The first NVIDIA RTX 50-Series GPUs for desktops are finally here! Were they worth the wait?

February 9, 2025 – NVIDIA has finally released two of its highly anticipated GPUs, the GeForce RTX 5090 and the GeForce RTX 5080, with the GeForce RTX 5070 and the GeForce RTX 5070 Ti soon to follow this month.

These new GPUs are characterized by the Blackwell architecture, which replaces the Ada Lovelace architecture from the previous series (RTX 40). They also come with the new DLSS4 technology for super-fast frame generation and improved AI capabilities.

DLSS is a technology that uses AI to boost frames per second (FPS), reduce latency, and improve image quality. ‌According to NVIDIA, the new DLSS4 technology utilizes Multi Frame Generation to boost FPS in graphically demanding games.

Previously, DLSS3 from the 40-series GPUs would use AI to generate every other frame. Now, DLSS4 with Multi Frame Generation can create up to three frames per rendered frame, increasing overall performance.

The new DLSS4 also brings enhanced Ray Reconstruction and Super Resolution, which means it will leverage AI to speed up ray tracing (a.k.a. RTX; i.e., the calculations to accurately simulate lighting in a virtual environment) and upscale frame resolution.

Moreover, the new 50-series GPUs come with NVIDIA Reflex 2 technology, which provides improved responsiveness and reduced latency in action-packed competitive games. Other GeForce RTX GPUs will also benefit from this new technology in a future update.

This is enough to get gamers excited, but game developers will also be able to leverage the greatly enhanced AI capabilities of these new GPUs to boost ray tracing, render immense scenes, and create photorealistic game characters, all thanks to the RTX Kit.

These new graphics cards are also accompanied by a suite of tools, such as NVIDIA Studio, which is capable of harnessing the power of RTX and AI for video editing, 3D rendering, and graphic design.

But let’s have a quick comparison of the main tech specs found in the GeForce RTX 5090, GeForce RTX 5080, the upcoming GeForce RTX 5070 and GeForce RTX 5070 Ti, and their predecessors from the 40-series.

GPU	DLSS	# CUDA Cores	Boost Clock (GHz)	AI TOPs	Memory (VRAM)	Memory Bandwidth	TGP	Price*
RTX 5090	DLSS4	21760	2.41	3352	32 GB GDDR7	1792 GB/s	575	$1999
RTX 5080	DLSS4	10752	2.62	1801	16 GB GDDR7	960 GB/s	360	$999
RTX 5070 Ti	DLSS4	8960	2.45	1406	16 GB GDDR7	896 GB/s	300	$749
RTX 5070	DLSS4	6144	2.51	988	12 GB GDDR7	672 GB/s	250	$549
RTX 4090	DLSS3	16384	2.52	1321	24 GB GDDR6X	1008 GB/s	450	$1599
RTX 4080	DLSS3	9728	2.51	780	16 GB GDDR6X	717 GB/s	320	$999
RTX 4070 Ti	DLSS3	7680	2.61	641	12 GB GDDR6X	504 GB/s	285	$740
RTX 4070	DLSS3	5888	2.48	466	12 GB GDDR6X	504 GB/s	200	$549

As you can see from the table above, aside from the already mentioned new architecture and DLSS4, the new GPUs also come with increased numbers of CUDA cores, AI Tops, and more VRAM than their predecessors.

The increased numbers of CUDA cores across the board indicate greater computing capabilities, as CUDA cores are the processing units within the GPU that can execute instructions simultaneously.

A bit surprisingly, the boost clock speeds (i.e., the rates at which the GPU’s cores execute instructions) haven’t increased significantly, and in the case of the RTX 5090 and RTX 5070 Ti they have slightly decreased.

The more than doubled AI TOPS show NVIDIA’s commitment to making these new GPUs AI powerhouses. “TOPS” stands for “Trillions of Operations per Second” and this metric indicates how fast the AI chips in NVIDIA GPUs have become.

Why are the AI capabilities of these new GPUs important, you may ask? Not only does this enhance the above-mentioned DLSS4 functionality and the RTX Kit for game developers, but also because we are on the verge of witnessing a revolution in game character behavior (Non-Player Characters or NPCs).

In fact, NVIDIA is redefining game AI with the introduction of NVIDIA ACE autonomous game characters. Think of NPCs that can perceive, plan, and act like human players. And have you heard of NVIDIA’s Audio2Face? This technology leverages AI to generate realistic facial animations based on audio files. These are advancements to look forward in future games.

Regarding VRAM, only the RTX 5090 and the RTX 5070 Ti have seen an increase in the amount of dedicated memory, rising from 24 to 32 GB and from 12 to 16 GB, respectively, compared to their RTX 4090 and RTX 4070 Ti predecessors. However, all 50-series GPUs sport the new GDDR7 VRAM, which is considerably faster than the previous GDDR6X.

