I will buy a frame.work laptop when it has ECC support. It need not come with ECC, but all of the current Ryzen PRO-based laptops have ECC support by default (in the CPU)… The problem is all the manufacturers don’t care to support it on their motherboards! So many would be frame.work fanboys if you added this in your first AMD laptop. Please consider it; it need not be default to support it!
Thanks!
P.S. My frame.work wish-list:
– ECC
– AMD CPU & GPU
– As much open source firmware as possible
– 2nd NVMe option
– As durable as possible build (drop testing)
– Beat thinkpad @ mechanical keyboards (perhaps even ortho-linear)
– One of these OLEDs will be available by this time, please support installation:
– not compromising on ports
– Careful of windows vs linux “sleep modes” in the BIOS! Causes problems in Linux!
Bonus:
– Separate internal battery that allows for hot swapping of the main battery
– Pre-overclocked RAM for Ryzen laptops. You can squeeze lots more performance without decreasing stability.
I’m for basically all of this. Especially Ryzen, ECC, Linux, BIOS, overclocking (advanced BIOS), open-source, durability and great keyboard.
I’m really intrigued by the whole DIY and repairability approach but it has some shortcomings that makes it hard to pull the trigger. Intel-only, no touch screen, no Ryzen, limited number of I/O ports…
– AMD CPU & GPU
– As much open source firmware as possible
Yeah, there is no open source firmware for AMD CPUs, System76 doesn’t have open source firmware for their AMD laptops.
Thought I’d throw in an edit, I’ve tinkered around a bit looking at AMD, they are good CPUs but Single Core performance is a bit behind. They make this up with more cores though.
Personally I do not care which platform it is on. Intel or AMD, ARM even, as long as it meets the performance and doesn’t suck too much. However, having the software advantage and the ability to personalize many things at a low level is extremely useful. ECC support is definitely a worthwhile feature, as current manufacturers often overlooks its importance on consumer devices. I am all for Ryzen and the AMD platform as long as nothing major is lost over it. I don’t know how it is in laptops but on the PC platform the switch to AMD resulted in me losing thunderbolts which was quite disappointing. Other than those issues I love the AMD chips and their capabilities. As long as there is going to be choices of platforms I believe developing a motherboard for AMD can go a long way.
@Water261 Hard to get any more obvious than that… Amazing how many customers argue in favor of the company actually not giving the other customers what they’re asking for. This can be reverse engineered, leveraged, leaked, promoted, petitioned, etc… But every time I see a discussion online about it - customers come out of the woodwork to make sure everyone knows we cannot have the information about how our product we own works. We are not allowed to know what the black box operating system in our CPUs is doing, what network components could connect to it, etc… I think the best strategy is to ignore such comments and push forward on multiple fronts. Sure, Power9/10 is an option, but not nearly as powerful as AMD Ryzen/EPYC would be if we could boot/run them without the backdoor(s).
I agree that ECC would be nice. Luckily going forward, DDR5 requires it, so we should see major adoption on many platforms. Doesn’t help at the moment, but at least it’s a good step.
It’s worth noting that what DDR5 requires is not ECC in the way you’re used to it in PC memory. What it requires is on-die ECC within the RAM chip, which does not protect against bits errors in transit. It’s also worth keeping in the reason it’s being required is due to the memory cells becoming increasingly unreliable at the smaller sizes they’re being shrunk to. You can fully expect manufacturers to push that to the extent they can, using smaller higher error rate memory cell designs because the ECC is sufficient to get away with it generally, but ECC has it’s limits, and it’s all a game of statistics.
In certain respects the ECC required for DDR5 memory, and the reasons for it, are much more akin to the ECC that every SSD ever has than the ECC that one is used to with ECC memory.
TL;DR: Make no mistake, DDR5’s on-die ECC is no replacement for the conventional types of ECC for RAM
I both agree and disagree. What you say is right, it doesn’t matter the name of the hardware (for the sake of the name) but in this case there are good reasons for favoring one over the other.
You want the most bang for your buck. While Intel’s CPUs are perfectly adequate for a lot of people and their needs, that isn’t true for many others. And then there’s the long-term performance longevity and the limitation of use (gaming, editing, anything that needs multithreading and graphics performance).
