One year later, the Intel-AMD Kaby Lake-G platform looks dead in the water

[Redaktion]

The future of Kaby Lake-G looks brim. Just as many have feared, the platform is looking more and more like a one-off design with no direct successor or long-lasting commitment from either Intel or AMD if the current product landscape is anything to go by.

https://www.notebookcheck.net/One-year-later-the-Intel-AMD-Kaby-Lake-G-platform-looks-dead-in-the-water.375301.0.html

[DavidC1]

I'm not surprised. I know Notebookcheck is the one that said its Nvidia's GPP crippling its prospects, but I doubt this.

The product itself is not compelling enough. It's not cheap, and despite being much more integrated, there's zero battery life gain compared to a system with a discrete GPU. To make matters worse, the CPU is a generation old with 4 cores meaning its Core i3 class. The GPU is acceptable.

I'm pretty sure if it was a Coffeelake 6-core then it might have found more markets. And they need to do something about the battery life because it sucks.

I think Intel's waiting for their own Gen 11-12 GPUs to get ready before trying. The rumors are that the future of KBL-G is Intel architecture GPUs.

[jeremy]

I'm not surprised. I know Notebookcheck is the one that said its Nvidia's GPP crippling its prospects, but I doubt this.

The product itself is not compelling enough. It's not cheap, and despite being much more integrated, there's zero battery life gain compared to a system with a discrete GPU. To make matters worse, the CPU is a generation old with 4 cores meaning its Core i3 class. The GPU is acceptable.

I'm pretty sure if it was a Coffeelake 6-core then it might have found more markets. And they need to do something about the battery life because it sucks.

I think Intel's waiting for their own Gen 11-12 GPUs to get ready before trying. The rumors are that the future of KBL-G is Intel architecture GPUs.

It's an interesting first generation product that at least shows Intel is interested in the market and has the technology, even if they don't have the GPU itself. I'd bet AMD is trying all avenues to see if TSMC or GF can replicate Intel's EMIB interposer. It makes HBM quite a bit more viable for smaller chips, since one doens't have to package a massive, fullsize silicon interposer in addition to all of the HBM stacks and the main IC itself. An effective HBM implementation would free AMD from Ryzen's DDR4 controller (since the only other alternative right now is GDDR5, and that has it's own drawbacks for laptops and socketed desktops).

[Shirley Dulcey]

Kaby Lake-G was a weird shotgun marriage between Intel and AMD. I don't expect to see another generation of that. It was mostly a proof of concept.

What I DO expect to see in 2020 is a multi chip module product using Intel's own GPU. I think there is a future for the approach because it has a potential advantage of faster communication between CPU and GPU, and because it will take up less motherboard real estate which will make it appealing for thin and light laptops.