Nvidia RTX 5090 Review: Ray Tracing, DLSS 4, and Raw Power

1. Introduction2. Meet the RTX 5090 Founders Edition3. Testing Methodology4. Alan Wake II5. Black Myth: Wukong6. Cyberpunk 20777. Final Fantasy XVI8. Forza Horizon 59. Ghost of Tsushima10. Horizon Forbidden West11. The Last of Us Part 112. A Plague Tale: Requiem13. Senua's Saga: Hellblade 214. Starfield15. Total War: Warhammer III16. Performance Summary (12-Game Average)17. Cost Per Frame Analysis18. Ray Tracing: Alan Wake II19. Ray Tracing: Black Myth: Wukong20. Ray Tracing: Cyberpunk 207721. Ray Tracing: F1 2422. Ray Tracing: Ratchet and Clank: Rift Apart23. Ray Tracing: Returnal24. Ray Tracing: Shadow of the Tomb Raider25. Ray Tracing: Star Wars Outlaws26. Ray Tracing Performance Summary27. DLSS 4 Testing28. AI & Productivity Benchmarks29. Clock Speed30. GPU Temperatures31. Acoustics32. Power Draw: 1080p33. Power Draw: 1440p34. Power Draw: 2160p (4K)35. Performance per Watt: 1080p36. Performance per Watt: 1440p37. Performance per Watt: 2160p (4K)38. Total System Power Draw39. Closing Thoughts40. View All Pages

Announced back at CES 2025, Nvidia's RTX 50 series is here, built on the latest Blackwell architecture. First up for review is the new flagship, the RTX 5090, which hits the market at an eye-watering £1939/$1999. Packing in a total of 92.2 billion transistors, 21760 CUDA Cores and 32GB of GDDR7 memory, Nvidia claims that the RTX 5090 is ‘up to 2x' faster than the out-going RTX 4090, but is that really the case? And what's power consumption like given the rated 575W TGP? All that and more is covered in today's review…

Timestamps

00:00 Intro
00:49 RTX 5090 recap, pricing + specs
02:50 Test setup
04:31 Alan Wake 2
05:11 Black Myth: Wukong
05:47 Cyberpunk 2077
06:17 Final Fantasy XVI
06:42 Forza Horizon 5
07:11 Ghost of Tsushima
07:38 Horizon Forbidden West
08:05 The Last of Us Part 1
08:30 Plague Tale: Requiem
08:53 Senua’s Saga: Hellblade 2
09:22 Starfield
09:50 Total War: Warhammer III
10:37 12-game average results
11:41 Cost per frame analysis
12:47 RT Alan Wake II
13:19 RT Black Myth: Wukong
13:47 RT Cyberpunk 2077
14:14 RT F1 24
14:45 RT Ratchet & Clank
15:07 RT Returnal
15:34 RT Shadow of the Tomb Raider
16:07 RT Star Wars Outlaws
16:33 RT 8-game average
17:31 DLSS 4 – new Transformer model
19:59 DLSS 4 – Multi Frame Generation
27:09 AI & productivity benchmarks
28:33 Closer look at the Founders Edition
29:44 Thermals & acoustics (with soundest)
31:31 Power draw and efficiency
33:33 Closing thoughts

If you missed the announcement earlier in the month, the Nvidia RTX 5090 is arriving as the first of four 50 series GPUs. The RTX 5080 is the next cab off the rank, launching on January 30th, while the RTX 5070 and RTX 5070 Ti are set to arrive some time next month. This review is firmly focused on the flagship SKU, though, and there is plenty to cover. We look at raster performance, ray tracing, new upgrades made to DLSS 4, alongside AI benchmarks, power, efficiency and more.

	RTX 5090	RTX 4090	RTX 4080 Super	RTX 4080	RTX 4070 Ti Super
Process	TSMC N4	TSMC N4	TSMC N4	TSMC N4	TSMC N4
SMs	170	128	80	76	66
CUDA Cores	21760	16384	10240	9728	8448
Tensor Cores	680	512	320	304	264
RT Cores	170	128	80	76	66
Texture Units	680	512	320	304	264
ROPs	176	176	112	112	96
GPU Boost Clock	2407 MHz	2520 MHz	2550 MHz	2505 MHz	2610 MHz
Memory Data Rate	28 Gbps	21 Gbps	23 Gbps	22.4 Gbps	21 Gbps
L2 Cache	98304 KB	73728 KB	65536 KB	65536 KB	49152 KB
Total Video Memory	32GB GDDR7	24GB GDDR6X	16GB GDDR6X	16GB GDDR6X	16GB GDDR6X
Memory Interface	512-bit	384-bit	256-bit	256-bit	256-bit
Memory Bandwidth	1792 GB/Sec	1008 GB/Sec	736 GB/Sec	716.8 GB/Sec	672 GB/Sec
TGP	575W	450W	320W	320W	285W

First, a quick spec recap. The RTX 5090 is built on the new GB202 die, measuring 750mm², though it's not quite a full implementation of the silicon. Instead we find a total of 11 Graphics Processing Clusters (GPCs), each holding up to eight Texture Processing Clusters (TPCs), for a total of 85. Each TPC is home to two Streaming Multiprocessors (SMs), giving us 170, and each SM still holds 128 CUDA Cores, meaning the RTX 5090 has an eye-watering total of 21760 shaders. We also find 170 RT cores, 680 Tensor cores, 680 Texture Units, and 176 ROPs.

This time around, however, there's no node-shrink, and GB202 remains fabricated on TSMC's N4 node, as per the RTX 40-series. As such, rated clock speed is not increased this generation and is actually touted slightly below that of the RTX 4090, with the RTX 5090 delivering a rated 2407MHz boost clock, compared to its predecessor's 2520MHz boost.

The memory configuration has seen significant upgrades, though. The RTX 5090 now comes equipped with a super-wide 512-bit memory interface, paired with 32GB GDDR7 memory running at 28Gbps, and that puts total memory bandwidth at a staggering 1792 GB/s. L2 cache is also increased to 98MB, up from the 74MB of the RTX 4090.

Considering the large increases to die size and core count, but with no node shrink, it's perhaps unsurprising to see power draw has increased, this time boasting a 575W TGP. This is something we focus on closely in this review, using our enhanced GPU power testing methodology, so read on for our most detailed power and efficiency testing yet.

The new RTX 5090 Founders Edition ships in a compact eco-friendly box. Inside, a smaller box houses the quick-start guide and 4x 8-pin to 12VHPWR adapter.

The graphics card itself remains as stylish as ever – if not more so this time around. The overall design language is similar to the likes of the RTX 4090 and RTX 4080 Super, but there are a few differences. For one, while mostly black, there are some dark grey accents on the shroud, like the x-shaped bracket that sits between the two fans.

