Video summary

AMD Radeon RX 6700 XT GPU Review: Literally Anything Will Sell

Main summary

Key takeaways

Product Review

Product Reviewed

AMD Radeon RX 6700 XT (partner models launch March 18, reviewed in advance)

  • Positioning: MSRP $480 (AMD “SCP” / suggested retail)
  • Architecture: 40 compute units (cut-down)
  • Memory: 12GB GDDR6
  • Board power: 230W advertised (target)

Review notes that the 40-CU design is a different die than the larger 6800 / 6800 XT parts.

Key Features Mentioned

  • 40 CU RDNA (cut-down GPU / different die vs 6800/6800 XT)
  • 12GB GDDR6
    • Reviewer argues 12GB is fine for this segment.
    • VRAM tools like GPU-Z / Hardware Info show allocated memory, not necessarily actual usage.
  • Board power target: 230W advertised
  • Reference behavior (thermals)
    • High junction temps (90s) with auto fan control
    • Tends to keep the core “hot” by default
  • Overclocking / undervolting
    • Slight gains are possible, but the OC process is “erratic” vs NVIDIA.
    • Validation is required, because some voltage changes can reduce performance.
  • Benchmark focus areas
    • Rasterization (1080p / 1440p / 4K)
    • Ray tracing
    • Thermals / fans
    • Power

User Experience: Thermals & Clocks

Frequency behavior / validation

  • Typical clock references:
    • RX 6800 XT: 2300–2360 MHz
    • RX 6800: ~2280 MHz (flat)
    • RX 6700 XT: closer to ~2560 MHz
  • Clocks bounce due to power limits
    • Can be flattened by boosting the power limit, but at the cost of higher power draw.

Thermals (reference-style cooler behavior)

  • Average edge: ~75–77°C
  • Junction stays in the 90s
  • Can hit ~100°C under certain conditions (still below throttling point)
  • Edge-to-junction delta: ~25–30°C
    • Described as undesirable and attributed to mounting / thermal solution limitations

Overclocking / Undervolting Results

Port Royal (ray tracing benchmark)

  • Baseline: 58.64
  • Slight power-limit change (~+1.15%): +1.3%
  • Undervolting outcomes:
    • 1140 mV attempt: regression
    • 1160 mV: gain
  • Reported overall uplift: +5.6% (Port Royal)

Stability limits

  • Higher memory OC → memory errors
  • Higher core clocks → stability issues
  • Conclusion: limited headroom and lots of tuning time required for more gains.

Benchmark Takeaways (Positioning vs Competitors)

General theme

At this performance tier, results are often framed as being between an RTX 3060 Ti and RTX 3070. However, the reviewer repeatedly argues the MSRP gap isn’t compelling.

Rasterization Examples

Rainbow Six Siege

  • 1080p: ~371 FPS avg
    • Between RTX 3070 and RTX 3060 Ti
    • Reviewer: below what it “needs to be” for its MSRP/value
  • 1440p: ~388 FPS avg
    • Overclock mentioned for the result
    • “Good for high refresh at 1440p,” still between 3060 Ti and “6800-class” cards (with AMD scaling comments)

Shadow of the Tomb Raider

  • 1080p: ~165 FPS avg
    • Above RTX 3060 Ti; functionally similar to RTX 3070
  • 1440p: ~119 FPS avg
    • Again between RTX 3070 stock and RX 6800

Red Dead Redemption 2

  • 1080p: ~134 FPS avg
    • About 9% ahead of RTX 3070
  • 1440p: ~101 FPS avg
    • Functionally equivalent to RTX 3070; slightly ahead of 3060 Ti
  • 4K: playable, but more sensitive to lows
    • RTX 3070 pulls away

Cyberpunk 2077 (older patch used)

  • 1080p: ~97 FPS avg
    • Above RX 5700 XT (~+19%)
  • 1440p: ~133 FPS avg
  • Frame-time plot: small early hitches, generally acceptable

GTA V (scaling / CPU bind notes)

  • 1440p: ~133 FPS avg
  • 4K: ~61 FPS avg
    • Where RTX 3070 surpasses RX 6700 XT
    • RX 6800 (non-XT) much higher

Strange Brigade

  • 1440p: ~185 FPS avg (higher with OC)
  • Placement: between RTX 3070 and RX 6800
  • Reviewer note: still too expensive for value

Ray Tracing Examples

Quake 2 (ray tracing)

  • ~52 FPS avg (between RTX 2060 KO and 2070 S)
  • Reviewer emphasizes RTX dominance for heavy RT
  • At 1440p: ~46 FPS avg for RX 6700 XT
    • RTX 3060 Ti much higher

Control (RT + DLSS discussion)

  • AMD can’t use DLSS; reviewer compares RT performance with DLSS disabled
  • Without DLSS:
    • ~49 FPS avg
    • Struggles at higher resolutions
  • At 4K with DLSS disabled: ~15 FPS avg

Minecraft RTX

  • Reviewer: AMD couldn’t run beyond 1080p as playable
  • At 1080p: very low performance for RX 6700 XT
  • RTX 3060 Ti with DLSS is far higher (example reported: ~67 FPS avg, ~114% ahead)

Power Consumption

  • “Power virus” workload:
    • RX 6700 XT maxes at ~219W
    • Slightly below RTX 3070 stock (reported as just under it)
  • PCIe slot power draw: checked and reported “fine”

Pros Mentioned

  • Strong rasterization at 1080p/1440p; often competitive with RTX 3070 / RTX 3060 Ti depending on the title
  • 12GB VRAM is not inherently problematic at this tier (per reviewer explanation)
  • Thermals reasonable on average edge temps (junction hot, but average is OK)
    • Fans around ~1700 RPM in test
  • Power usage roughly in line with expectations (~219W)

Cons Mentioned

Price / value criticism

  • Reviewer repeatedly argues performance isn’t exciting enough at MSRP ($480)
  • Value framed in terms of diminishing returns (example: “$1 per ~1% improvement”)
  • Supply scarcity reduces real price pressure, so conclusions depend heavily on MSRP rather than market availability.

