AMD Radeon RX 9070 Specifications
- Architecture: RDNA 4
- GPU: Navi 48 (XT variant)
- Manufacturing Process: TSMC 4nm (N4)
- Transistor Count: 53.9 billion
- Die Size: 357 mm²
- Compute Units (CUs): 56
- Stream Processors: 3,584 (56 CUs × 64 stream processors per CU)
- Ray Accelerators: 56 (one per CU)
- AI Accelerators: 112
- Base Clock: 2,070 MHz (Game Clock)
- Boost Clock: 2,540 MHz
- Memory: 16 GB GDDR6
- Memory Bus Width: 256-bit
- Memory Speed: 20 Gbps
- Memory Bandwidth: 640 GB/s
- Infinity Cache: 64 MB (3rd Generation)
- Render Output Units (ROPs): Likely 128 (consistent with RX 9070 XT, though not explicitly confirmed for non-XT)
- Texture Mapping Units (TMUs): Likely 256 (based on leaks and consistency with RX 9070 XT)
- Peak Single-Precision Performance: Approximately 36.4 TFLOPS (calculated as 2,540 MHz × 3,584 SPs × 2 operations per cycle ÷ 1,000)
- Total Board Power (TBP): 220W
- Recommended PSU: 650W
- PCIe Interface: PCIe 5.0 x16
- Display Outputs: HDMI 2.1b, DisplayPort 2.1a UHBR13.5
- Power Connectors: Traditional 8-pin connectors (exact number may vary by AIB model, typically 2x 8-pin based on leaks)
- Launch Price: $549 USD (MSRP)
Key Features
- RDNA 4 Architecture: Offers optimized compute units for better efficiency and performance per CU compared to RDNA 3, along with improved ray tracing capabilities (3rd-generation ray accelerators) and AI enhancements via matrix cores.
- Ray Tracing: Doubled throughput per compute unit compared to RDNA 3, significantly boosting ray tracing performance.
- AI Capabilities: Includes 112 AI accelerators, enhancing support for features like FidelityFX Super Resolution 4 (FSR 4), which is exclusive to the RX 9000 series at launch.
- Memory Configuration: 16 GB of GDDR6 memory provides a strong advantage over competitors like the NVIDIA RTX 5070 (12 GB), especially for high-resolution gaming and future-proofing.
- Power Efficiency: At 220W TBP, it’s notably efficient for its performance tier, requiring a modest PSU compared to higher-end GPUs.
Additional Notes
- Availability: The RX 9070 will launch exclusively through AMD’s board partners (e.g., ASUS, Gigabyte, Sapphire, XFX) with no reference design from AMD itself. It will be available starting March 6, 2025, alongside the RX 9070 XT.
- Performance Positioning: AMD claims the RX 9070 is approximately 20% faster than the RX 7900 GRE at 4K resolution and 38% faster than the RX 6800 XT, with significant gains in ray tracing workloads (up to 38% over the 7900 GRE).
- Comparison to RX 9070 XT: The non-XT variant has fewer compute units (56 vs. 64), lower clock speeds (2,540 MHz boost vs. 2,970 MHz), and a reduced TBP (220W vs. 304W), making it a more affordable yet still potent option.
These specifications align with AMD’s focus on the mainstream market with RDNA 4, targeting gamers looking for strong 1440p and capable 4K performance at a competitive price point. For the most accurate and up-to-date details, keep an eye on AMD’s official announcements or reviews following the launch on March 6, 2025.