The fans are obviously now both positioned on the front side of the card, rather than one on the top and one on the bottom. Nvidia calls this the ‘double flow through' design, as the PCB has been moved to the dead centre of the card itself to allow maximum airflow to pass through the heatsink.

Image: Nvidia.

Above you can see an ‘exploded' view of the design – it's quite a feat of engineering. By shrinking the PCB to something barely bigger than my palm, Nvidia has been able to place in centrally, with the GPU and memory cooled by a 3D Vapour Chamber. Five heat-pipes extend from either side of the vapour chamber, and the two fans are able to blow air directly through the fin stack with far less resistance than the previous Founders Edition design.

Nvidia also claims that this new innovative thermal design has allowed them to shrink the card so it's now only a dual-slot thickness again, when the 4090 was a triple-slot card. It measures 304mm long and 137mm tall.

We can also get a look at the back of the card – it doesn't feature much of a backplate in the traditional sense given the flow through design, but it feels very robust in the hand and continues the industrial design.

It's fascinating to see that Nvidia has also adjusted the height of the fin stack on the back of the card – you can just about make it out from the above images, but essentially they've ‘scooped' out the central area of the fins, making it concave. It looks very cool but is also designed to improve heat dissipation.

As expected, power is delivered by a single 12VHPWR, AKA PCIe 5.0, connector, rated for up to 600W.

Video outputs are also upgraded for this generation, with the RTX 5090 offering three DisplayPort 2.1 UHBR20 connectors and one HDMI 2.1.

Driver Notes

• AMD GPUs were benchmarked with the Adrenalin 24.12.1 driver.
• Nvidia GPUs (except for RTX 5090) were benchmarked with the 566.14 driver.
• RTX 5090 was benchmarked with the 571.86 driver supplied to press.

Results are only directly comparable where this exact configuration has been used.

Test System:

We test using a custom built system powered by MSI, based on AMD's Zen 5 platform. You can view the Powered by MSI store on AWD-IT's website HERE.

CPU	AMD Ryzen 7 9800X3D
Motherboard	MSI MPG X870E Carbon WiFi
Memory	64GB (2x32GB) Kingston Fury Beast DDR5 6000MT/s CL30
Graphics Card	Varies
SSD	4TB Kingston NV3 Gen 4 PCIe NVMe
Chassis	MSI MPG Gungnir 300R Airflow
CPU Cooler	MSI MAG CoreLiquid i360
Power Supply	MSI MEG Ai1300P
Operating System	Windows 11 23H2
Monitor	MSI MPG 321URX QD-OLED
Resizable BAR	Enabled for all supported GPUs

Comparison Graphics Cards List

• AMD RX 7900 XTX 24GB
• AMD RX 7900 XT 20GB
• Sapphire RX 7900 GRE Nitro+ 16GB
• AMD RX 7800 XT 16GB
• Sapphire RX 7700 XT Pulse 12GB
• Sapphire RX 7600 XT Pulse 16GB
• Nvidia RTX 4090 FE 24GB
• Nvidia RTX 4080 Super FE 16GB
• MSI RTX 4070 Ti Super Ventus 3X 16GB
• Nvidia RTX 4070 Super FE 12GB
• Nvidia RTX 4070 FE 12GB
• Nvidia RTX 4060 Ti FE 8GB

All cards were tested at reference specifications. For factory overclocked cards, this means we manually ‘undo' the overclock via MSI Afterburner or AMD/Intel's built-in tuning tools. Or, for cards like the Sapphire RX 7900 GRE Nitro+, we enable the reference-clocked BIOS instead of the default OC BIOS.

Software and Games List

• Alan Wake II (DX12)
• Black Myth: Wukong (DX12)
• Cyberpunk 2077 (DX12)
• F1 24 (DX12)
• Final Fantasy XVI (DX12)
• Forza Horizon 5 (DX12)
• Ghost of Tsushima (DX12)
• Horizon Forbidden West (DX12)
• The Last of Us Part 1 (DX12)
• A Plague Tale: Requiem (DX12)
• Ratchet and Clank: Rift Apart (DX12)
• Returnal (DX12)
• Senua's Saga: Hellblade 2 (DX12)
• Shadow of the Tomb Raider (DX12)
• Starfield (DX12)
• Star Wars Outlaws (DX12)
• Total War: Warhammer III (DX11)

We run each benchmark/game three times, and present mean averages in our graphs. We use FrameView to measure average frame rates as well as 1% low values (99th percentile) across our three runs.

Alan Wake 2 is a 2023 survival horror game developed by Remedy Entertainment and published by Epic Games Publishing. A sequel to Alan Wake, the story follows best-selling novelist Alan Wake, who has been trapped in an alternate dimension for 13 years, as he attempts to escape by writing a horror story involving an FBI special agent named Saga Anderson. The game was released for PlayStation 5, Windows, and Xbox Series X/S on 27 October 2023. (Wikipedia).

Engine: Northlight. We test using the High preset, FSR set to native resolution, DX12 API.

#

Kicking off with Alan Wake 2, we did encounter some frame time issues in this game which affect the 1% lows. We've spoken with a couple of other reviewers who also experienced similar, so it's likely a driver issue that just needs to be ironed out. The average frame rate still scaled in-line with our expectations though, with the RTX 5090 hitting 129fps at 1440p, making it 29% faster than the RTX 4090.

Up at 4K, we're now looking at 75fps on average, so almost a 20fps increase over the RTX 4090, and that works out as a 31% improvement.

Black Myth: Wukong is a 2024 action role-playing game developed and published by Game Science. The game is inspired by the classical Chinese novel Journey to the West and follows an anthropomorphic monkey based on Sun Wukong from the novel. Black Myth: Wukong was released for PlayStation 5 and Windows on August 20, 2024, with an Xbox Series X/S version to be released at a later date. (Wikipedia).

Engine: Unreal Engine 5. We test using the Cinematic preset, TSR set to 100% render scale, DX12 API.

Black Myth: Wukong is next and it shows you how demanding this game is when the fastest consumer graphics card can only manage 87fps at native 1440p using the Cinematic preset! That works out as a 24% bump over the 4090, or a 57% uplift compared to the 4080 Super.

Up at 4K, even the 5090 can't even secure a locked 60fps, we fall just shy to 58fps with the 1% lows at 47fps. That makes the Blackwell flagship 30% faster on average when compared to the 4090.

Cyberpunk 2077 is a 2020 action role-playing video game developed and published by CD Projekt. The story takes place in Night City, an open world set in the Cyberpunk universe. Players assume the first-person perspective of a customisable mercenary known as V, who can acquire skills in hacking and machinery with options for melee and ranged combat. Cyberpunk 2077 was released for Microsoft Windows, PlayStation 4, Stadia, and Xbox One on 10 December 2020. (Wikipedia).

