Asus ROG Strix Scar 18 G834JYR (2024) review

Frame rates

1920×1080, High preset, upscaling off.

Measured on this machine. Source: Notebookcheck, 2024-02-08.

In use

The short version

The Asus ROG Strix Scar 18 G834JYR is the 2024 revision of Asus's largest gaming laptop, and on paper it is close to the complete article for this generation. Our record lists an RTX 4090 at a 175W power limit — the top of the band this graphics card is sold in — a Core i9-14900HX, 32GB of memory, an 18in 2560x1600 Mini LED panel at 240Hz, a 90Wh battery and a 3.1kg chassis, at a UK price of £4,000. There is no obvious component in that list being held back by another.

Notebookcheck's measured frame rates, taken at 1920x1080 on the High preset with upscaling switched off, corroborate it. The Witcher 3 at 343fps and Strange Brigade at 341fps head the table, The Finals runs at 214fps, and the slowest result in eleven games is Starfield at 84fps — a title notorious for punishing every machine it ran on. Most importantly for a machine at this price, our record holds a Cyberpunk 2077 result of 118fps, and that is the single most useful number here because Cyberpunk 2077 is the closest thing our scoring suite has to a pure test of graphics throughput.

The comparison data is where a buyer's expectations need adjusting. Across three shared games the Scar 18 averages one frame per second ahead of a Lenovo Legion Pro 5 carrying an RTX 4070 at 140W. Across four shared games it averages one frame per second behind a Lenovo Legion 7 16 IRX G9, also on an RTX 4070. A fully configured RTX 4090 tying two RTX 4070 machines is a result that demands an explanation rather than a shrug — and the explanation is not that the hardware is equivalent.

The same comparison table contains the proof. Against a Schenker XMG Apex 15 with an RTX 4050 at 140W, the Scar 18 is 94fps ahead across six shared games. Against an Asus ROG Strix Scar 17, also on an RTX 4090, it is 35fps behind across three. A method that opens a 94fps gap in one direction and concedes 35fps in another is plainly capable of separating machines. When it returns a one-frame difference, that near-tie carries specific information: at 1920x1080, in the small shared sets those two comparisons rest on, the games were waiting on the processor rather than the graphics card, and processor-limited games compress genuine graphics differences to nothing.

There is no score on this page. Our scoring suite is five games and we hold four: GTA V at 181fps, Dota 2 at 174fps, X-Plane 11 at 137fps and Cyberpunk 2077 at 118fps, averaging 152fps. Baldur's Gate 3 is missing. Every scored laptop on the site is averaged over all five, and ranking a four-game average against those would quietly distort the comparison, so we publish the numbers and withhold the ranking.

Buy this machine if you want an 18in Mini LED display, a graphics card configured at its ceiling and a processor that will not bottleneck anything, and the laptop is going to live on a desk. Look elsewhere if you expect £4,000 to buy the fastest RTX 4090 laptop in our records — the Scar 17 comparison says otherwise — or if the machine has to be carried.

Performance

175W, and why it is the first number to read

Nvidia's laptop graphics cards are configurable products rather than fixed ones. The same silicon is sold to manufacturers within a defined power band, and each manufacturer chooses a point in that band according to what its chassis can cool, how much fan noise the target buyer will accept, and where the machine sits in its own range. The consequence governs everything below: two machines carrying the same RTX 4090 badge can deliver measurably different performance because one is permitted to draw more power than the other, and the difference between a restrained configuration and a generous one is often larger than the difference between adjacent model numbers.

Our record puts this machine's RTX 4090 at 175W, which is the top of the band for this part. That is the specification a buyer at this price should want to see, and it is consistent with the physical machine: an 18in chassis at 3.1kg has the internal volume — fan diameter, heat pipe cross-section, fin stack area, exhaust aperture, thermal mass — to sustain a graphics card at that draw alongside a Core i9-14900HX, which brings substantial thermal demands of its own. There is nothing in this specification that looks strained or internally inconsistent.

