Nvidia RTX 5080 allegedly adopts faster 30 Gbps GDDR7 modules, delivering 960 GB/s of bandwidth — The remaining Blackwell lineup is anticipated to stick with slower 28 Gbps memory
Rumors counsel that Nvidia will launch the RTX 5080 with 30 Gbps GDDR7 memory, whereas the remaining Blackwell lineup will comply with slower 28 Gbps modules, by means of Benchlife as seen by Harukaze on X. Do bear in mind that this moreover consists of the flagship RTX 5090. Nonetheless, given the substantial disparities in CUDA core rely and totally different specs between the RTX 5090 and RTX 5080, a small memory nerf is unlikely to make a distinction to the flagship.
Nvidia’s RTX 50 assortment is anticipated to be the first GPU lineup to debut with GDDR7 memory. GDDR7 touts an a variety of benefits over its predecessor similar to bigger speeds and higher density. Upcoming GDDR7 decisions from Samsung promise speeds of as a lot as 42.5 Gbps in 24Gb (3GB per module) capacities. Mainstream and funds GPUs bear from slim bus widths that prohibit VRAM capacities to a paltry 8GB or 12GB (128-bit/192-bit) using 16Gb modules. Whereas newer 24Gb modules can theoretically improve these numbers as a lot as 12GB and 18GB respectively, don’t anticipate availability until 2025 or later, likely inside the kind of a SUPER refresh.
With that context in ideas, Benchlife’s report implies that Nvidia will use slower 28 Gbps GDDR7 modules all through all its Blackwell assortment of GPUs, other than the RTX 5080. Like its predecessor, the RTX 5080 is rumored to personal the quickest memory in all of the lineup, this time at 30 Gbps, allowing it to dish out 960 GB/s of bandwidth if we assume a 256-bit memory interface. Earlier leaks counsel the RTX 5080 will ship with 16GB of VRAM and that configuration is simply attainable with 16Gb modules and a 256-bit interface, so our calculations stand fixed.
Together with to the mix, AMD’s RDNA 4 allegedly makes use of a barely conservative memory setup with 18 Gbps GDDR6 modules. Likewise, Intel’s Battlemage B580 adheres to a 12GB configuration that features 19 Gbps GDDR6 memory. Nvidia’s dominant place on the market offers them the leverage to value a reasonably penny for the latest tech, along with GDDR7 VRAM.
Yesterday, a leak from Zotac confirmed that Nvidia is prepping to unveil the RTX 5090, RTX 5080, and RTX 5070 assortment at CES subsequent month. What this implies is that the RTX 5060 lineup could very effectively be pushed to Q2 25, aligning its launch with the availability of 24Gb modules (a minimum of from Samsung). This might allow for bigger capacities, as a lot as 12GB for the RTX 5060 nevertheless that’s wishful contemplating.
