[CES 2025] NVIDIA Announces GeForce RTX 50 Series, Launching January 30th

NVIDIA has just announced the GeForce RTX 50 series powered by Blackwell architecture, and beyond that there’s a significant amount of new technologies presented to further improve gaming performance, and applications in non-gaming scenarios. There is truly a lot of things to go through, and to keep things still readable, we’ll be making this part 1 of the NVIDIA GPU launch coverage, focusing on the hardware first.

Meet The NVIDIA GeForce RTX 50 Series

The new Blackwell-powered RTX 50 series features four new models: the flagship RTX 5090, then RTX 5080, RTX 5070 Ti, and RTX 5070. The new Founders Edition cards feature a slightly different look but maintains the same general shape, though internally it is now packing a pair of flow-through fans instead of the push-pull setup seen in the previous generation.

To achieve that, the PCB has to be shrunk down to the size of a palm, and somehow NVIDIA is able to contain all 575 watts of power in a GPU just two slots thick. Yes, you’re looking at a GPU that guzzles even more power than the RTX 4090, yet still managed to be smaller (at 304mm long and 137mm wide) and be technically SFF-compatible per NVIDIA’s standards. Wild stuff.

Specs

So, specs: the RTX 5090 packs a whopping 21,760 CUDA cores, 680 Tensor Cores, 170 RT cores, paired with 32GB of GDDR7 memory via a 512-bit bus that can deliver a theoretical maximum bandwidth of 1.8 terabytes per second. As mentioned, it consumes up to 575 watts of power, and NVIDIA recommends 1000W PSUs to keep it fed – all that power will be sent through the 12V-2×6 connector, with 4×8-pin adapter still available if your PSU is not ATX 3.1-compliant.

For display outputs, all GPUs in this lineup can now display up 4K 480Hz 12-bit HDR via DisplayPort 2.1b + DSC (digital stream compression), or 8K 120Hz 12-bit HDR via both DisplayPort and HDMI 2.1a (also with DSC). For multi-monitor setups, they can power up to four 4K 165Hz monitors concurrently, or two 4K 360Hz (or 8K 100Hz) monitors using either outputs.

Besides the uber-fast RTX 5090, there’s also the RTX 5080 that comes with just 10,752 CUDA cores with 16GB GDDR7 VRAM while consume 360 watts at full tilt. Despite that, it shares the same card dimensions as the RTX 5090 FE at 304mm long and 137mm wide. Meanwhile, the RTX 5070 Ti features 8,960 CUDA cores, 16GB GDDR7, and 300W TGP – this will only be available in AIB variants; the RTX 5070 has 6,144 CUDA cores, 12GB GDDR7 memory, and 250W TGP. The FE card also measures smaller, at 242mm long, 112mm wide, and 2 slots thick.

These GPUs will be supported by new technologies including DLSS 4 that enables multi-frame generation, Reflex 2 which further cuts down on input latency, and upgrades in encoding/decoding capabilities to support more efficient encoding and more simultaneous playbacks.

Performance Claims

As usual, take the first-party numbers with a good degree of skepticism, but here’s what NVIDIA’s numbers say. All RTX 50 GPUs, when compared to RTX 40 counterparts, are 2x faster – with the big caveat being the game in question supports DLSS 4. This new version of DLSS has a new RTX 50-exclusive featured called Multi Frame Generation (MFG), which generates three interpolated frames in between natively-rendered frames. In conditions where all equal features are enabled (RT or DLSS 3), all RTX 50 GPUs lead their counterparts by roughly 30-40% based on the charts shown.

RTX 50 Laptops

NVIDIA has also announced that laptops with RTX 50 series GPUs are coming as well, with prices starting from $1,299 for RTX 5070-powered laptops (the lowest-end model announced so far). From the top end, the RTX 5090 Laptop GPU has 10,496 CUDA cores and 24GB of GDDR7 memory; this is followed by RTX 5080 with 7,680 CUDA cores + 16GB GDDR7, RTX 5070 Ti with 5,888 CUDA cores + 12GB GDDR7, and RTX 5070 with 4,608 CUDA cores + 8GB GDDR7 memory.

Availability & Pricing

NVIDIA said the RTX 5090 ($1,999 / RM10,390) and RTX 5080 ($999 / RM5,190) will both be available starting January 30th; while both RTX 5070 Ti ($749 / RM3,890) and RTX 5070 ($549 / RM2,850) will be available sometime in February. All laptops featuring RTX 50 series GPUs, meanwhile, will be available starting this March, with the exception of RTX 5070 laptops – these will be available starting this April.

Pokdepinion: Ignoring all the AI stuff, the raw performance uplift seems fairly limited especially given the increased power draw.