Nvidia’s RTX 5080 graphics card will supposedly be loaded with faster video memory than expected, in a new twist in the tale of rumors being spun around Team Green’s next-gen GPU.
Wccftech has heard from sources (engage skeptical mode now) who claim that the RTX 5080 is set to be equipped with 16GB of GDDR7 VRAM (as previously rumored), but that this RAM will be running at 32Gbps, a faster speed than the 28Gbps memory that speculation suggested in the past. (Samsung has 28Gbps and 32Gbps GDDR7 modules available initially, as you may recall).
With a 256-bit memory bus and 32Gbps VRAM, as Wccftech points out, the total bandwidth in this case will (just) exceed the magic 1TB/s mark (1024GB/s to be precise). To put that in perspective, the RTX 4080 offers a total memory bandwidth of 736GB/s.
Another interesting wrinkle here is the assertion that the RTX 5090 is set to use 28Gbps memory, but it’ll have a way wider memory bus at 512-bit, meaning its total bandwidth will still far exceed the RTX 5080 as you’d expect (it’ll be closing on double, in fact – plus the 5090 will supposedly offer 32GB of VRAM).
Eventually, the likelihood is that the RTX 5090 will also use 32Gbps video RAM, but initially, this is going to be the domain of the RTX 5080 – assuming this info from the grapevine is correct (and all the other past rumored specs, for that matter – add seasoning as ever).
Analysis: Memory matters – some compensation from Nvidia?
Okay, so assuming this is on the money, why would Nvidia make this VRAM speed choice for the RTX 5080, then? Well, in case you didn’t notice, there’s already been quite the reaction to the (purported) memory loadout with the RTX 5080. Firstly, with the choice to use that 256-bit bus (sticking with the same as on the RTX 4080, and not upgrading there), and more recently, the 16GB of VRAM configuration has left many PC gamers displeased.
Nvidia likely anticipated this, and so perhaps this is Team Green’s way of compensating for these shortcomings to some extent, and ensuring overall memory bandwidth is suitably peppy. At 1TB/s, or just over, you can’t really complain – this represents an almost 40% generational uplift for video memory speed.
Another positive bit of chatter is that while the 16GB loadout of RAM has been widely viewed as Nvidia cheaping out on video memory – once again – there could be a 24GB version of the RTX 5080 launched down the line.
Worries still remain about the RTX 5080 overall, though, with its rumored core count representing a big drop compared to the RTX 5090. The former will have fewer than half the CUDA Cores of the next-gen flagship, supposedly, a larger gulf than seen in the RTX 4080 pitted against the RTX 4090 (where the difference is more like 40%, rather than 50%).
Are we going to get a nerfed xx80 series graphics card for the Blackwell generation? That’s what some are suggesting, but in all honesty, such speculation is rather meaningless until we know what Nvidia’s pricing plans are for the RTX 5080. Is the idea to angle the GPU a bit more affordably than the RTX 4080?
The trouble is, though, it’s difficult to envisage Team Green pinning a more reasonable price tag on a 5080 given past form – it’s just feels like this is seriously wishful thinking. Still, all we can do is wait and see. In theory the RTX 5090 and 5080 could be unleashed together, with an initial revelation at CES 2025 (we might even see Blackwell laptop GPUs, too).