The Witcher 3 Remaster Costs the Steam Deck a Quarter of Its Frame Rate. Its Preset Also Changed Under You

By Indie Kings | September 28, 2026

Updated September 28, 2026: The Witcher 3 Remastered launches tomorrow and it costs roughly a quarter of the Steam Deck's frame rate. A pre-release hands-on measured the stock Deck preset dropping from 50 to 60 fps before the remaster to about 42 to 44 fps after, averaging near 44, with one dip to 39. That makes a 40 fps lock the realistic ceiling, where CD Projekt Red's published Deck tier claims 800p at 30 fps. The new preset also turns FSR on by default, which the old one did not, and raises ambient occlusion to High. Changing settings to High crashed the game at least twice.

Image: Steam Deck. Credit: Valve via Steam.

What was measured, and on what

First, the honest caveat, because it changes how much weight the numbers carry. This is a single pre-release build tested on a single device, by one person, over a playthrough, with no frame capture tooling mentioned and no settings sweep. It is a hands-on impression with numbers attached, not a benchmark. Nothing here should be read as a measurement of a production build.

With that stated, the figures were reported as follows. On the stock Steam Deck graphical preset, areas that were running at 50 to 60 fps before the remaster sat at roughly 42 to 44 fps after. The report describes this as consistent throughout the game rather than confined to particular scenes, and puts the average at around 44 fps against high 50s previously. The single lowest frame rate observed was 39 fps.

So the shape of the change is a fairly even reduction of about 8 to 16 fps depending on where you were, landing the whole game in the low 40s instead of the mid 50s. That is not a catastrophic result on a device that was already managing the original, but it is a real cost, and it arrives alongside a substantial visual upgrade rather than instead of one.

The comparison is the interesting part. The tester's stated prior expectation was that "with the updated visuals, I expected it would be harder to run," and the finding is that the worry was partly warranted. The result is not "the Deck can no longer run it." It is "the Deck now runs it about 12 frames per second slower than it did, and the way to get that back is to accept a lower frame rate or lower settings."

Why 40 fps is the number to plan around

The frame rates in this range are where the choice becomes real, and it is worth being concrete about why 40 rather than 30 or 45.

The tester concluded that the game "sticks above 40 FPS consistently, with only 1 drop that I saw go to 39 FPS," and judged that a 40 fps lock with the stock Steam Deck preset should be solid and enjoyable. That is a deliberate recommendation rather than a description, and the reasoning is straightforward: 40 divides evenly into 120 Hz panels, giving consistent three-to-three frame pacing, whereas 45 does not divide into 120 at all and would produce visibly uneven pacing.

That is the same observation this site made when the Switch 2's 40 fps tier was announced, that 40 is the legitimate middle rung above 30 in feel and below 60, and it applies identically to the Deck's 120 Hz display. The Deck LCD is 60 Hz, so on that panel 40 fps does not divide evenly and the 3:3 argument does not carry, which is a distinction worth keeping in mind. The Deck OLED is 120 Hz.

So the answer depends on which Deck you own, and that is worth saying plainly. On the OLED, lock at 40. On the LCD, where 40 does not pace cleanly, the sensible options are a 30 fps lock if you want the settings headroom, or accepting the low 40s knowing the pacing will be less even than the number suggests.

The preset changed under you, and that matters more than the frame rate

The frame rate is the number people will quote. The preset change is the one that will change how the game looks without anyone deciding it should.

The Deck preset on the Remastered version is not the old preset with better textures. Per the report, it now has more enabled by default: ambient occlusion set to GTAO on High rather than off or low, screen space shadows on, and FSR turned on by default, which it was not previously. Most other settings remain on Low and ray tracing is off.

Each of those is a defensible default on its own, and taken together they explain a meaningful part of the performance drop. Screen space ambient occlusion is one of the more expensive post-process effects in the deferred renderer, and GTAO specifically costs more than the older simpler ambient occlusion. Adding it at High on top of a lighting rework is not going to be free, and the lighting rework is itself the point of the remaster.

What makes the FSR default worth flagging separately is that it changes the image. FSR upscaling trades native resolution for frame rate with a reconstruction pass, and it was previously something a player opted into on the Deck. Now it is on out of the box. A player who launches the remastered game expecting the sharp look they had and finds it softer, with aliasing the tester specifically described as the old Deck's original problem, has run into a default rather than a mistake on their part.

