Ever since Intel and AMD confirmed that they’d be bringing a new chip to market with an Intel CPU and AMD GPU, there’s been speculation about where we might see these parts. The higher power consumption on AMD’s Vega, relative to a simpler, lower-end integrated GPU, suggested the AMD+Intel solution might be confined to small desktops or larger laptops. Now, there’s at least a chance that we’ll see the chip in a socket form factor, or at the very least, that it might be used a bit more broadly.
Leaked roadmaps from earlier this year suggested that Intel’s Hades Canyon platform would include two SKUs — a 66W part with an integrated Vega GPU and a locked clock speed, and a 100W TDP part with a “K” suffix (visible below, barely, at the top line) marketed as being VR-capable and sitting at the top of Intel’s product stack.
New data from Intel’s overclocking website (screenshotted in case they remove it) now shows a new, unlaunched CPU — the Core i7-8809G — as being on Intel’s list of overclocking chips. The i7-8809G is a quad-core/eight-thread CPU with a base clock of 3.1GHz, and a (rumored) boost clock of 4.1GHz. The 4GB HBM2 buffer on the Core i7-8809G is rumored to be clocked at 800MHz vs. the 700MHz on the (not listed) Core i7-8705G that supposedly fills in the bottom of the roadmap.
If we plug these figures in, the Core i7-8705G would have 180GB/s of memory bandwidth for its GPU, while the Core i7-8809G would offer 204GB/s. That’d put the upper-end GPU on the 8809G in range of the full-size RX 470 (211GB/s). The rest of the chip is firmly anchored to Kaby Lake, not Coffee Lake, with a quad-core CPU, eight threads, and formal support for DDR4-2400 as opposed to DDR4-2666.
What we’ve seen at this point suggests that earlier rumors were on-track. A 3.1GHz base/4.1GHz turbo clock on an unlocked quad-core CPU with a 100W target TDP matches the specs that leaked in early November. If this chip proves to have the 24 CUs it’s supposed to pack, that works out to an on-die GPU with 1,536 GPU cores. No, it won’t take on a $ 200+ midrange GPU, but it could easily compete with hardware below that point, and a quad-core/eight-thread CPU will have few problems with gaming workloads that suit its on-board GPU. Also, interestingly, this CPU will apparently ship with both Intel integrated and AMD’s on-board GPU solutions, which could make this the first CPU to support hooking up six monitors out of the box.
I suspect there’s a great deal of curiosity from Intel and AMD alike about whether customers will bite on this sort of GPU configuration. Integrated graphics occupies a tricky spot in company product stacks. Consumers want the advantages an on-die GPU delivers, particularly in mobile, where the power savings are most welcome. But attempts to create a specific premium market for iGPUs haven’t had good results, and there are fundamental limits to how much power you can pack into an integrated GPU when it’s sharing a socket with the CPU. If these Intel chips take off, it could spur new interest in the product category from both companies.