NVIDIA's RTX 5070 Laptop GPU gets a 12GB upgrade — and it's about time
NVIDIA has added a 12GB GDDR7 variant of the RTX 5070 Laptop GPU — and it did so by slipping the announcement into a driver update rather than a proper press release. The new SKU carries 50% more video memory than the existing 8GB model. For anyone shopping a mid-range gaming laptop in 2025 or early 2026, the distinction matters.
The announcement arrived inside the GeForce Game Ready Driver 596.36 release notes, per Notebookcheck — an unusual move for a GPU launch. NVIDIA cited strong demand and constrained memory supply as the rationale, and confirmed the 8GB version stays in production alongside the new 12GB configuration.
The spec
Nothing changed inside the GPU itself. The Blackwell-architecture GB206 chip is identical: 4,608 CUDA cores, 144 texture units, 48 raster operators. The extra memory comes entirely from denser GDDR7 modules — 24Gb (3GB) chips instead of the original 16Gb (2GB) ones. Because the memory bus stays at 128 bits, four of those new modules hit exactly 12GB. Bandwidth holds at 384GB/s, which is reasonable for this tier, though the narrow bus can still be a bottleneck at higher resolutions.
The value question
Systems carrying the 12GB variant are expected to ship in June 2026, with OEM partners including Lenovo Legion, MSI Crosshair, and ASUS ROG already listed. UK pricing for current RTX 5070 laptops sits around £1,299–£1,600; no official price delta has been announced for the 12GB SKU. The concern flagged by Tom's Guide is real: if 12GB RTX 5070 systems land close to RTX 5070 Ti pricing, the value proposition collapses.
The 8GB model will linger on spec sheets as a cheaper option, which creates a new shopping hazard — two laptops that look identical on a retailer shelf but differ by 4GB of VRAM. Anyone buying before June should check carefully whether they're getting the old configuration.
The practical upside beyond gaming: 4GB extra headroom helps when running local AI models, which increasingly need more than 8GB to load without cutting quality. It's a small but real improvement for a GPU that was otherwise undersupplied in memory from the start.