The tester's own read on the visual result was that the game "still looks massively different, thanks to the new lighting rework and shading model," which is the point, and that they would consider the change a worthwhile improvement. That is a judgement about the remaster rather than about the Deck, and it should be held separately from the question of whether 44 fps is enough. Both can be true, and in the reviewer's view both were.

The settings crash, and what it does to your confidence in the settings

This is the most actionable item in the whole piece and the easiest to skim past, because it is reported as two words in a paragraph about graphics options.

The report states plainly: "I did experience a crash or two when trying to change settings to High", and that a reboot brought the game back with those settings turned on. So the crash occurred on attempting the change, the process recovered by restarting the machine, and the settings the player wanted were present on the next launch.

Two things follow, and the second is the one that matters. The obvious reading is that this is a launch-day pre-release quirk and day-one patch 1.01 may or may not address it. The subtler reading is what it does to a player's willingness to experiment: if raising a graphics option can crash the game, players on a low-memory device who are already considering whether they can afford higher settings are now weighing a setting change against a possible reboot. On an eight gigabyte handheld running a Windows game through a compatibility layer, that is a real hesitation, and it is not the hesitation anyone would have had before this remaster.

The safe advice that follows is unglamorous. Take the settings change in one direction at a time rather than a batch, and if the game crashes on a change, reboot rather than assuming the setting is now permanently applied, because the report's own observation is that it was applied on the next launch but that is a single data point from a pre-release build and not a guarantee. Nobody should be told the crash is fixed, because no evidence exists either way yet.

Load times, and one load time that is not fine

Load behaviour was checked and produced one good result and one clear problem.

The good result: loading times between the two versions were about the same. For a free update that adds path tracing, higher resolution textures and a rebuilt asset pipeline, unchanged load times are a genuine achievement, and they are consistent with the DirectX 12 work this site covered, where the shader cache is prebuilt on first boot rather than compiled at load.

The problem: loading into a recent save game after dying took way longer than it should have, while first-time loading was described as not too bad. That is a specific and reproducible-sounding complaint rather than a general one, and it points at save loading specifically rather than the game's asset pipeline.

It matters more than a normal load time complaint because of where it happens. Loading a save after a death is what a player does dozens of times in a difficult boss fight, and it is the moment you least want a stall. If the pattern holds into the launch build, it is worth knowing before you attempt anything hard, and it is the kind of thing that gets reported as "the game is slow" by people who never isolated it to post-death loading.

Also in the positive column, and more than a minor change: the text size of some of the UI has gotten bigger, which the tester found makes quest details and the current time noticeably easier to read on a seven inch screen. That is a real Deck-specific quality of life gain that has nothing to do with frame rates, and on a handheld it may matter as much as the graphics do.

The published target versus the measured result

Putting the two together produces a defensible conclusion and an honest caveat about both sides of it.

CD Projekt Red's published Steam Deck tier, as covered before launch, is 800p at 30 fps with no ray tracing, FSR on Performance, on both the LCD and OLED models, with the publisher's own assessment that it "runs like the original."

The measured result from a pre-release build is roughly 42 to 44 fps with the stock preset, averaging near 44, which is comfortably above the published 30 fps target. So the target is not wrong, and CDPR's assessment that the Deck can handle it holds up. The remaster does not make the game unplayable on the device, and the published target is a conservative floor rather than a stretch.

The two figures are measuring different things, and that is the honest reconciliation. A published target is what the developer will support, chosen with margin for the worst case. A measured average is what one playthrough produced, with the game reaching areas it may not always return to. A 30 fps supported target and a 44 fps measured average are consistent with each other, and treating the measured figure as a new promise would be a mistake given it is one build and one playthrough.

What can be said with confidence is narrow and worth stating narrowly. The Deck still runs this comfortably. The frame rate is lower than it was, by roughly a quarter. The way to recover frames is a 30 fps lock with settings pushed up, with the crashes described above as the cost of trying. And every one of those numbers came from a pre-release build that day-one patch 1.01 may change, so they are a starting expectation rather than a result to hold the publisher to.

