Samsung Displays GDDR7 Memory Modules At GTC: Designed For Next-Gen Gaming GPUs, 16Gb Densities & 32 Gbps At First

NVIDIA & AMD To Utilize 16 Gb Dies on First-Gen GDDR7 GPUs, 2 GB Memory Capacity & 32 Gbps Speeds 1

While HBM memory such as HBM3E dominates the AI data center landscape, it's GDDR7 that will be powering the majority of the client segment including gaming GPUs & Samsung has shown its next-gen memory at GTC.

Samsung is going to be one of the key suppliers of the next-generation GDDR7 memory standard which is going to be the staple for next-gen gaming GPUs. The company announced development back in 2023 and will soon be talking about even faster variants though the one that is going to be the starting point of this new standard will stick with 32 Gbps speeds.

The initial GDDR7 memory modules from Samsung will come in 16 Gb densities with each module offering 2 GB capacities.

The speeds will be set at 32 Gbps (PAM3) and can be optimized down to 28 Gbps to enhance yields and the overall performance per cost during its infancy. Samsung also states that GDDR7 memory will offer 20% better power efficiency while operating at a low voltage of just 1.1V (lower than the 1.2V standard). Other features included reduced heat output in high-speed operations thanks to newer packaging materials and an optimized circuit design which is going to reduce thermal resistance by 70% compared to GDDR6.

JEDEC recently published the specifications for GDDR7 memory with the main highlights including:

Core independent LFSR (linear-feedback shift register) training patterns with eye masking and error counters to improve training accuracy while reducing training time.

Doubles the number of independent channels doubles from 2 in GDDR6 to 4 in GDDR7.

Support for 16 Gbit to 32 Gbit densities including support for 2-Channel mode to double system capacity.

Address the market need for RAS (Reliability, Availability, Serviceability) by incorporating the latest data integrity features including on-die ECC (ODECC) with real-time reporting, data poison, Error check and Scrub, and command address parity with command blocking (CAPARBLK).

Although there are much faster specifications for GDDR7 in the works, those will come in time and only by 2025-2026. That brings us back to the 28 Gbps GDDR7 memory dies which have been cited as the product of choice for NVIDIA's next-gen GeForce RTX 50 "Blackwell" Gaming GPUs. So far, the rumors have mentioned that the top three dies will utilize GDDR7 memory which means that entry-level SKUs might stick with GDDR6(X) for now and move to GDDR7 later on as it becomes a viable alternative.

512-bit / 28 Gbps / 32 GB (Max Memory) / 1792 GB/s (Max Bandwidth)

384-bit / 28 Gbps / 24 GB (Max Memory) / 1344 GB/s (Max Bandwidth)

256-bit / 28 Gbps / 16 GB (Max Memory) / 896.0 GB/s (Max Bandwidth)

192-bit / 28 Gbps / 12 GB (Max Memory) / 672.0 GB/s (Max Bandwidth)

128-bit / 28 Gbps / 8 GB (Max Memory) / 448.0 GB/s (Max Bandwidth)

One thing that we have noticed with the GDDR6 and GDDR6X memory modules is that overclocking them has become very easy. You can easily set these chips at much higher frequencies manually & they work like a charm.

The first GDDR7 graphics cards are probably going to arrive later this year as production on the new memory ramps up. It will be a new beginning for the whole client graphics memory market and we look forward to how it progresses.

News Source: HardwareLuxx