Verdict
The RTX 5090 produces some absolutely striking operation figures . In our examination , it sack up an unbelievable 90 fps on median at 4 K across all of our five form of address , and it does that without DLSS or Frame Generation thrown into the mixture . likewise , AI performance goes above and beyond , and it ’s one of the few cards that have see a substantial ironware uptick equate to last gen .
The problem , if you had n’t already guessed , is stock , and although the RTX 5090 was initially debut at $ 2,000 you’re able to now only get them a lot high than that .
Pros
Cons
Key Features
Introduction
This , right now , is the enceinte art lineup Nvidia has ever released . At least from the point in time of view of pure public presentation . It has utterly no adequate , and even its herald , the RTX 4090 , pales in compare to this goliath .
The spec relegate alone from a gen - on - gen standpoint is just astronomical ; CUDA cores alone have shot up by 32.8 % .
equate to the “ improvement ” with the RTX 5080 compared to the4080 Super , well , it ’s an increment of a factor of six , and you’re able to separate , because this thing just smash almost every psychometric test you bewilder at it .
Image Credit (Trusted Reviews)
But there ’s a taking into custody , because of course there is . This is the RTX 50 serial after all , and even ignoring some of the ROP issues plaguing certain manufacturing business , the reality is that strain is n’t available . Even at launch , it was incredibly limited , and although the MSRP was initially pitched at $ 2000 , some resellers now have these listed at about 85 % high than that .
It ’s not exactly safe time value per build at its retail price , but at that patsy - up , it make for an abyssal purchase . So yes , it is by far thebest graphic cardyou can get today , at least in purport ; it ’s just a shame you ca n’t in reality get it .
Specs
The hardware here is utterly wild . Unlike with theRTX 5080and the lower - specced cards , the 5090 has realize an unbelievable increment in the overall number of cores uncommitted to it . This is in no small part thanks to a massively supersized GPU dice leveraging 92.2 billion transistors with an area of 750 mm² . In contrast , the RTX 4090 , running off a similar 5 nm process ( albeit slightly less optimised ) , come in at 608.5 mm² and 76.3 billion transistors . That is a wild increment .
Top - line spectacles for the 5090 include 21,760 CUDA centre , 680 TMUs , 176 ROPs , 170 ray tracing cores , and a banging 680 Tensor cores .
Across the circuit board , that ’s an increase of 32.8 % compare to every core count on the RTX 4090 . No small material body for certain . On top of that , Nvidia has paired thatBlackwell - power central processing unit with 32 GB of Samsung ’s tardy GDDR7 retentivity on a 512 - bit bus , have an astonishing 1,792 GB / s of total memory bandwidth .
That ’s an extra 8 GB on top of last gen for capacity and 77.8 % more speed as a solution of the GDDR ascent .
There are a few more gain here deserving mentioning , as it ’s also apply a full - fatty PCIe 5.0 x16 connector plus a 575W TDP versus the 4090 ’s 450W cap . It ’s still using the 12VHPWR connector , although it ’s a polished interlingual rendition designed to give it a lilliputian bit more robustness and hopefully keep any unwanted cable burnouts ( do n’t worry , your old 12VHPWR cable are still compatible ) .
One affair worth noting , however , is that the TDP is sitting pretty at 575W. Nvidia has directly recommended a minimum of a 1000W PSU to run its RTX 5090 .
During my time examination , I find out pattern extend as eminent as 865W and even higher under some load conditions . I highly recommend you go slightly above that 1000W bull’s eye and aim for a 1200W whole or else , strictly just for the sake of headroom and comfortableness .
Test Setup
When it hail to really fight the RTX 5090 to its limits , I desire to vamp Trusted Reviews ’ computer graphic card benchmarks to showcase just what this card and its surrounding competitors can really deliver in modern titles and background .
I ’ve been testing , benchmarking , and playing around with these GPUs for the last few month , really getting to clutches with them and understand their architecture in all manner of scenarios , so I had a sound idea of the game that would really push these things to their demarcation
To that final stage , I ’m using a combining of both synthetic and in - biz benchmark for the majority of the testing . Five game total across three separate resolutions , with minimum and average frame rate report .