Where the power limit shows up is at the demanding end of the measured set. Light titles are decided by other things: The Witcher 3 at 343fps and Strange Brigade at 341fps would be fast on a considerably lesser machine, and those results are health checks confirming nothing is wrong rather than demonstrations of what 175W buys. The power expresses itself in Cyberpunk 2077 at 118fps and Starfield at 84fps, which are the two titles in this table that will consume every watt made available to them.

The measured spread, top to bottom

Eleven games are recorded, all at 1920x1080 on High with upscaling disabled, running from 343fps down to 84fps. That is roughly a four-to-one spread, and it tells you immediately that the test set is not homogeneous — it contains titles that barely trouble a modern graphics card and titles that will take everything available.

The Witcher 3 at 343fps and Strange Brigade at 341fps lead the table. Strange Brigade's Vulkan renderer has been a benchmark fixture for years precisely because it scales predictably and asks little of the processor; a figure at that level confirms the graphics card is being properly fed. The Witcher 3 is a well-optimised older engine at settings that stopped being challenging some time ago. Both are useful as verification and useless as differentiators.

The Finals at 214fps is the most interesting result in the upper band. It is a modern competitive shooter with destructible environments and a genuinely demanding renderer, and 214fps at 1080p High is a strong showing — the kind of number that a competitive player choosing this machine would actually use.

The upper-middle band is GTA V at 181fps, Dota 2 at 174fps, Final Fantasy XV at 166fps, Far Cry 5 at 156fps and Palworld at 151fps. Two of those need flagging immediately because they are scoring-suite titles. GTA V at 181fps is an old engine that leans heavily on single-thread processor performance and ceased to be graphics-limited on hardware of this class years ago; that number is a statement about the Core i9-14900HX. Dota 2 at 174fps is the same story — nothing about Dota 2's visual complexity troubles an RTX 4090 at 175W, and its ceiling is set elsewhere entirely. Palworld at 151fps sits in a similar category: an early-access title whose performance characteristics owed more to its simulation than to its rendering.

X-Plane 11 at 137fps is the most extreme processor-bound case in the set. Its simulation loop is famously heavy on the processor and famously indifferent to graphics hardware; it is in the scoring suite because flight simulation is a genuine category of software people buy machines like this for, not because it separates graphics cards.

The bottom of the table is where the RTX 4090 is actually working. Cyberpunk 2077 at 118fps is the single most valuable measurement in this machine's record and it is treated separately below. Starfield at 84fps is the floor, and its position needs context: that release was heavily processor-dependent and notoriously indifferent to graphics hardware in its own right, so 84fps there is not straightforwardly a statement about the graphics card either. It is the worst case in the set and it is still comfortably playable.

The Cyberpunk 2077 result is the one that matters

Of the eleven measured titles, Cyberpunk 2077 at 118fps is the only one that provides a clean read on graphics performance, and its presence in this machine's record is a significant advantage over the many machines of this generation that lack it.

Cyberpunk 2077 will consume whatever graphics capability is placed in front of it. It responds to power limits, cooling and memory bandwidth in a predictable way, and it does not run into a processor ceiling on hardware of this class at 1080p in the way that GTA V, Dota 2 and X-Plane 11 all do. 118fps at 1920x1080 on High with upscaling switched off is a genuinely high-end result, and it is the strongest available evidence that the 175W power limit in our record is real and is being sustained rather than being a peak figure the cooling cannot hold.

It also has a practical dimension. Achieving 118fps without upscaling means DLSS remains available as reserve capacity rather than as a requirement, and that reserve is what determines how comfortably a machine of this generation ages. A laptop that needs upscaling to reach a playable frame rate at 1080p today has nowhere to go; one that produces 118fps natively has substantial room to raise resolution, raise settings, or absorb heavier releases.

Four suite games, and the one that is missing

Our scoring suite exists so that machines can be compared on identical software rather than on whichever titles were current when each was reviewed. For this machine we hold four of five: GTA V at 181fps, Dota 2 at 174fps, X-Plane 11 at 137fps and Cyberpunk 2077 at 118fps, a mean of 152fps. Baldur's Gate 3 is absent.

