Samsung pronounces new GDDR6W reminiscence, rivals HBM2 | Tech Lada

In context: As producers proceed to squeeze each final drop of efficiency out of in the present day’s GDDR6 and GDDR6X reminiscence modules, Samsung has introduced a brand new and improved entry to the household: GDDR6W. Samsung claims that GDDR6W can compete with the bandwidth and speeds of HBM2.

In 2016, Samsung and different producers started making the successor to quick (however flawed) high-bandwidth reminiscence modules (HBMs). Excessive Bandwidth Reminiscence 2 (HBM2) seemingly fastened all the issues with the earlier era, rising capability, speeds, and bandwidth. Sadly, HBM2 by no means noticed vital success within the desktop graphics card market.

The Fury and Vega traces of playing cards used HBM and HBM2, respectively. Sadly, each flopped and AMD went again to GDDR6 reminiscence once more, beginning with the RX 5000 line. Some customers have been understandably upset by the fast abandonment of HBM2.

On Tuesday, Samsung revealed its newest addition to the GDDR6 household, GDDR6W. The South Korean tech large needed to convey a few of the advantages of HBM2 to the already profitable GDDR6 platform, particularly the elevated bandwidth. Based mostly on the small print and numbers offered by Samsung, GDDR6W may very well be a sport changer in future GPUs.

Samsung positioned a heavy emphasis on digital actuality and “Metaverse” functions. Nonetheless, there is not any purpose why GDDR6W will not convey advantages to future discrete graphics playing cards usually.

Samsung began by taking its current GDDR6 platform and implementing what it calls “fan-out wafer-level packaging” (FOWLP). As an alternative of putting reminiscence dies on a PCB, they’re mounted on to a silicon wafer. The redistribution layers permit for “finer wiring patterns” and since there aren’t any PCBs concerned, the modules will likely be thinner total and have higher warmth dispersal.

The reminiscence modules are double-stacked, which Samsung notes can allow double the capability and double the reminiscence bandwidth. The reminiscence restrict is 24GB per die for traditional GDDR6, so in concept the brand new platform may help 48GB on a single graphics card.

These modifications in module placement and total die measurement lead to GDDR6W being 36% shorter than its GDDR6 counterpart. As a result of unchanged footprint, these modules may very well be “put into the identical manufacturing processes” utilized in present GDDR6 merchandise.

As seen above, the bandwidth of GDDR6W may be very near that present in HBM2E dies. The present GDDR6X bandwidth restrict is round 1TB per second, and GDDR6W considerably exceeds it by round 400MB/s.

Whereas we nearly definitely will not see GDDR6W on any Lovelace or RDNA3 graphics card, there’s at all times the possibility that Nvidia or AMD will go head-to-head and undertake GDDR6W in a subsequent era. Regardless, Samsung’s new brainchild is a large step ahead for graphics processors, whether or not discrete or “for the Metaverse.”