Information about Intel’s Arc GPUs, which will be manufactured by TSMC
Ghana Education News | August 21, 2021, | More information about Intel’s Arc GPUs, which will be manufactured by TSMC
Intel’s Arc GPUs – New gaming GPUs
Intel unveiled Arc earlier this week, the name for the new gaming GPUs that will compete with Nvidia’s GeForce and AMD’s Radeon cards when they ship in early 2022. Today, Intel revealed a few more information about Arc, including its underlying Xe-HPG architecture, graphics drivers, and the “XeSS” upscaling technology, which will operate with both Intel and Nvidia and AMD GPUs.
The HPG in Xe-HPG stands for “high-performance gaming,” and it builds on Intel’s 11th-generation Tiger Lake laptop processors’ Xe-LP (“low-power”) graphics technology and the low-end DG1 dedicated GPU. Xe-HPG, like Nvidia’s RTX 2000 and 3000-series GPUs and AMD’s RDNA2-based Radeon RX 6000-series cards, will feature all of the capabilities in DirectX 12 Ultimate at a substantial stage.
Support for hardware-accelerated ray tracing, variable rate shading, and mesh shaders are among the highlights of the new Intel gaming GPUs.
Intel has been referring to its GPU hardware as “execution units,” or EUs, since at least 2006. The moniker “Xe-HPG” is being phased out in favor of the concept of the “Xe-core,” according to the corporation. Each Xe-core contains 16 vector engines, 16 matrix (or XMX) engines, as well as an L1 cache and additional hardware.
A “render slice,” made up of four Xe-cores, ray-tracing units, and other fixed-function hardware, is the very minimum any Xe-HPG GPU will require to work (along with L2 cache and a memory interface).
Alchemist-based chips can have “up to” eight of these render slices, implying that we’ll see at least a few distinct Arc GPUs with varying compute capability. However, Intel has not revealed any precise hardware configurations, nor has it provided any information on clock speeds, memory interfaces, or the amount or type of RAM that the initial Arc GPUs will have.
(A quick refresher on jargons worth knowing: Xe-HPG is the name of the GPU microarchitecture, Alchemist is the codename for Xe-HPG-based chips, and Arc is the consumer branding you’ll find on shop shelves.)
Intel’s AI-accelerated upscaling technology, also known as Xe Super Sampling or XeSS, will rely heavily on the XMX hardware in Xe-HPG GPUs. The goal, similar to Nvidia’s DLSS, is to upgrade lower-resolution pictures with little quality loss. As a result, a game may be played at 4K resolutions without requiring the graphics horsepower required to generate a native 4K image.
If you use Intel’s XMX hardware, you’ll get the greatest picture quality and performance out of XeSS, according to Intel. A different version of XeSS is available that leverages the DP4a instructions found in contemporary Nvidia and AMD GPUs. Intel didn’t name particular GPU models, although DP4a instructions have been found in Nvidia’s still-popular GTX 1060 and 1050 series cards since the Pascal architecture.
Arc GPUs will utilize the same driver package as Intel’s integrated GPUs when they are launched, so any enhancements made to one design may help all of them (Intel boasted that it had overhauled its graphics driver in the past year to improve output in CPU-bound games running on Xe-LP GPUs by between 15 and 80 percent).
Finally, there’s the big news as we talk about the : Intel isn’t currently producing Alchemist chips in its own chip facilities. Intel’s GPUs will have to compete for capacity with AMD, Apple, and everyone else producing cutting-edge chips at TSMC, thanks to the company’s announcement that it will utilize TSMC’s 6 nm process for Alchemist chips.
READ: Crowdsourcing platform Toloka rolls out 20,000 new jobs
Read: Pokémon Go Zacian raid guide: Best moveset and counters
“The reason [for using TSMC for Xe-HPG] is simple,” wrote Intel SVP Stuart Penn inan accompanying press release. “Just as our designers use the right architecture for the right workload, we also choose the node that best fits that architecture. At this point in time, these foundry nodes are the right choice for our discrete graphics products.”