Nvidia GeForce RTX 5090 vs. AMD Radeon RX 7900 XTX

It’s time for the ultimate showdown between AMD and Nvidia, as Radeon RX 7900 XTX takes on GeForce RTX 5090. While this is technically a battle of old versus new, there’s currently no modern replacement for the RDNA 3 graphics card at present nor will there be until later this year. As such, there’s no opponent more worthy for the Blackwell architecture in its mightiest form.

Much separates these two flagships with the odds unavoidably stacked in GeForce RTX 5090’s favour in part thanks to its more-advanced specifications. However, its advantages come with an intimidating four-figure cost and Radeon RX 7900 XTX could offer greater value in some areas. Pitting them against one another, let’s see which is more worthy of serving as king of the GPU hill.

Specs

Make no mistake, the GB202 GPU at the heart of GeForce RTX 5090 is enormous at 744mm² and appears all the larger put next to Navi 31 XTX’s 529mm² die. This size of Nvidia’s champion affords it more room to cram transistors, almost doubling its rival with 92.2b to the 57.7b offered by AMD’s contender. Manufacturing processes naturally play a part too, with Team Green benefitting from TSMC’s more advanced 4N node compared to Team Red’s mix of older N5 and N6 nodes.

Much as battles of specification can boil down to a numbers game, it’s difficult to quantify the quality of particular components. There’s no avoiding that Nvidia has history on its side here, with more mature designs for its ray tracing cores and AI accelerators. For context, this is a battle of GeForce’s 4th Gen RT cores and 5th Gen Tensor Cores against Radeon’s 3rd Gen and 2nd Gen equivalents. This translates to some profound differences in performance as illustrated by benchmarks below.

	GeForce RTX 5090	Radeon RX 7900 XTX
Released	Jan 2025	Nov 2022
Codename	Blackwell	Plum Bonito
GPU	GB202	Navi 31 XTX
Transistors	92.2 billion	57.7 billion
Die size	744mm²	529mm²
Cores	170 SMs	96 CUs
RT units	170	96
AI engines	680 5th Gen Tensor	192 AI Accelerators
Base clock	2,017MHz	1,929MHz
Boost clock	2,407MHz	2,498MHz
Memory	32GB GDDR7	24GB GDDR6
Mem. clock	28Gb/s	20Gb/s
Mem. interface	512-bit	384-bit
Mem. bandwidth	1.7TB/s	960GB/s
TOPS	3,352	–
TBP	575W	450W
Launch MSRP	$1,999	$999

There’s simply no beating GeForce RTX 5090’s memory configuration, try as Radeon RX 7900 XTX might. Nvidia’s graphics card boasts 32GB of GDDR7 VRAM clocked at 28Gb/s, besting the size and speed of its rival’s 24GB of GDDR6 running at 20Gb/s. Compounding its lead, the Blackwell flagship rides on a wider 512-bit memory interface too, resulting in a total bandwidth of 1.7TB/s. By comparison, RDNA 3’s best arrives with a 384-bit bus and can only muster half the bandwidth at 960GB/s.

One category in which Radeon RX 7900 XTX enjoys a win over GeForce RTX 5090 is TBP (Total Board Power). The former’s request of 450W is downright frugal next to the latter’s demands of 575W. This makes AMD’s star player much easier to satiate that Nvidia’s, although you will need a suitably beefy power supply for either pixel pusher in the region of 750-1,000W.

Finally, let’s discuss price. Featuring top-of-the-line specs from their respective time periods, it should come as no surprise that GeForce RTX 5090 and Radeon RX 7900 XTX come at substantial cost. As the new kid on the block, you won’t find Nvidia’s latest and greatest graphics card anywhere remotely below its eye-watering $1,999 MSRP. However, you’ll often find AMD’s alternative below its launch price, boosting its relative value.

Performance

Both GeForce RTX 5090 and Radeon RX 7900 XTX have proven their mettle in the Club386 test bench, complete with a 7950X3D that provides ample room for each graphics card to show their stuff. These scores and frame rates are as fresh as they come, using the latest available drivers for each at the time of writing.

In addition to a more robust helping of synthetic tests, the Club386 benchmark suite below also includes gaming performance at FHD (1080p), QHD (1440p), and UHD (2160p). I’ll also be examining how well AMD and Nvidia’s upscaling solutions fare in one of the most demanding ray traced workloads on the market today.

Applications & AI

Combining the samples per minute of all three scenes in Blender’s benchmark suite, GeForce RTX 5090 takes a clear lead over Radeon RX 7900 XTX. With respective totals of 15,101 and 4,330, Nvidia’s performance almost quadruples AMD’s making the former the preferred choice for creatives in the field of 3D graphics.

