Video summary
RTX 5070 - Cyberpunk 2077 - Ultra, Ray Tracing & Path Tracing Tested!
Main summary
Key takeaways
Product reviewed
- GeForce RTX 5070 Founders Edition (tested in Cyberpunk 2077)
- No manual overclock, using latest NVIDIA drivers
- Test platform: Ryzen 7 9800 X3D + 32GB RAM
- Target testing flow:
- Start at 1440p
- Then evaluate 1080p and 4K
- Across multiple RT modes and DLSS / Frame Generation combinations
Key settings & outcomes (performance / visuals)
Baseline Ultra (1440p, native)
- Ultra preset at 1440p (native) with resolution scaling disabled
- Performance:
- ~90–100 FPS while driving fast
- Dips into the 80s in heavier areas (e.g., “Bob’s roundabout” / medical center)
- Experience: described as stable and smooth (“buttery”)
- Noted downside: the reviewer would prefer more high-refresh headroom, but it’s still acceptable
User experience notes
- The game is GPU-bound, not CPU bottlenecked.
- Driving fast causes larger drops than walking or driving slowly.
DLAA / resolution scaling at 1440p Ultra (sharper, FPS hit)
- Improves image vs built-in TAA:
- Uses 100% resolution with Nvidia DAA (sharper than in-game TAA)
- Performance impact:
- Commonly 70s FPS, sometimes into the 60s
- Pros:
- Image becomes sharper and cleaner
- Cons:
- 60 FPS dips are less desirable (especially for first-person responsiveness)
Frame Generation interaction (1440p)
- With DLSS Frame Generation (2x) on top of the DAA / ~60–70 FPS baseline:
- Minimal artifacting
- Latency increases but is “acceptable” (roughly 40–50 ms, depending on area)
- Reviewer stance:
- 2x Frame Gen is recommended
- 4x Frame Gen is not recommended at 1440p:
- too many dips near/under the ~200 FPS threshold
- causes visible artifacting
Ray Tracing (RT) Ultra (1440p)
RT Ultra + no ray reconstruction (with “crappy TAA” noted)
- At native 1440p RT Ultra:
- Around ~40 FPS
- Visuals:
- Fuzzy / noisy appearance because ray reconstruction isn’t used
RT Ultra + DLAA (ray reconstruction enabled)
- Improves clarity:
- Reflections sharpen
- Less fuzzy / noisy output
- Performance:
- Described as similar overall to the earlier RT Ultra / DAA RT approach:
- often 30s in heavy areas
- Described as similar overall to the earlier RT Ultra / DAA RT approach:
- Tradeoff for Frame Generation:
- Not great for frame generation because starting FPS is often <60
- Reviewer explicitly: don’t stack Frame Gen here (bad latency / experience)
With DLSS Super Resolution Quality (1440p)
- Makes RT more playable:
- Reflections stay high quality and less noisy
- Performance:
- ~60+ baseline (around 72 FPS average) in intensive areas
- Reviewer conclusion at 1440p:
- RTX 5070 can play RT on if you use DLSS
- Use Frame Gen at 2x for smoothness
RT Ultra + 2x Frame Gen (1440p)
- Described as “buttery smooth”
- Artifacting:
- stays limited at 2x
- Latency:
- roughly 40-something ms
- 4x Frame Gen:
- near the ~200 FPS threshold
- artifacting appears when the camera moves faster and/or FPS dips
Overdrive / Path Tracing (1440p)
- Without DLSS / without ray reconstruction / no heavy RT assist:
- ~20 FPS
- Very noisy and unplayable
- With DLAA and ray reconstruction:
- around 40-ish FPS
- still not meeting the reviewer’s expectation for a $550-ish GPU
- 2x Frame Gen:
- can be used
- latency cited as ~50–60 ms and “not too terrible”
- but less suitable for faster twitch / muscle-memory situations (especially multiplayer shooters)
Overall Path Tracing verdict at 1440p
- RT Overdrive / path tracing is possible, but not ideal for a sustained “best experience.”
1080p results
1080p native (RT off, Ultra preset)
- ~150 FPS
- Reviewer calls it near-flawless for non-RT scenarios (strong headroom)
1080p with LAA/DAA (sharper)
- FPS cut roughly ~30%
- drops into 120s / 110s
- Still buttery smooth
- Reviewer emphasizes:
- On 1080p, LAA helps visuals a lot
- Still recommend 1440p upgrades overall because it looks better
1080p RT Ultra with DLSS / DLAA strategies
- Native RT Ultra initially:
- 30+ FPS, but looks bad without DLAA/DLSS cleanup
- With DLAA + DLSS quality:
- target ~60+ FPS
- example: ~62 FPS in the “medical center” test
- Frame Generation at 2x:
- described as sharp/clean
- low extra latency
- Path tracing at 1080p internal resolution:
- requires DLSS to become playable
- Frame Gen “on top of ~30 FPS” starts to feel bad
- due to input lag + artifacts
4K results (main final recommendation)
RT off at 4K
- Native 4K Ultra with TAA:
- drops into the 30s
- DLSS is necessary for 4K playability.
