NVIDIA GeForce RTX 50: New memory technology and bus interface for a leap in performance

NVIDIA’s next-generation gaming GPU, the GeForce RTX 50 “Blackwell”, is expected to offer support for GDDR7 memory, claims leaker Kopite7kimi. His latest revelations thus indicate that the NVIDIA GeForce RTX 50 flagship gaming GPU will offer GDDR7 memory support. The NVIDIA GeForce RTX 50 series is expected to utilize the Blackwell gaming GPU architecture, albeit slightly different from the one coming out next year for HPC and AI customers. There are currently roadmaps pointing to a 2025 launch, but some rumors suggest we may see the latest graphics lineup in late 2024. Christmas sales, top models and skimming the cream..,

The next generation of gaming graphics cards will require equally advanced memory technology with significantly more bandwidth. In this context, reference is made to the memory configuration which, according to Kopite7kimi, will be based on the latest GDDR7 memory standard. The GDDR7 memory will replace the GDDR6 and GDDR6X standard with even faster pin speeds and higher capacities. The first chips will use up to 24 Gb modules and speeds of up to 32 Gbps. These will debut in 2024. An even faster revision is planned for 2026, but it is unlikely that NVIDIA would wait that long for the latest memory technologies. Instead, these would be used by a new or updated product family. Great, isn’t it?

Below is the bandwidth that the 32Gbps pin speeds would offer in various bus configurations:

• 128-bit @ 32 G bps: 512 GB/s
• 192-bit @ 32 Gbps: 768 GB/s
• 256-bit @ 32 Gbps: 1024 GB/s
• 320-bit @ 32 Gbps: 1280 GB/s
• 384-bit @ 32 Gbps: 1536 GB/s
• 512-bit @ 32 Gbps: 2048 GB/s

Although the exact memory configurations for the NVIDIA GeForce RTX 50 flagship GPU are not mentioned, it is rumored that the card will have a 384-bit bus interface. It therefore appears that 512-bit is no longer on the planning list or is only intended for an even higher-end variant. The 384-bit memory bus suggests that we will get either the same 24 GB of VRAM or 36 GB depending on the DRAM density used. At a speed of 32 Gbps, the 384-bit bus interface would offer a bandwidth of up to 1.5 TB/s, which is an increase of over 50% over the existing RTX 4090 GPUs.

So Kopite7 thinks its 384-bit bus and GDDR7 too, which matches my rumor

So that's about 1.5TB/s bandwidth, depending on R7 clocks. https://t.co/YDP3TWo0cb

— RedGamingTech (@RedGamingTech) November 15, 2023

Similar Founders Edition design as the leaked “4090 Ti” prototype

Kopite7kimi claims that the flagship NVIDIA GeForce RTX 50 graphics card will have a Founders Edition design similar to the leaked “GeForce RTX 4090 Ti” that has surfaced several times before. The graphics card features quad-slot cooling with a unique side-mounted PCB design and power cables routed through large copper connectors around the body. If the same design is used for the RTX 50 GPUs, we can definitely expect design updates in the final model compared to the prototype solution seen below.

The information also mentions that the GB202 could be released as early as next year. The GB203 could use a 256-bit bus interface, while the GB204 and GB205 GPUs are rumored to be mutually exclusive. Yesterday, RedGamingTech also highlighted similar specs, which have now been confirmed by Kopite. It seems that things are really shaping up for NVIDIA’s next-gen lineup, although the company is also preparing its GeForce RTX 40 “SUPER” refresh, which is set to debut at CES in January 2024 and launch the same month if all goes according to plan. We definitely shouldn’t expect anything official on NVIDIA’s RTX 50 GPUs at this early stage, but we can expect more updates and leaks as we get closer to the next generation of the gaming family.

Source: YouTube