Unsurprisingly, Nvidia performs well in Geekbench AI. GeForce RTX 5090 delivers a half-precision score of 66,821, besting Radeon RX 7900 XTX’s result of 43,047 and giving it a lead of 55%.

If there was any doubt that the Blackwell flagship was the card to get for the best performance in all things artificial intelligence, this should put it to bed. It’s worth noting, however, that this performance bump comes with a 100% increase in cost.

The gaps in AI between architectures become all the more apparent in Procyon’s AI Text Generation benchmark. Using a Llama 3.1 model, GeForce RTX 5090’s score of 6,477 almost doubles the 3,525 result of Radeon RX 7900 XTX. Factoring cost into the equation, Nvidia’s higher pricing scales almost linearly with performance.

In our final set of artificial tests, 3DMark paints a particularly painful picture for AMD. Regardless of rendering, rasterised or ray traced, GeForce RTX 5090 is leagues ahead of Radeon RX 7900 XTX.

Looking more closely at the results, Nvidia enjoys its biggest win in the ray traced realms of 3DMark Speed Way with a superior score that’s 118% higher than AMD. While the gap narrows in 3DMark Steel Nomad, it still amounts to a triple-digit difference of 105% in rasterised capabilities.

These victories sadly aren’t indicative of average differences in real-world frame rates, but the general notion that GeForce RTX 5090 is a much-more capable graphics card than Radeon RX 7900 XTX holds true. You’d certainly hope so given the difference in age and price between the two.

Gaming

Assassin’s Creed Mirage gives GeForce RTX 5090 and Radeon RX 7900 XTX a suitable playground to showcase their rasterised rendering might. Running this game with these graphics cards at FHD almost feels like a waste, as each punch north of 200fps on average. It’s also the resolution that sees AMD achieve frame rates closest to Nvidia, as just 32fps separates the two. This pattern will repeat itself throughout the suite.

Moving up to QHD, GeForce RTX 5090 maintains an average frame rate north of 200fps while Radeon RX 7900 XTX turns in a lower but nonetheless respectable 167fps. Unsurprisingly, both cards deliver a suitable frame rate for a high-refresh rate gaming experience without need for upscaling at this resolution.

Running through Baghdad at 4K creates a more pronounced difference in performance. GeForce RTX 5090 strides confidently in the triple-digits with a 150fps average, while Radeon RX 7900 XTX just about manages the same feat with 104fps.

These numbers naturally show Nvidia’s card is faster but the value of its uplift, ranging between 16-44%, does come into question considering it’s much more expensive.

Booting up Final Fantasy XIV: Dawntrail delivers differences in frame rate similar to Assassin’s Creed Mirage. Across all three resolutions, GeForce RTX 5090 dominates with leads of 16-64% relative to Radeon RX 7900 XTX’s frame rates.

AMD does once again manage to keep its head above triple-digit water but only Nvidia can keep both minimum and average frame rates greater than 100fps at higher resolutions.

GeForce RTX 5090 sets pace as we shift gears to the ray traced races of Forza Motorsport. The difference in calibre between these cards at this kind of rendering is apparent at FHD, with 57fps separating Radeon RX 7900 XTX from its rival.

The gap widens as the weight of the workload at QHD and UHD presses down on the pixel pushers. Nvidia’s contender keeps a cooler head, outright lapping AMD’s champ at the latter resolution and coming perilously close to doing so again at QHD.

Take note that Radeon RX 7900 XTX falls short of what I deem a playable frame rate in racing games, failing to hit that all-important 60fps average. Meanwhile, GeForce RTX 5090 barrels ahead with a comfortable 87fps result.

It’s almost dizzying seeing average FHD frame rates of 400fps+ from GeForce RTX 5090 and Radeon RX 7900 XTX in Mount & Blade II Bannerlord. Each card performs excellently at QHD too, delivering 367fps and 300fps respectively.

The battle naturally turns bloodier at UHD, where GeForce RTX 5090 proudly beats its chest with a 270fps result to Radeon RX 7900 XTX’s 163fps. This glorious victory of sorts makes Nvidia’s card the better by a whopping 65%.

More horsepower only matters if there’s road to run on and it’s clear GeForce RTX 5090 is hungry for more in Rainbow Six Extraction. The card’s 1% lows fall slightly between FHD and QHD, but average frames are practically identical with just 3fps separating the two resolutions. By comparison, Radeon RX 7900 XTX is breathing down its rival’s neck at first but the higher resolution creates more space between each card.

