Next AMD CPU

So which of amd’s next CPU’s are we going to get. are we going to get the new AI chips or will you wait for the Strix Halo 16 core cpu’s and launch those across all levels or do you plan on both.

I can’t tell if you’re asking what mainboards will be released in the future, which is not public knowledge, or what mainboards the forum members would be interested in buying.

Nobody outside of FW will know this. StrixnPoint will be released next month by AMD so those are the logical next step for the AMD Framework.

But we will never know until framework reveals the next motherboards

Framework has not made any announcements.

That being said, Strix Halo is expected to have power levels and heat outputs far more than the Framework Laptop 13 is designed to handle and even above what the Framework Laptop 16 is designed to handle.

The lower power variants of Strix Halo might be plausible for the Framework Laptop 16, however I don’t believe it is a great fit for the Framework Laptop 16. Strix Halo is designed to be a high-power CPU+GPU combo unit (which uses exclusively what would normally be the CPU cooling system), however the Framework Laptop 16’s expansion bay system makes a separate medium-power CPU and high-power GPU a much better fit (both the IO and cooling are designed around them being separate).

I think Strix Point is the most likely next AMD generation for the Framework Laptop 16. The Framework Laptop 13 may see a combination of Strix Point and the upcoming Kraken Point.

4 Likes

Framework said they skipped the Ryzen 8000 series mobile chips because they offered no real performance uplift over the 7000 series except for AI tasks. This year the Ryzen AI 300 chips have been detailed and they might offer enough that Framework would use them for a refresh. They offer up to 12 cores(from the maximum of 8 currently available, though it is a mix of performance and efficiency cores), up to 16 GPU cores(Up from a maximum of 12 currently available), a 16% uplift in IPC, 50 TOPS of AI performance is that is something Framework wants to start chasing, and AMD is advertising a 28 watt TDP which can be configured up or down as needed.

A 28 watt TDP chip could even make sense in the Framework 16 because it would help those use cases some have reported where their laptops are power draining while under heavy load with the current 180 watt charger.

1 Like

The thing is do they go for the AI chips or the yet to be unreleased Halo version that goes to 16 cores. also the efficiency cores just mean less level 3 cash still same processing power. very different from intel E cores.

Not exactly. On chiplet CPUs (usually desktop and server) that is indeed what AMD does. Not on monolithic CPUs (usually laptop).

For example the 7545U, which has 2 full-size Zen4 cores and 4 smaller Zen4c cores, has the same 16 MB of L3 cache as the 7640u (6 full-size Zen4 cores) as well as the 7840u/7840hs/7940hs (8 full-size Zen4 cores).

On monolithic CPUs AMD seems to follow a follow a pattern where the amount of cache is equal to the number of cores (regardless of core type) multiplied by 2 MB and then rounded up to the nearest multiple of 8 MB. All of which can be accessed by all of the cores.

The new Ryzen AI Strix Point CPUs follow that same pattern. They have 12 cores (4 full-size Zen5 cores and 8 smaller Zen5c cores), which applying that above rule calculates to 24 MB.

So on monolithic CPUs the only difference between AMD’s big and small cores is that the big cores can achieve higher clock speeds (at much higher power consumption). In the case of Ryzen 8000 series the Zen4 cores peak at 5 GHz while the smaller Zen4c cores cap out at 3.7 GHz), which resulted in an ~30% boost to Zen4’s performance in single core benchmarks.

The reason for this is because chiplet CPUs (CPUs that are internally split into multiple chips, which means the memory controller is on a separate chip from the cores) have a much more severe performance penalty from cache misses (a cache miss is where the CPU needs data that isn’t in the cache so it has to access memory). So on full sized Zen4 cores on chiplet CPUs AMD doubles the L3 cache in order to reduce cache misses. But not on monolithic CPUs (which are what most of their laptop CPUs are).