Nvidia GeForce RTX 50-series
The biggest, best, most exciting products in technology – reviewed and rated
AI and Multi Frame Generation are the stars of this release rather than the silicon, but these alone give reasons to buy
Looking for a particular past review? Visit our index at pcpro.link/index
RTX 5080
SCORE
PRICEFounders Edition, £816 (£979 inc VAT)from nvidia.com
RTX 5090
SCORE
PRICEFounders Edition, £1,616 (£1,939 inc VAT)from nvidia.com
W
ith a trio of major architectural changes, you can see why Nvidia kept us waiting almost two years for an update to its RTX 40-series. First up is its move to a 4nm process from 5nm, while it has also made the jump from PCIe 4 to PCIe 5 (see “Why PCI Express rules the world”, p40).
But the biggest hardware change the new Blackwell architecture brings is GDDR7 video memory – and it’s far faster than GDDR6 and GDDR6X. For example, the RTX 5090 has a roughly 33% increase in Gbits/sec over the RTX 4090. Add the much wider 512-bit memory interface and you have a memory bandwidth of 1,790Gbits/sec for the 5090 compared to 1,010Gbits/sec for the 4090. The new RTX 5070 Ti and RTX 5070 cards will reap similar benefits.
This, together with an enormous 32GB of VRAM, means the 5090 is the first 8K-capable graphics card on the market. 8K textures have an enormous footprint in memory, so moving them through the rendering pipelines to generate playable frame rates wasn’t feasible until now – unless you used aggressive upscaling. The Nvidia RTX 5080’s memory also has a much faster effective speed of 30Gbits/sec compared to the 23Gbits/sec of the RTX 4080 Super and 22.4Gbits/sec of the RTX 4080. Again, that matters when it comes to rendering complex scenes at high resolutions: you get that much more headroom before memory becomes a brake on performance.