Intel Unveils Process Roadmap, 10nm Shipping in June 2019, 7nm in 2021 – Intel 7nm Xe GPUs For Datacenter in 2021, 10nm Xe Consumer GPUs in 2020
Intel has just unveiled their latest process roadmap which confirms the launch of their 10nm node based products in 2019 and introduction of their next-gen 7nm node based products in 2021. The company has not only revealed their process roadmap but also unveiled actual product lines that would be based on the new process nodes.
Intel's Big Game Plan Officially Unveiled: 10nm To Target TSMC 7nm in 2019, 7nm To Target TSMC 5nm in 2021
The new information comes from Intel's 2019 Investor Meeting webcast during which Intel showed off their latest process node roadmap which gives us a glimpse of their strategy for future product lineups based on 10nm and 7nm process nodes.
Intel 10nm, 10nm+ and 10nm++
Starting off with the 10nm family, Intel has clarified that their 10nm process node can deliver some major enhancements to performance per watt. Compared to 14nm++, the first iteration of 10nm is shown to be a good leap in efficiency and Intel plans to provide enhanced variants of 10nm moving forward with 10nm+ planned for 2020 and 10nm++ planned for 2021. Some of the major upgrades that 10nm would deliver include:
- 2.7x density scaling vs 14nm
- Self-aligned Quad-Patterning
- Contact Over Active Gate
- Cobalt Interconnect (M0, M1)
- 1st Gen Foveros 3D Stacking
- 2nd Gen EMIB
10nm Process Technology: Intel’s first volume 10nm processor, a mobile PC platform code-named “Ice Lake,” will begin shipping in June. The Ice Lake platform will take full advantage of 10nm along with architecture innovations. It is expected to deliver approximately 3 times faster wireless speeds, 2 times faster video transcode speeds, 2 times faster graphics performance, and 2.5 to 3 times faster artificial intelligence (AI) performance over previous generation products.
While we are talking about 10nm, Intel has assured that their first 10nm product, codenamed Ice Lake is shipping in June. The first die shot of the Ice Lake chip has also been released, showing off four cores, Gen 11 GPU and more. The chip is expected to feature approximately 2x the graphics performance, 2.5-3X AI performance, 2x video encode and 3x wireless speeds compared to existing Coffee Lake Refresh based processors.
Intel's 10nm wouldn't be restricted to just Ice Lake mobile platforms and Lakefield SOCs. The 10nm node would be utilized in multiple products throughout 2019-2020 which would include Xeon CPUs for HPC, FPGA, 5G/Networking, General-Purpose GPUs and AI inference. Intel also confirmed that they are expecting 14nm supply issues to fully resolve by Q4 2019.
Intel has also disclosed some early configurations on Lakefield SOCs with a 5W & 7W configuration featuring 5 cores and 5 threads (1x Sunny Cove Big Core and 4x Small Core), 2 x 4GB of LPDDR4X memory and a clock speeds of up to 4267 MHz.
Intel's 10nm Ice Lake Die Shot:
Intel 10nm+ Tiger Lake CPUs in 2020 For Mobility Platforms - New CPU Architecture, Xe GPU and More
Intel has also confirmed that one of their key 10nm+ products for 2020 would be Tiger Lake which is designed to feature a new core architecture that succeeds Ice Lake's Sunny Cove core. We know from Intel's own CPU core roadmap that this core architecture would be known as Willow Cove & is expected to arrive in 2020 so Tiger Lake would be one of the key processor lineup using it.
The Tiger Lake CPUs would also be leveraging from Intel Xe Graphics Engine which will be a good upgrade over the Gen 11 Graphics Engine featured on the Ice Lake CPUs. Intel is claiming a 4x boost in performance over Whiskey Lake processors that feature the Gen 9.5 graphics chip. The Gen 11 graphics features a 2x performance increase over Gen 9.5. The Intel Xe GPU would also allow for the latest display technologies and the CPUs themselves would feature next-gen I/O capability.
Overall, the CPUs would offer a 2.5-3x performance boost over Whiskey Lake processors in a 15W package. There is also a 9W variant (4+2 config) which delivers a 2x productivity boost over 5W Amber Lake in 2+2 config. Encode performance would be boosted by 4x while allowing encoding up to 8K at 60 FPS. The configuration disclosure lists down the Tiger Lake-U series with a total of 96 execution units on GT2 (Xe) GPU and a 4 core / 8 thread CPU package.
Intel 7nm, 7nm+ and 7nm++
At the same time as Intel introduces their 10nm++ products, they will also have production and launch planned for their next-gen 7nm process node. Later in this article, you will note a major product that Intel has planned to launch on 7nm process node by 2021. Intel would continue to offer optimizations of the 7nm process node with 7nm+ in 2022 and 7nm++ in 2023. Just like 10nm, 7nm would deliver a good list of enhancements over 10nm which would include:
- 2x density scaling vs 10nm
- Planned intra-node optimizations
- 4x reduction in design rules
- Next-Gen Foveros & EMIB Packaging
7nm Status: Renduchintala provided first updates on Intel’s 7nm process technology that will deliver 2 times scaling and is expected to provide approximately 20 percent increase in performance per watt with a 4 times reduction in design rule complexity. It will mark the company’s first commercial use of extreme ultraviolet (EUV) lithography, a technology that will help drive scaling for multiple node generations.
Intel also unveiled their leading-edge packaging process which is both PC-Centric (transistor scaling and monolithic integration) and data-centric (Heterogenous processes and integration). This allows Intel to scale and package their chip architectures into various designs by utilizing their Foveros & EMIB interconnect technologies.
Intel 7nm Xe GPUs For Datacenter (HPC/AI) Launching in 2021, 10nm Xe Lineup For Consumers Arriving in 2020
Intel has also confirmed that the lead product for their 7nm process node would be their Xe GPU architecture based General Purpose graphics card lineup for data centers. The new Xe architecture based GP-GPU would be built using Foveros 3D stacking architecture which means we are going to see a one-chip, single-stacked-chip harmony between the GPU, memory and interconnect. This will allow High-Bandwidth memory to be stacked right on top of the GPU die, featuring a package size that is much smaller than existing GPUs but at the same time, denser than anything we have got so far.
Intel has confirmed that their Xe architecture based 7nm graphics card would be introduced first on Datacenter markets, targeting AI and HPC workloads by 2021 which is also the shipping date of their 7nm process node.
While Datacenter would be first to use 7nm Xe GPUs, Intel's 10nm Xe GPU lineup would be making its way to the mainstream and enthusiast gaming market in 2020. The first confirmed product to feature the 7nm datacenter Xe GPUs would be the Aurora supercomputer, more on that here.