Aaron44126

New post on the Framework Laptop 16 chassis design & manufacturing process. https://frame.work/blog/framework-laptop-16-deep-dive---enclosure

Huh, maybe I will give GPU C-states another try.

There's rather limited space for direct uploads. You can see what you have posted and your quota here: https://notebooktalk.net/attachments/ I find it best to just host images externally. OneDrive is easy enough (use the "Embed" option in the web UI to get a URL to an image that you have uploaded). If you paste a URL to an image in the compose box here, it will automatically be converted to an in-line picture.

Display panel issue. Again! No idea why this suddenly popped up. (I'm sure I would have noticed it quickly so it hasn't been around for long.) (Dark spot in the middle of the circle here, not quite centered.) I think it's not stuck pixels but rather some kind of display defect. It seems to be "behind" the pixel grid and moves around slightly if I change my viewing angle. I was a bit bummed because the dark pixel guideline is more loose than the bright pixel guideline (I had a bright pixel last time), requiring six messed up pixels to qualify for an automatic replacement rather than just one. So, I was prepared to get some grief about that from the tech. But, I measured it and found that it does cover six pixels, so there should be no reason for them not to replace it. Call me persnickety, but I do have sensorimotor OCD and this sort of thing is very bothersome once I notice it. I called Dell tech support and the tech agreed to replace the display. (...I had to talk him out of replacing the graphics card instead???) I asked for parts only this time so I'll do the replacement myself and won't have to worry about scheduling a tech. (I'm going to be traveling soon, so if there is a parts delay that could lead to a scheduling headache like I ran into when having the panel replaced last year.) My last two parts for this laptop faced weeks-long shortages. I wonder if I will actually get this one to me tomorrow... @Ionising_Radiation Have you been liking the Honeywell PTM7950 pad thing? Since I have to take off the heatsink to replace the display enclosure, I'll have to repaste so I'm thinking about trying this if it actually works decently... [Edit] (One hour later) At this moment, it is looking like both the display and the PTM7950 pad will be here tomorrow. [Edit 2] (Seven hours later) It's on the way, FedEx has picked up the display panel with estimated delivery of tomorrow.

I don't know if there was a NVPCF error in Device Manager beforehand, but I can say that disabling the NVPCF device from safe mode was enough to allow the system to boot normally.

I think all of the displays are IPS at this point, other than the Precision 7670/7680 OLED display, so I guess they don't bother specifically stating it.

Immediately at boot.

Only Windows 10 systems here. Experienced the BSOD on two of them (so far).

BIOS update warning. The nvpcf.sys BSOD from earlier rears its head. I had a coworker upgrade the BIOS today and then got stuck in a BSOD loop. Fixing it required going to safe mode, disabling the NVIDIA Platform Controllers & Framework device, and then installing an older NVIDIA driver. Too late to check now but I suspect turbo boost 3.0 was enabled which is what was "saving" him from the BSOD before the BIOS update was applied. If you have Intel Turbo Boost Technology 3.0 enabled, I would suggest that you try disabling it before applying the BIOS update and make sure that your system can boot up without a BSOD.

I think the BIOS update effectively disabled that option and removes the option to enable it again, so you should be fine if you already have it disabled; it will be basically no change for you.

P4000 is an 8GB card. P5000 is 16GB.

It can be an issue if you want to use that drive in some sort of RAID or other "combination" setup; often you'll not be "allowed" to include a drive that is showing a SMART warning and you might have to jump through hoops to work around that.

Normally the touchscreen glass is not easily separable from the panel; the glass, digitizer, and display panel are adhered together. You'd probably have to replace the entire display enclosure with panel included.

The physical connector is the same but the pinout is not the same. Framework is using a custom pinout. It is not expected that any Dell DGFF GPUs will work in this thing.

...Is there some rule that you have to put games on the app store if they are converted with this toolkit? There are lots of Mac programs & games distributed outside of the app store. They could just distribute through Steam like they do with the Windows version.

