These modules […] will run at up to 7467MT/s, the maximum supported by Intel® Core™ Ultra Series 3 processors.
which is simply not true, because these CPUs support up to 9600 MT/s. Other vendors sell them with (soldered) 9600 MT/s memory and there have been announcements for 96GB LPCAMM2 modules running at 9600 MT/s that target Panther Lake.
It is reasonable to assume, that:
the processor is paired with memory that matches its spec sheet
that people can upgrade to higher density and higher speed memory later
Unfortunately none of this holds true. It feels very dishonest of Framework to hide the fact, that their platform runs way slower than what can be expected. This fact is not mentioned in the FW13 Pro specs tabs. It is a known strategy of other OEMs to not mention specs, when they are bad. You’ll only find this info on the memory module page, which is an unlikely place to visit.
I’d really like to give Framework the benefit of the doubt, but I have a hard time doing so. I’d wished for transparency on such fundamental decisions. I’m not saying that this is Frameworks fault or even Frameworks decision that the memory runs at 7467MT/s, but it is Frameworks failure not to point this out. Cleary they know, and certainly they are aware that they sell to a tech-savvy audience. I’d be very disappointed if I paid extra for the X9 388H which is up to 10% (CPU) (GPU=same) faster on paper than the X7 368H, just to find out, that its GPU is actually up to 22% and its CPU 11% slower, because it is strangled by slow memory.
--
(@mods: please relabel, there doesn’t seem to be a Core Ultra 3 tag yet)
I highly suspect that the omission is intentional.
Of all other computers I have used or searched, they are either limited by Mainboard speed (FL12 and an old desktop I own), or RAM speed, most other laptops (like FL13 and other socketed RAM computer with compatible faster soldered RAM for the same CPU). The only exception I found other than FL13P is this one, a $500 super budget laptop, with 6400MT/s LPDDR5 soldered RAM, N150 CPU(supports up to 4800MT/s), but the RAM is detuned to 4000MT/s but can be restored to 4800MT/s using a custom BIOS setting. Being a super cheap laptop, cutting corners is excepted. However, on a premium “pro” laptop like FL13P that is more than four times the price, I expect the manufacturer, Framework, to make the RAM runs on 8533MT/s during the development process and not cut corners like a budget brand
If you don’t trust Framework the company, or Framework the staff, fine: move on, and go buy a big-brand laptop. Employees here should not have to wade through passive-aggressive nonsense about who is being “dishonest” or making “intentional” omissions.
For what it’s worth, I trust Framework. This may be because the CEO helps makes the product launch videos, or perhaps it’s because the folks who run the blog are exceptionally honest about how difficult trade-offs are made about passing on costs during the RAM pricing crisis. Not everyone will make the same decisions as me, and that’s fine.
Please consider using a neutral wording, even if you think you’ve found incorrect specification data on the website. It is generally healthier to prefer cock-up to conspiracy unless proven otherwise.
You should carefully read what you just wrote; You are aggressively defending Framework like they can do no wrong.
I placed my order for FW 13 Pro Batch 5 and found out about this issue only by accident, when I was looking for a deal for LPCAMM2 memory and wanted to check which memory speeds Frameworks is selling themselves.
Framework directly addressed the elephant in the room when they introduced the FW Desktop, explaining why the RAM is non-user-upgradable and why they still think it is a good idea to have this device and how FW can have an impact with it nevertheless.
Slow memory speeds are a known issue, especially on processors with high performance integrated graphics. That’s why there is coverage on this topic by tech outlets. And this is what they will look for in their tests (aside from TDP limitations).
Framework knows about this fact, but they choose (either actively or by negligence) to not disclose and address it properly. This a valid criticisms.
As for your suggestion: I am here because I want to buy a Framework, but I also want to know what I am buying and not find things out by accident. Especially when the elephant in the room is massive. We are talking about a performance impact that is greater than the gains you typically get a from generational upgrade.
As a company, we place a high value on transparency. Across all five of our product lines (and their various generations), one of our priorities is to provide detailed information and make it easy to access. We know many of our customers are especially interested in technical specifications, so we do our best to share as much relevant detail as possible.
Framework is relatively more transparent than Lenovo in this instance if we talk about 1. Where the information is (PSREF vs purchase website), and 2. How large/small the font is:
Man, talk about reading the fine prints in the PSREF.
There’s also this back in Feb. Not sure about pricing, availability, and which platform can truly make use of the 9600MT/s speed (implying overcoming the ‘limitation’).
Your statement is false. Obviously, I am asking you to mind your manners. Employees of any company are under an implicit pressure to tolerate all sorts of unpleasantness from customers, and customers sometimes exploit that; that’s sometimes the unfortunate nature of being in (pre) sales. So I am calling you out, and perhaps a bit more directly this time.
A neutral tone would have been more appropriate, because as we discover, there is a perfectly valid reason for the confusion. Please try to take this feedback in the spirit that it is intended; it is not suggesting, obviously, that questions may not be asked.
Tone can be difficult to judge, and culturally dependent. Personally, I found the initial comments to be reasonable, and your responses, especially this second one, to be insulting and unconstructive. But I come from an academic culture where direct, even overreaching criticism on topics is acceptable, whereas patronizing personal criticism is not.
