AMD on FSR 4 for RDNA 3 and Older GPUs: "No Updates to Share at This Time."

AMD says it has "no updates to share" on bringing FSR 4 support to RDNA 3/2/1 GPUs, citing architectural differences—RDNA 4 adds FP8 Wave Matrix Multiply-Accumulate instructions while older generations lack that hardware. An accidental AMD GPUOpen leak revealed an INT8 variant of FSR 4 that ComputerBase tested on RDNA 3/2 hardware; in Cyberpunk 2077 at 4K Ultra on an RX 7900 XTX FSR 4 ran ~11% faster than native but ~16% slower than FSR 3.1, suggesting performance and deliberate product separation may explain AMD's current public stance.

Analysis

Market structure: AMD’s choice to hold back FSR 4 for RDNA3/2 preserves pricing power for Radeon RX 9000 (RDNA4) hardware while ceding near-term software parity to competitors — NVIDIA (NVDA) benefits if gamers perceive AMD as artificially segmenting features. The leaked INT8 library and tests (FSR4: +11% vs native but −16% vs FSR3.1 in Cyberpunk 4K on 7900 XTX) show a mixed value proposition that will slow broad upscaling-driven refresh cycles and favor incumbents with superior frame-generation stacks.

Risk assessment: Near-term (days–weeks) the story creates volatility risk in AMD (AMD) equity and options (IV could rise 10–30% around developer/driver announcements); short-term (3–6 months) the bigger tail is product cannibalization — a release of INT8 FSR4 to older GPUs could shave mid-single-digit percentage points off RX9000 ASPs over 12 months. Hidden dependencies include game-engine uptake and OEM/system integrator bundling; catalysts are AMD GPUOpen disclosures, GDC announcements and the next quarterly call (next 60–90 days).

Trade implications: Tactical positioning: favor relative-strength in NVDA vs AMD — NVDA likely to capture developer mindshare if AMD restricts software across install base. Use options to size conviction: buy AMD 3-month put spread (roughly 25–30 delta short put, 15–20 delta long put) sized to 1% portfolio risk and establish a 2–4% overweight in NVDA for 3–6 months. If long AMD, sell 6–8 week covered calls to harvest elevated IV ahead of driver releases.

Contrarian angle: Market consensus assumes permanent product separation; that’s not inevitable — a measured INT8 rollout would boost AMD’s software ecosystem value and installer base, potentially increasing service/licensing optionality and used-GPU demand. If AMD releases FSR4 to older cards, expect initial negative ASP reaction followed by a stabilizing software-driven retention boost over 6–12 months; mispricing likely in short-dated AMD options and in overly bearish short positions.

Key Decisions for Investors

• Establish a 2–4% portfolio overweight in NVDA (NVIDIA) for a 3–6 month horizon to capture share of developer mindshare and potential market flight from perceived AMD feature withholding; target relative outperformance of +8–15% vs AMD over 6 months.
• Enter a protective bearish option on AMD: buy a 3-month put spread sized to 1% total portfolio risk (sell ~25–30 delta put, buy ~15–20 delta lower strike) to profit if AMD shares gap down after driver or earnings announcements within 60–90 days.
• If you hold AMD stock, sell 6–8 week covered calls at ~10–15% OTM to monetize elevated IV ahead of expected GPUOpen/GDC/earnings catalyst; if called away, redeploy proceeds into NVDA or chip-equipment names (TSM, ASML) for cyclic exposure.
• Allocate 0.5–1% to an idiosyncratic long in AMD software/ecosystem beneficiaries (game engine partners or middleware vendors) on a 6–12 month view if AMD signals intent to broaden FSR4 support — exit if confirmation not delivered within 90 days.