- Nvidia’s neural texture compression function became just more effective
- It got this lift thanks to Microsoft’s new cooperative vector in DXR 1.2
- Testing with a demo showed almost a reduction of 90% in video RAM -Footprint for textures, but this does not apply in all scenarios (in any way)
Concerned your NVIDIA graphics card can be underpowered with regard to video RAM (VRAM) it has on board? You may be aware that Team Green is working on tech to mitigate such concerns, and it seems to be taking significant progress in that regard.
WCCFTECH reports that Nvidia’s RTX Neural Texture Compression (or NTC) has received a big boost thanks to Microsoft’s new cooperative vector in Directx Raytracing 1.2.
Does that sound like a lot of techie -wrap? Well, yes, and if you dare look at the in -depth explanation of what’s on foot here, you’ll probably be very lost in a country with confusion -see the post on X below from Osvaldo Pinali Doederlein (who tested the feature with a Prer Lease graphics driver).
See first look at NVIDA’s neural texture compression with DXR1.2 Cooperative Vector! First, this needs a preview driver (590.26), I installed that you do not have the broken screen, only after a few hard resets decided to work.July 15, 2025
So, in short, Nvidia’s texture compression law, which makes it possible to make textures less, which means that more can fit into a modest help with VRAM, now works much more effectively thanks to the cooperative vector.
As Doneederlein observes, in the demo he ran as a test, NTC offered a saving of almost 90% in reducing VRAM footprint, which is a massive deal. If and this is an important warning, the game supports NTC (so developers must code it in, as is the case with NVIDIA DLSS, of course).
Given that textures generally occupy half of VRAM in a given gaming session (or a lot more), the benefits should be translated into a significant victory for games where grabs textures outside the video memory becomes a task that slows everything.
Analysis: The Wider Image
NTC uses AI (neural network, and thus the name: Neural texture compression) to intelligently compress graphics structures – in theory without much of a noticeable difference from native structures when decompressed to show to the player – and as mentioned, Microsoft’s cooperative vectors in DXR 1.2, this process ensures more effective.
In fact, much more effective, and I think NVIDIA had an eye on how this kind of technology would move down the line as Team Green made decisions to implement some current Gen Blackwell graphics card and stick to 8 GB video RAM.
It has been a very unpopular decision in some quarters, so maybe Nvidia – and counts vigorously – on the NTC to prove a big difference in future -proofing some of these GPUs. (Such as RTX 5060 TI 8GB and Vanilla RTX 5060 – RTX 5050 also has 8 GB, but this memory configuration is more understandable on a budget graphics card).
That said, we need to be pretty careful about relying on some early results in a single demo as a reason to celebrate – as promising as they can look. The number of a (close) 90% reduction will not be the case in each scenario you can knock on it, but even more modest boost (to the tune of, for example, 50%) will still be good news.
The second point to keep in mind is that the NTC could have a side effect of a certain performance of the image velocity – at least in certain scenarios. Hopefully, any such image speed decreases – or effects on the image quality for that matter – be mild and presumably worth exchanging in making life easier for vram (otherwise what would be the poenget?).
Although I give more cautious sounds here, as already mentioned, textures are not everything in terms of what is packed on VRAM in a game session, and then wider concerns remain around the future of 8 GB graphics cards in that light.
Still, the NTC could be an important piece of the puzzle for Nvidia, and it looks more at these test results. As you might guess, this technique for Team Green’s graphics card, and only the more modern offers on it. (While NTC in theory can be used by rival GPUs, such as AMDS Radeon boards, realistic NVIDIA recommends that only its RTX 4000 or 5000 graphics cards have the necessary OOMPH and hardware to run it).
So where does it leave AMD and Intel’s products? These GPU producers have their own neural compression technology on the boil, and the good thing is that they will also benefit from Directx Cooperative Vectors (everyone will). Although Nvidia right now has the lead in this department – but we are still a fair way from seeing this technique realized. As noted, game developers need to support the technique in their titles -this is not a driver -level function that can be used across all games. (Well, not unless NVIDIA is planning a diluted version at a later date – a kind of equivalent of what smooth movement is for DLSS).
In short, Nvidia has some persuasions to do, again very much in the same way as DLSS, because if developers do not adopt this, it is a non-starter. However, the sales points appear to be seriously convincing.