Here, I think what would be particularly helpful from Framework would be not so much additional listings of the speed (though listing it in the configuration options would be helpful) but an explanation of why the speed is so much lower than might be expected. I do trust Framework enough to expect that they had reasons other than cost-cutting for the choice, and tradeoffs are something they are often up-front about, for example, with the soldered RAM in the Framework Desktop, or RAM prices, or chassis colors.
In the linked reddit thread, Destroya actually does mention that there are difficulties with speed and socketed RAM, even with LPCAMM2, and that there might be a video or blog post about this at some point. It would certainly be helpful to see that.
We may have to agree to disagree, not least on the basis that you don’t seem to be applying your maxim equally. I thus remain mostly of the same view, but I accept there may be a gentler way to ask people to moderate their tone, especially when employees are unable to.
(I should be happy to continue the meta-discussion in PMs if you’d find it interesting from a philosophical perspective, but otherwise I am wary of derailing the thread).
Thank you. I would like to see the actual RAM running speed of the Core Ultra 3 mainboard without clicking the FAQ, the same form as the FL12 to make them consistent.
I have a few questions
The Ultra 5 325’s maximum supported RAM speed is 7467MT/s, what makes the X7 358H and X9 388H run at 7467MT/s despite their maximum supported speed is 9600MT/s? Memory controller? It’s on the CPU itself so it’s unlikely. Development cost? It’s time consuming and costly to design a mainboard for two memory speeds without rearranging the components from the ground up. Signal integrity? if so the RAM itself shouldn’t labeled as 8533MT/s if it can’t be achieved via a socketed form. BIOS programming?
There are 7500MT/s LPCAMM2 available, will it still run at 7467MT/s?
What speed will it run when a 9600MT/s (if available) is plugged in?
I have the same question and also what’s the limitation (since the memory controller is on the CPU itself) and what should we do to overcome said limitation.
Edit: Found an explanation on Reddit, the so-called high-density interconnect(HDI) technology
The reason is the insufficient motherboard performance of the Framework 13 Pro (and the TP T14 G7). To fully utilize the speeds of LPDDR5X-9600/8533, a motherboard with HDI (high-density interconnect) technology is required, but the motherboards in the Framework 13 Pro and TP T14 G7 do not meet that requirement. In contrast, models like the ThinkBook 14+ and ThinkPad P1 Gen 8 use motherboards with this technology, so they are not limited by the platform.
Basically, the FW13 “Pro” only supports the slowest LPCAMM2 modules. That’s disappointing, especially considering how heavily the SoC depends on memory performance. Especially for iGPU and NPU.
The whole point of LPCAMM2 is to deliver next-generation, high-speed RAM. Instead, we’re limited to slower variants, missing out on a significant opportunity to improve overall performance.
Would be great to have an explanation from framework engineers
Or maybe it’s just a poorly worded statement from Lenovo,I hope so (I think?)
I’m just curious if you have links or anything to any of the machines or further details showing that they use LPCAMM2 and run at 8533/9600. All the machines I can find that actually specify that they run at those speeds have soldered RAM. Again, that’s just what I can find.
The Thinkpad P1 Gen 8 that I can find that uses LPCAMM2 memory, runs at 7467MT/s. Granted, it’s a Core Ultra 9 285H, but that CPU supports up to 8400MT/s RAM when soldered.
The Thinkbook 14+ I can only find listed in places like eBay and Alliexpress. While it shows 8533MT/s LPCAMM2 modules, it doesn’t say they run at that speed. I suspect not, since the T series that includes 8533MT/s LPCAMM2 modules state that they run at 7467MT/s (“due to platform limitations,” which is similar to what Framework says).
It’s very possible I’m just missing something. But I’ve yet to find a Panther Lake machine that uses LPCAMM2 modules and specifically says it runs at anything higher than 7467MT/s. So I’m wondering if that’s just an unfortunate limitation at the moment, rather than something specific to Framework.
Yes, but that’s just the memory type that is planned to be supplied with the unit. That gen 9 P1 doesn’t actually exist yet. At the top of the specs page the “models” section isn’t even there yet. LPCAMM2 at 8533MT/s is also what comes with the current gen P1 (gen 8) and the gen 7 T series. But they both run at 7467MT/s.
That page says that it is available for early viewing, but actual model details will be unveiled soon. So it’s definitely not confirmed to be running at that speed.
Those Chinese pages are a similar story. They list the specs of the supplied RAM modules, but I don’t see any confirmation that they will actually run at those speeds.
I guess my point is that the Framework 13 Pro also comes with “LPDDR5X-8533” memory. But it doesn’t run at those speeds. So far I’ve seen no confirmation that any of those models will actually run at those speeds. The two Lenovo models using 8533MT/s LPCAMM2 modules that are fully released and have full specs for the specific models listed, both say they run at 7467MT/s.
You might be right that those machines run at higher speeds. But as you say, I think we’ll have to wait to see for sure, because the wording is a bit ambiguous. I think the fact that they all say “up to” a certain MT/s makes it seem like they are talking about the memory module specs and not what it actually runs in a specific machine.
