The Witcher 3 Remastered Benchmarked on 14 GPUs. The RTX 5090 Equals the 5080

By Indie Kings | September 29, 2026

Updated September 29, 2026: TechSpot benchmarked The Witcher 3: Wild Hunt Remastered across fourteen graphics cards this morning, and the data answers the question our own coverage had to leave open. Our 161 FPS piece measured one card, on a pre-release build, above the official ceiling, and said so plainly. These fourteen were measured with frame generation off, which makes them directly comparable to CDPR's official target in a way our single figure never was. Four findings are at zero across this site: an RTX 5090 and an RTX 5080 are indistinguishable in this game at several settings, Radeon cards hit a hard ceiling far earlier than GeForce ones, the Remastered runs 36 to 45 percent faster than the original at 4K Medium, and path tracing frequently looks worse than the ray tracing it replaces.

Image: The Remastered on PC, the version benchmarked across fourteen cards. Credit: Wccftech.

What was measured, and what was left out

Two methodology facts decide how much weight this carries, and both cut in the reader's favour.

First, frame generation was not used anywhere in this test. That is not stated as a limitation, it is simply absent: the article contains no mention of frame generation, DLSS, FSR, XeSS, Reflex or V-Sync, and the upscaling is described only as Quality, Balanced and Performance at the three resolutions. The consequences are direct. A reader comparing these numbers to a requirements table is comparing like with like, and a reader comparing them to our 161 FPS figure is not, because that run had Multi Frame Generation on Dynamic.

Second, the AMD driver was a pre-release build. TechSpot states that all AMD testing used version 26.20.15.02, developed specifically for the Remastered, and that the public Adrenalin 26.9.1 was "effectively unusable in our testing, producing numerous rendering glitches." Every Radeon number below is therefore a best case that a player on the public driver may not reproduce, and it is the single biggest caveat on the dataset.

The rest of the setup is a Ryzen 7 9800X3D with 32GB of DDR5-6000, tested in Novigrad, which TechSpot chose deliberately because it has historically been among the most demanding locations in the game. That makes these numbers pessimistic relative to open world, and TechSpot quantifies the gap: moving outside the city raises the path traced ceiling from roughly 60 FPS to roughly 80. Published September 29 at 05:08 CDT, so this is a launch-day read on pre-release silicon and drivers, not a settled verdict.

The 5090 and the 5080 are the same card here

This is the finding that should change buying behaviour, and it is not a close call.

TechSpot's own summary is blunt about the open-world case: outside Novigrad the path traced ceiling rises to roughly 80 FPS, "but even there the RTX 5090 is no faster than the RTX 5080."

The pattern is consistent enough to name. At the ray traced and path traced presets, in the city, the top two NVIDIA cards converge on an identical number at every resolution tested. That is the signature of a bottleneck that is not the GPU. It is consistent with a CPU-side limit, with a frame-time cap, or with a game-side limiter, and TechSpot does not claim to know which. What it does establish is that at the settings where path tracing is meant to shine, the most expensive card in the range buys nothing over the one below it.

One piece of supporting evidence for a non-GPU limit, offered as an indication rather than a conclusion: when path tracing was tested at 1080p on the RTX 5090, CPU utilization peaked at around 35 percent, with no core of the 9800X3D near saturation. On a GPU-bound path traced workload that number is hard to explain, and it is the first thing worth raising with CD Projekt Red.

Radeon hits a wall that GeForce does not

The second ceiling is vendor-specific and it is the reason several otherwise capable cards post misleading numbers.

TechSpot attributes the Radeon flattening to a driver-related limitation rather than silicon, and the tell is that the ceiling moves with preset rather than with card. Four different Radeon cards landing within two frames of each other, across 1080p and 1440p, is a cap rather than a hardware ranking. They also note that at 1080p Ultra the RX 9070 XT and the RX 9070 post the same average frame rate despite the XT producing 12 percent stronger 1 percent lows, which is another shape a limiter takes.

The practical consequence is that a Radeon card's headline number in this game currently says more about the driver than the card, and the ranking only becomes trustworthy at 4K, where TechSpot says the ceilings finally disappear and the RX 9060 XT 8GB and 16GB both reach 64 FPS on bandwidth alone.

The remaster is faster than the game it replaces

The most useful result for anyone deciding whether to upgrade, and the one that runs against the assumption that a remaster costs performance.

At 4K on the Medium preset, TechSpot measured gains of 36 percent for the RTX 5070 Ti and 45 percent for the RX 9070 XT over the original game, which is now labelled Classic. The 5070 Ti at Medium is 5 percent faster at 1080p and 23 percent faster at 1440p on the Remastered than on the original, despite the Remastered looking considerably better at that preset. In the conclusion the range is given as roughly 45 to 58 percent in the example being discussed.

TechSpot also states plainly that the Remastered can run faster than the original in some cases, which is the sentence most likely to be dropped in coverage and the one a buyer most needs.

The reason is visible in the scaling rather than stated outright. Medium and Ultra are not GPU-bound on these cards in Novigrad, they are hitting the caps described above, so raising the visual settings costs little. That also means the Medium result should not be generalised to the ray traced presets, where the same test run shows the 5070 Ti taking a 23 percent hit at 1440p and a 7 to 12 percent hit elsewhere. The faster-than-original finding is a Medium and Ultra finding.