Engine: REDengine 4. We test using the Ultra preset, FSR disabled, DX12 API.

As for Cyberpunk 2077, at 1440p the RTX 5090 delivers 190fps exactly, so about another 40fps on top of the RTX 4090, and that makes it 25% faster. It's also 59% ahead of AMD's current flagship, the RX 7900 XTX.

Stepping up to 4K reveals the single biggest performance increase over the 4090 that we will see all day – the 5090 averages 101fps, making it 43% faster than the 4090.

Final Fantasy XVI is a 2023 action role-playing game developed and published by Square Enix. The sixteenth main instalment in the Final Fantasy series, it was released for the PlayStation 5 in June 2023, with a Windows version released in September 2024.

Engine: Square Enix in-house engine. We test using the Ultra preset, DLSS/FSR disabled, DX12 API.

Final Fantasy XVI is another very GPU-heavy title, and at 1440p the RTX 5090 hits 92ps, giving it a 25% lead over the 4090, while it's 47% faster than the 7900 XTX.

We then average 61fps at 4K, a 28% improvement over the 4080, or a 61% lead over the 7900 XTX.

Forza Horizon 5 is a 2021 racing video game developed by Playground Games and published by Xbox Game Studios. The twelfth main instalment of the Forza series, the game is set in a fictionalised representation of Mexico. It was released on 9 November 2021 for Microsoft Windows, Xbox One, and Xbox Series X/S. (Wikipedia).

Engine: ForzaTech. We test using the Extreme preset, DX12 API.

#

Forza Horizon 5 doesn't scale quite as well as the other titles we have assessed so far. We're still able to hit 235fps using the Extreme preset at 1440p, but that's just an 18% gain over the 4090.

Even at 4K, the jump in performance is relatively subdued, with another 18% margin on show given the RTX 5090 manages 183fps on average. That also makes it just 36% faster than the 7900 XTX.

Ghost of Tsushima is a 2020 action-adventure game developed by Sucker Punch Productions and published by Sony Interactive Entertainment. The player controls Jin Sakai, a samurai on a quest to protect Tsushima Island during the first Mongol invasion of Japan.Ghost of Tsushima was released for the PlayStation 4 in July 2020, and an expanded version for PlayStation 4 and PlayStation 5, subtitled Director's Cut and featuring the Iki Island expansion, was released in August 2021. A Windows version of Director's Cut, developed by Nixxes Software, was released in May 2024. (Wikipedia).

Engine: Sucker Punch in-house engine. We test using the Very High preset, DX12 API.

Ghost of Tsushima does fare better and at 1440p the RTX 5090 manages to average 164fps, giving it a 25% lead over the RTX 4090, and a 48% advantage over the 7900 XTX.

Stepping up to 4K gives the RTX 5090 more breathing room, its average of 111fps works out as a 31% gain over the 4090, and it's now 60% ahead of AMD's fastest consumer graphics card.

Horizon Forbidden West is a 2022 action role-playing game developed by Guerrilla Games and published by Sony Interactive Entertainment. The sequel to Horizon Zero Dawn (2017), the game is set in a post-apocalyptic version of the Western United States, recovering from the aftermath of an extinction event caused by a rogue robot swarm. The game and the Burning Shores expansion were collected together, re-released as Horizon Forbidden West Complete Edition for PlayStation 5 in October 2023, and ported to Windows by Nixxes Software in March 2024. A sequel is in development. (Wikipedia).

Engine: Decima. We test using the Very High preset, DX12 API.

 

Up next is Horizon Forbidden West, where we again saw lower-than-expected performance scaling. It's likely CPU bottlenecks are at play, even on the Ryzen 7 9800X3D, as the 5090 delivers just a 17% gain over the RTX 4090 at 1440p.

Now that does increase to 23% at 4K, though that's still on the lower side compared to the rest of our games tested. It's also a 64% gain over the RTX 4080 Super.

The Last of Us Part I is a 2022 action-adventure game developed by Naughty Dog and published by Sony Interactive Entertainment. A remake of the 2013 game The Last of Us, it features revised gameplay, including enhanced combat and exploration, and expanded accessibility options. It was released for Microsoft Windows in March 2023. (Wikipedia).

Engine: Naughty Dog in-house engine. We test using the Ultra preset, DX12 API.

 

Moving onto The Last of Us Part 1, this does get the RTX 5090 back on track with an average of 176fps at 1440p, making it 29% faster than the 4090 and 65% ahead of the 7900 XTX.

Up at 4K, the 5090 is able to stretch its lead to 36% over the 4090, a healthy performance increase, with an average frame rate of 101fps.

A Plague Tale: Requiem is an action-adventure stealth game developed by Asobo Studio and published by Focus Entertainment. It is the sequel to A Plague Tale: Innocence (2019), and follows siblings Amicia and Hugo de Rune who must look for a cure to Hugo's blood disease in Southern France while fleeing from soldiers of the Inquisition and hordes of rats that are spreading the black plague. The game was released for Nintendo Switch, PlayStation 5, Windows, and Xbox Series X/S on 18 October 2022. (Wikipedia).

Engine: Asobo Studio in-house engine. We test using the Ultra preset, DX12 API.

As for A Plague Tale: Requiem, the 5090 averages 164fps at 1440p, which equates to a 25% uplift over the 4090, or a 59% gain over the 4080 Super.

4K resolution once again sees Nvidia's new flagship extend its leads over the rest of the field, this time it's 34% ahead of its direct predecessor and 78% faster than the 4080 Super.

Senua's Saga: Hellblade II is a 2024 action-adventure game developed by Ninja Theory and published by Xbox Game Studios. The game serves as the sequel to Hellblade: Senua's Sacrifice (2017) and is set in 9th century Iceland, drawing inspiration from Norse mythology and culture. Senua's Saga: Hellblade II was released for Windows and Xbox Series X/S on 21 May 2024.

Engine: Unreal Engine 5. We test using the High preset, TSR set to 100% render scale, DX12 API.

Despite being a very GPU-heavy Unreal Engine 5 title, Senua's Saga: Hellblade 2 shows lower performance scaling than I was expecting from the 5090. It's 21% faster than the 4090 at 1440p, for instance, and 56% ahead of the RX 7900 XTX.

Even at 4K, though, it can't do better than a 23% uplift over the 4090, with an average frame rate of 66fps. That means it's also 61% faster than the 4080 Super.