Four of five is a better position than most machines of this generation manage, and the missing title is not the most damaging one it could be. Baldur's Gate 3 is a mixed test — genuinely demanding of the graphics card, but with a well-documented dependence on processor performance in its later act — so its absence removes a data point rather than removing the suite's only graphics test. Cyberpunk 2077 is present, and that is the one that would have been most costly to lose.

Still, look at the composition of the four we hold: three of them are substantially processor-limited at 1080p on hardware of this class, and one is not. The mean of 152fps is therefore a weighted blend in which the graphics card gets a single clear vote out of four. That is not a criticism of the suite, which has to represent what people actually play rather than consisting solely of graphical showcases, but it is essential context for the comparisons below — because a shared-game average built on those same titles will report ties between machines with very different graphics hardware.

The procedural consequence is that we publish no score. A four-game average ranked against machines averaged over five introduces a systematic bias whose direction depends on which title is absent. We show the four figures, show their mean, and decline to convert that into a position on a leaderboard it does not belong on.

What this means when you play something

Practically, this is at the top of what laptops of this generation do. Every measured title runs at 84fps or better. Nine of the eleven run at 137fps or better. Five run at 166fps or better. If your library is competitive multiplayer, older single-player work, strategy or simulation, this machine produces frame rates well beyond what most players can distinguish and has capacity left over for raising settings or resolution.

If your library is the heaviest end of modern releases, Cyberpunk 2077 at 118fps and Starfield at 84fps are the numbers that describe your experience, and both are comfortable. Neither required upscaling to get there. For a machine measured in February 2024 and expected to remain viable through several years of heavier releases, that native headroom is the most important thing on this page after the power limit.

Our record contains no measurements at the panel's native 2560x1600. That resolution carries roughly two-thirds more pixels than the 1920x1080 the benchmarks use, and almost all of the extra load falls on the graphics card rather than the processor. In GTA V, Dota 2 and X-Plane 11 the drop would be modest, because the graphics card had spare capacity to begin with. In Cyberpunk 2077 and Starfield it would be substantial. Those are structural expectations rather than measurements, and this review will not attach numbers to them.

The screen

18in at 2560x1600, and a Mini LED backlight

The panel is an 18in 2560x1600 Mini LED running at 240Hz, and the backlight technology is the most consequential difference between this machine and the IPS-panelled alternatives at similar prices.

A Mini LED backlight divides the illumination behind the panel into a large number of independently controlled zones rather than lighting the whole surface uniformly. The practical effect is a dramatic improvement in contrast and black level over a conventional IPS backlight — dark scenes look genuinely dark rather than dark grey — and typically much higher peak brightness in small highlights, which is what makes HDR content meaningful rather than nominal. For dark games, for film, and for HDR material generally, that is a real and immediately visible upgrade.

The trade-off is equally real and worth stating plainly. Zoned backlights produce blooming: a bright object on a dark background illuminates its whole zone, creating a visible halo. How objectionable that is depends entirely on how many zones the panel has and how well the local dimming algorithm is tuned, and our record holds no zone count, no brightness measurement, no contrast figure and no response time for this display. That is a significant gap for a panel whose entire selling point is a characteristic that varies enormously between implementations. A good Mini LED panel at this price is a genuine luxury; a poorly implemented one can be more distracting than the IPS panel it replaced.

What can be said is that Mini LED at 240Hz in an 18in gaming laptop was, in this generation, a specification found only at the top of the market, and its presence here is consistent with a £4,000 machine rather than an anomaly.

18in as a format

The 18in diagonal is the reason to choose this machine over a 16in one with comparable internals. It is a materially different surface — closer to working at a small desktop monitor than at a laptop — and for strategy games, flight and vehicle simulators, and anything with dense interface elements, the difference is functional rather than cosmetic. Given that this machine's record includes X-Plane 11 at 137fps, the pairing of an 18in display with simulation performance of that order is a coherent proposition.

2560x1600 across 18in gives a lower pixel density than the same resolution on a smaller panel. That is a trade rather than a fault: interface elements and text are physically larger and more comfortable at typical viewing distances, at some cost in fineness. The 16:10 aspect ratio provides more vertical space than 16:9, which shows up in every application with a toolbar.