8GB cards collapse at 4K and nobody will say so

The third finding is the one with the clearest money attached, because the cards affected are current-generation retail products rather than old silicon.

Two entries deserve emphasis. At 4K with ray tracing and Performance upscaling, the RTX 5060 Ti 16GB delivers a somewhat playable 37 FPS while the 8GB version falls to 13 FPS, trailing even the standard RTX 5060 and the Arc B580. Those two cards are the cheaper purchase, so the 8GB model is not a budget option at that setting, it is the worst card in the chart.

The bandwidth story runs the other way for AMD. At 1440p Ultra the 8GB RTX 5060 Ti was almost 10 percent slower in average frame rate and a little over 20 percent slower in 1 percent lows than the 16GB model, which TechSpot attributes to reduced PCIe bandwidth rather than VRAM. At 4K Medium the RX 9060 XT 8GB and 16GB both reach 64 FPS, which TechSpot credits to their greater bandwidth. Same nominal VRAM, opposite outcome, and the variable is the interface.

Path tracing is the worst visual upgrade in the game

The fourth finding is a judgement rather than a measurement, and it is reported as one. TechSpot states that for its review path tracing frequently looks worse than conventional ray tracing, that characters can become overexposed, that other scenes take on a strange blue tint, and that Yennefer's face in particular can look flatter and more blown out with path tracing on. Their conclusion after examining numerous scenes is that they generally found ray tracing more visually pleasing, and they describe that as disappointing given the extent of the remaster.

This sits awkwardly against the hardware, and the hardware is the interesting part. Path tracing is the mode CDPR reserved a dedicated, NVIDIA-only hardware tier for, and the mode where the published requirements ask for an RTX 5080. It is also the mode with the tightest ceilings in the entire dataset, where the 5090 and 5080 are pinned together at 62 and then 59 FPS, and where the Arc B580 manages 14 FPS at 1080p and 7 FPS at 4K.

So the most expensive configuration in the game is attached to the visual mode the reviewer preferred less, on the settings where the hardware is least differentiated. That is a coherent thing for CD Projekt Red to change, and it costs them nothing to change the recommendation.

What this says about CDPR's published targets

This is the section that connects to our earlier coverage, and the connection is favourable rather than corrective.

Our September 28 coverage of CDPR's requirements table established that the Path Tracing tier is defined as an RTX 5080 at 2160p holding 60 FPS with frame generation, and noted that this implies a base frame rate meaningfully below 60 being topped up. Our 161 FPS piece went further, stating in its FAQ that the only rigorous measurement then published was a single pre-release run on an RTX 5090 above the official ceiling.

TechSpot's 43 FPS for the RTX 5080 at 4K with path tracing and Performance upscaling is not a contradiction of that 60 FPS target, and the reason is the one established above: CDPR's figure includes frame generation and this test does not. Reading it as a failure of the official spec would be exactly the kind of relay error this site exists to avoid. The honest reading is narrower and more interesting: the base frame rate on the card CDPR names for its top tier is 43 FPS in a demanding city at 4K, and the published 60 is a synthesised figure.

Two things follow. A player buying to the Path Tracing tier is buying a card that needs frame generation to reach its target, which our earlier coverage already flagged and which this data now quantifies rather than infers. And a player who skips path tracing entirely, on the strength of the visual finding above, gets a materially better experience: the RX 9070 at roughly 650 dollars averaged 70 FPS at 1440p ray traced with Balanced upscaling, and the RX 9060 XT 8GB averaged 121 FPS at 1440p on Medium at native resolution.

One comparison is worth flagging for our own Steam Deck coverage. This site reported the Remastered costs the Steam Deck a substantial fraction of its budget, from a pre-release build. TechSpot's 1440p Medium result of 121 FPS for the RX 9060 XT 8GB and 102 FPS for the Arc B580 is the closest current-generation analogue to handheld-class hardware in this dataset, and it says the game is far less demanding without ray tracing than the path traced tier implies.

What is still unverified

Collecting the open items, because this is a single-source dataset and that has to be said plainly rather than buried.

• Every frame rate here is one reviewer's one machine. TechSpot says it settled on a smaller initial sample because of the short lead time, and that the game will appear in its regular test suite later. No second lab has reproduced these numbers.
• The AMD figures depend on a driver the public cannot install. 26.20.15.02 is described as developed specifically for this game, and the public 26.9.1 as producing rendering glitches. Every Radeon number is provisional.
• The 5090-equals-5080 ceiling has no confirmed cause. The 35 percent CPU utilisation is an indicator, not a diagnosis. TechSpot does not attribute it, and neither can we.
• Path tracing looking worse is a reviewer's opinion. It is TechSpot's judgement, stated as such, and reasonable people will differ. It is not a measurement.
• No frame time, no 1 percent low dataset, and no latency figures were published for these fourteen cards. TechSpot refers to 1 percent lows in the Ultra comparisons, but the underlying per-card lows are in chart images rather than in the article text, so no per-card low is quoted here.
• Frame rate versus CDPR's table is our comparison, not TechSpot's. The article does not reference CDPR's requirements.

