On the graphics card front, things are getting a little more hectic, with GDDR7 memory due to arrive by the end of the year. This means that Intel, AMD or NVIDIA should be able to use them from then on. However, don’t expect greatly improved memory capacities. The first generations of chips will retain a density of 16 Gbit!
GDDR7: no, capacities won’t explode… Not at first!
If you find that your 16 GB graphics card lacks VRAM, that’s not going to change with the new chips. In fact, according to Kopite7kimi, graphics cards will initially use 16 Gbit memory chips. In other words: no improvement in memory capacities, 16 Gbit = 2 GB density, which is already in use today.
What’s interesting, however, is the mention of chips with asymmetrical capacities in 24 Gbit, i.e. 3 GB. Roughly speaking, with a 128-bit memory bus, like the RTX 4060, we could have a VRAM capacity of 12 GB. On a card with a 256-bit bus, like the RTX 4080, capacity could reach 24 GB, the same as an RTX 4090!
Finally, in terms of speed, the first chips, those available in late 2024 or early 2025, will offer speeds of 32 Gbps. This contrasts with the 20 Gbps of current GDDR6. So, if they don’t have more memory, they’ll be faster!
Of course, as time goes by, speeds and capacities will improve. The pro sector could then benefit from chips offering densities of 32 Gbit (4 GB) or 64 Gbit (8 GB), but that will be for cards designed for AI, for example. As for speeds, remember that the first cards equipped with GDDR6 saw their memory run at between 14 and 16 Gbps… And that’s not counting NVIDIA, which could, in collaboration with a chip manufacturer, develop GDDR7X, for example.
