We’ve known that NVIDIA was definitely going to be unveiling new GPUs, and the CES 2025 keynote made it official with the all-new GeForce RTX 50-series for both desktops and laptops. Based on the Blackwell architecture, these new GeForce RTX 50-series GPUs boost both gaming and creative creative work with plenty of AI love (ICYMI, NVIDIA had updated the RTX badge with an additional tagline as well).
Just how much of a leap are we looking at? In the words of CEO Jenson Huang: “The RTX 5070, 4090 performance at $549“.
Beefy performance leaps on desktop
How do I even start with breaking down their CES 2025 keynote? According to NVIDIA, there are significant leaps spanning the entire RTX 50-series family that looks to leapfrog the Ada Lovelace-based RTX 40-series.
First up is DLSS 4, that looks to further build on leveraging AI for helping to generate even higher frame rates for all the games with improved details. But I’m going to save this for a separate write-up, as it’s quite a significant upgrade which also benefits previous RTX GPUs in some form.
These GPUs require plenty of transistors to help render all those pixels, and the flagship RTX 5090 has 92 billion of them. This easily makes the flagship RTX 5090 outperform the previous RTX 4090 by 2x. Pair it with fifth-generation Tensor Cores for AI workloads and fourth-generation RT cores for improved raytracing, and you’re looking at an AI-empowered silicon across every stack.
Rather than bore you with more details, I’ve nicely tabulated the specs for each SKU with relevant numbers.
| GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 Ti | GeForce RTX 5070 | |
|---|---|---|---|---|
| NVIDIA CUDA Cores | 21760 | 10752 | 8960 | 6144 |
| Tensor Cores (AI) | 3352 | 1801 | 1406 | 988 |
| Ray Tracing Cores | 318 TFLOPS | 171 TFLOPS | 133 TFLOPS | 94 TFLOPS |
| Memory config + memory Interface width | 32GB GDDR7 (512-bit) | 16GB GDDR7 (256-bit) | 16GB GDDR7 (256-bit) | 12GB GDDR7 (192-bit) |
| Memory bandwidth | 1792 GB/s | 960 GB/s | 896 GB/s | 672 GB/s |
| NVIDIA Encoders | 3 x 9th Gen | 2 x 9th Gen | 2 x 9th Gen | 1 x 9th Gen |
| NVIDIA Decoders | 2 x 6th Gen | 2 x 6th Gen | 1 x 6th Gen | 1 x 6th Gen |
| Required system PSU | 1000W | 850W | 750W | 650W |
As a standard default, the desktop versions support up to 4 multi-monitor outputs, with 3x DisplayPort 2.1b and 1 HDMI (2.1a). They all run on PCI Express Gen 5, providing even more bandwidth to help render all those details.
Packing performance in portability
The laptop versions of GeForce RTX 50 Series also get significant improvements based on the desktop versions, tailored and optimized to fit thin chassis with much better power draw:
- new Max-Q technologies (suite of AI-powered tech optimizing performance, power and acoustics for efficiency) such as Advanced Power Gating,
- Low Latency Sleep,
- Accelerated Frequency Switching increases battery life by up to 40%,
- Voltage optimized GDDR7 for optimal power efficiency with ultra low voltage states
| GeForce RTX 5090 Laptop GPU | GeForce RTX 5080 Laptop GPU | GeForce RTX 5070 Ti Laptop GPU | GeForce RTX 5070 Laptop GPU | |
|---|---|---|---|---|
| NVIDIA CUDA Cores | 10496 | 7680 | 5888 | 4608 |
| Tensor Cores (AI) | 1824 TOPS | 1334 TOPS | 992 TOPS | 798 TOPS |
| Ray Tracing Cores | 318 TFLOPS | 171 TFLOPS | 133 TFLOPS | 94 TFLOPS |
| Memory config + memory Interface width | 24GB GDDR7 (256-bit) | 16GB GDDR7 (256-bit) | 12GB GDDR7 (192-bit) | 8GB GDDR7 (128-bit) |
| NVIDIA Encoders | 3 x 9th Gen | 2 x 9th Gen | 2 x 9th Gen | 1 x 9th Gen |
| NVIDIA Decoders | 2 x 6th Gen | 2 x 6th Gen | 1 x 6th Gen | 1 x 6th Gen |
| GPU Power draw | 95-150W | 80-150W | 60-115W | 50-100W |
If you’re one of the many that prefers a powerful gaming laptop and also use various creative production apps, you’ll be running laps around work while on the move.
Creative work and Media Consumption
If you’re heavily involved in professional video work, these RTX 50 Series GPUs on both desktop and laptops deliver significant improvements in offloading heavy-duty tasks.
Generative AI performance is 2x compared to the previous RTX 40-series, while also reducing the required memory requirement. This will prove useful for those using apps that integrate it.
Each model includes a new sixth-generation of NVIDIA encoders, helping to export video much faster while also offering a 5% improvement in video quality on HEVC and AV1. Even the sixth-generation video decoder doubles the decoding speed for H.264 videos (the standard codec for decoding all those .mp4/.mov files).
To deliver all these benefits, NVIDIA has been working with the likes of Adobe Premiere Pro, Blackmagic Design’s DaVinci Resolve, ByteDance’s Capcut and Wondershare Filmora to integrate them. Expect to see app updates for them starting in February.
Streamers will also get to use enjoy new AI-powered beta effects being added to the NVIDIA Broadcast app (which has helped to turn personal spaces into studios using AI-powered features to improve audio and video quality – without even investing in expensive specialized equipment). The update is expected to be available in February, and the initial support for these effects will require a GeForce RTX 5080 or higher model; other models will be supported in future updates.
Release/Availability
Following their tradition, NVIDIA is releasing the flagship RTX 5090 and RTX 5080 near the end of January 2025 across the Middle East; this will be followed by the mid-tier RTX 5070 TI and RTX 5070 starting in February 2025. Pricing will follow from the US$ then converted to the respective local currencies from the third-party OEMs
| GeForce RTX 5090 | GeForce RTX 5080 | GeForce RTX 5070 TI | GeForce RTX 5070 | |
|---|---|---|---|---|
| Price | US$ 1,999 | US$ 999 | US$ 749 | US$ 549 |
If you’re looking to buy the Founders Edition of the RTX 5090/5080/5070, they’ll be available from NVIDIA’s website and select retailers in the Middle East (likely the official partners who sell their GPUs in each country).
Laptops featuring the RTX 50 series are expected to be available for purchase starting from March. I believe they’ll push it to April, given that it’ll be Ramadan and Eid is the best time when people will definitely be buying them.
Image credit: NVIDIA