FAQ

What frame rate does The Witcher 3 Remastered run at on the Steam Deck?

A pre-release hands-on measured roughly 42 to 44 fps on the stock Deck graphical preset, averaging around 44, down from 50 to 60 fps before the remaster, with one dip to 39 observed. That was a single playthrough on a single device with no settings sweep, so treat it as an expectation rather than a benchmark. CD Projekt Red's published target is more conservative at 800p and 30 fps on FSR Performance with no ray tracing.

Should I lock the Steam Deck to 40 fps or 30 fps?

Lock to 40 on the stock preset if the game is holding above it, which the test found it did consistently. On a 120 Hz panel such as the Deck OLED, 40 divides evenly into 120 for clean three-to-three frame pacing, while 45 does not divide into 120 at all and paces unevenly. Drop to a 30 fps lock if you want to push settings up. On the 60 Hz LCD, 40 does not pace cleanly, so 30 is the tidier choice there.

What changed in the Steam Deck graphical preset?

More is enabled by default than before. FSR upscaling is now on by default where it previously was not, ambient occlusion is set to GTAO on High, and screen space shadows are on. Ray tracing remains off and most other options stay on Low. The FSR default is the one that changes the image, softening the too-sharp look the Deck preset previously had and its visible aliasing.

Does changing settings crash the game on the Deck?

A pre-release build crashed once or twice when settings were changed to High, and a reboot restored the game with those settings applied. This was observed on a pre-release build and day-one patch 1.01 may or may not address it, so no claim is made either way. Take changes one at a time rather than in a batch, and if the game crashes on a change, reboot rather than assuming the setting is stuck.

Are load times any different after the remaster?

Broadly the same, which is a genuine result for an update that adds path tracing, higher resolution textures and a rebuilt asset pipeline, and it is consistent with the prebuilt shader cache the DirectX 12 port introduced. One exception was reported: loading a recent save after dying took far longer than it should, while first-time loads were acceptable. That one is worth watching in the launch build, since post-death loading happens repeatedly in difficult fights.

Did CD Projekt Red's promise that it runs like the original hold up?

On the evidence available, yes. The published target of 800p and 30 fps with FSR Performance is a conservative floor, and a measured 44 fps average sits comfortably above it. A target chosen with margin for worst cases and a measured average from one playthrough are consistent with each other rather than in conflict, though the publisher's assessment of running like the original is unmeasured and the frame rates come from a pre-release build.

Bottom Line

The number to take from the pre-release testing of the Witcher 3 Remastered on the Steam Deck is 40, and it is a frame rate rather than a feature. The game holds roughly 42 to 44 fps on the stock graphical preset, averaging near 44, down from 50 to 60 fps before the remaster, with a single observed dip to 39. That is a cost of about 8 to 16 fps, or roughly a quarter of the frame rate, and it is the price of a free update that brings path tracing, a lighting rework and higher resolution textures to an eleven year old game. Nobody is suggesting the Deck cannot run it. The suggestion is that the headroom that existed before is gone, and that anyone who was running the original at a comfortable 50 to 60 and assumed they would keep that is now budgeting for the low 40s.

Forty is the right number to plan around rather than thirty, and the reason is arithmetic before it is preference. Forty divides evenly into a 120 Hz panel for clean three-to-three frame pacing, forty five does not divide into 120 at all and produces visibly uneven pacing, and sixty is simply not reachable on the stock preset any more. So on an OLED Deck, lock at 40. On the 60 Hz LCD, where 40 does not divide evenly, a 30 fps lock is the tidier option and it is also the route to pushing settings up. CD Projekt Red's published Deck target is 800p at 30 fps on FSR Performance with no ray tracing, and the measured average sitting well above that is a point in the publisher's favour rather than a contradiction. A supported target is chosen with margin for the worst case, a measured average is what one playthrough produced, and the two being different numbers is what a target and a measurement are supposed to look like.