Starfield is a 2023 action role-playing game developed by Bethesda Game Studios and published by Bethesda Softworks. Announced in 2018, Starfield was delayed several times. The game was released for Windows and Xbox Series X/S on September 6, 2023. (Wikipedia).

Engine: Creation Engine 2. We test using the Ultra preset, with 100% resolution scale, dynamic resolution disabled, DX12 API.

Starfield is another one where we see disappointing improvements compared to the 40-series, or specifically the 4090. At 1440p for instance, the 5090 is just 16% faster, and while it is possible, I don't think we are CPU limited here, with GPU utilisation reporting at 98%.

Even at 4K the gain is just 21% over the 4090, which works out as an extra 16fps in our benchmark run in the forest area on the planet Jemison.

Total War: Warhammer III is a turn-based strategy and real-time tactics video game developed by Creative Assembly and published by Sega. It is part of the Total War series, and the third to be set in Games Workshop's Warhammer Fantasy fictional universe (following 2016's Total War: Warhammer and 2017's Total War: Warhammer II). The game was announced on February 3, 2021 and was released on February 17, 2022. (Wikipedia).

Engine: TW Engine 3 (Warscape). We test using the Ultra preset, with unlimited video memory enabled, DX11 API.

 

Lastly, we close with Total War: Warhammer III which actually scaled better than I thought it would, being a slightly older DX11 title. At 1440p the RTX 5090 can crank through 254fps on average, making it 22% faster than the 4090.

At 4K however, the average of 153fps works out as a 36% improvement over the 4090, while it's also 66% faster than AMD's RX 7900 XTX.

Here we present frame rate figures for each graphics card, averaged across all 12 rasterised games on test today. These figures can disguise significant variations in performance from game to game, but provide a useful overview of the sort of performance you can expect at each resolution tested.

Looking at the 12-game average results, for rasterisation the RTX 5090 comes in 22% faster than its direct predecessor, the RTX 4090, when testing at 1440p. That also makes it 54% faster than the RX 7900 XTX and a 57% uplift compared to the 4080 Super.

Of course, this GPU is really designed for 4K and it scales best at this resolution, now delivering an extra 29% performance over the 4090, while it's 64% and 71% faster than the 7900 XTX and 4080 Super, respectively.

Just to break down the performance versus the RTX 4090, here we can see the relative performance gains for the 5090 over its predecessor. At 1440p it ranges from 16% up to 29% faster.

As mentioned, those margins do increase at 4K, as we see just one title showing a gain of less than 20%, while six of the twelve games tested show improvements of at least 30% over the 4090.

The MSRPs

Using the average frame rate data presented earlier in the review, here we look at the cost per frame using the UK MSRP launch prices for each GPU. Please note this only compares rasterised performance and value.

With an asking price of £1939 I don't think anyone was expecting the RTX 5090 to come in as a cost per frame champion, so it's not surprise to see that overall value is fairly poor here. I really don't think ‘value' is what the RTX 5090 is all about though, so I don't see this as much of an issue. Even considering the price increase over the RTX 4090, cost per frame does still see an improvement, 11% at 4K, to be exact.

Current retail pricing

It's a similar story when looking at cost per frame using current retail prices. RTX 4090 stock is increasingly hard to find, so prices have crept up to the £2000 mark here in the UK. As such, the RTX 5090 looks like even better value, though it's still second bottom on our chart.

Alan Wake 2 is a 2023 survival horror game developed by Remedy Entertainment and published by Epic Games Publishing. A sequel to Alan Wake, the story follows best-selling novelist Alan Wake, who has been trapped in an alternate dimension for 13 years, as he attempts to escape by writing a horror story involving an FBI special agent named Saga Anderson. The game was released for PlayStation 5, Windows, and Xbox Series X/S on 27 October 2023. (Wikipedia)

Engine: Northlight. We test using the High preset, High Ray Tracing preset, FSR set to Quality upscaling, DXR API.

Moving onto ray tracing performance, there's a clear gulf in class when looking at Alan Wake 2 – only a handful of GPUs can deliver actually playable performance in this game with RT enabled. Of course the 5090 is top of the pile and it comes in 28% faster than the 4090, though the 1% lows are in need of fixing as we mentioned earlier in the review.

Up at 4K, the 5090 manages 55fps and that makes it 30% faster than the 4090, while it's 83% ahead of the 4080 Super.

Black Myth: Wukong is a 2024 action role-playing game developed and published by Game Science. The game is inspired by the classical Chinese novel Journey to the West and follows an anthropomorphic monkey based on Sun Wukong from the novel. Black Myth: Wukong was released for PlayStation 5 and Windows on August 20, 2024, with an Xbox Series X/S version to be released at a later date. (Wikipedia).

Engine: Unreal Engine 5. We test using the Cinematic preset, Ray Tracing set to Very High, TSR set to 67% render scale, DXR API.

 

Black Myth: Wukong is similarly punishing and there's not many GPUs that deliver playable performance with ray tracing at Very High settings. Here the 5090 delivers 95fps on average, giving it a 30% lead over the 4090 at 1440p.

Up at 4K, the RTX 5090 now hits 57fps, stretching its lead to 37% over the 4090. Compared to the next fastest GPU on the chart, the 4080 Super, the Blackwell flagship is 86% more performant.

Cyberpunk 2077 is a 2020 action role-playing video game developed and published by CD Projekt. The story takes place in Night City, an open world set in the Cyberpunk universe. Players assume the first-person perspective of a customisable mercenary known as V, who can acquire skills in hacking and machinery with options for melee and ranged combat. Cyberpunk 2077 was released for Microsoft Windows, PlayStation 4, Stadia, and Xbox One on 10 December 2020. (Wikipedia).

Engine: REDengine 4. We test using the Ray Tracing: Ultra preset, DXR API. DLSS/FSR are disabled.

As for Cyberpunk, at 1440p the 5090 is capable of 102fps using the RT Ultra preset, though do note upscaling is disabled. That makes it 24% faster than the 4090, while it's well over double the performance of the RX 7900 XTX.

4K resolution once agains sees the 5090 extend its lead over the 4090, coming in 30% faster with an average frame rate of 54fps.

F1 24 is a racing video game developed by Codemasters and published by EA Sports. It is the sixteenth entry in the F1 series by Codemasters. It holds the official licence for the 2023 Formula One and Formula 2 championships. The game was released for Microsoft Windows, PlayStation 4, PlayStation 5, Xbox One, Xbox Series X/S, and Linux (through Valve's Proton compatibility layer) on 16 June 2023. (Wikipedia).

Engine: EGO. We test using the Ultra High preset, High ray tracing quality, DXR API. DLSS/FSR are disabled.