240Hz, and what actually reaches it

Working through the measured set at 1920x1080: The Witcher 3 at 343fps and Strange Brigade at 341fps clear 240fps comfortably. Nothing else does. The Finals at 214fps comes closest and falls short. GTA V at 181fps, Dota 2 at 174fps, Final Fantasy XV at 166fps, Far Cry 5 at 156fps and Palworld at 151fps sit in the band where the panel is displaying everything the machine produces without capping it. X-Plane 11 at 137fps, Cyberpunk 2077 at 118fps and Starfield at 84fps are well below.

At the panel's native resolution those figures would all fall, and the two titles that clear 240fps at 1080p would probably stop doing so. The realistic reading is that 240Hz is reachable in light titles at reduced resolution and aspirational in everything else — the normal position for a gaming laptop of this generation rather than a fault specific to this machine.

The value of the refresh rate is not that games hit the ceiling. It is that the display never becomes the constraint: whatever frame rate the machine produces is shown without tearing or judder, motion clarity improves progressively across the whole range rather than only at the top, and a competitive player who lowers settings deliberately has somewhere for the extra frames to go. The Finals at 214fps is precisely the case where that matters, and it is the sort of title people buy a 240Hz panel for.

Build, size and portability

3.1kg, and what it is for

At 3.1kg, plus the substantial charger required to feed a 175W graphics card and a Core i9-14900HX simultaneously, this is a desktop replacement rather than a portable computer. It is designed to live on a desk and be moved occasionally, and buyers should evaluate it on that basis. If the machine needs to be carried regularly, this is the wrong category of product, not a heavy example of the right one.

The size is not a cost paid for nothing, though — it is the precondition for everything in the performance section. Cooling capacity scales with internal volume, and an 18in chassis at 3.1kg has more fan diameter, heat pipe cross-section, fin surface area and exhaust aperture than anything smaller can offer. Running an RTX 4090 at the top of its power band alongside a top-of-stack HX processor requires all of it. The 118fps Cyberpunk 2077 result is what that cooling volume converts into, and a smaller machine carrying the same components would very probably run the graphics card at a lower budget.

This is the version of the argument that works in this machine's favour. It has accepted every portability penalty available and, unlike some large machines, it is extracting the corresponding performance benefit: the power limit in our record is at the ceiling for this card, not below it.

The 90Wh battery

Our record lists a 90Wh cell. In a 3.1kg 18in chassis that is a moderate allocation rather than a generous one — smaller machines routinely carry more — and it reinforces the reading that this laptop is intended to be plugged in. Our record holds no battery runtime measurement and this review will not guess at one.

Structurally, no laptop of this generation runs an RTX 4090 at 175W from a battery; the draw exceeds what any cell can supply, so unplugged gaming means substantially reduced performance regardless of capacity. On a machine of this type that is barely a criticism, because nobody buys an 18in desktop replacement to game away from a socket. The 90Wh cell exists to cover a move between rooms and some light work, and it is adequate for that.

32GB of memory

The configuration in our record carries 32GB, which is correct for a machine at this price. It is enough for a demanding game running alongside streaming software, a browser and communication tools, and it removes the constraint that 16GB machines of this generation have increasingly run into. At £4,000 anything less would have been a legitimate complaint; as specified, memory is simply not a consideration.

How it compares

Four machines from the same graphics generation are set against the Scar 18 G834JYR, chosen to give a spread rather than only the nearest neighbours: the two closest results, the machine it beats by the widest margin, and the machine that beats it by the widest margin. Each comparison uses only the games both machines hold, preferring scoring-suite titles. None of the four has a confirmed UK price in our record, so nothing here is a value comparison in cash terms.

The sample sizes vary and that matters. The Schenker comparison covers six shared games, the Legion 7 comparison four, and the other two three each. A three-game sample drawn from processor-weighted titles is the weakest evidence in the set, and this review treats it accordingly.

Schenker XMG Apex 15 (Late 23) — the widest win

The XMG Apex 15 carries an RTX 4050 at a 140W power limit, and across six shared games the Scar 18 averages 94fps ahead of it. Six games is the largest sample in this set and 94fps is an emphatic margin.

