Early Tuesday morning, Nvidia revealed a new generation of GPUs with the Blackwell architecture during the opening presentation at CES 2025. We already wrote about desktop graphics, but at the same time the company also published GeForce RTX 5000 graphics models, which will be available for purchase in notebooks. These are based on the same architecture and the same chips (with the exception of the largest GB202), so like the desktops they will offer DLSS 4 with interpolation of several frames.
GeForce RTX 5090 (notebooks)
The most powerful notebook model of the GeForce RTX 5000 will be a model called the GeForce RTX 5090. However, the notebook version is essentially a different graphics (the time when Nvidia relied on the fact that GPUs for notebooks would supposedly not be slower than for desktops is long forgotten), based on chip GB203. So it is closer to the RTX5080 desktop model.
The chip has active 10,496 shaders (82 SM blocks). However, Nvidia does not specify the frequency of the GPU. This will be much lower in load than the desktop GeForce RTX 5080, because the notebook version will have consumption set in the range of 95–150 W. The performance will vary according to this setting, in any case it will be significantly lower than the desktop RTX 5080, as it has a multiple times higher consumption of 360W TDP (not to mention the 5090, which has more than 2x more units and approx. 4x consumption).
One of the most interesting things is that this graphic has 24 GB memoirs GDDR7. This is not because it has wider than 256-bit memories (256-bit width is a feature of the GB203 GPU), but because GDDR7 chips with a capacity of 24Gb will be used. The desktop RTX 5080 as well as other RTX 5000 graphics could also have a 50% higher capacity with them, but for now it seems that they will only be used in this one mobile model. Unfortunately, Nvidia does not disclose the memory frequency (and therefore bandwidth). However, it will almost certainly be above 640 GB/s (which would be equivalent to 20.0 GHz).
GeForce RTX 5080 (notebooks)
The GeForce RTX 5080 model should still include the GB203 chip, as evidenced by the memory size GDDR7 – 16 GBwhich requires 256-bit memory. However, the chip has significantly reduced computing units to yen 7680 shaders (less than the desktop 5070 Ti), i.e. 60 SM. Consumption should be adjustable between 80–150 Wwhich again in practice means a considerable dispersion of performances.
GeForce RTX 5070 Ti (notebooks)
The notebook GeForce RTX 5070 Ti should probably already have a chip one number smaller than the desktop model, as 5888 shadersit should therefore be the GB205 GPU otherwise used in the desktop RTX 5070. This also corresponds to the fact that this graphics only has 12 GB memoirs GDDR7i.e. a 192-bit bus. Consumption can be set between 60–115 W. Against the desktop RTX 5070 Ti (300 W, 8960 shaders), but also the desktop RTX 5070 (250 W, 6144 shaders), the performance will be much lower again.
GeForce RTX 5070 (notebooks)
The RTX 5070 continues this trend, where mobile graphics are somewhat more modest in nature than their names suggest. It has active 4608 shaders (36 SM). These should already be the parameters of the GB206 chip, which we will probably see in the desktop GeForce RTX 5060 Ti. This is also related to the fact that this GPU only has a 128-bit bus, and only on it 8 GB memoirs GDDR7. The consumption can again be set to a large extent (50–100 W), but even at the higher end the performance will probably be a long way from the 250W desktop RTX 5070.
PCI Express 5.0, DisplayPort 2.1b
While Nvidia doesn’t reveal the frequency (nor the memory frequency), we have confirmation that all these GPUs should be able to use PCI Express 5.0. There is a difference in multimedia acceleration between individual GPUs. The same generation of codecs as the desktop versions is present, but the two cheaper models have only one encoder and one decoder each. Up to the GeForce RTX 5080 has two encoders and two decoders, and the GeForce RTX 5090 adds a third encoder (two encoders remain). These are new generations, but it seems that they can’t handle new formats (VVC acceleration is missing, which only the Intel Lunar Lake / Core 200V processor still has). Nvidia only added YUV 4:2:2 chrominance subsampling support for some formats. This encoding is used by some professional camera outputs.
But even these mobile graphics should enable image output via DisplayPort 2.1b, like desktop cards. HDMI is only supported in version 2.1b, as expected graphics cannot be updated yet, just a preliminary announcement of HDMI 2.2. Which output laptops will have out of the chassis, however, depends on the manufacturer. Therefore, it is not guaranteed that you will always be able to use DP 2.1 with these graphics.
The first laptops in March
Notebooks with these GPUs will begin to appear on the shelves for the first time in March – these should be the first models with three more powerful graphics. The GeForce RTX 5070 should be available later, with the first laptops only in April. This is probably another effect of the fact that it contains a chip from the desktop RTX 5060 Ti, which has a later release and will therefore probably also be ready for release later.
Source: Nvidia
Source: www.cnews.cz