Framework Laptop 16 will feature swappable rear modules. It's not really MXM, they are using a customized DGFF connector, but it will allow for either a discrete GPU or extra battery to be installed. They have not yet given details for what types of modules will be available at launch. (We can assume that dGPU will be there, but anything else is up in the air.) ...I am sort of thinking that swapping them won't be as easy as just pulling one module out and snapping in another. They use a DGFF connector on the motherboard side, so you'll probably have to snap out the keyboard and remove a couple of screws on the motherboard side holding the DGFF connector in place before a module can actually be removed.

Today, Framework released details on the power adapter for Framework Laptop 16, which will be 180W. https://frame.work/blog/framework-laptop-16-deep-dive---180w-power-adapter Some tidbits I saw: Under high load with a dGPU in play, the system can draw from battery power to push total power use above 180W. Framework Laptop 16 supports 240W over USB-C, and while Framework will not be providing such a power adapter, you can source your own and use it with this system. (You can also order the system without a power adapter at all, if you plan to do that.) The cables on both sides of the "brick" can be unplugged/replaced.

Saw this. Intrigued. But right now, the offering from Apple is pitched as being for developers or publishers, not users, to evaluate how their games might run on macOS and help with a port. It is still up to the publisher to leverage this and push out a Mac version of their app. This as opposed to say the Steam/Proton situation on Linux where end users can just download any Windows game in the Steam store, without the developer/publisher necessarily taking any action to make it available on Linux, and it will "probably work". Now, since maybe someone will slap together a solution to "easily" allow users to fire up any old Windows game using this framework (more like Lutris on Linux) and that would be pretty cool. I think I saw mentioned that Apple contributed some of their work back to Wine so it should help other products (like CrossOver) that use that stack.

Yes, I have experienced the same with Dell. They won't swap one part for another, even with an issue like this. You'd have to source the part and replace it yourself, or return the system and order a different one with the configuration that you want.

You should be able to override the board ID mismatch, try adding options "-4" and "-5". (I think it is only blocked for Turing and up.) Make sure you save the original vBIOS in case it doesn't work at all and you need to flash back.

I missed this post. They have detailed the display in the Framework Laptop 16. https://frame.work/blog/framework-laptop-16-deep-dive---display It looks like they were planning to allow preorders around Computex but they have had to delay and now it will be pushed into summer. With regards to the display... The Framework Laptop 16 has 2560x1600 resolution, a 100% DCI-P3 color gamut, variable refresh rate up to 165Hz, 9ms rise+fall time, 1500:1 contrast, and an unusually high 500 nit brightness. I was hoping for full 4K (3840×2400) so that 200% scaling can be used while maintaining a FHD desktop working space, but this would still be a fine display panel for gaming in a laptop. VRR and 165Hz is pretty nice and 500 nits is plenty bright. And, it has a matte finish. As you would expect, it is designed to be easy to swap out, though I don't know if you'd be able to transfer in another "generic" 16" eDP panel... We'll have to see what the mounting and rear display cable attachment look like. LTT also posted a video showing an upgrade of a Framework Laptop 13 to the new AMD motherboard, and putting the leftover 12th-gen Alder Lake motherboard in a standalone Cooler Master case.

Quadro P5000 is the best one that you could possibly expect to work well (with the ES vBIOS) but I do not know if it is a sure bet... I tried one myself in the M6700 and did get it working but I was plagued by periodic BSODs, so I ended up going back to Maxwell, and @jeamn experienced something similar with GeForce 1070 in Precision M6800. However, NBR user @DynamiteZerg and I think also @RMSMajestic had P5000 working in Precision M6800 with no trouble. Quadro P5200 is an unknown. If it has the same vBIOS issues booting Windows that earlier Pascal cards had, it is unlikely that there would be a vBIOS that could be installed that would work around it. Quadro RTX series (Turing generation, HP version) will work but requires a pretty substantial mod to the heatsink and even a minor mod to the chassis to fit in.

Did you flash the engineering sample vBIOS? It is required for both M6700 and M6800. The link is in a post by @TheQuentincc just a little ways up from here.

