Video summary

AMD Radeon RX 6900 XT GPU Review & Benchmarks: Undervolting, Gaming, Power, Noise

Main summary

Key takeaways

Product Review

Product overview / key specs

  • GPU: AMD Radeon RX 6900 XT (RDNA2 architecture, same family as the 6800 XT).
  • Memory: 16 GB GDDR6 (vs. NVIDIA RTX 3090’s 24 GB).
  • Cooler: same reference cooler design as the 6800 XT (PCB changes on the VRM side).
  • Positioning/MSRP: marketed as a halo-class product, roughly $500 less than the RTX 3090.

Main features & behavior

  • 6900 XT is the higher “bin” of the 6800-series: same architecture and memory configuration, but better silicon quality lets it run the same clocks at lower voltages.
  • AMD VBIOS/driver behavior: frequency and fan control follow junction temperature (Tjunction) rather than edge temperature.
  • Auto fan behavior: in the reviewer’s sample the 6900 XT runs quieter and at lower RPMs in auto mode versus the 6800 XT due to lower junction temps and better mount/bin.
  • Software/driver caveat: tools may accept very high clock settings that are silently ignored or can cause performance regressions without crashes — validate performance after changing clocks.

Performance — rasterization (selected highlights)

  • General behavior: AMD gains relative ground more often at 1080p/1440p; NVIDIA often leads at 4K. Results are highly game-dependent.
  • Selected numbers (4K unless noted):
    • Shadow of the Tomb Raider: 6900 XT stock ≈ 98 FPS; 6800 XT ≈ 93 FPS; 6900 XT OC+UV ≈ 102 FPS (+4%); RTX 3090 ≈ 104 FPS.
    • Total War: Three Kingdoms (Ultra): 6900 XT ≈ 45 FPS; RTX 3090 ≈ 54.3 FPS (large NVIDIA lead). 6900 XT OC ≈ +5% in this title.
    • Red Dead Redemption 2: 6900 XT ≈ 84 FPS; RTX 3080 slightly higher; RTX 3090 ≈ 92 FPS.
    • Rainbow Six Siege (1080p): extremely high framerates — 6900 XT ≈ 453 FPS (leader vs 3090 in that test).
  • Overall: the 6900 XT gives small-to-modest gains over the 6800 XT in many rasterized tests, but rarely enough to justify the price premium for most gamers.

Ray tracing & DLSS

  • Ray tracing is a weakness for AMD versus NVIDIA; NVIDIA holds a large advantage, especially where DLSS is available.
  • Examples:
    • Shadow of the Tomb Raider (RT shadows): 6900 XT trails RTX 3080/3090; 3090 ≈ 25% ahead at 4K in the RT test.
    • Minecraft Path Tracing (4K): 6900 XT ≈ 15 FPS; RTX 3080 without DLSS ≈ 33 FPS; RTX 3080 with DLSS ≈ 87 FPS. DLSS provides a very large practical advantage.
    • Control (RT): 6900 XT is playable in places but falls significantly behind NVIDIA once DLSS is enabled.
  • Conclusion: AMD is effectively non-competitive in heavy path-traced or DLSS-boosted workloads.

Overclocking, undervolting & stability

  • Overclocking model / software limits:
    • AMD’s software enforces a min/max core clock delta of 100 MHz.
    • Memory frequency caps: max software memory setting = 2150 MHz; GB clock limited to 3000 MHz.
    • Drivers/tools may accept settings that are silently ignored or that cause performance regressions — validate with benchmarks.
  • Practical tuning found by the reviewer:
    • Best practical core range: ≈ 2450–2550 MHz for consistent uplift. Higher ranges (2590–2690 MHz) sometimes regressed performance.
    • Voltage: stock ≈ 1175 mV; reducing to ≈ 1050–1080 mV often produced better results. Going too low (≈ 1020–1030 mV) caused instability.
    • Final reported uplift: ≈ 8.3% improvement versus stock after OC + undervolt tuning (game/benchmark dependent).
  • Validation: Port Royal was used as a consistent metric to validate stability and performance changes.

Thermals, noise & power

  • Temperatures (auto settings): edge ≈ 80°C; junction ≈ 93–95°C (junction is the control target).
  • Fans / noise:
    • Reviewer observed much lower initial auto fan RPM on 6900 XT (example ≈ 300 RPM) vs 6800 XT (≈ 1600 RPM in that test environment).
    • Measured noise (20” distance): 6900 XT ≈ 36.1 dBA (auto VBIOS); 6800 XT ≈ 40 dBA in the same test.
    • Max noise quoted ≈ 57 dBA (if incorrectly mounted or at max manual fan).
  • Power:
    • Total card draw ≈ 300 W stock (FurMark).
    • When OC/UV in games, measured ≈ 340–350 W.
    • Similar to 6800 XT and ≈ 20 W below the RTX 3080 FE in the reviewer’s measurements.

Pros

  • Strong rasterization performance, especially at 1080p/1440p in many titles.
  • Quieter at stock/auto settings vs the reviewer’s 6800 XT sample (lower auto fan RPM).
  • Better silicon binning allows efficient undervolting for some uplift.
  • Lower power draw than some NVIDIA counterparts at stock.

