Apple M5 Ultra (36-Core) vs AMD Ryzen 9 9950X3D2
We compared two desktop CPUs: the 4.61 GHz Apple M5 Ultra (36-Core) with 36-cores against the 4.3 GHz AMD Ryzen 9 9950X3D2 with 16-cores. On this page, you'll find out which processor has better performance in benchmarks, games and other useful information.
Review
Single-Core Performance
Performance in single-threaded apps and benchmarks
Multi-Core Performance
Measure performance when all cores are involved
Power Efficiency
The efficiency score of power consumption
Integrated Graphics
iGPU capabilities for gaming and 3D-intensive tasks
The "iGPU" and "Power Efficiency" have a minor impact on the NanoReview Score for desktop CPU.
Choose your main usage scenario to recalculate the Final Score and see which CPU suits your needs.
Value for money (Beta)
Key Differences
Advantages of Apple M5 Ultra (36-Core)
- More powerful Apple M5 Ultra GPU (80-core) integrated graphics: 33.2 vs 0.6 TFLOPS
- Around 1139.2 GB/s (1271%) higher theoretical memory bandwidth
- Newer - released 5 months later
- Has 20 more physical cores
- Supports up to 512 GB LPDDR5X-9600 RAM
- More modern manufacturing process – 3 versus 4 nanometers
- 20% faster in a single-core Geekbench v6 test - 4331 vs 3614 points
Advantages of AMD Ryzen 9 9950X3D2
- Unlocked multiplier
Test in Benchmarks
Comparing the performance of CPUs across various tasksCinebench
Cinebench 2024 (Single)
Cinebench 2024 (Multi)
| Cinebench 2026 (Single) | - | 558 |
| Cinebench 2026 (Multi) | - | 10459 |
Cinebench Scores (1st and 3rd place)
GeekBench v6
Geekbench 6 (Single-Core)
Geekbench 6 (Multi-Core)
| File compression | - | 2030 MB/sec |
| Clang compilation | - | 240.9 Klines/sec |
| HTML 5 Browser | - | 460.9 pages/sec |
| PDF Renderer | - | 570.9 Mpixels/sec |
| Text processing | - | 361 pages/sec |
| Background blur | - | 90.2 images/sec |
| Photo processing | - | 160.1 images/sec |
| Ray tracing | - | 56.9 Mpixels/sec |
PassMark
Passmark CPU (Single-Core)
Passmark CPU (Multi-Core)
Blender
Blender CPU
Performance Per Watt
| Cinebench 2024 / Watt | - | 12.6 PPW |
| Geekbench 6 Multi / Watt | - | 135.6 PPW |
| Passmark CPU Mark / Watt | - | 363.9 PPW |
| Blender / Watt | - | 3.74 PPW |
These are rough calculations, as most CPUs surpass PL1 TDP wattage even in multicore workloads.
Specifications
Full technical specification of Apple M5 Ultra (36-Core) and AMD Ryzen 9 9950X3D2General
| Vendor | Apple | Amd |
| Released | August 25, 2026 | March 26, 2026 |
| Type | Desktop | Desktop |
| Instruction Set | ARMv9 | x86-64 |
| Codename | Apple M5 | Zen 5 (Granite Ridge) |
| Model number | - | 9950X3D2 |
| Integrated GPU | Apple M5 Ultra GPU (80-core) | Radeon Graphics |
AMD Ryzen CPU ranking (#1st place)
CPU
| P-Cores | 24 | 16 |
| P-Threads | 24 | 32 |
| Base Frequency (P) | 4.61 GHz | 4.3 GHz |
| Turbo Boost Frequency (P) | - | 5.6 GHz |
| E-Cores | 12 | - |
| E-Threads | 12 | - |
| Base Frequency (E) | 4.38 GHz | - |
| Total Cores | 36 | 16 |
| Total Threads | 36 | 32 |
| Bus Frequency | 100 MHz | 100 MHz |
| Multiplier | 46x | 43x |
| L1 Cache | 192K (per core) | 80K (per core) |
| L2 Cache | - | 1MB (per core) |
| L3 Cache | - | 192MB (shared) |
| Unlocked Multiplier | No | Yes |
Package
| Fabrication Process | 3 nm | 4 nm |
| TDP (PL1) | - | 200 W |
| Max. Boost TDP (PL2) | - | 270 W |
| Socket | Apple M-Socket | AM5 |
| Peak Temperature | 100°C | 95°C |
Desktop CPU ranking (1st and 3rd place)
iGPU
| Integrated Graphics | Apple M5 Ultra GPU (80-core) | Radeon Graphics |
| GPU Base Clock | 500 MHz | 400 MHz |
| GPU Boost Clock | 1620 MHz | 2200 MHz |
| Shading Units | 10240 | 128 |
| TMUs | 640 | 8 |
| ROPs | 320 | 4 |
| Execution Units | 1280 | 2 |
| TGP | - | 15 W |
iGPU FLOPS
Memory Support
| Memory Types | - LPDDR5X-9600 | - DDR5-5600 |
| Max. Memory Size | 512 GB | 256 GB |
| Memory Channels | 64 | 2 |
| Max. Memory Bandwidth | 1228.8 GB/s | 89.6 GB/s |
| ECC Support | No | Yes |
Misc
| Official Site | - | AMD Ryzen 9 9950X3D2 official page |
| PCI Express Version | - | 5.0 |
| PCI Express Lanes | - | 28 |
Cast your vote
Total votes: 30