Samsung announces new GDDR6W memory to compete with HBM2

[ad_1]

In the context of: As manufacturers continue to squeeze every last drop in performance out of existing GDDR6 and GDDR6X memory modules, Samsung has announced a new and improved entry to the family—GDDR6W. Samsung claims that GDDR6W can compete with the bandwidth and speeds of HBM2.

In 2016, Samsung and other producers began manufacturing the successor to the fast (but flawed) High-Bandwidth Memory Modules (HBM). High Bandwidth Memory 2 (HBM2) seems to have fixed all the issues with the previous generation, increasing capacities, speeds, and bandwidth. Unfortunately, HBM2 was not very successful in the desktop graphics card market.

The Fury and Vega lines of cards used HBM and HBM2, respectively. Alas, all of them failed, and AMD returned to GDDR6 memory again, starting with the RX 5000 lineup. Some users were understandably disappointed by the quick abandonment of HBM2.

On Tuesday, Samsung open Its latest entry into the GDDR6 family, GDDR6W. The South Korean tech giant wanted to bring some of the advantages of HBM2 to its already successful GDDR6 platform, particularly the increased bandwidth. Based on the details and numbers provided by Samsung, GDDR6W could be a game-changer in future GPUs.

Samsung places a heavy emphasis on virtual reality and “Metaverse” applications. However, there is no reason why GDDR6W should not bring benefits to future discrete graphics cards in general.

Samsung started by taking the existing GDDR6 platform and implementing what it calls Fan Chip Level Encapsulation (FOWLP). Instead of placing the die memory on the PCB, it is installed directly on a silicon wafer. The redistribution layers allow for “fine wiring patterns”, and since there is no PCB involved, the modules will generally be thinner and feature better heat dispersion.

The memory modules are double stacked, which Samsung notes can allow for twice the capacity and double the memory bandwidth. The memory limit is 24GB per standard GDDR6 die, so the new platform can theoretically support 48GB on a single graphics card.

These changes in module placement and overall die size cause GDDR6W to be 36% shorter than its GDDR6 counterpart. Since the footprint does not change, these modules can be “placed in the same production processes” used in existing GDDR6 products.

As we saw above, the bandwidth of the GDDR6W is very close to that of the HBM2E. The current bandwidth limit of GDDR6X is about 1TB per second, and GDDR6W is greatly higher than 400MB/s.

While we almost certainly won’t see GDDR6W on any Lovelace or RDNA3 graphics card, there’s always the chance that Nvidia or AMD could take a swing at the other and build GDDR6W into the next generation. Regardless, Samsung’s new creation is a huge step forward for graphics processors, whether discrete or “for the Metaverse.”

[ad_2]

Source link

Related Posts

Precaliga