What is verified: all fourteen card results, at the stated presets, resolutions and upscaling modes, on a Ryzen 7 9800X3D in Novigrad; the pre-release driver string and the public driver's condition; the CPU utilisation figure; the path tracing visual judgement; and the 36, 45 and 45 to 58 percent gains over Classic at Medium.

FAQ

Is the RTX 5080 really as fast as the RTX 5090 in The Witcher 3 Remastered?

At the ray traced and path traced presets in Novigrad, yes, they tie. Both hit 93 to 94 FPS at 1440p on RT Ultra, both hit 62 FPS at 1080p path traced, and both hit 59 FPS at 1440p path traced. Outside the city the path traced ceiling rises to roughly 80 FPS and they still tie. They separate at 4K on Ultra and Medium, where the 5090 leads by 62 percent and about 44 percent respectively. So the answer depends entirely on which preset you intend to run.

Does this mean CDPR's RTX 5080 path tracing requirement is wrong?

No, and the distinction matters. CDPR's published tier is 60 FPS at 2160p with frame generation. TechSpot measured 43 FPS for that card at 4K path traced with Performance upscaling and did not use frame generation anywhere in the test. The two figures are measuring different configurations, so 43 against 60 is not a failure of the spec. What it does show is that the base frame rate on CDPR's own top-tier card is 43 FPS in a demanding location, and the published 60 is topped up by synthesis.

Why do several Radeon cards post nearly identical frame rates?

TechSpot attributes it to a driver-related performance limitation rather than a hardware ranking, and the evidence is that the ceiling tracks the preset rather than the card. At 1080p Medium the RX 9070 XT, RX 9070, RX 9070 GRE and both RX 9060 XT models all land between 147 and 149 FPS while four different GeForce cards sit between 194 and 198. The rankings only become meaningful at 4K, where the ceilings disappear. Note also that all AMD testing used a pre-release driver, version 26.20.15.02, because the public 26.9.1 was producing rendering glitches.

Is the Remastered faster than the original game?

At Medium, yes, and by a wide margin. TechSpot measured 4K Medium gains of 36 percent for the RTX 5070 Ti and 45 percent for the RX 9070 XT over Classic, and put the range in one example at roughly 45 to 58 percent. The RTX 5070 Ti is 5 percent faster at 1080p and 23 percent faster at 1440p on the Remastered than the original. This does not generalise to the ray traced presets, where the same cards lose between 7 and 23 percent moving to Remastered.

Should I avoid the 8GB version of the RTX 5060 Ti?

At 4K with ray tracing, yes, unambiguously. The 16GB model delivers 37 FPS with Performance upscaling where the 8GB model manages 13, which is worse than the standard RTX 5060 and the Arc B580, both cheaper cards. At 4K Medium the gap narrows to 67 against 47 FPS, still large. The AMD comparison inverts: the RX 9060 XT 8GB and 16GB both reach 64 FPS at 4K Medium, which TechSpot credits to their greater PCIe bandwidth, so VRAM capacity is not the whole variable.

Is path tracing actually better looking?

TechSpot's assessment is no, or at least not reliably. They report that path tracing frequently looks worse than conventional ray tracing, with characters becoming overexposed and some scenes taking on a blue tint, and that after examining numerous scenes they generally preferred ray tracing. They call that disappointing given the scope of the remaster. It is a reviewer's judgement rather than a measurement, and it is the mode CDPR reserved a dedicated NVIDIA-only hardware tier for, which is what makes it worth knowing.

Bottom Line

Fourteen cards, one city, one CPU, and a set of results that rearrange the buying question in a way the requirements table does not.

The headline is not that the game runs well, though at Medium it runs extremely well and sometimes faster than the eleven-year-old original it replaces. The headline is that the game's most expensive configurations are the least differentiated. An RTX 5090 buys nothing over an RTX 5080 at ray traced and path traced settings, four Radeon cards sit pinned to the same artificial ceiling, and the mode CDPR reserved a whole hardware tier for is the one the reviewer preferred less. A 35 percent CPU utilisation reading on a path traced 5090 suggests the limit is not the GPU at all, and that is the question worth putting to the developer.

For this site's own earlier coverage, the honest update is a confirmation rather than a correction. We reported on September 28 that CDPR's top tier is defined with frame generation on, and that the 60 in that column is a synthesised figure rather than a rendered one. This dataset puts a number on it: 43 FPS base on the card CDPR names, 60 published. We also said the only rigorous measurement available was a single pre-release run on a 5090. That is now superseded, and the reason it took a launch-day review rather than a week of testing is worth remembering when deciding what to wait for.

Related: The Witcher 3 Remastered Averages 161 FPS. The Same Run Reveals Where 66 Milliseconds Goes | CDPR Published Six Sets of Witcher 3 Requirements. Steam Still Shows the Old Ones | CDPR's Own Change List Corrects the Remastered Coverage. FSR 4 and Frame Generation Are Not the Same | The Witcher 3 Remaster Costs the Steam Deck a Quarter of Its Budget