Thermal behavior

  • Junction temps in the 90s; can reach ~100°C
  • Large edge-to-junction delta (~25–30°C)

Ray tracing weakness

  • In heavy RT titles, RX 6700 XT is far behind NVIDIA
  • DLSS advantage further worsens comparisons in supported games

Overclocking limitations

  • OC process is harder/erratic; limited stable headroom

4K mismatch

  • Some games remain playable, but often too weak compared to RTX 3070 / 6800-class options

Comparisons Explicitly Made

  • Versus: RTX 3060 Ti and RTX 3070 (main raster competition)
  • Versus: RX 6800 and RX 6800 XT (scaling and higher-tier positioning)
  • Versus: RX 5700 XT (generational uplift discussion; reviewer frames it as limited vs price)
  • Ray tracing comparisons against NVIDIA cards (e.g., RTX 2060 / 2060 KO / RTX 3070 / 3060 Ti)
  • Resizable BAR (rebar) discussion:
    • Reviewer notes small gains
    • Available to both AMD and NVIDIA in the reviewer’s context

Overall Verdict / Recommendation (From the Conclusion)

  • The card is described as “fine” and not “garbage.”
  • But it’s hard to get excited because:
    • Value at MSRP isn’t compelling versus RTX 3060 Ti / RTX 3070
    • Ray tracing is not its strength
    • Market conditions reduce the likelihood of aggressive pricing competition
  • Practical implication:
    • Buy based on what’s available in stock, and avoid problematic variants (review mentions some “hard avoid” examples generally), though this particular card wasn’t positioned as one of them.

Unique Points Mentioned (Deduplicated)

  1. Launch timing: reviews March 18, one day early due to supply/market behavior.
  2. Partner models launch same day; MSRP expected $480.
  3. Architecture: 40 CU cut-down die, different silicon vs 6800/6800 XT.
  4. Test plan includes gaming (raster + ray tracing), thermals, frequency validation, and power.
  5. Market critique: GPUs may “sell out anyway,” limiting price competition.
  6. Spec emphasis: 12GB GDDR6; reviewer argues it’s sufficient.
  7. VRAM tools explain allocated vs actual usage.
  8. Future note: VRAM may matter more under ray tracing.
  9. Advertised board power 230W, with evaluation grounded in MSRP/SCP context.
  10. Frequency behavior: lower-tier AMD parts can run higher clocks; power-limit causes bouncing.
  11. Increasing power limit can flatten clocks but increases power.
  12. Thermals: edge ~75–77°C, junction in 90s, can approach ~100°C; throttling expected above ~110°C (core).
  13. Edge-to-junction delta (~25–30°C) is undesirable; attributed to mounting/thermal solution limitations.
  14. OC is “erratic” vs NVIDIA; limited stable headroom.
  15. Port Royal: power-limit +1.15% ≈ +1.3%; undervolt results net overall +5.6%, with bad behavior around 1140 mV.
  16. Stability: memory errors from higher memory OC; core instability from higher clocks.
  17. Rainbow Six Siege 1080p: ~371 FPS avg, between RTX 3070 and 3060 Ti.
  18. Rainbow Six Siege 1440p: ~388 FPS avg, overclocked, still between tiers.
  19. R6 Siege 4K: ~100 FPS, playable; AMD scaling framed as inverse to NVIDIA behavior.
  20. Shadow of the Tomb Raider: 165 FPS @1080p, ~119 FPS @1440p.
  21. Overclock gains framed as small (~3%) due to stability constraints.
  22. RDR2: 134 FPS @1080p (+9% vs RTX 3070), 101 FPS @1440p; 4K playable with low sensitivity concerns.
  23. Frame times: generally consistent; small early hitches (max ~7–8 ms changes).
  24. Cyberpunk 2077: older patch; 97 FPS @1080p, 133 FPS @1440p.
  25. GTA V: CPU-bound notes; 61 FPS @4K, where RTX 3070 gains vs RX 6700 XT.
  26. Strange Brigade: 185 FPS @1440p, still “too expensive” for value.
  27. Ray tracing: Quake 2 not competitive; ~52 FPS avg (and ~46 FPS @1440p).
  28. Control: DLSS absence hurts AMD; ~49 FPS no-DLSS RT, and ~15 FPS @4K no-DLSS.
  29. Minecraft RTX: AMD limited to 1080p playable, with very low performance.
  30. Power: workload max around ~219W; PCIe draw reported fine.
  31. Final framing: “whatever/fine,” but disappointing due to MSRP value and RT limitations; buy depends on stock price.

Speaker-Specific / Different Perspectives

  • Single primary reviewer voice (not multiple speakers).
  • The review consistently combines:
    • Market/value critique (MSRP vs limited competition during scarcity)
    • Technical validation tone (clock validation, junction/edge thermals, OC stability)
    • Value judgment tone (not compelling vs RTX 3070-class alternatives; RT weakness)

Verdict framing: Buy RX 6700 XT only if you find a good in-stock price, because at $480 MSRP the review finds it not compelling versus RTX 3060 Ti / RTX 3070, and particularly weak for ray tracing compared to NVIDIA.

Original video