I recently read a few articles on upcoming panther lake. Confirmed stuff, not leaks.
So the top end SKUs are locked to LPDDR5 but LPCAMM is supported and might be an option. If Framework doesn’t invest in LPCAMM then the better iGPU is just totally off limits. There hasn’t been any Lunar Lake option and I don’t think there ever will be. By the same token, without LPCAMM, high end Panther Lake won’t happen.
Then there is the issue with Thunderbolt. Thunderbolt 4? Really Intel? Still relying on an integrated TB4 controller when you have created and released TB5?! Somehow, I don’t think Framework is going to put in the extra cost to make the next Intel version TB5, that goes double if they aren’t going to be using top end Panther Lake as that would be a weird pairing.
AMD isn’t much better for Thunderbolt either which is just a huge let down.
The only bright spot is that if Framework does decide to go CAMM, they might as well do it for every board going forward.
Actually, I see top end Panther Lake doesn’t have enough PCIe lanes to serve all the TB ports and a 4 lane SSD.
Great, so no top end Panther Lake. What a disappointment. Good job Intel, you played yourself.
I have a work-supplied Lunar Lake device, and I like its power efficiency. I suspect that power budgets are what limits Thunderbolt4 and too-few PCIe lanes than are needed for full-bandwidth operation.
I might also speculate about the error rate actually producing the 18A tiles, Intel needs to show they’re a success and that they’ve rebounded from the last decade of under-investing in their future silicon production processes.
I was waiting for Panther Lake since I have a ton of thunderbolt peripherals and am unwilling to risk compatibility by moving to an AMD board…that and none of the AMD boards give me a compelling enough reason to move off of the 12th gen Intel I am currently on. If Framework releases a Panther Lake board I will likely move to it, however I am beginning to suspect that I will actually be waiting for Nova Lake.
AMD is not blowing away Intel in the mobile CPU realm by any stretch of the imagination.
Intel Core 300 series looks amazing. The 18A process promises great power efficiency improvements and Intel might actually make drivers for the NPU which would really help battery use for LLM on the go. The 325 with 4 GPU cores seems especially interesting for a laptop.
I have a 7640u and I look forward to upgrading to a potential Framework 13 motherboard.
The AI 5 340 built on TSMC 4nm is compelling, but has the issue that AMD isn’t really good at drivers for ML or their iGPUs.
The Core 5 325 promises to be better across all metrics using Intel 18A process, and Intel does have proprietary software that make use of that iGPU and NPU silicon.
I do have 32GB of laptop DDR5 5600 and I plan to reuse that. I would be surprised if Framework makes CAMM modules, and it would be really expensive to upgrade to that right now.
Considering Micron has shuttered their Crucial division, I don’t think CAMM is on the table at all. Literally no reason to update to the latest AMD processors when it’s just a reheated version of last generation with bumped clock speeds. I maintain there is little logic to move to the new Panther Lake processors and regress on iGPU performance. I get nobody here advocates gaming on these things but it was nice to have the option for light gaming. Regressing there is not awesome. Same goes for reductions in thread count. It just doesn’t look good from a marketing perspective. I guess the regression in thread count is inevitable but the regression in iGPU performance stings far more to me. I’ll wait until Feb. to pick up a Panther Lake system…maybe. It’ll replace my wife’s laptop and I very much want a haptic trackpad and a VRR OLED display. Ideally the new Thinkpad X1 Carbon Aura Edition (who came up with that name…) will meet these requirements. It’s substantially more repairable to boot so I imagine it matches Framework on that aspect and I trust Lenevo to deliver better firmware far more than FW.
I’m a little confused by this statement. From what I’ve read, one of the main reasons to want a Panther Lake system would be the iGPU, which is trading blows or exceeding Strix Halo iGPU performance on the bigger Panther Lake systems.
Even setting the bigger Panther Lake systems aside though, won’t even the smaller Panther Lake outperform the current Framework 13 iGPU options? (honest question, not rhetorical).
I don’t really follow the logic here. SO-DIMM appears to be on it’s way out. More and more mobile CPUs are dropping support for it because it just can’t maintain signal integrity or timings. Sure, Gorgon Point isn’t a big deal, but it’s not like the memory requirements for future AMD chips are going to get more relaxed. They’re going to get stricter. Investing in CAMM2 seems like a no-brainer.