F1 24 delivers similar gains, though the ray traced effects aren't as demanding, so the 5090 is able to hit 155fps with everything maxed out at 1440p. That puts it 30% ahead of the 4090, and it's 82% faster than the 7900 XTX, but with substantially better 1% lows.

As we step up to 4K, the gap between the 5090 and the rest of the pack only widens, as its now 34% faster than Nvidia's previous flagship and almost double the performance of the 4080 Super.

Ratchet & Clank: Rift Apart is a 2021 third-person shooter platform game developed by Insomniac Games and published by Sony Interactive Entertainment for the PlayStation 5. It is the ninth main instalment in the Ratchet & Clank series and a sequel to Ratchet & Clank: Into the Nexus. Rift Apart was announced in June 2020 and was released on June 11, 2021. A Windows port by Nixxes Software was released on July 26, 2023. (Wikipedia).

Engine: Insomniac Games in-house engine. We test using the Very High preset, Very High ray tracing, DXR API. DLSS/FSR are disabled.

Meanwhile, in Ratchet and Clank, it's a similar story – the RTX 5090 comes in 26% faster than the 4090 at 1440p, while AMD's fastest consumer GPU, the RX 7900 XTX, is nowhere to be seen.

At 4K, the 5090 is now 28% more performant than the 4090, giving it a 79% lead over the 4080 Super.

Returnal is a 2021 roguelike video game developed by Housemarque and published by Sony Interactive Entertainment. It was released for the PlayStation 5 on April 30, 2021 and Windows on February 15, 2023. The game follows Selene Vassos, an astronaut who lands on the planet Atropos in search of the mysterious “White Shadow” signal and finds herself trapped in a time loop. (Wikipedia).

Engine: Unreal Engine 4. We test using the Epic preset, Epic ray traced shadows and reflections, DXR API. DLSS/FSR are disabled.

Next up is Returnal which only includes ray traced shadows and reflections, allowing the 5090 to cruise to 155fps on average at 1440p, though that's a relatively small 22% gain over the 4090.

Up at 4K it manages 95fps, a slider wider margin of 28% compared to its predecessor, while it's 66% ahead of the 4080 Super.

Shadow of the Tomb Raider is a 2018 action-adventure video game developed by Eidos-Montréal and published by Square Enix's European subsidiary. It continues the narrative from the 2015 game Rise of the Tomb Raider and is the twelfth mainline entry in the Tomb Raider series, as well as the third and final entry of the Survivor trilogy. The game was originally released worldwide for PlayStation 4, Windows, and Xbox One. (Wikipedia).

Engine: Foundation Engine. We test using the Highest preset, RT Ultra Shadows, DXR API. DLSS/FSR are disabled.

Shadow of the Tomb Raider feels like an ancient relic at this point – I've been benchmarking it since the days of the RTX 20-series, but it's still interesting to see how performance scales in games that only have a single RT effect. At 1440p, the 5090 hits an eye-watering 235fps – no DLSS required! – and that makes it 25% faster than the 4090.

Even at 4K, we're looking at 140fps on average with no upscaling, and that translates to a 27% gain over the 4090.

Star Wars Outlaws is a 2024 action-adventure game developed by Massive Entertainment and published by Ubisoft. Set in the Star Wars universe between the events of The Empire Strikes Back (1980) and Return of the Jedi (1983), the story follows Kay Vess, a young scoundrel who assembles a team for a massive heist in order to pay off a crime syndicate. Star Wars Outlaws was released for PlayStation 5, Windows, and Xbox Series X/S on August 30, 2024.

Engine: Snowdrop. We test using the Ultra preset but with Nvidia RTXDI disabled, DXR API. DLSS/FSR are disabled.

Lastly, we finish with Star Wars: Outlaws, a game with ray tracing at the heart of its engine. Even then, the 5090 displays relatively meagre gains over the 4090 at 1440p, given it is just 21% faster at 112fps on average.

4K does scale slightly better, this time showing the 5090 with a 25% advantage over the 4090, but that's still one of the less impressive results we have seen today.

Ray tracing 8-game average FPS

Here we present frame rate figures for each graphics card, averaged across all 8 ray traced games on test today. These figures can disguise significant variations in performance from game to game, but provide a useful overview of the sort of performance you can expect at each resolution tested.

Overall performance with ray tracing enabled scales very similarly to what we saw from the rasterised numbers. I was expecting a slightly bigger gain, relative to the 4090, when enabling RT, but at 1440p the 5090 is 25% faster, and that increases to 29% at 4K, which is the exact same margin of difference we saw in our 12-game rasterisation average.

Cost per frame, launch MSRP data

Please note this only compares ray traced performance and value.

Interestingly, overall value does look considerably better when looking at the data based on our ray traced numbers. Granted, the 5090 is still hardly jaw-dropping in this regard, but its performance being so much better than anything AMD has to offer makes it a much better value if ray tracing is your primary concern. It remains to be seen if RDNA 4 is able to change that!

Cost per frame, current pricing data

The picture does change slightly when looking at the data based on current pricing, mainly because a lot of GPUs have received hefty discounts over the course of their life cycles, but the 5090 is still middle of the pack at 4K.

On this page we present a number of different charts relating to Nvidia's DLSS 4 suite of technologies. I do recommend watching the video as image quality analysis is important when considering the DLSS 4 numbers, but I'll explain each chart here.

New Transformer DLSS model

The first thing to look at is the introduction of a new Transformer-based DLSS algorithm. Nvidia has suggested that it pushed the old DLSS upscaler as far as it can go in terms of image quality, and that switching to the new, more complex Transformer model instead of the older Convolutional Neural Network (CNN) model uses ‘2x more parameters and 4x more compute to provide greater stability, reduced ghosting, higher details and enhanced anti-aliasing in game scenes.' (Source).

The good news here is that this actually applies to all RTX GPUs! That's right, the Transformer model is compatible with cards back to the 20-series, and it works with both Super Resolution and Ray Reconstruction.

In the video I do spot visual benefits compared to the older model, but it's also worth making it clear there is a slight performance cost compared to the CNN model. In Cyberpunk 2077 – Ultra settings, path tracing on, 4K with Super Resolution Performance and Ray Reconstruction enabled – I saw a 4fps reduction on the 5090. However, the cost is greater on older GPUs, with the RTX 3090 dropping from 35fps to 29fps.

Putting it another way, the cost for a 5090 to use the Transformer model is just 4% of your fps, whereas for a 3090 it increases to a 17% cost over the CNN model. That may mean it isn't always viable for those on older GPUs, but it's great to see improvement being made for all RTX owners.

DLSS Multi Frame Generation

Multi Frame Generation (MFG) is what's new and currently exclusive to the 50 series however. As we know, Frame Gen debuted with the 40-series and it uses AI to insert one generated frame between each traditionally rendered frame. Now, Nvidia has not only overhauled the Frame Gen model itself, but it can generate up to three additional frames for every traditionally rendered frame.

