Whether the 42.5 Gbps GDDR7 module will end up on graphics cards or not, is still a technological breakthrough for Samsung.
Samsung to present its ultra-fast GDDR7 memory at ISSCC, which is roughly 77 percent faster than the flagship GDDR6
The ISSCC special events will take place in California from February 16 to February 20, in San Francisco, where chip manufacturers will showcase their flagship products. The event will include a dedicated "Nonvolatile Memory and DRAM" event that will be held on February 19. Even though it won't be a public event, the schedule is available to the public on the ISSCC website.
At the event, Samsung is about to present its flagship GDDR7 DRAM, which can run at a whopping 42.5 Gbps speed. It will be a 24 Gb or 3 GB module that is aimed at the best possible GDDR7 performance, which will also be more power efficient than its predecessors. GDDR7 DRAM is already quite faster than GDDR6 but at the moment, a 42.5 Gbps module is too high to be released for GPUs.
Still, it could make its way into future products, such as NVIDIA's RTX 60 series cards. Right now, NVIDIA is going to equip its Blackwell RTX 50 series GPUs with up to 28 Gbps of GDDR7 modules. The speed is already 4 Gbps higher than the flagship 24 Gbps GDDR6 memory speed present on the RTX 4090. However, the 42.5 Gbps memory speed is roughly 77% higher than the 24 Gbps GDDR6 DRAM.
Unfortunately, we didn't see 24 Gbps of GDDR6 memory speed on most GPUs, including the flagship RX 7900 XTX and XT GPUs from AMD and even NVIDIA's flagship RTX 4090 GPU. These modules run at 20 Gbps and 21 Gbps memory speeds, respectively. Hence, 42.5 Gbps may not necessarily happen even on future GPUs.
Right now, NVIDIA's flagship Blackwell GeForce RTX 5090 GPU will utilize 28 Gbps memory, which could be improved to over 30 Gbps with the future generations, just as we saw the evolution of GDDR6 modules. The RTX 5090 is reportedly going to feature 32 GB of GDDR7 memory on a 512-bit memory bus, which results in 1.7-2.0 TB/s of memory bandwidth. This is already 1.7x to 2.0x higher than the RTX 4090, and theoretically, if the VRAM on the RTX 5090 could reach 42.5 Gbps, the memory bandwidth would come close to 2.5 Gbps.
The ISSCC Nonvolatile and DRAM event will also host a presentation by SK Hynix, which will showcase its state-of-the-art 321-layer 4D NAND, which recently entered the mass production state.
News Source: ISSCC
About the author: Sarfraz Khan is a hardware reporter with a focus on PC components and the builder community. With years of experience writing about PC hardware and laptops, his work has been featured on several reputable technology publications. Sarfraz's hands-on experience is demonstrated through his first-person accounts of using and comparing different hardware configurations, providing practical and relatable insights for everyday users. His technical analysis is respected by peers in the enthusiast community and has been cited by specialized hardware sites such as Germany's Igor's Lab.
Follow Wccftech on Google to get more of our news coverage in your feeds.