Cons

  • Weak ray-tracing performance compared with NVIDIA; DLSS gives Nvidia a large practical advantage in RT-heavy games.
  • Overclocking is confusing and can cause silent regressions; software limits (memory cap) reduce headroom.
  • Only 16 GB VRAM (vs 24 GB on RTX 3090) — may matter to creators and some high-res workloads.
  • Price/performance: hard to justify over the 6800 XT or RTX 3080 for most gamers — a halo product with limited real-world gaming value.
  • Likely constrained stock availability at launch.

Comparisons summary

  • vs AMD 6800 XT: small rasterization uplift; better binning and quieter auto behavior, but not usually enough to justify the price for most buyers.
  • vs NVIDIA RTX 3080: often similar or slightly behind in 4K rasterization, sometimes ahead at lower resolutions; the 3080 is frequently the better all-around gaming value.
  • vs NVIDIA RTX 3090: 3090 leads at 4K and in RT; 3090’s 24 GB is useful for creators. The 6900 XT competes closer at lower resolutions and is AMD’s halo SKU.

The 6900 XT is primarily a halo card: competitive in rasterization at 1080p/1440p but not compelling enough versus the 6800 XT or RTX 3080 for most buyers.

Notable numerical datapoints (selected)

  • Memory: 16 GB.
  • Noise: ≈ 36.1 dBA (6900 XT stock, auto).
  • Temperatures: edge ≈ 80°C; junction ≈ 93–95°C.
  • Power draw: ≈ 300 W stock (FurMark); ≈ 340–350 W OC/UV in games.
  • OC uplift: measured up to ≈ 8.3% versus stock after tuning.
  • Example game numbers: Tomb Raider 4K ≈ 98 FPS (6900 XT stock); Total War 4K ≈ 45 FPS (6900 XT) vs RTX 3090 ≈ 54.3 FPS; Minecraft RT 4K ≈ 15 FPS (6900 XT) vs RTX 3080 ≈ 33 FPS (no DLSS) / ≈ 87 FPS (DLSS).

Reviewer verdict / recommendation

  • For pure rasterized gaming (especially 1080p/1440p) the 6900 XT is competitive and can lead in some titles, but it does not provide enough practical benefit over the 6800 XT or RTX 3080 for most buyers.
  • If you prioritize ray tracing performance or want DLSS benefits, choose NVIDIA (3080/3070/3090 depending on needs). If you want high rasterization performance and will avoid ray tracing, a 6800 XT or RTX 3080 is typically the better value.
  • Conclusion: unless you have excess budget, specific 1080p/1440p competitive needs, or you want the halo SKU, buy a 6800 XT or RTX 3080 instead of the 6900 XT.

All unique points mentioned about the product

  • Same RDNA2 architecture as 6800 XT; largely same behaviors.
  • Same cooler as 6800 XT (PCB/VRM changes).
  • 16 GB memory; 3090 is the only current competitor with larger 24 GB framebuffer.
  • AMD positions the 6900 XT as near parity with the 3090 in rasterization (marketing claim).
  • 6900 XT is a better silicon bin vs 6800 XT; runs at lower voltages for given clocks.
  • VBIOS controls follow junction temp, not edge temp.
  • Auto fan RPM observed significantly lower on the reviewer’s 6900 XT sample (≈ 300 RPM vs ≈ 1600 RPM on 6800 XT).
  • Edge temp ≈ 80°C; junction ≈ 93–95°C under load.
  • Overclocking software limits: min/max delta 100 MHz; memory max setting 2150 MHz; GB freq limited to 3000 MHz.
  • Overclocking can be deceptive — drivers/software may accept settings but silently lower performance.
  • Best stable OC range found ≈ 2450–2550 MHz core.
  • Performance often improved by undervolting (reducing mV from ≈ 1175 to ≈ 1050–1080).
  • Port Royal used to validate OC stability/performance.
  • SAM (Smart Access Memory) impacts are game-dependent: small in some titles (Tomb Raider), large in others (Hitman 2).
  • Rasterization: 6900 XT often gains vs 6800 XT, and can beat 3090 at lower resolutions; 3090 frequently leads at 4K.
  • Ray tracing: AMD lags, especially in full path-traced titles (Minecraft) and DLSS-enabled titles.
  • Power: ≈ 300 W stock, ≈ 340–350 W when OC/UV (game-dependent).
  • Practical buying advice: 6900 XT is a halo part — better to buy 6800 XT or RTX 3080 for most gamers.
  • Likely constrained stock / not easy to buy at launch.

Other contributors / voices noted

  • Primary reviewer: GamersNexus narrator (measurements and analysis).
  • Andrew Coleman (team member) — noted as someone who can use extra VRAM (3090) in Blender/sculpting.
  • Sponsor mention: Arctic Cooling (Liquid Freezer 2 cooler recommended) — unrelated to card performance but included in the video.

Rate this summary

Your feedback will help improve summaries.

Improve this summary

Reprocess with a stronger model when the summary feels incomplete or inaccurate.

Pro

Translate summary in another language

Pro

Ask questions to this video

Chat for follow-up questions, clarifications, and source-backed answers.

Coming soon

Share this summary

Original video