Samsung GDDR6W Graphics Memory Unveiled: Doubling The Capacity & Performance With Increased Densities & 64-Bit DRAM - Wccftech

Samsung has conscionable unveiled its next-generation graphics representation known arsenic GDDR6W which doubles the show & capableness versus GDDR6 DRAM.

Samsung GDDR6W Memory For Next-Gen Graphics Solutions Unveiled: 64-Bit I/O, 4 GB Capacity Per DRAM

Press Release: As precocious graphics and show technologies develop, they are blurring the lines betwixt the metaverse and our mundane experience. Much of this important displacement is being made imaginable by the advancement of representation solutions designed for graphics products. High-bandwidth Graphics Memory Solution: The Key to Hyper-real Gaming and Digital twin One of the biggest challenges for improving virtual world is taking the complexities of real-world objects and environments and recreating them successful a virtual space.

Doing truthful requires monolithic representation and accrued computing power. At the aforesaid time, the benefits of creating a much true-to-life metaverse volition beryllium far-reaching, including real-life simulations of analyzable scenarios and more, sparking innovation crossed a fig of industries. This is the cardinal thought down 1 of the astir fashionable concepts successful virtual reality: the integer twins. A integer duplicate is simply a virtual practice of an entity oregon space. Updated successful real-time successful accordance with the existent environment, a integer duplicate spans the lifecycle of its root and uses simulation, instrumentality learning, and reasoning to assistance decision-making.

While until precocious this was not a feasible proposition owed to limitations connected information processing and transference, integer twins are present gaining traction acknowledgment to the availability of precocious bandwidth technologies. Like different tech innovations, the gaming manufacture thrives connected changeless innovation, with caller updates successful velocity and show driving the marketplace guardant twelvemonth aft year. Thanks to the improvement of technologies similar Ray Tracing successful 3D rendering, which traces the reflection of airy successful a fixed scene, graphics successful high-end AAA gaming are becoming hyper-realistic and progressively immersive. Ray tracing enables the postulation of airy accusation to find the colour of each pixel done real-time calculation.

This benignant of calculation requires the near-simultaneous computation of important amounts of information — betwixt 60 to 140 pages worthy for 1 2nd of an in-game scene. What’s more, show prime is rising fast, with resolutions rapidly transitioning from 4K to 8K standard, portion framework buffers are expanding to grow 2 times much than existing ones successful response. That’s wherefore precocious capableness and precocious bandwidth are indispensable to gathering the increasing representation request arsenic games proceed to develop. 

_{Frame buffer: Memory that temporarily stores representation accusation that appears connected the screen.} Developing 'GDDR6W' Graphics Memory, with Doubled Capacity and Performance Based connected the Cutting-edge Fan-Out Wafer-Level Packaging (FOWLP) Technology High performance, precocious capacity, and precocious bandwidth representation solutions are helping bring the virtual realm to a person lucifer with reality. To conscionable this increasing marketplace demand, Samsung Electronics has developed GDDR6W (x64): the industry’s archetypal next-generation graphics DRAM technology. GDDR6W builds connected Samsung’s GDDR6 (x32) products by introducing a Fan-Out Wafer-Level Packaging (FOWLP) technology, drastically expanding representation bandwidth and capacity.

Since its launch, GDDR6 has already seen important improvements. Last July, Samsung developed a 24Gbps GDDR6 memory, the industry’s fastest graphics DRAM. GDDR6W doubles that bandwidth (performance) and capableness portion remaining the identical size of GDDR6.

Thanks to the unchanged footprint, caller representation chips tin easy beryllium enactment into the aforesaid accumulation processes customers person utilized for GDDR6, with the usage of the FOWLP operation and stacking technology, cutting manufacturing clip and costs.

As shown successful the representation below, since it tin beryllium equipped with doubly arsenic galore representation chips successful an identical size package, the graphic DRAM capableness has accrued from 16Gb to 32Gb, portion bandwidth and the fig of I/Os person doubled from 32 to 64. In different words, the country required for representation has been reduced by 50% compared to erstwhile models.

Generally, the size of a bundle increases arsenic much chips are stacked. But determination are carnal factors that bounds the maximum tallness of a package. What’s more, though stacking chips increases capacity, determination is simply a trade-off successful vigor dissipation and performance. In bid to flooded these trade-offs, we’ve applied our FOWLP exertion to GDDR6W. FOWLP exertion straight mounts representation dice connected a silicon wafer, alternatively of a PCB. In doing so, RDL (Re-distribution layer) exertion is applied, enabling overmuch finer wiring patterns. Additionally, arsenic there’s nary PCB involved, it reduces the thickness of the bundle and improves vigor dissipation.

The tallness of the FOWLP-based GDDR6W is 0.7mm – 36% slimmer than the erstwhile bundle with a tallness of 1.1mm. And contempt the spot being multi-layered, it inactive offers the aforesaid thermal properties and show arsenic the existing GDDR6. Unlike GDDR6, however, the bandwidth of the FOWLP-based GDDR6W tin beryllium doubled acknowledgment to the expanded I/O per azygous package. Packaging refers to the process of cutting fabricated wafers into semiconductor shapes oregon connecting wires. In the industry, this is known arsenic a 'back-end process.'

While the semiconductor manufacture has continuously developed towards scaling circuits arsenic overmuch arsenic imaginable during the front-end process, packaging exertion is becoming much and much important arsenic the manufacture approaches the carnal limits of spot sizes limits. That’s wherefore Samsung is utilizing its 3D IC bundle exertion successful GDDR6W, creating a azygous bundle by stacking a assortment of chips successful a wafer state. This is 1 of galore innovations planned to marque precocious packaging for GDDR6W faster and much efficient.

The recently developed GDDR6W exertion tin enactment HBM-level bandwidth astatine a strategy level. HBM2E has a system-level bandwidth of 1.6TB/s based connected 4K system-level I/O and a 3.2Gpbs transmission complaint per pin. GDDR6W, connected the different hand, tin nutrient a bandwidth of 1.4TB/s based connected 512 system-level I/O and a transmission complaint of 22Gpbs per pin. Furthermore, since GDDR6W reduces the fig of I/O to astir 1/8 compared with utilizing HBM2E, it removes the necessity of utilizing micro bumps. That makes it much cost-effective without the request for an interposer layer.

● System Level: Using graphics cards arsenic an example, mounting notation points of 8 bundle GDDR6W and 4 bundle HBM. “By applying an precocious packaging exertion to GDDR6, GDDR6W delivers doubly the representation capableness and show of similar-sized packages,” said CheolMin Park, Vice President of New Business Planning, astatine Samsung Electronics Memory Business.

With GDDR6W, we’re capable to foster differentiated representation products that tin fulfill assorted lawsuit needs – a large measurement towards securing our enactment successful the market."

Samsung Electronics completed the JEDEC standardization for GDDR6W products successful the 2nd 4th of this year. It has besides announced that it volition grow the exertion of GDDR6W to tiny signifier origin devices specified arsenic notebooks arsenic good arsenic caller high-performance accelerators utilized for AI and HPC applications, done practice with its GPU partners.