Project Helix leaks RTX 5090 power
A GDC leak says Microsoft’s next Xbox—Project Helix—offers RTX 5090‑level ray tracing and native PC game support, positioning it as a generational leap for console and PC gamers (noobfeed.com). Nvidia used GTC to push new inference tech tied to partners like Groq, underscoring how ray tracing and AI inference are converging in next‑gen gaming hardware ( ).
Microsoft used its GDC keynote on March 11, 2026 to publish the first official Project Helix details, saying the system is built around a custom AMD SoC and that alpha hardware will be shipped to developers beginning in 2027. (news.xbox.com: ) Multiple reporting outlets and leaks frame Project Helix as a high‑end console–PC hybrid co‑designed with AMD, with industry coverage naming RDNA5 GPU and Zen6 CPU elements and price estimates clustering around $999–$1,200 in early rumor cycles. (wccftech.com: ) (geeky-gadgets.com: ) Microsoft told developers Project Helix will “integrate intelligence directly into the graphics and compute pipeline” and deliver an order‑of‑magnitude uplift in ray‑tracing capability compared with current gen consoles, framing the hardware as tightly coupled to next‑gen DirectX and AMD’s FSR Diamond stack. (news.xbox.com: ) (tomshardware.com: ) At Nvidia’s GTC 2026 (March 16–19), the company productized Groq’s inference tech with the Groq 3 LPU and Rubin/Vera system announcements — including LPX racks built around 256 LPUs and product claims about extreme low‑latency inference throughput targeted at agentic AI applications. (nvidia.com: ) (storagereview.com: ) Nvidia’s December 24, 2025 licensing/hiring deal with Groq — widely reported as a ~$20 billion strategic transaction for Groq IP and talent — underpins the LPU integration shown at GTC and signals a vendor move to embed dedicated inference silicon in full data‑center stacks. (cnbc.com: ) (tomshardware.com: ) Public technical materials for the Groq 3 LPU detail hardware specs that target inference density and low latency — each LPX accelerator is listed with hundreds of megabytes of on‑chip SRAM and terabytes‑per‑second class bandwidth — a profile companies say is designed to feed real‑time, multi‑token AI workloads that could be repurposed for graphics‑adjacent tasks. (nvidia.com: ) (techzine.eu: ) Combined timeline: Microsoft will distribute alpha Project Helix hardware to studios in 2027 and retains multi‑year AMD co‑engineering commitments, while Nvidia’s GTC product roadmap and Groq integration were unveiled mid‑March 2026 — setting clear milestone dates for developers and data‑center customers to test emergent inference‑plus‑graphics workflows. (news.xbox.com: ) (storagereview.com: )