Once again, UHD creates a chasm between the cards as GeForce RTX 5090’s 248fps average runs circles around the 156fps of Radeon RX 7900 XTX. Every frame matters in fast-paced first-person shooters, particularly of the competitive variety. This makes the additional 59% offered by Nvidia quite attractive, particularly in combination with Nvidia Reflex tech.

DLSS vs. FSR

After experiencing the likes of Alan Wake 2 and Cyberpunk 2077’s path tracing modes, I’m an unabashed champion of ray tracing. However, the breath-taking qualities this kind of rendering demands place enormous strain on even the very best graphics cards. Running the likes of CD Projket Red’s Ray Tracing: Overdrive preset at native 4K isn’t something I’d recommend, even on a GeForce RTX 5090. It surprises me to learn that such a feat is very much doable at FHD and even QHD, with average frame rates above 60fps.

Sadly, Radeon RX 7900 XTX simply doesn’t have the mustard for this kind of task. Cyberpunk 2077 is playable at a push at FHD with this graphics card pushing 33fps on average, but QHD and UHD are out of the question as the pixel pusher sputters single-digit frame rates at the latter resolution.

What strikes me most about these results is that GeForce RTX 5090 can deliver comparable performance to Radeon RX 7900 XTX while handling four times the pixels (UHD vs. FHD). Running the same resolution, Nvidia enjoys a 215% lead. Staggering stuff.

Employing each graphics card’s super resolution solutions massively improves Radeon RX 7900 XTX performance but FSR simply isn’t enough to keep up with GeForce RTX 5090 using DLSS. Upscaling finally earns AMD its first 60fps victory at FHD but it also turns Nvidia’s two-pair hand into three-of-a-kind.

Much as these graphs illustrate the quantity of frames GeForce RTX 5090 offers, they don’t demonstrate their quality. DLSS remains a step ahead in this regard boasting superior image quality to FSR due to fewer artefacts and a generally sharper image. This in part thanks to machine learning and the results above use the longstanding ‘Convolutional Neural Network’ model but Nvidia is introducing a new ‘Transformer’ model that further boosts image quality with a performance penalty, widening the gap in technologies further.

Club386 will take a more in-depth look at the differences between the two models in due course, so stay tuned.

As a final coup de grace, GeForce RTX 5090 deploys MFG (Multi Frame Generation) to raise its frame rates into the stratosphere. Radeon RX 7900 XTX’s single frame generation does give its performance a welcome lift but there’s simply no catching up to its rival now.

These frame generation technologies incur a small penalty to latency, affecting the perceived smoothness of the overall framerate. Features like Nvidia Reflex and AMD Anti-Lag attempt to minimise this but they can’t remedy the other soft requirement of a high base frame rate. With these variables in mind, GeForce RTX 5090 once again emerges as the only appropriate graphics card for all three resolutions while Radeon RX 7900 XTX only manages to achieve an ideal experience at FHD.

Finally, a quick word on percentages: GeForce RTX 5090 leads Radeon RX 7900 XTX by 306-491% in this scenario.

Vitals

Both AMD and Nvidia build excellent reference design coolers for their respective graphics cards that keep any sign of thermal throttling well at bay. However, Radeon RX 7900 XTX is the cooler of the two pixel pushers, as GeForce RTX 4090 runs 9°C under load.

Despite running hotter, GeForce RTX 5090 is the quieter graphics card next to Radeon RX 7900 XTX. There’s 5.5db between them but you shouldn’t hear either kicking up a fuss louder than your system’s fans.

Given the differences in temperature observed earlier, I am curious if it’s possible to push fan speed more aggressively on GeForce RTX 5090 to make it more of a match.

With great performance comes great power consumption, as GeForce RTX 5090 demonstrates with its astonishingly large draw of 707W under load. By comparison, Radeon RX 7900 XTX’s 482W reading is 32% lower. Beefy power supplies are a must on either graphics card but especially so with Nvidia’s new flagship.

Thankfully, both GeForce RTX 5090 and Radeon RX 7900 XTX don’t pull a massive amount of watts while idle.

Conclusion

It should come as no surprise that GeForce RTX 5090 is more performant than Radeon RX 7900 XTX, as its predecessor was before it. However, the gap between AMD and Nvidia’s respective flagships is now even larger in terms of raw performance and features.

If you can realistically afford either graphics card, there’s little reason to opt for Radeon RX 7900 XTX over GeForce RTX 5090. The latter is more performant in applications and games, boasts more robust specifications, and offers a more advanced feature set through DLSS 4. Just note that the additional $1,000 expenditure won’t always translate into a 1:1 boost in frame rates.

Radeon RX 9000 Series graphics cards are en route, but it seems doubtful that its highest SKU will be able to challenge GeForce RTX 5090. For the foreseeable, Nvidia’s flagship is the undisputed king of the hill.