I heard that AMD plans to bring a 16-core chip to laptops at CES, and there’s also a plan to include 3D V-cache in these CPUs.
Will this laptop brand fall behind again?
The potential of this company is huge, and it could shake up many others (for example, by implementing cooling kits), and even introducing a PCI Express 5.0 port (I’m not sure if they already have it). Sometimes I consider buying something here, but honestly, when I see what’s available in the store, my enthusiasm drops due to the mediocrity of their catalog (processors from 3 years ago, what’s the point of having an upgradable laptop if there are no parts to upgrade?).
But by any chance, are there plans to change this this year?
Which kernel are you using?
Which BIOS version are you using?
Which Framework Laptop 16 model are you using? (AMD Ryzen™ 7040 Series)
These AMD 16 core are the fire range chips and comes with no iGPU so I doubt Framework will select them for the framework 16.
The most obvious choice are the AMD strix point chips that are already available.
Yes you are correct, there is also Strix Halo with 16 cores. However this one is designed to not have a dGPU as the included iGPU is fast enough for most usage. I suppose it would work as well even if it’s more expensive because made of several chiplet.
There is also Kraken Point which is limited to 8 cores.
Strix Point fits right in between these two with 12 cores.
There have been leaks about two 16-core laptop chips from AMD: Ryzen AI Max+ 395 and Ryzen 9 9945HX3D.
The Ryzen AI Max+ 395 is a CPU+GPU combo unit with a very fast integrated GPU . There are 3 problems with using this CPU in a FWL16:
Combining a high power CPU and high power GPU into a single chip means a lot of heat (and therefore a lot of cooling needed) for that chip. The FWL16 is designed to have separate cooling for the CPU and GPU. This CPU would either be bottlenecked by the FWL16’s cooling system or require major design changes (ex. to combine the cooling systems into one) which would probably break compatibility between generations of FWL16 components.
This CPU is expected to have significantly fewer PCIe lanes (since AMD isn’t expecting a separate GPU to be connected). This would reduce the overall PCIe bandwidth available, and as a reduce how much can be done with Framework’s modular expansion bay.
Combining a big/expensive CPU and a big/expensive GPU into a single unit means that if you want to replace or upgrade one then you actually need to replace/upgrade both, which goes against Framework’s missions of repairability and upgradeability.
The 9945HX3D on the other hand has its own problems, mainly that it is actually the same chips as Ryzen 9000 desktop except repurposed into a laptop form factor, which comes with the following consequences:
Desktop CPUs (and by extension the 9945HX3D) typically have higher power draw (due to different design decisions and manufacturing processes), especially under light loads (such as web browsing). As a result chips that were designed for desktop but repurposed into laptops tend to have much worse battery life.
The 9945HX3D has a very slow integrated GPU. This makes the FWL16’s modular expansion bay less versatile due to more users needing the dGPU module in order to get satisfactory performance (and only one expansion bay module can be used at a time, so other modules like the new SSD bay can’t be used).
The 9945HX3D has fewer features. It lacks integrated USB4 support and it lacks an NPU (I don’t care about the NPU, but some people do).
So both the Ryzen AI Max+ 395 and the Ryzen 9 9945HX3D are expected to have compromises. Framework might offer one of them (personally I think the 9945HX3D is more likely, although they previously stated that they didn’t do the prior 7945HX3D for some of the same reasons as I listed above), or they might not. We don’t know (those CPUs haven’t even been announced by AMD, an announcement by Framework likely wouldn’t be until a little while later).
The current processors that Framework offers were first announced by AMD exactly 2 years ago, not 3.
Strix halo will be great, powerful iGPU means the expansion bay space will be more useful, and I don’t need carrying a fxxking heavy gpu with me in my college life. And current the cpu part thermal design can actually hold ~70w which is ok for me.
The standby power consumption of 16 core is enormous, the battery life will actually be a disaster. Strix halo with 8 core and 32 cu which tdp at 70w will be a great one.
But I believe the best way is provide both 9945 for upgradable and performance and 8 core strictly version for daily using( although it is impossible)
Ryzen AI Max+ 395 and the Ryzen 9 9945HX3D are 2 different options for me.
9945HX3D need a dGPU => more for gaming laptop.
395 => do not need a dGPU so the expansion bay is usable for more ssd.
For power, I do not think that the Ryzen AI Max+ 395 need more power than the actual CPU+dGPU => so not an issues for the battery.
