Spinning off graphics (ATI, the best tech/IPs in the company) to alleviate the debt of the processor segment (AMD, the worst performing segment and outright uncompetitive) is a short term fool's errand IMO. AMD trashed ATIs position in the market so unless someone like Samsung, Intel, Apple, or MS are picking up ATI the spun off entity is doomed to decline. And AMD's processor offerings... uninspiring.
I don't think there is much hope for the company. I think their best market play is HBM with a midrange graphic performant APU. Basically a PS4 level PC with 8GB of high speed memory on an interposer. AMD would hit a number of check boxes cut out a lot of middle men.
# CPU. Solid, but not a world beater. NV and ARM still are not competitive in the massive x86 space so the goal would be to carve out solid margin segments from Intel. As AMD cannot best Intel in sheer x86 for a solid margin the CPU offering is only "price of entry" to support Windows. The compelling aspect has to come from...
# GPU. If given time Intel with eDRAM will blunt a major move by AMD with HBM. AMD needs to move. With a solid GPU on die AMD raises the bar for the non-discrete APU market. Offerings with NV require a discrete GPU which cannot be cost competitive compared to an APU and Intel isn't there (yet) in performance. OEMs are averse to pushing volume models with GPUs as it adds costs parents/students are rarely willing to splurge for as extras. But having a performant GPU "SOC" gives AMD a price advantage and "performance/features" win that Intel cannot match at this time. With 4K being the new "thing" AMD is in a great place to attack the must have feature with significant priceerformance Intel and CPU+NV machines cannot match.
# Memory. By going with HBM AMD also captures some of the memory margin. The key is hitting 8GB of memory. If 4GB is too far off they should look at tiered memory (4GB HBM, 4-8GB of cheap system memory).
In total, if properly executed, AMD could offer system designers and builders a CPU+GPU+MEMORY bundle the competition cannot match in price/performance. AMD would capture the margins on the CPU+GPU+MEMORY and the 'package' would be a compelling product category. Pitched at the $600-$1000 desktop market and the $800-$1300 laptop market AMD could capture volume sales.
Being able to have an affordable solution that essentially has dGPU performance and best in class 4K experience would make a compelling sales point against the cheaper Desktop/Laptop offerings that cannot compete in performance. And against more expensive models with dGPU this offering has the advantage on price.
Cheap PCs << AMD HBM
AMD HBM >> CPU + dGPU
Big picture AMD has a compelling offering no one can match right now at their price. They can truly console level performance and best in class 4K at a great price. They are able to bundle/pitch the CPU/Memory/GPU and create buzz in a boring market segment.
If AMD cannot leverage ATI's technology for a compelling offering AMD has nothing of significance to be competitive long term. They are fabless. They have an x86 license that only allows them to compete for table scraps against Intel. Intel kills AMD in serve CPUs, big/fast CPUs, and small/power conscious x86. ARM has the low power segment covered.
The ATI IPs are their best option for a hail marry.
Selling them off pretty much tells me AMD is raising the flag and going to go silently into night without much of a struggle. A HBM APU, on the other hand, scratches a market itch and would offer something new and better not seen in the space for a long time.--for those reasons AMD will screw the pooch.
Maybe Intel will buy ATI ...