Apple M3 Ultra
Overview
Two M3 Max dies joined by Apple's UltraFusion packaging over more than 10,000 connections, giving a 32-core CPU, up to 80 GPU cores and a 32-core Neural Engine. Configurable to 512 GB of unified memory, more than any other personal computer, which is why it is the desk-side machine of choice for running large models locally. Apple publishes no FLOPS figure; the values here are theoretical peak, derived as twice the M3 Max in line with the two-die design. Note that Apple's "up to 2x faster than M2 Ultra" claim describes workload performance on an architecture that added hardware ray tracing, not raw FP32 throughput, where the gain over M2 Ultra is nearer 1.2x.
Performance Metrics
Peak theoretical throughput by precision type
| Precision | Bits | Peak TFLOPS | |
|---|---|---|---|
| INT8 | 8 | 131.1 | |
| FP16 | 16 | 65.5 | |
| FP32 | 32 | 32.8 |
Power Specifications
TDP
--
Max Power
--
Power Connector
PCIe Slot
Cooling
Air
Memory Specifications
Capacity
512 GB
Type
LPDDR5
Bandwidth
819 GB/s
Interface
--
Hardware & Design
Form Factor
SoC
Architecture
Apple M3
Process Node
3nm
Launch Year
2025
Variant
Standard
Market Segment
Professional
Transistors
184.0 billion
Full Specifications
| Compute Engine | |
|---|---|
| GPU cores | 80 |
| Chip Design | |
| Transistors | 184.0 billion |
| Process Node | 3nm |
| Memory | |
| VRAM | 512 GB |
| Memory Type | LPDDR5 |
| Bandwidth | 819 GB/s |
| Power & Thermal | |
| General | |
| Form Factor | SoC |
| Architecture | Apple M3 |
| Launch Year | 2025 |
Documentation & Resources
Common Use Cases
The Apple M3 Ultra is optimized for high-performance computing tasks with Apple M3 architecture delivering 32.77 TFLOPS of compute power.
Where to Rent
Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.
Browse GPU Cloud ProvidersSimilar GPUs
Stay Updated on GPU Releases
Get notified when new GPUs are added or specifications are updated.
No spam, unsubscribe anytime.