This is the result that makes everything else in the section interpretable. A comparison method that opens a gap this large when the hardware genuinely differs is demonstrably capable of resolving differences — which means the near-ties below are findings rather than measurement failure.

It is also worth noting what the Apex 15 represents. An RTX 4050 at 140W is a generously configured entry-tier card — a machine making the most of what it has. Even so, the gap is 94fps. That is a useful boundary on the argument, made correctly elsewhere, that power limits matter more than model numbers: they matter enormously within a tier, and they do not close a three-tier gap. A well-fed 4050 is still a 4050.

Lenovo Legion Pro 5 — level, against an RTX 4070

The Legion Pro 5 carries an RTX 4070 at a 140W power limit, and across three shared games the Scar 18 averages one frame per second ahead. On the face of it, a 175W RTX 4090 failing to separate from a 140W RTX 4070 is startling.

It should not be taken at face value, and the reason is the sample. Three shared games is a very small basis for any conclusion, and the identity of those games determines everything. Of this machine's four suite results, three — GTA V, Dota 2 and X-Plane 11 — are substantially processor-limited at 1080p on hardware of this class. In a processor-limited game, both machines are waiting on their processors, and the graphics card's capability is not the variable being tested. Two machines with fast processors will report the same frame rates there whether one carries a 4070 and the other a 4090.

The correct conclusion is narrow and useful: in the three games this comparison rests on, the difference between an RTX 4090 at 175W and an RTX 4070 at 140W is invisible, and a buyer whose library resembles those three games will not perceive it. In Cyberpunk 2077 — where this machine produces 118fps and the graphics card is genuinely the constraint — the gap would be real and substantial. Our record does not hold that comparison, so this review cannot quantify it, but the tie should emphatically not be read as evidence that the hardware is equivalent.

Lenovo Legion 7 16 IRX G9 — level, across four games

The Legion 7 16 IRX G9 carries an RTX 4070 and across four shared games the Scar 18 averages one frame per second behind. Our record holds no power limit for the Lenovo's graphics card.

Four games is a slightly better sample than three and the same reasoning applies. A four-game shared set drawn from a suite in which three of the four titles are processor-bound will report a tie between machines whose graphics hardware differs by two full tiers, and that is very likely what has happened here. The Legion 7 is a machine built around a large chassis and a serious processor, so in exactly those titles it has everything it needs to match anything.

There is an important asymmetry worth spelling out. A tie in processor-limited games tells you nothing about the graphics cards; it does tell you something about the processors, namely that they are comparable. For a buyer whose actual use is strategy games, simulators and older titles, that is genuinely relevant information — it means the £4,000 machine and the cheaper one will feel the same in those games. For a buyer playing Cyberpunk-class titles, it is irrelevant, and the 118fps figure is the number to reason from instead.

Asus ROG Strix Scar 17 — the widest loss, and it is in-house

The Scar 17 carries an RTX 4090 and across three shared games the Scar 18 averages 35fps behind it. Our record holds no power limit for that machine's graphics card.

This is the result a £4,000 buyer should sit with, for two reasons. The first is what it is: a machine from the same manufacturer, in the same product line, one size smaller, carrying the same graphics card, and ahead across the shared games. The second is the conditions under which it appears. Thirty-five frames per second is a substantial gap, and it emerges from a three-game sample weighted towards the processor-limited titles where differences are hardest to detect. A gap that survives those conditions is likely to be larger, not smaller, in genuinely graphics-limited software.

What the Scar 17 is doing differently, our record does not say — it holds no power limit for that machine. The plausible explanations are a configuration difference we have not captured, or a cooling design that sustains its limits better in the measured workload. Either way, the practical conclusion is that on our figures this is not the fastest machine in its own family, and a buyer whose priority is frame rate should look at the smaller model before committing.

What the spread proves

In order: 94fps ahead of a 4050 across six games, level with a 4070 across three, level with a 4070 across four, 35fps behind another 4090 across three. The pattern is coherent once sample composition is taken into account, and it makes two points that a buyer should carry away.