Image Credit (Trusted Reviews)
Each plot at each solution is lean three meter , to average out results , and I ’m test using the highest possible graphical preset available , with minimal changes , bar a few outliers . Cyberpunk 2077 , for example , I ’ve disable DLSS for its testing , and in Black Myth Wukong I ’m only using the “ Ultra ” graphical preset not the “ Extreme ” one , with beam trace additionally disabled . As for Total War : Warhammer 3 , I ’m using the Mirrors of Madness bench mark to stress the CPU load at the same prison term .
On top of that , I ’ve also run some additional testing for both DLSS and Frame Generation tech at 4 K as well . This is use Cyberpunk 2077 and Black Myth Wukong again ( with electron beam tracing enabled this sentence ) . In those scenario , I ’m testing using DLSS set to “ Quality ” ( 67 % in Black Myth ) , and then a separate set of run with Frame Generation enabled as well . To keep it a fair playing field , and so we can liken against extra erstwhile notice , I ’m only tolerate Cyberpunk to use shape - gen at a x2 ratio , not the maximum x4 .
celluloid testing , stem mostly around UL ’s 3D Mark and Procyon bench mark suites . For 3D Mark that admit Steel Nomad and Speedway . Procyon is all about AI computational carrying out for us , and for that , Computer Vision , and Image Generation are the name of the secret plan . I ’ve examine all the cards , with either the default APIs , or with Nvidia ’s TensorRT .
As for the trial bottom , I ’ve resolve to go with AMD this meter around , specifically with the Ryzen 9 9900X.
Although potentially an X3D chip may have delivered higher frame - rates , the fact its 3D V - Cache tend to vary in carrying into action degree dependent on the form of address that ’s run it , and whether or not that can take vantage of it , I ’ve opt to stay with the stock X variation instead , prefer to lean on greater multi - core performance and higher clockspeeds as a issue .
Both Intel and AMD ’s latest C.P.U. are beastly when it come to the competitiveness to becomethe right gambling CPU , so you should n’t see too much of a dispute going from one chip to the other ( in fact my own personal testing can back that up as well ) .
Gaming Performance
Nvidia ’s RTX 5090 is one of those nontextual matter posting that just hold up all expectation . Yes , the Mary Leontyne Price point is high-pitched , the ability thread transcendent , and its girth ( certainly with the Zotac Solid variant I have for testing ) abysmal . And yet , despite those minor foibles ( andstock availableness , but we ’ll get onto that afterward ) , it just perfectly eradicate the competition at the gamy - ending .
There ’s no line ; from a pure new performance viewpoint , this is the greatest graphics placard of all time , at least by 2025 ’s standards .
Let ’s take those 4 K numbers as an example . Average skeleton pace across all five title at 4 K was 89.13 fps ; the lower limit was 57.98 .
Image Credit (Trusted Reviews)
Yeah , its minimal frame rate was intimately hitting that fabled 60 frames we all know and love as received at 4K. And the affair is , our examination suite is n’t kind by design . Both Cyberpunk and Black Myth Wukong are massive frame - rate drain liken to most games , and none of the games on test are employ DLSS to mother those numbers .
Cyberpunk 2077 , for instance , land at 54.9 Federal Protective Service , and Black Myth at 60.3 . Jump into Metro Exodus or Total War : Warhammer 3 , and you ’re talking 93 Federal Protective Service and 80 fps , respectively . Again , it ’s aim near enough 8.3 million pixels that rapidly . It ’s just astounding .
At 2560×1440 , average skeletal system rates ride well at 127.7 fps , and 1080p sees a less functioning bump , with modal frames across all title landing at 140.77 fps as well . Those are solid numbers . Yes , that 1080p number is seemingly quite humiliated , but I mistrust that ’s more probable down to the CPU becoming the bottleneck for the legal age of those game .
Gaming Benchmarks
Ray Tracing, DLSS, & AI Performance
Of course , in the Din Land of AI upscaling and frame coevals , DLSS 4really beam , and its new transformational model takes a tidy sum of the fogginess and latency out of the par that its old convolutional mannequin used to put in . I certainly do n’t tolerate with it any longer . Although I will say I am still a big advocate of traditional rasterisation and MSAA , even still .
nevertheless , with DLSS on quality , Cyberpunk 2077 at 4 K saw underframe rates increase by 26 Federal Protective Service . Add in frame generation to the mixture , on an x2 constituent , and that DLSS score closely doubles , with the RTX 5090 shore an outstanding 146.1 fps .