In terms of power consumption (Total Graphics Power or TGP), all the 50-series models require more power compared to their predecessors. The increase is a modest 15W for the RTX 5070 Ti, but a whopping 125W for the RTX 5090. The RTX 5070 and RTX 5080 fall somewhere in between, requiring an additional 50W and 40W compared to the RTX 4070 and RTX 4080, respectively.

Except for the RTX 5090, all other 50-series GPUs come with a price tag that is very close to (if not identical to) their 40-series predecessors. This is certainly welcome news.

But how do the enhanced tech specs of the NVIDIA GeForce RTX 50-series translate into actual gaming performance? let’s find out below.

Benchmarking Results

RTX 5090 vs RTX 4090

Benchmarks for Gamers have compared the performance of the RTX 5090 and RTX 4090 across several graphically demanding games at max settings and 4K resolution. Even without multi-frame generation (DLSS4), the RTX 5090 consistently outperformed its predecessor.

The observed differences in average frames per second (FPS) reportedly ranged between 20% and 40%, with the RTX 5090 always on top. Tom’s Guide found similar results, reporting that the RTX 5090 delivered 20% to 30% higher frame rates in the tested games.

The FPS difference becomes even more pronounced when frame generation is enabled (DLSS3 in the RTX 4090 and DLSS4 in the RTX 5090). For example, in Cyberpunk 2077 the RTX 5090 achieved up to a 155% boost in average FPS.

However, this performance increase comes at a cost—power consumption. In all tested games the RTX 5090 consistently consumed significantly more power (without frame generation enabled), averaging around 160W higher and peaking at 228W higher in Ghost of Tsushima.

This inevitably translates to higher electricity bills. It’s up to gamers to decide whether the extra FPS are worth it, particularly when the difference may not be noticeable—especially in games where the performance gap is only around 20 FPS or less (e.g., Indiana Jones and the Great Circle, Cyberpunk 2077, Black Myth: Wukong, Stalker 2: Heart of Chernobyl, Senua’s Saga: Hellblade II, and Starfield).

Of course, as mentioned earlier, FPS gaps become much larger when multi-frame generation is enabled. However, it’s important to note that DLSS4 is not yet available in many current games and does not reduce power consumption.

RTX 5080 vs RTX 4080

How does the RTX 5080 compare to its predecessor, the RTX 4080? Testing Games evaluated these two GPUs in graphically demanding games at high settings, both in 4k and 1440p, with interesting results.

At 4K, the RTX 5080 delivered higher average FPS, but the increase was modest, ranging from 11% to 18%. In Forza Horizon 5, for example, the 18% boost translates to just 21 more FPS (114 vs. 135 average FPS).

Additionally, at 4K and high settings, both the RTX 4080 and RTX 5080 struggled to exceed 60 FPS in most games when frame generation was turned off.

Performance at 1440p was better, with the RTX 5080 achieving average FPS above 60 in all games on high settings except Alan Wake 2, where it managed only 35 average FPS without DLSS4.

Overall, FPS differences between the RTX 4080 and RTX 5080 at 1440p were similar to those at 4K, ranging between 8% and 19% in favor of the RTX 5080. Notably, power consumption between these two GPUs was nearly identical in most tested games both at 4K and 1440p.

Conclusions

While the new technology in NVIDIA’s latest GPUs is undoubtedly impressive and aligns with the trend of AI-driven advancements, the first two RTX 50-series models may not fully meet gamers’ expectations.

Even though gaming performance has improved, the leap from the previous generation doesn’t seem substantial enough to justify the RTX 5090’s significantly higher price and power consumption—at least not until more games fully utilize multi-frame generation (DLSS4).

The RTX 5080, which has not seen a price increase compared to its predecessor and has a more contained rise in power consumption, still delivers only modest FPS gains—up to 18% without DLSS4.

However, in demanding games where GPUs struggle to render frames at high settings and 4K resolutions, DLSS4 could be a game changer, as demonstrated in Testing Games‘ benchmarks.

Perhaps the true potential of the RTX 50-series will only be realized as AI-driven games become the industry standard. As mentioned earlier, an AI revolution is coming to gaming, affecting not only rendering and graphics but also NPC behavior. These GPUs are built for what’s ahead.

While we wait for AI-driven gaming to take shape, we look forward to the release of the NVIDIA GeForce RTX 5070 and 5070 Ti, which may offer a better balance between price, performance, and power consumption for today’s games.

And What About the Laptop versions?

Following the unveiling of the GeForce RTX 50-series, NVIDIA confirmed that both desktop and laptop versions of these GPUs will be available. The GeForce RTX 5090, RTX 5080, and RTX 5070 Ti Laptop GPUs will launch in March, with the RTX 5070 Laptop GPU arriving in April.

If you’re interested, keep an eye on major laptop manufacturers, including Acer, ASUS, Dell, GIGABYTE, HP, Lenovo, MECHREVO, MSI, and Razer. Some of these brands are already listing models with the latest NVIDIA GeForce RTX 50-series GPUs for preorder.

back to top