The two sentences of AMD and CAMM were unrelated to one another.
It appears I was in error regarding iGPU performance. 4 Xe cores has the same number of execution units as current iGPUs in FW products. With u-arch improvements, we can assume at least small improvement in iGPU. The logic of having a small iGPU makes sense when you have a dGPU but that isn’t present in the FW13. So it’s an odd pairing to use those non-X CPUs and no dGPU. FW could choose to do it, it just would further reduce competitiveness from a price/performance ratio. Which is something FW already struggles with. How much performance, battery life, etc are you willing to trade away for some benefits in repairability? The performance delta is too great for me to consider losing out on that much GPU compute for such an expensive laptop. MSI has a $1,200 laptop with an X class Panther Lake. That really doesn’t look good compared to FW.
The big iGPU is out since it requires soldered memory afaik. Not sure if LPCAMM would function as a replacement. It didn’t work out for the FW desktop so ¯\_(ツ)_/¯ .
Regarding CAMM, find me a manufacturer. CAMM was already more expensive than SO-DIMM; with the current RAMpocalypse, I don’t see costs shifting downward, I see them going even farther up. FW has already raised prices on the SODIMM modules sold, CAMM will be prohibitively expensive.
I mean devils advocate, they may as well offer a soldered sku if they were willing to do it in a desktop of all things for some reason.
I have read that panther lake supports lpcamm2 as well as soldered, but the only lpcamm2 modules I’ve ever seen that you could actually buy were crucial / micron who’ve decided to exit the client business entirely because they can sell as much as they can make to farms generating fake puppy videos at massive margins so who knows if/when we’ll ever see broad market availability. Micron leaving client might have pushed it back years.
I’ve been hoping for FW adoption for years because 5600 MT sodimms even bottleneck the 780m but the last 6 months have made me pretty pessimistic.
Yeah, I referenced that a couple comments above. I don’t think CAMM is likely at all. If Framework decides to pivot to soldered RAM…well, it’s just one more differentiating factor lost.
I think Panther lake is a bust for Framework. At least the high end SKUs.
Honestly a real shame since panther lake looks really good - honestly the first actual tempting upgrade from Phoenix.
If the only way to get that is soldered memory, I’d say just do it. They’ve already made a desktop that’s less modular, upgradable, and repairable than what existed on the market already, may as well do it in mobile where it at least makes sense (can’t easily add GPU perf with aic, low load and idle power matters a lot more because people use them unplugged).
There are preliminary testing for the 4 Xe core part. Despite being half the compute units of Lunarlake and 1/3 of the top PTO device it performs similar to the predecessor Lunar lake and AMD Strix Point competition. So it shouldn’t be slower, at least not significantly.
Yes, that’s true, hence the lower performance on the regular Framework too. You’ll still get faster graphics performance than Intel 13 and better battery life than whatever is available. FW13 will do much better.
The Iris Graphics Xe G7(which is the dual channel version, the single channel version is just HD Graphics) gets 1700-1800 in Time Spy Graphics. Core Ultra Gen 1’s graphics is nearly 2x as fast at 3000-3400 points.
The 4Xe core version(which is actually only 2/3rd of the compute units) is also getting 3400 Time Spy. So it should be relatively significantly faster in graphics. Due to architectural improvements, it’ll do better than Time Spy compared to Core Ultra Gen 1. Until Xe2(Core Ultra Gen 2 in Lunarlake), it suffered in games. Core Ultra Gen 1 is only 1.5x faster in games over Iris Xe G7.
Pantherlake also inherits the big advantage Lunarlake has over predecessor Intel and competitor AMD, which is great battery life.
Yes, PantherLake supports CUDIMM. JEDEC basically added additional clock to help with stability on higher speeds.
Intel shipped reference PantherLake boards with 6400MT/s SO-DIMMs and getting memory stable is part of the reason why PantherLake was delayed.
I personally worked on PantherLake and had to keep my best to not talk about it until CES 2026 (when embargo was lifted). It’s honestly the first x86_64 CPU that got me excited since the release of Ryzen CPUs nearly a decade ago
