Intel have debut some new GPU-adjacent designs at the International Solid-State Circuits Conference (or ISSCC for short), but before you get too excited this announcement isn’t the result of Intel’s new graphics division condensing years of work into only a few months, it’s purely an unrelated research project on GPU power efficiency.
Intel have plenty of competition for high-end GPUs. Here are the best graphics cards around.
Intel are expected to one day put together some form of discrete GPU since they acquired AMD’s top GPU-mind. Raja Koduri may have joined Intel to head up the newly-formed Core and Visual Computing Group, but even for an experienced GPU architect, three months on the job may be asking a little much to prototype an entire functioning GPU architecture.
The announcement from ISSCC came via PCWatch, who rightfully reported that this prototype GPU isn’t going to be on the market as a standalone product anytime soon. The design is built upon the 14nm fabrication process, and features largely the same technology that Intel currently utilised within their current iGPU designs. At 1.2V, the GPU would reach a blistering speed of… 400MHz.
The design features 18 of Intel’s Execution Units (EUs), and totals just over 1.5 billion transistors in total. For a little reference, AMD’s Vega 64 has 12.5 billion transistors and Nvidia's GTX 1080 Ti 12 billion of the little 16nm suckers.
You can take a look at the prototype GPU schematics right here if you like intense block diagrams.
The chip’s function, however, wasn’t to unleash this underwhelming iGPU-fueled graphics power on the unsuspecting world. No, it was actually meant to show greater granular voltage control and power efficiency applied to a GPU, leading to greater power efficiency for their own iGPU chips.
“Last week at ISSCC, Intel Labs presented a research paper exploring new circuit techniques optimizes for power management”, Intel say in a statement to Liliputing, “The team used an existing Intel integrated GPU architecture (Gen 9 GPU) as a proof of concept for these circuit techniques. This is a test vehicle only, not a future product. While we intend to compete in graphics products in the future, this research paper is unrelated. Our goal with this research is to explore possible, future circuit techniques that may improve the power and performance of Intel products.”
Intel still have a long way to go with any discrete GPU products, and that’s probably a good thing - we don’t want another Larrabee, after all. The GPU design published at ISSCC is pretty low-power and low-performance, and I don’t know about you, but I’d hope that when Intel do launch a discrete graphics product it packs a little more punch than this.
When Intel do announce their own discrete GPU - however far away that may be - I imagine they would see fit to organise their own event to do so, rather than let slip their entire design at an electrical engineering conference in San Fran.