Core Architecture & AI Enhancements
Built on NVIDIA Blackwell architecture, featuring 4th-generation ray tracing cores and 5th-generation Tensor Cores with FP4 precision and AI frame generation via DLSS 4. This enables clean performance boosts in supported titles using Multi-Frame Generation technologies.
Includes enhanced responsiveness and latency reduction with NVIDIA Reflex 2 and Frame Warp.
Compute Power & Memory
CUDA Cores: 8,960—the highest among RTX 5070-class GPUs to date.
Clock Speeds: Factory base clock ~2,295 MHz, with a boost clock around 2,572 MHz in “OC Plus” factory overclocked configuration.
Memory: Massive 16 GB GDDR7 at 28 Gbps, across a 256-bit interface, delivering up to 896 GB/s bandwidth.
Thermal & Power Design
TDP: Rated at 300 W, suggesting more headroom and overclocking potential compared to the non‑Ti variant.
Uses a single 16-pin power connector (adapter splits into three 8‑pin) and recommends at least a 750 W (typical) or 850 W PSU depending on system config.
Cooling is handled via a robust triple‑fan “OC Plus” cooler, essentially the same beefier cooler used on flagship RTX 5080/5090 models.
Display & Physical Attributes
Outputs: 3× DisplayPort 2.1b (UHBR20) + 1× HDMI 2.1b, supporting simultaneous run of up to four displays.
Max Resolutions: 4K @ 480 Hz or 8K @ 165 Hz via DSC support.
Form Factor: Approximately 12.93″×5.42″×2.35″ (3‑slot wide), typical triple-slot card size.
Package & Compatibility
Ships with a 3‑Year limited manufacturer warranty, auxiliary power cable, and support kit.
Compatible with Windows 10/11 (64-bit) and Linux-equipped systems using a PCIe 5.0 x16 slot. Internet connection needed for driver installation.