Yeah, so my perspective is that you're giving it too much credit. It's not an intelligence at all, just a machine. A sophisticated one, sure, but also one that was built by people who can understand how it works. My background — When I was finishing up my master's degree (≈15 years ago) I did a fair amount of studying of AI. I've did some projects that had to do with predicting the weather, and teaching it to get better at board games, and such. I built a neural network literally from scratch (not using "off-the-shelf" libraries) that could identify the number of open spaces in a parking lot given a photo, so I do understand the fundamentals. And in learning about this stuff, my general impression of AI went quickly from "Wow, computers are really smart" to more like "Wow, it's actually pretty simple once you know how it works, it just looks smart because of the massive scale". That sentiment has been echoed to me by multiple colleagues who have done work in AI. The techniques have gotten more sophisticated since then, sure, but the bulk of the advancement has come not so much from radically new methods of building AI, but rather from basically increasing hardware power and the general passage of time allowing for the training of larger and larger AI models. Now, AI has made some notable mistakes. With literally millions of neural network node weights it is generally difficult to figure out why exactly the network came to the conclusion that it did by examining the network directly, but at the same time it is not by just taking the time to think about it. A chatbot becomes racist because it "reads" racist content online or people interact with it that way. Or more critically, a Tesla crashes into the side of a semi-truck because Elon Musk (brilliantly) thought that it was fine to use visual data rather than LIDAR for self-drive and because of the lighting or whatever it couldn't differentiate the side of the truck from clear sky, which is a mistake that could kill someone. Garbage in, garbage out, as they say. In the end, it was a human that made the mistake and the machine just did its thing. I'm not trying to dismiss the dangers of AI. It will definitely be able to do some things better/faster than people can. I'm really worried about jobs and misinformation as I stated before. And the other thing would be letting people who don't have a full understanding of the capabilities/non-capabilities allow AI to be used for critical decision making. I'm just trying to point out while it is getting better, there are very real limitations to what it can do and what it can't do that will take a long time yet to overcome (if they ever are). Those limitations become more clear if you take the time to understand how this stuff actually works, and I would recommend that anyone who is "worried" about AI take the time to do that. With regards to the supposed drone simulation, that's just not making sense. With this sort of thing you would give the AI a "goal" (maximize enemy kill count?) by assigning points to various outcomes. This is exactly what they did when teaching AI to master chess, Go, and StarCraft II. Then, you would run an absurd amount of simulations with the AI taking random approaches, scoring itself based on the rules set, and over time "learning" which approaches take you from the "current state" to a "more favorable state" and eventually your desired outcome ("win the game" / "kill the enemy") with a high degree of certainty. Maybe the AI in such a training scenario would "discover" by random chance that destroying the communications tower and eliminating the pilot would lead to a better outcome. This would be quickly discovered by analyzing the results of the training runs and then the engineers would correct the "rules" by assigning massive negative points for this behavior and start the training from scratch. It would not be discovered by chance during a random one-off real-time simulation. The AI has not had a chance to "learn" that taking these actions resulted in a better outcome if it has not been trained. It's not a human so it can't just determine these things through intuition. I rather suspect the guy who gave the presentation was on the non-technical side, interpreted something he saw or heard wrong, or basically just didn't understand the difference between a thought experiment and a real simulation.

This appears to be a bogus story. (Indeed, it seems to be something more out of science fiction and it is easy to poke holes in.) Some AI stories/videos I've run across recently. AI in education (...Actually pretty cool, if it can be pulled off.) AI researcher on tripping up ChatGPT 4 with simple queries (Despite how "smart" and "confident" it sounds, ChatGPT is basically spitting out amalgamations of its training data – that's exactly how AI neural networks work – and not actually "thinking" like a person does. Throwing more training data at it won't necessarily fix this because "common sense" isn't something that is spelled out on the Internet for a machine to pick up and "learn".) Geoffrey Hinton (long-time researcher, at Google from 2013 until pretty recently) warning about dangers of AI Right now, I'm less worried about "death and destruction" caused by AI and more worried about it displacing real human jobs in certain sectors, and also stirring up the misinformation scene even more than it is already. These are real problems happening now and likely to get significantly worse over the next few years. There doesn't seem to be anything in the works to seriously address either problem, either.