock a strong value play.
Server: AMD EPYC Steamboat
For datacenters and high-performance computing, AMD’s EPYC Steamboat variant pushes core density to new heights.
- Up to 264 cores (8 × 33-core chiplets).
- Cache: 7–8 MB L3 per core.
- Total L3: 1.8–2.1 GB thanks to a dedicated L3 die chiplet (not V-Cache).
This configuration is designed for AI, cloud, and data-intensive applications where parallelism is most critical.
Key Takeaway: Two massive generations in a row
AMD Zen 6 and Zen 7 form a one-two punch similar to Zen 2 and Zen 3.
- Zen 6 lays the groundwork for new core designs and higher core counts.
- Zen 7 Grimlock builds on that foundation with fundamental architectural improvements, AI acceleration, and unprecedented scalability.
Together, they represent AMD’s most ambitious CPU evolution yet, a clear statement of intent for the AI-driven era of computing.
If AMD maintains its momentum, the AMD Zen 7 Grimlock series could define the next decade of CPU performance. From 7 GHz desktop chips to 264-core servers and ultra-efficient laptop APUs, AMD’s engineering ambitions are on full display.