Again, this has implications for image quality given we know an AI-generated frame isn't always a perfect match for one that's been cranked out on the CUDA cores, so be sure to check the video for the visual comparison. Anyone who's used frame gen in the past, however, will know what sort of thing to expect – certain UI elements don't always play nice, there is still some fizzling around fast-moving objects, and translucent objects or things intersecting with transparencies can cause visual difficulties.

For the most part however, I'd say the increase to visual smoothness is well worth the occasional hiccup, though everyone will have their own preference when it comes to the visual quality. In terms of frame rate numbers, above I tested with MFG disabled using DLAA, MFG disabled using Super Resolution Performance, and then MFG 2X, MFG 3X and MFG 4X, all using Super Resolution Performance.

It's worth stressing that the idea is not to just whack on MFG 4X for the highest frame rate. Rather, the 2X, 3X or 4X settings are designed to give you options to get as close to your monitor's maximum refresh rate – with frame gen and the added latency (which we will get to shortly), there's no longer a benefit to exceeding your monitor's max refresh, so if you have a 144Hz display, it's likely that MFG 2X is all you need. For those using 4K/240Hz screens, MFG 4X could be required to hit that refresh rate target.

Speaking of latency however, this is one key area of discussion as Frame Gen diverges from what we'd typically expect from a frame rate increase. That's because, despite frame rate going up, latency does not go down – it actually increases given how the algorithm has to buffer one traditionally rendered frame in order to squeeze the AI generated frames in between.

To show how latency scales in a complete apples-to-apples test, I used DLAA – instead of Super Resolution – first with MFG Off, and then the 2X, 3X and 4X modes. You can see that latency is only going up, despite the frame rate increasing.

That's a paradigm shift compared to what we've had before, as typically latency goes down as frame rate goes up. We can see a clear example of that when looking at Super Resolution scaling in the chart above – the lower the internal resolution, the higher the frame rate, the lower the latency. That's why Nvidia claims DLSS 4 will improve latency versus native, because Super Resolution can drop it significantly, to the point where even with Frame Gen added in the mix, it's lower than what we saw from a true native render.

In a nutshell, while MFG 4X gives you the visual fluidity of 240FPS, it does not improve the latency to match – so you are left with a similar, if not slightly worse, ‘feel' than if the technology wasn't enabled to begin with.

Just to show another example of MFG in action, I also tried out Alan Wake 2. Now, as mentioned earlier in the review, I have been having frame time issues with the 5090, but it still shows the raw average fps increases that can be had with DLSS. Once more, though, latency is increased with the technology enabled – we see 51ms of PC latency when using MFG 4X alongside Super Resolution Performance mode, compared to 38ms when just using Super Resolution Performance. Again, that means there is a disconnect between the visual fluidity and the input – how sensitive you are to this can vary from person to person.

I do think its easier to get away with the added latency in slower, third-person games like Alan Wake – for first person games, there feels like less of a disconnect between you and the character's movements, so any extra latency is more easily felt. Third-person games seem to give you more of a ‘buffer zone' when it comes to the feel of the game, though maybe that's just me!

Hogwarts Legacy is a slightly different example and it's very interesting because it gives us an insight as to what MFG can do in CPU-bound games. As we can see in the above chart, there is no performance gain to be had when moving from DLAA to DLSS Super Resolution Performance on the RTX 5090 simply because we are so CPU bottlenecked, even at 4K. It's only when enabling MFG that frame rates increase, first to 115fps with MFG 2X, up to 229fps with MFG 4X. And as the 5090 is so CPU bottlenecked, latency increases by much smaller amounts – though it's still going up, rather than down as we'd typically expect of a frame rate increase.

The snag is that the frame pacing in Hogwarts Legacy has always been terrible, and MFG cannot fix that. Here, for instance, we're comparing Super Resolution Performance mode without MFG, to the same upscaling but with MFG 2X enabled. Yes, the overall frame times are lower, as the frame rate has increased, but the consistency is still very poor and there's a number of large spikes. To be clear, that's not caused by MFG, that's just the game – hopefully it'll get fixed one day.

Performance versus RTX 4090

As a final example, I've included some comparisons to the RTX 4090, which only supports ‘Frame Generation' rather than ‘Multi Frame Generation' – i.e. the insertion of one AI-generated frame rather than up to three on the RTX 5090. As such, that means you can get significantly higher frame rates with the 5090, which is how Nvidia devised its ‘up to 2x faster' marketing slides that generated all that discussion earlier in the month. I'm presenting these charts with otherwise no comment – if the idea of generating more frames is valuable to you, then great, the 5090 can often double the frame rate of a 4090 with MFG 4X. If not, we can move on!

Here we present a range of AI and productivity benchmarks, designed to offer insight into the sort of workloads that might be undertaken by someone purchasing this calibre of GPU.

Starting with our Stable Diffusion XL FP16 image generation test, as expected the RTX 5090 is decently faster than the RTX 4090, to the tune of 33%, which scales in-line with our game testing.

We also run the AI Text Generation benchmark designed to measure LLM performance, with four AI models we can test in Procyon. Gains over the 4090 varies slightly depending on the model, but Llama 2 performance has improved by 37% to pick out one example.

I also include the Geekbench AI Pro benchmark. The Half Precision metric shows the largest gains for the 5090, with a 25% increase, though the other two categories show improvements of less than 20% over the 4090.

Blender benchmark makes a return, measuring the performance of the Classroom render scene. This shows a 31% gain over the 4090.

Finally, we run the V-Ray GPU benchmark, using the RTX path. This shows another 36% gain for the 5090 over its predecessor.

Here we present the average clock speed for each graphics card while running A Plague Tale: Requiem for 30 minutes. We use GPU-Z to record the GPU core frequency during gameplay. We calculate the average core frequency during the 30 minute run to present here.

Looking now at clock speed of the Founders Edition model, it ran at a pretty consistent frequency during the 30-minute stress test. It initially boosted to 2670MHz, before settling down at around 2630MHz after three or-so minutes, where it held for the remainder of the test. This is about 50MHz slower than what my RTX 4090 typically achieves, so it's interesting to see slightly lower clock speeds with Blackwell.

For our temperature testing, we measure steady-state GPU temperatures under load. A reading under load comes from running A Plague Tale: Requiem for 30 minutes.

As for thermals, it's quite something to see a dual-slot graphics card deal with 575W of heat in such an effective manner. Over the thirty minute stress test, the GPU core settled at just 72C, while the VRAM was warmer – but still well within safe limits – at 88C.