The finding that will matter more day to day is not the frame rate at all. The stock Deck preset is not the old preset with better textures, and three defaults changed. FSR upscaling is now on out of the box where it was previously opt in, ambient occlusion is set to GTAO on High where the old preset had effectively none, and screen space shadows are on. The GTAO change is a large part of why the frame rate fell, since it is one of the dearer post-process effects in a deferred renderer and the lighting rework it sits on top of is the point of the whole update. But the FSR default is the one that alters the image without anyone choosing to, and a player who launches expecting the look they had and finds it softer, with the aliasing that the tester described as one of the old Deck preset's worst traits, has met a default rather than done anything wrong. Read that as a deliberate trade for the frame rate and it is defensible. Meet it as a bug report against your own settings and it is a bad afternoon.

Then there is the crash, which is two words in the source and the most actionable thing in the whole test. Changing settings to High crashed the game once or twice, and a reboot brought it back with those settings applied. Nobody should be told this is fixed, because the build was pre-release and the day-one patch may or may not touch it. What can be said is what it does to behaviour. A player on an eight gigabyte handheld who is deciding whether they can afford higher settings now has to weigh a settings change against a possible reboot, and that hesitation did not exist before this remaster. Change one option at a time, and if it crashes, reboot rather than assuming the setting is broken or stuck.

Load times are the pleasant surprise, with one exception. Times between the two versions were about the same, which is a real achievement for an update that changed the renderer and the asset pipeline, and it fits the DirectX 12 work where the shader cache is prebuilt at first boot instead of compiled at load. The exception is the one that will generate complaints: loading a recent save after dying took far longer than it should, while first-time loads were acceptable. That is a narrow, specific fault pointing at save loading rather than the asset pipeline, and it happens exactly when a player is least patient, in the middle of a difficult fight. If it survives into the launch build it will be reported as the game being slow by people who never isolated it. The other change worth crediting has nothing to do with performance, since the UI text is larger and quest details and the in-game time are noticeably easier to read on a seven inch screen, which on a handheld may be worth more than the graphics.

The honest limit on all of this is that it is one pre-release build, one device, one playthrough, no frame capture tooling and no settings sweep, and day-one patch 1.01 exists. Nothing here is a benchmark and it should not be quoted as one in either direction. What survives that caveat is the shape rather than the decimals. The Deck still runs this comfortably. The frame rate is lower than it was, by about a quarter. The recovery route is a 30 fps lock with settings pushed up, at the cost described above. The preset has changed under anyone who does not read the changelog. And the publisher's claim that it runs like the original survives contact with the measurement, which is the single most useful thing a pre-launch test can tell you.

Source: A Steam Deck HQ pre-release hands-on with the Remastered build is the source for every measured and observed figure in this article: the drop from 50 to 60 fps to 42 to 44 fps, the average of around 44, the single dip to 39 fps, the consistency of the drop throughout the game, the 40 fps lock recommendation for the stock preset, the option to push some settings to High at a 30 fps lock, the one or two crashes when changing settings to High and their resolution by reboot, the unchanged load times, the slower post-death recent save loading, the larger UI text, the Deck preset defaults of FSR on, GTAO on High and screen space shadows on with ray tracing off and most other settings on Low, the observation that the old Deck preset's lack of an upscaler made the game too sharp at distance, and the $24.99 sale price noted by that source. This site's earlier coverage of the September 7 announcement is the source for the published Steam Deck tier of 800p at 30 fps on FSR Performance with no ray tracing, CD Projekt Red's assessment that it runs like the original, the DirectX 12 prebuilt shader cache, and the preload range. The 3:3 frame pacing argument for 40 fps on a 120 Hz panel is this article's, and the same argument was made in this site's September 11 Switch 2 coverage, while the distinction between 120 Hz and 60 Hz panels is noted because the LCD is 60 Hz. The comparison of published target against measured average, and the assessment that the target is a conservative floor, is this article's reading of the two sets of figures. No retailer link is reproduced, no frame rate is estimated, and no claim is made that the settings crash is or is not fixed in the launch build

Related: Witcher 3 Remastered Sept 29: Free Upgrade, 30 Changes [2026] | Witcher 3 Remastered Switch 2 Locks Sept 29: 42GB, 40FPS, Free Upgrade [2026] | Witcher 3 Remastered: Free Upgrade, Sept 29 Launch [2026] | The Witcher 3 Remastered: Release Date, Features, & Upgrade Details | Intel Driver 101.8531: 35% Witcher 3 FPS Boost on Core Ultra 3