The first is that this machine's hardware is not in doubt. The 175W power limit, the 14900HX and the 118fps Cyberpunk 2077 result all point the same way, and the 94fps margin over the entry tier confirms the measurement can see what it needs to see. The ties against RTX 4070 machines are artefacts of small, processor-weighted shared sets — a limitation of what our record holds in common, not a finding about the silicon.

The second is that the machine's position within the top tier is less commanding than the specification suggests. The Scar 17 comparison is the only head-to-head against equivalent graphics hardware in this set, and this machine loses it. That deserves more weight than the two ties, because it is the only comparison in the group where both machines carry the same class of graphics card and a difference nonetheless appeared.

Who it's for

Buy it if

You want an 18in Mini LED display. This is the strongest and most specific argument for the machine. An 18in 2560x1600 panel at 240Hz with a zoned backlight is a combination that existed only at the top of this generation's market, and the contrast advantage over IPS is immediately visible in dark games and HDR material. No 16in machine offers the size; most machines at any size offer only IPS.

You want a graphics card configured at its ceiling. 175W is the top of the band for an RTX 4090, and the Cyberpunk 2077 result at 118fps is the evidence that it is being sustained. A buyer at this price should not accept a restrained configuration, and on our record this machine does not have one.

The machine will live on a desk. 3.1kg and an 18in footprint is a desktop replacement, and if that matches how it will be used, the size is a benefit — it is what makes the cooling, and therefore the power limit, possible.

You want the heavy titles to be comfortable without upscaling. Cyberpunk 2077 at 118fps and Starfield at 84fps were both measured with upscaling switched off. That means DLSS is available as reserve rather than as a requirement, which is what determines whether a machine of this generation stays comfortable as software gets heavier.

You want a processor that will not become the bottleneck. GTA V at 181fps, Dota 2 at 174fps and X-Plane 11 at 137fps are strong results in titles that are decided by processor throughput, and they are the work of the Core i9-14900HX. Simulator and strategy players benefit from that directly.

Buy something else if

You want the fastest RTX 4090 laptop in our records. On our figures this is not it — the Scar 17 is 35fps ahead across three shared games, in conditions that understate rather than exaggerate graphics differences. If maximum frame rate is the objective, that is the machine to investigate first.

The machine has to be carried. 3.1kg plus a high-wattage charger is not a portable computer, and there is no configuration of this laptop that makes it one. Buyers who need portability should be shopping in a different category rather than looking for a lighter version of this one.

Your library is dominated by processor-limited titles. This is the uncomfortable but honest conclusion of the comparison section. If what you actually play is strategy games, simulators, older single-player titles and Dota-class competitive games, our record shows this machine level with RTX 4070 laptops in exactly those titles. You would be paying £4,000 for graphics capability that your software will not ask for.

You are uncertain about Mini LED. Our record holds no zone count, brightness, contrast or response time measurement for this display, and the quality of a zoned backlight varies enormously between implementations. Blooming around bright objects on dark backgrounds is inherent to the technology and its severity depends on the specific panel. At this price it is worth seeing one before committing rather than buying on the specification.

You want a published score to compare against the rest of the site. There is not one, because Baldur's Gate 3 is missing from our figures. Four of the five suite games are present and their mean is 152fps, but that number cannot be honestly ranked against machines averaged over all five.

What our record does and does not settle

This machine's specification is internally coherent in a way that not every laptop at this price manages. A 175W RTX 4090 alongside a Core i9-14900HX in an 18in 3.1kg chassis is a design where each element supports the others, the Mini LED panel is a genuine step above the IPS displays that dominate this class, and the Cyberpunk 2077 figure confirms the graphics card is doing the work the power limit implies. There is nothing here that looks like a specification error or a marketing number.

What our record does not settle is why the Scar 17 is ahead of it. That comparison holds no power limit for the rival, covers only three shared games, and yet produces a 35fps margin in favour of the smaller machine. Whether that reflects a configuration difference, a sustained-load difference, or something in the composition of those three titles, we cannot say from what we hold. It is the one loose thread in an otherwise clean record, and at £4,000 it is worth pulling on before buying.