4K with DLSS (Performance / Quality) to reach playable RT
- 4K + DLSS Performance (RT off / ray reconstruction not the main point):
- ~80 FPS
- described as “60+ all of the time”
- 4K RT Ultra + DLSS Performance (ray reconstruction + DLSS):
- around 60s FPS baseline
- heavy zones (e.g., “Bob’s roundabout”):
- ends up in 50s
- 2x Frame Generation at 4K:
- smooths dips
- described as high-quality and responsive
- latency around 40–50 ms
Best overall configuration (reviewer’s claim)
- 4K + DLSS Performance + RT Ultra (ray reconstruction) + 2x Frame Gen
- Called the best overall experience among everything tested
Tradeoffs
- VRAM usage hits the edge:
- ~11–11.5 GB VRAM at peak
- warning about VRAM pressure at 4K with these settings
- 4x Frame Gen at 4K:
- increases artifacting
- not recommended unless starting FPS is high enough (generally needs very high baseline FPS)
Path Tracing (4K)
- 4K path tracing + Performance DLSS (no frame gen):
- 30+ FPS
- Enabling Frame Gen:
- pushes into the 60s
- but:
- input lag becomes noticeable (~60+ ms)
- artifacts become insane at higher FG factors (especially 4x)
Pros (as stated)
- Strong 1440p Ultra performance:
- typically 80–100 FPS, stable
- RT ray reconstruction + DLSS makes RTX 5070 capable of playable ray tracing
- Frame Generation at 2x is the sweet spot:
- smoother experience
- latency still considered acceptable
- minimal artifacting at 2x in recommended scenarios
- 4K can work with the right setup:
- best experience claimed: 4K + DLSS Performance + RT Ultra + 2x FG
- 1080p is described as excellent/high headroom, with easy scaling to very high FPS using FG
Cons (as stated)
- TAA in Cyberpunk is criticized heavily (“crappy TAA”):
- blurry / fuzzy / soft without DLAA / DLSS
- At RT Ultra native settings, FPS drops are steep:
- often unsuitable without DLSS
- 4x Frame Generation often causes artifacting:
- because FPS dips below the reviewer’s desired threshold (~200 FPS)
- Path Tracing / Overdrive remains very demanding:
- even with DLSS it’s not consistently “great”
- Frame Gen can worsen latency/artifacts
- At 4K, VRAM is near capacity (~11–11.5GB), potentially limiting configurations
Comparisons / reasoning against other approaches
- TAA vs DLAA:
- DLAA makes the image sharper/cleaner, but can reduce FPS
- RT without ray reconstruction vs with ray reconstruction:
- ray reconstruction significantly improves clarity (less fuzz/noise)
- Frame Gen 2x vs 4x:
- 2x is recommended
- 4x is mainly for flex/testing and tends to artifact
- RTX 5070 vs higher-end GPUs:
- path tracing is difficult even for top cards
- (reviewer notes even a 5090 has trouble with path tracing)
- Overall implication:
- DLSS is essential for making RT/path tracing feasible on this GPU
Verdict / recommendation (based on the video)
Best general playstyle (reviewer’s pick)
- 4K + RT Ultra (ray reconstruction) + DLSS Performance + 2x Frame Generation
Best 1440p playstyle
- 1440p Ultra RT with the appropriate DLSS setting (Quality/near-best) + 2x Frame Gen
- Avoid 4x Frame Gen unless starting FPS stays high
Path tracing (Overdrive)
- Prefer DLSS + careful Frame Gen, but the reviewer finds it less satisfying overall due to:
- latency
- artifacting when pushed
Overall: RTX 5070 is positioned as a strong 1440p card with excellent RT via DLSS, and it can even deliver convincing 4K RT—but path tracing is the weak point, especially at higher settings without heavy upscaling and conservative Frame Gen.
Unique points by speaker
- Only one main speaker (Chris) is present in the provided subtitles.
- No distinct additional speakers with separate technical conclusions are clearly indicated beyond incidental “Bob”/NPC references (not separate reviewers).