DRAM, or dynamic random-access memory, is the type of random access memory that “stores each bit of data in a separate capacitor with an integrated circuit”. DRAM is the main type of memory found in desktops, laptops, workstation computers, and even some video game consoles. It carries the advantage of being simpler than other options, while allowing for more dense levels of memory. Billions of transistors and capacitors can fit in a single memory chip, and each are placed on top of each other, not side by side. Although it is efficient, DRAM does use large amounts of power. Having said ALL of that, and thanks Wikipedia for the rundown, our good friends at Samsung have come out with a new type of DRAM, dubbed the HBM2 module. Said to power through data “at the rate of 256GBps, these new chips are up to seven times faster than any DDR5, a type of DRAM, or graphics random-access memory, with a double data rate bandwidth”.
According to An And Tech, Samsung’s new high-bandwidth memory (HBM) technology will “substantially increase bandwidth available to computing devices (GPUs) and reduce power consumption”. This is the second generation of HBM, as the first (found in AMD’s Fury graphics cards) had some issues related to actual capacity and clock-rates. These new HBM2 stacks are based on the 20nm architecture, and offer twice the gigabytes per second, 256GB/s, that the first generation of HBM allowed (first gen HBM allowed for just 128GB/s). Moreover, current HBM-equipped GPUs will be able to hold as much as 16GB memory, a huge advantage over the 4GB shown in AMD’s version of the process. Getting even more technical, Samsung’s device consists of four 8Gb chips (or 4GB stacks total), which, with said 256GB/s of bandwidth, attach to the 1024-bit interface at 1.2 volts. The company is already planning on producing its own 8GB stacks, based on the notion that stacks can be two, four, or eight chips high (Ars Technica).
HMB2 being brought to graphics will mean a world of difference for our devices. Bandwidth will be faster, all while fixing that issue of power efficiency. Think VR. One thing Virtual Reality headsets need is high frame rates. It also would help for that pair of goggles attached to your head to not get hot from it sucking at the battery too fast. Let us not forget how much faster and better next gen devices are going to be if we also take into account GPU’s transition from 28nm to 14nm.
Despite the fact that this is an incredibly brand-new solution, Samsung obviously wants to be ahead of the pack by diving right in. Soon, we will see HBM2 technology used by graphics card makers all over, as well as big supercomputer and data centers; and smartphones and tablets will surely enjoy the amount of space saving we’re talking about. Samsung may be the first to come out and shout their plan from the rooftops (they are already in mass production), but to no surprise, AMD is already on it (as far as updating its current HBM devices), and Intel is surely expected to integrate the technology on their own chips.
Read Related Articles
- NVIDIA Announces Two New Graphics Cards to Fit Any Gamer’s Budget
- AMD Delivers High-End, VR-Capable Graphics with the $200 Radeon RX 480
- Sulon Q VR Headset Powered by AMD, Completely Tetherless
- AMD’s Radeon R9 Fury X: Most Say Don’t Waste Your Time with 4K Gaming
- Acer Adds Tear-Free Gaming to a Monstrously Large Curved Display