Intel is like someone who only train their arms so they have a huge upper body, small legs and weak core muscles. So they’re good at particular things but struggles with combined tasks and special tasks.
Arguably not that big of a deal (to be fair): Intel has struggled more with security (and been affected a lot more by performance degradation in regards to their security fixes (5 - 30% reduction depending on workload (gaming unaffected afaik)).
AMD has both solid CPUs and GPUs. Good single threaded performance, great multithreaded performance and great at graphical workloads (relative to mobile offerings).
Thunderbolt can be added with a controller (although there could be challenges or expenses doing it this way, obviously Intel wants you to buy their CPUs only). Personally I’d rather have AMD without Thunderbolt than Intel with.
AMD is more power efficient and their solution is quite solid. Their hardware video decoding/encoding picture quality is crap, though, so that’s a bummer. Nvidia (and I think Intel) shines here. Disclaimer: details might be wrong here, I’m going off memory. It might differ on mobile graphics solutions or be rectified in the latest chips.
ECC is a very good reason, especially for a laptop where you don’t have insanely high clock frequencies anyway.
Point is, these are reasons many want AMD. More: Intel arguably doesn’t deserve it (AMD has made strides and really accelerated performance progression and put an end to the horrid monopoly resulting in expensive hardware with minimal progress and features being locked into higher-end SKUs). AMD has always been the one to support open (source) standards and have been an important piece for open-source technology as well strengthening the foundation for Linux.
For me it’s about balanced performance (office, media, gaming - allround use), supporting the brand that actually deserves it (definitely not Intel which has been the case for years, even when they were holding on to the single one advantage that won them benchmark scores even though real-life workloads are quite multithreaded.
For network controllers, though, I’d definitely want Intel! Not that there aren’t other good controllers, just that Intel has affordable yet solid controllers and the likes of Realtek is often associated with unstable network and issues. Might be an enthusiast’s “anchient rule” but we all know Intel is cheap and solid for network.
Non-angry (tedious-to-read) somewhat relevant “rant”: Nobody has a near-wiped computer with only games and benchmark programs running. It’s filled up with junk and in the background are browser processes, Discord, anti-malware, game launchers, streaming clients, editing programs, bloatware and various other programs. So the real-world workload will quickly demand a strong multitasking performance and that single-threading advantage will quickly fade as their computer gets more bloated and games continue to better take advantage of multiple threads. Same story back when dual cores were recommended. It was clearly the best in the short-term, but then quad cores with lower clocks lived longer because the dual cores just collapsed under the increasingly multithreaded workloads.
Yet, reviewers test products as if we are buying new hardware every year and they always ask the idiotic question of “is it worth it upgrading from last gen?” and of course they aren’t enthusiastically recommending it because of course it’s not worth it. With the progress we were used to a few years back it still wasn’t worth it upgrading after 5 generations.
If we can get more DIY enthusiasts using AMD, then perhaps we can get enough interest to reverse engineer them enough to disable the PSP. All modern CPUs have block box OSs running inside your CPU that have access to memory & registers. They’re called “management engines” - and some are even network accessible. The Ryzen architecture will be around for a while - and is already too ubiquitous to pass up. Would love to see FOSS firmware development on frame.work
See coreboot.org for how much these efforts have not progressed. I’ve talked to execs at CPU companies and always find it interesting that this is the only question they will not give any attention to, despite how much security vulnerabilities can affect their revenue. Why not open source their management engine so users can do their own security reviews?
Disclaimer: I don’t know what I’m talking about! Assume any of this to possibly be incorrect.
Things like HDCP are handled by PSP. Probably other DRM schemes, maybe even Widevine relies on it? What about anti-cheat? Enterprise-software?
So the motivations might be pressure from large industry players. There might be pressure from the government even. Not to delve into conspiracies - but backdoors aren’t exactly UFOs. It’s a proven thing, a thing you’d be naive not to suspect.
Perhaps the industry is afraid of piracy or software locks being bypassed? We all know they want controll over all chains of the product even though we paid for it. You don’t own nearly as much as you think you do when you pay for things.