Black Myth Wukong might depend like DLSS is having no effect , but you have to remember that in its standard examination scenario , it does n’t have shaft trace enable . Whereas in our DLSS tests , it does . With TSR at 100 % and DLSS at 67 % , frame rate remain within the border of misplay at 4K.
That in of itself is telling , as ray tracing well stymie frame charge per unit when enabled here . Frame - gen also bumps the figures up too , vomit up another 39.7 Federal Protective Service on top for the 5090 .
As for our AI test , UL Procyon ’s Computer Vision , scored a neat 4,879 point with Nvidia TensorRT . That ’s not really as high-pitched as I ’d expect . The RTX 5080 , for quotation , landed a grievance of 4272 , and the RTX 4080 Super , 3728 .
Image generation , mind you , was a night - and - day difference . With Tensor , the 5090 scored 8333 points , finish the run in just 12 seconds and develop an AI - generated image in 0.75 seconds per image . That just dwarfs the 5080 and the 4080 Super by a Brobdingnagian margin .
Image Credit (Trusted Reviews)
Power Consumption & Temperature
Throughout my testing , I employ a combination of superpower meters and HWMonitor , plus a room thermometer to give me a good agreement of how these lineup run .
My test layer includes 11 120 mm fans in a positive air flow setup , with the GPU being fed directly by three inlet to keep things chilled . My RTX 5090 is Zotac ’s Solid AIB fashion model , which is a severely chunky triplex - slot design complete with a heavyset backplate and its own treble devotee cooling result as well .
temp were very upstanding , maxing out at 76 C under load . I never see it top much higher than that either , even during seriously farsighted gambling sessions . It does , however , tend to superheat your office a tad , so I ’d extremely advocate either a secure window , air conditioning , or just be unforced and ready for it to be toasty , even under unfounded temps . That ’s a sight of GPU die to cool after all .
Power lot capped at 865.3W under maximal cargo ; ironically , it was Final Fantasy XIV that drew the eminent peak impressively . Again , if you ’re look at this card , you want to get the best PSU you could budget for , preferably at around 1200W or so .
Should you buy it?
60 Federal Protective Service at 4 K without gimmicks , without DLSS , no AI upscaling , nothing . It just dominates those frame rates .
meld with 32 GB of VRAM and some serious AI technical school backing it all up , it ’s a circuit board that ’s going to carry you forward for many years to come . The only downside is if you ca n’t find inventory of it or it ’s at that gaga $ 3700 price point .
Not only do you need the graphics tease itself , but you need to see you ’ve father one of the honorable monitors you’re able to get to really takeadvantage of it . flux that with a ample PSU and a seriously quick CPU to back that all up , and all of a sudden it ’s not quite the small $ 2000 ascent you first thought it was .
Final Thoughts
There ’s no dubiousness in my brain that the RTX 5090 is one of the best graphics cards out there on the mart today .
For pure performance , it has no adequate . In fact , the second - near GPU out there is arguably the RTX 4090 , and that ’s a two - year - old wit at this stage . It ’s just unparalleled ; that 32.8 % boost to the die size , the hardware on it , and the overall carrying into action that you get at 4 K with no upscaling is just outstanding .
But that come with some fair major caveat . No matter how you seem at it , even at its retail price , the monetary value - to - performance ratio is seriously lacking . If you ’re go into this await for a in force value card or trying to convince yourself of that , you ’re croak to be in for a bad time . Particularly to get the most out of this , you really do need a high-pitched refresh rate 4 K show to drive it on .
Image Credit (Trusted Reviews)
It is by far thebest graphic cardyou can get today , at least in feeling ; it ’s just a shame you ca n’t actually get it easy .
Trusted Score
How we test
Each and every computer graphic card that come in for review gets cast off into its own fix , dedicated test bench to be benchmarked and compared against other cards in the GPU results entourage .
There ’s a total of 67 data point we apply to analyze each GPU , cover all three cardinal resolutions , along with GPU temperature , system power draw , $ per frame , and many , many more .
We use a compounding of synthetic and in - game benchmarks to give us a good apprehension of how these circuit board perform , both in isolation and against each other , before making any judgements , while also consider veridical - reality ancestry story and pricing .
Test Data
Full Specs
Image Credit (Trusted Reviews)
Image Credit (Trusted Reviews)
Image Credit (Trusted Reviews)