It's also worth adding that no software I tried was able to accurately read a hot spot temperature for the GPU, likely due to a bug. I asked Nvidia about this and they suggested any software tools will need an update, but most interestingly, Nvidia said ‘GPU hotspot temperature has never been a relevant metric for GeForce GPUs, it does not affect GPU Boost clocks or behaviour and never has. Instead, it is the “GPU Temperature” that affects boost clock functionality.'

We take our noise measurements with the sound meter positioned 1 foot from the graphics card. I measured the noise floor to be 32 dBA, thus anything above this level can be attributed to the graphics cards. The power supply is passive for the entire power output range we tested all graphics cards in, while all CPU and system fans were disabled. A reading under load comes from running A Plague Tale: Requiem for 30 minutes.

Despite running at very reasonable temperatures, the 5090 Founders Edition does not compromise on noise levels either. I mean, it's not silent by any means, hitting 39dBa on the sound meter, and that was with fan speed reported at 46%, or 1585rpm. Again though, for a dual-slot card that draws 575W of power (!) I think this is very impressive indeed.

The only snag is I did hear some coil while during my testing. It wasn't awful and was most obvious at lower resolutions or during loading screens, but I've included a soundtest in the video.

Here we present power draw figures for the graphics card-only, on a per-game basis for all twelve games we tested at 1080p. This is measured using Nvidia's Power Capture Analysis Tool, also known as PCAT. You can read more about our updated power draw testing methodology HERE.

Per-Game Results at 1080p:

Click to enlarge.

12-Game Average at 1080p:

The power draw data for 1080p isn't especially relevant considering you'd have to be mad to pay nearly two grand for an RTX 5090 to game at Full HD resolution, but we include it nonetheless. As expected, the 5090 does not draw close to its rated TBP at this resolution as GPU load is much lower than 1440p or especially 4K.

Ray tracing results

As a bonus, we now include the power figures for all eight games we test with ray tracing enabled.

8-Game ray tracing average:

Power does creep up to 450W with ray tracing enabled, even at 1080p, due to the increased load placed on the GPU.

Here we present power draw figures for the graphics card-only, on a per-game basis for all twelve games we tested at 1440p. This is measured using Nvidia's Power Capture Analysis Tool, also known as PCAT. You can read more about our updated power draw testing methodology HERE.

Per-Game Results at 1440p:

Click to enlarge.

12-Game Average at 1440p:

Up at 1440p we do see power increasing and a couple of games now show over 500W of power draw, but most are still below that. Still, the average of 463W is still very high – over 100W above the RTX 4090, working out as a 29% increase. Remember, the 5090 proved 22% faster in our testing, so that gives you an idea of the efficiency, which we will look at shortly…

Ray tracing results

As a bonus, we now include the power figures for all eight games we test with ray tracing enabled.

8-Game ray tracing average:

As expected, power draw is higher still with RT enabled, this time averaging 491.9W at 1440p.

Here we present power draw figures for the graphics card-only, on a per-game basis for all twelve games we tested at 2160p (4K). This is measured using Nvidia's Power Capture Analysis Tool, also known as PCAT. You can read more about our updated power draw testing methodology HERE.

Per-Game Results at 2160p (4K):

Click to enlarge.

12-Game Average at 2160p (4K):

Up at 4K, power draw gets to its highest level from the RTX 5090, as we'd expect. Now a number of games pull well over 520W, though there's still a few that are closer to 400-450W. Even so, that is quite a remarkable amount of juice for a single graphics card, marking a 30% increase in power draw versus the 4090.

Ray tracing results

As a bonus, we now include the power figures for all eight games we test with ray tracing enabled.

8-Game ray tracing average:

With ray tracing enabled, that figure gets even higher! Now we're looking at 546.3W on average, with Star Wars: Outlaws even hitting 580W power draw…

Using the graphics card-only power draw figures presented earlier in the review, here we present performance per Watt on a per-game basis for all twelve games we tested at 1080p.

Per-Game Results at 1080p:

Click to enlarge.

12-Game Average at 1080p:

We'll briefly skip over the 1080p data, but you can see the RTX 5090 is not at the top of the chart – performance per Watt has regressed from the previous generation.

Ray tracing results

As a bonus, we now include the power figures for all eight games we test with ray tracing enabled.

8-Game ray tracing average:

That's the same when looking at RT workloads too – the RTX 5090 is slightly less efficient that its 40-series forbears.

Using the graphics card-only power draw figures presented earlier in the review, here we present performance per Watt on a per-game basis for all twelve games we tested at 1440p.

Per-Game Results at 1440p:

Click to enlarge.

12-Game Average at 1440p:

At 1440p, we can see performance per Watt is on par with the RTX 4060 Ti, but slightly behind the RTX 4090. Compared to the RTX 4080 Super, it marks an 8% reduction in efficiency which is not something I was expecting to see, especially given Nvidia's usual strong messaging around efficiency and performance per Watt.

Ray tracing results

As a bonus, we now include the power figures for all eight games we test with ray tracing enabled.

8-Game ray tracing average:

It's much the same at 1440p with RT enabled. The 4080 Super remains the most efficient GPU we've tested.

Using the graphics card-only power draw figures presented earlier in the review, here we present performance per Watt on a per-game basis for all twelve games we tested at 2160p (4K).

Per-Game Results at 2160p (4K):

Click to enlarge.

12-Game Average at 2160p (4K):

Now, at 4K, the RTX 5090 does close the gap slightly, but it still lags behind the 4090 and 4080 Super – this time it offers performance per Watt that's 5% lower. Of course you may not be bothered by that, but I find it fascinating that Nvidia has clearly had to push things further than ever before to squeeze out that extra performance over the 4090, to the point where performance per Watt has actually gone backwards due to such an increase in power draw!

Ray tracing results

As a bonus, we now include the power figures for all eight games we test with ray tracing enabled.

8-Game ray tracing average:

Likewise, we see the same behaviour at 4K with ray tracing enabled.

We measure system-wide power draw from the wall while running A Plague Tale: Requiem for 30 minutes (4K resolution).

Lastly, we close with a look at total system power draw. This scales in line with what we saw earlier in the review – the RTX 5090 is clearly notably more power hungry than the RTX 4090, with the whole system pulling 171W more.

After a slightly longer than usual wait, Nvidia's next-generation Blackwell architecture has arrived and it's the flagship RTX 5090 that is leading the way. RTX 5080 is due to land in a week's time at a slightly more attainable £979 asking price, but today's review is all about the £1939 behemoth that is Nvidia's RTX 5090.