Then there’s their own motivations. Security? Could bad PR stem from the PSP being open-sourced, modified, disabled? What does it let people do that they don’t want you to? Will it reveal information? Then there’s the actual effort for AMD to do it (why would they bother?). What about the closed-source OS created by (I believe) Trustonic (OS called Kinibi)? They certainly don’t want their IP to be available for free, they want to profit off of it.
It seems as if you can disable (parts) of PSP but as far as I know it can basically still perform operations in the background if it so desires. It full access to all memory data. It runs on its own processor too and what it’s doing is (afaik) invisible to the system.
That is easy to rationalize, but why don’t people rationalize from the other likely more damaging PoV? That is, if AMD is using “security through obscurity” and then has a meltdown (pun intended) with processor architecture bugs, then they lose more money than if they would have been more open with their spec and allows researchers to more easily analyze it. And furthermore, to patch it themselves instead of waiting for an official response if they find it inadequate. For example OpenBSD has had a different way to deal with those CPU bugs for a long time now (just disabling hyperthreading, for one).
A few in this thread pointed out there is no open source firmware for AMDs… But that shouldn’t be true for long. Please checkout Project X. I am sure Project X is open to collaborating with frame.work. They have resources and they are trying to coreboot modern AMDs.
Regarding ECC: The support is already there in the Ryzen hardware. But it has to be enabled at the motherboard, as far as I know the cost of enabling this should not be different in a meaningful way so I do not see this having to even affect the price.
It would be a great advantage point and selling point and hopefully pushing the industry to follow suit.
To those who might say it is not needed for most people, it should have been the standard from the beginning. People are affected by not having ECC all the time without knowing. To say that people do not need it is to say that their work, even if it is just personal projects, is not of importance. It is to say that their time is not valuable and that support tickets that would otherwise be avoided does not matter.
The advancing hardware is increasing the need for this and the fact that the support is already there on the Ryzen platform makes it a no-brainer. People would definitely buy the ECC memory in the configuration checkout.
So for the AMD version, please, do the no-brainer thing and enable ECC support. There is no reason not to. Those who do not want it can choose not to, but those who want ECC are not overruled by those who do not care about it.
Hi all,
Thank you for building and sharing such a laptop!
I just looked your configuration options, looking for reliable configuration which I can rely on as a software developer.
The most important for me is to have a reliable Linux system.
This includes having ECC support. Do you plan making such option. May be a small hdd support would be also nice.
Thank you!
Just adding to the above answer, since the i7-1165G7 is a mobile CPU it’s safe to assume that Framework will only use Mobile CPUs and after a quick check, none of the 11th Gen Processors support ECC memory since it’s more or less designed for Desktop use cases. Unless Framework moves to support the other categories, having ECC is unlikely as Intel probably won’t ship it on their mobile CPUs.
Maybe if/when they make AMD laptops. Supposedly the memory subsystem of Ryzen technically all supports ECC. Even if they don’t publicly talk about it on the APU’s. Would be cool to have a laptop with ECC memory; however ECC DIMMs do technically use more power since they are self error correcting.
Yep, it is a no-brainer for the AMD version. The support is already there. It has to be enabled, perhaps with minor hardware adjustments.
The advantages are there even for regular people and adding support for this only makes the laptop so much more capable and expands the customer base. It does not take away non-ECC support, so when it is already supported by the AMD platform, any argument against it is weird and non-sensical.
Framework has the potential to push the industry to the better. It would only have advantages, minimizing data corruption, crashed and what would otherwise be unecessary troubleshooting where a solution is impossible because bit flips happens at random and cannot be prevented in software.
Ignoring ECC for an AMD-version is like putting functional rear doors on a station wagon, with hinges and door handles, but then spot weld a small spot between the door and chassis so the door cannot be opened and the car is limited to being a 2-door station wagon.
Linus should really step up on one of these efforts promoting ECC in Ryzen laptops. Either Thinkpad and/or frame.work… He has plenty of influence but these untouchables are so damn hard to contact - it would be an easy thing that he would want to do (promoting an ECC-compat platform that he’d definitely buy & use)…