For termal dissipation, Strix halo with 8 core and 32 cu may be usable without cooling re-design . for the hight Ryzen AI Max+ 395 it may be needed to have a new cooling .
My interest for Strix halo is the memory. It look to be a dual LPCAMM2 so ~ quad channel and 128+Go of fast RAM. It may be really good for big local LLM. So can be a good developer laptop with local coding help and more… and compete with Apple M3/M4
But even the “small” 8 core and 32 cu can be nice for that.
I don’t even remember when the (beefy 45-watt) laptop CPU was not enough for me to do my job or game comfortably. Sure, new ones are faster, and so you can optimize your processes a bit, play games at a better framerate, but are those upgrades essential? I’d say, not. Add to that that most people don’t upgrade every year, meaning that they are comfortable with 2-3-4 year old CPUs no problem.
Also, early adoption as a customer in recent times is a tedious and dangerous endeavor. BIOS and drivers always require fixes after stuff releases, early hardware can also have issues fixed only through warranty, and you are a guinea pig used to find issues and then to test the fixes for them. You’re especially screwed if you’re experiencing something rare - the manufacturer might not get to your thing even a year after release. And no, big co’s like Dell are not better at it. I think they’re worse because Framework at least notices if people are being loud enough on the forum. With Dell, unless there’s a lot of negative press coverage about an issue, they are basically fine.
Framework themselves stated that they release products with a delay to let the bigger fish work out kinks in initial design, so for me waiting is not that big of a deal. I even think I haven’t waited long enough to get the FL16 as I’ve still got the one with liquid metal. And I can’t install Windows 24H2 on mine due to a bug in AMD drivers…
more likely kracken point due to the non-stupid price compared to strix point… imo framework wouldve announced something if they were releasing strix point laptops
kind of… but a few more subtle differences too.
if they continue the past tradition the difference will be…
HX will have a full number of PCIe lanes reachable to any dGPU. higher clocks and more TDP directed to your CPU with only a basic RDNA2 onboard graphics. Since its a FW you will likely be able to even go from an rDNA 3.5 based GPU to a full RDNA4 based GPU.
x95 will have half the lanes set aside for a dGPU it pulls off several lanes to the more robust iGPU that should be able to complete with the 4060 at a minimum. Since the rumored mobile ‘S’ series dGPUs are RDNA 3.5 based you wont be loosing out on much… but if you want all of the RDNA 4 niceties (maybe in an RDNA 4.5) refresh you could theoretically leaving some performance on the table. (CPU thermals for HX likely superior, greater bus width, etc in return for a more tightly controlled [but more shared like you pointed out] power package)
So it kind of depends what your looking at. If you think your platform will be stable for 4 years and only upgrade dGPUs in the slot the HX path seems better. If you want to upgrade every 2 years… you might not get that much better frames per watt and save a little money doing the ‘MAX’ approach. Unless your going to do really one sided workloads (e.g. heavy CPU and minimum GPU) you might not even see double digit % differences between the two platforms.
Added note: I just remembered the top seat mainboards Framework likes to go with are often part of the ‘pro’ series for enterprise management and other features. Those don’t get released for several more months. (April ship times are normal for the enterprise SKUs these days for notebooks)
Regardless of what is released by intel/amd/nvidia the point remains that the 16" 7040 AMD Ryzen CPUs released 2 years ago, it’d be nice to see new hardware. The point of these laptops was repair ability that could potentially be upgradable, so new hardware again would be nice. I personally don’t want to buy a 2 year old PC for “full price” when I can spend the same amount of money or less for a semi-repairable <6 month old PC or buy something that is being offloaded for less.
Because at the end of the day whatever I buy will probably be with me for about 7-10 years.
The 7000 series AMD processors have hit a performance wall with the 5600 Sodimm ddr5 ram. Framework will have to design boards with soldered on ram to see generational improvements and in doing so make their products less upgradeable. Even now with AMD’s newest Ai 300 series, ddr5 isn’t fast enough.
8000 series saw no benefit with 5600 speeds and every other laptop manufacturer had to make their computers with soldered on ram to show a visible upgrade from last generation.
This might be the Achilles heel of attempting to make a company with upgradeable laptops possible. Technology doesn’t advance in predictable ways.
No need for soldering RAM, there is now LPCAMM2 that start a LPDDR5X-7500 up to 8000 (an more?)
And the new Ryzen AI Max now use quad channel (dual for actual 7840HS.)
Both can make RAM speed x3 (and may be more).
: AMD Ryzen AI MAX+ Pro 395 Strix Halo