Much was made of the performance ahead of launch, people were breaking out rulers and pixel counting Nvidia's bar charts, but after thorough testing today we can confirm native rendering performance has increased in the ballpark of 30% over the RTX 4090 when testing at 4K. That makes the RTX 5090 64% faster on average compared to AMD's current consumer flagship, the RX 7900 XTX, while it's a 71% uplift over the RTX 4080 Super. Ray tracing also scales similarly, given we saw the exact same 29% margin over the RTX 4090 in the eight RT titles we tested.

Those are the sort of performance increases you can expect at 4K, but the uplift does get progressively smaller as resolution decreases. Versus the RTX 4090, for instance, we saw more limited gains of 22% at 1440p and 18% at 1080p. Now, I don't expect many people will be gaming at native 1080p on an RTX 5090, but it's worth bearing that in mind if you'd typically game with DLSS Super Resolution. After all, using its performance mode at 4K utilises a 1080p internal render resolution. Clearly this is a card designed for 4K – and perhaps even above – but that performance scaling at lower resolutions could be something to bear in mind.

Of course, whether or not you are impressed by those generational gains depends entirely on your perspective – an extra 30% over the 4090 could sound great, or it could be a disappointment. The main thing from my perspective as a reviewer is to give you, the reader, as much information as possible to allow you to make an informed decision, and I think I have done that today.

However, it's certainly fair to say the uplift gen-on-gen is smaller than what we've come to expect in this market segment. The RTX 3090 proved about 50% faster than the 2080 Ti, for instance, while the RTX 4090 was more like 80% faster than the RTX 3090 and 60% ahead of the RTX 3090 Ti. That does make the 30% gain seem less impressive, but do we need to adjust our expectations? It's no secret that each new node is getting more and more expensive to manufacture, with technological advancements getting harder and harder to come by.

It would certainly seem the lack of a node shrink is the limiting factor for the RTX 5090, with Blackwell remaining on TSMC's N4 process. That means, to get the extra performance, Nvidia increased die size by 23% over AD102, adding in 33% more cores, while also making the change to GDDR7 memory. Perhaps most eye-catching is the increased Total Graphics Power, hitting an eye-watering 575W, a 28% increase over the RTX 4090's 450W TGP.

In practice, we saw 4K power draw averaging 515W at 4K, though several games do get very close to that 575W figure. Power also tends to go up when ray tracing is enabled, where we recorded an average of 546.3W over the eight RT titles tested. As such, the increase in power draw is actually larger than the increase to performance, resulting a performance per Watt regression – certainly not something I was expecting to see! How much you care about that is entirely up to you, but considering the strides Nvidia usually makes in terms of efficiency gen-on-gen, this is certainly going to raise an eyebrow or two.

The same can also be said for pricing. At £1939/$1999, it is certainly not for the faint of heart, but I'm also not sure the target audience will care too much. Sure, it's an extra £300-400 over the RTX 4090, but the way I see it, anyone buying this sort of product either wants the outright fastest gaming GPU possible, or it's for professional use in which case the extra compute power will pay for itself over time. The reality is that Nvidia simply has no competition in this space so they can charge what they like – because if you want the best of the best, this is it, so be prepared to pay for the privilege.

Gamers do get the extra value add of DLSS 4, specifically Multi Frame Generation (MFG), which is a new feature exclusive to the RTX 50 series. I spent a fair bit of time testing MFG as part of this review and I think if you already got on with Frame Generation on the RX 40-series, you'll probably find a lot to like with MFG. It's been particularly useful in enabling 4K/240Hz gaming experiences that wouldn't otherwise be possible – such as high frame rate path tracing in Cyberpunk 2077 – and with the growing 4K OLED monitor segment, that's certainly good news.

However, it's definitely not a perfect technology as the discerning gamer will still notice some fizzling or shimmering that isn't otherwise there, while latency scaling is still backwards compared to what we've come to expect – in the sense that latency actually increases as frame rate increases with MFG, rather than latency decreasing. That means some will find it problematic as the feel doesn't always match up to the visual fluidity of the increased frame rate.

It is great to see Nvidia is improving other aspects of DLSS, though, with its new Transformer-based models of Super Resolution and Ray Reconstruction. Not only do these improve things like ghosting and overall level of detail compared to the previous Convolutional Neural Network (CNN) model, but this upgrade actually applies to all RTX GPUs, right the way back to the 20-series. There's even a possibility that Multi Frame Gen might come to older cards given that Nvidia hasn't explicitly ruled it out, but personally I'd be surprised to see that happen given it currently acts as an incentive to upgrade to the latest and greatest.

We can't end this review without a discussion of Nvidia's Founders Edition design, either. This is a highly impressive feat of engineering, considering it's a mere dual-slot thickness yet it is able to comfortably tame 575W of power. We saw the GPU settling at 72C during a thirty-minute 4K stress test, while the VRAM hit 88C, which is slightly warmer but still well within safe limits. I love to see the innovation in this department as, when pretty much every AIB partner is slapping quad-slot coolers onto their 5090s, this is a refreshing step back to a time when GPUs didn't cover the entire bottom-half of your motherboard.

Overall then, I do think the RTX 5090 is a solid card. Yes, there are some valid concerns around pricing and power draw increases, but I just don't think the target audience for this card will be particularly bothered – it's ultimately all about the increased performance, and while circa 30% gains won't blow your mind like the jump from an RTX 3090 to the RTX 4090, it's still a decent generational uplift considering that Moore's Law, if not completely dead, is certainly well out the door. It does remain to be seen how performance will scale as we head down the 50 series stack, though, where it's not always possible to pump core counts and power draw by 30% – but for that analysis, we will have to wait and see what the likes of the RTX 5080 and RTX 5070 can bring to the table.

One final note on availability, we've been told by a couple UK retailers to expect an incredibly tight supply at launch, with demand set to far outstrip what is actually available. Matters are complicated by the fact Chinese New Year is just around the corner, so it could take several weeks for things to improve. As a reminder, the retail launch is set for January 30th.

Discuss on our Facebook page HERE.

Pros

• It's the new fastest consumer GPU, with circa 30% gains over the RTX 4090.
• Dual-slot Founders Edition is an incredible feat of engineering.
• Huge 32GB GDDR7 frame buffer.
• DLSS 4 has improved Ray Reconstruction and Super Resolution scaling.
• Multi Frame Generation enables higher frame rates than would otherwise be possible.

Cons

• Incredibly expensive at £1939.
• Very high power demands.
• Performance per Watt is actually a step back compared to the RTX 4090 and RTX 4080 Super.
• Some may be disappointed with the raw performance uplift.

KitGuru says: The RTX 5090 sets a new performance standard by pushing both price and power consumption higher than ever. It's clearly only targeting a certain niche, but if you want the best of the best – this is it.

Become a Patron!