48 comments

[ 1.4 ms ] story [ 100 ms ] thread
It's hard to tell, but it looks like the laptop is effectively a shell for a compute element that can go in a NUC too.

It's a neat idea, and I recall there being a push for an open source device that was a computer in a PCMCIA card https://www.crowdsupply.com/eoma68/micro-desktop which unfortunately looks like it hasn't been updated in a while.

It would be amazing to replace the battery and get an upgraded Cumpute Element every few years while reusing the same basic hardware.

CPUs tend to get tied +/- 1 or 2 chipset generations, but given that this uses a PCIe interface, it may actually be plausible.

Then we can be stuck with the same low-res <200ppi screen for many years to come!

(My personal pet peeve is the stagnation of pixel density in displays outside of Apple hardware and phones.)

What's wrong with replacing the display and keyboard? They're the cheapest parts of the computer.

(I think the compute module includes the chipset, too, given that it includes the ram and wifi. It probably exposes pci to the laptop which is at that point just a bunch of peripherals.)

My Thinkpad T490 has 210ppi and it's nothing special. Runs Debian like a champ (I am addicted to TrackPoint so I'm stuck with Lenovo these days). The picture quality on this laptop is far superior (contrast, color saturation, black levels) to my 16" MBP (the LCD panel on the 16" MBP is shameful, it is the worst Apple display I've had in 15 years. If you calibrate it so that colors aren't washed out it is so dark it's impossible to use. If it wasn't a work laptop I would sell it, but luckily I have a good external 4k monitor).

So I call shinanigans on what you are saying here.

That said, I'm staying with the T490 because the best generation AMD thinkpads aren't available with high resolution panels yet, so I will grant you that you have to be patient to get a good laptop display. OTOH, you are wrong about retina displays, they aren't uniformly good so you have to be selective there as well.

This would be a bit more interesting at about half the price. I could think of some fantastic applications at that point.
So this is a rehash of the Intel Compute Card: https://www.tomshardware.com/news/intel-compute-card-bad-ide...

The problem here is the Compute Card was killed before one could meaningfully upgrade the internals of hardware using it. What commitment is there Intel won't do the same for the NUC Compute Element?

Putting the NUC brand on it doesn't change that you need a commitment to the form factor for people to buy into it.

I suppose the idea is that if the element board is used in multiple product lines (this laptop, and desktop NUCs as well) Intel is less likely to kill it suddenly?
I don't think putting it in a NUC gives it the groundswell it sounds like. There's no reason the 12th gen couldn't have a processor on a dinky motherboard just like every generation before it.

What the compute element really needs is a solid commitment. If they can promise me the next 3 generations will be drop-in compatible, there's value to it. If all we get is "in theory the next generation could be drop-in compatible, but Intel's track record says otherwise", less so. very much less so.

(comment deleted)
Why do they bother with this repackaged vintage technology? Intel should focus on new products, that could compete with M1 and Ryzen, instead they waste time on another generation of the same old underwhelming tech.
Removable CPU modules make no sense for a laptop. I do want a more serviceable laptop but I want that accomplished via an open form factor. Modules just limit the thermal performance and take up valuable interior volume that could be better filled with a bigger battery, discreet GPU, or more useful ports.
The article is showcasing a laptop that already has delivered on your request. Interior volume is optimized onto a single board. You no longer have various headers on the motherboard because of that. You can easily put a heatsink/thermal pipe on too.

Laptops are perfect for this type of application. We have yet to be able to modularize smartphones in a way that will satisfy our current idea of form-factor but that will be coming too.

> Interior volume is optimized onto a single board. You no longer have various headers on the motherboard because of that.

The module must plug into the motherboard via a set of card edge connectors so that statement is not true. Those connectors consume board space.

To be clear, I don't dislike this idea. I just don't see the need to split the motherboard in two to allow easily swapping a part that 99% of users don't need to swap often, if ever.

I imagine that a huge percentage of laptop failures come from physical damage to the screen, hinges, keyboard, or battery. It would be nice to be able to swap the body out and be able to reuse the existing core components
Or vice versa. Think about how many fully functional keyboards, screens and laptop cases are tossed because someone wants a slightly faster CPU.

I like the idea of a SoM in a laptop shell, but TBH it doesn't look like this laptop shell (as a MBP knockoff) will be particularly repairable.

True. If the compute module interface manages to become popular and somewhat standardized, we might start to see an investment in modular keyboards and other parts. Having reuse between models is difficult when so many parts are similar but one-off.
Reselling a used fully functional laptop isn't a problem. If someone wants to waste money chasing GHz status, that's fine.
(comment deleted)
It's sort of interesting but also sort of very impractical. As a buyer the only thing you should care about inside a laptop is the effectiveness of the thermal design. The ability to remove heat from the CPU decides whether your laptop is fast or slow, over a range that is much, much larger than the generation-to-generation improvement in CPU speed. A ThinkPad X1 with an 8th-generation Core CPU is going to outperform a Asus ZenBook with an 11th-generation CPU. So when I look at this replaceable computer on a stick, it just makes me wonder how hostile it is to proper mechanical engineering.
Simula's portable Linux VR headset is also planning on including an Intel NUC 11 Compute Element: www.simulavr.com

The modularity of Intel NUCs is nice. We'll be able to offer a seamless upgrade to Alder Lake when it arrives.

This looks neat as a laptop, but I worry how much I'd care to use a laptop that doesn't have S3 suspend - is this still the case for all Tiger Lake CPUs, or am I misunderstanding? Also, I guess I'm curious just how bad lacking S3 suspend is, are there workarounds? Does it end up working fine in Windows for "reasons"? It's my understanding that Linux drivers aren't yet able to work around the lack of S3, but again, I could be misinformed!

edit:

I assume this is not really available/details are off, but looking at the comparison page for the 15 inch laptops, it says they are recommended to be priced around $1500 and include discrete RTX 3070 graphics cards... I find that unlikely, but if it's true I'd definitely buy one :)

https://www.intel.com/content/www/us/en/products/compare.htm...

S0ix is the replacement, which allows for connected standby.

(see https://01.org/blogs/qwang59/2020/linux-s0ix-troubleshooting)

Connected standby has been the norm on mobile since ages, quite popular in laptops now and is making its way to desktops.

The OEM cannot both provide S0ix and S3 suspend at the same time though. (togglable through UEFI firmware options on some systems, but that's getting rare)

S0ix is a pretty clear regression compared to plain S3 suspend, it drains more power in pretty much any OS. People can try to make it work a bit better at the OS/kernel level, but it will never be as good.
It's really not that bad. Saying this as someone with a ThinkPad Nano running Fedora 34. I really don't notice a difference.
What makes you say that? My laptop can switch between S3 (for Linux) and S0ix (for Windows). I've measured power usage and it's almost identical.
Why are you averse to turning off you computer?

I understand saving state but I never let my workflow depend on program execution state. Power on, log in, do work, sip coffee, browse net, finish work, shutdown computer. I like to keep things simple.

Frankly, I've become VERY accustomed to quickly closing my lid and expecting to open it later with about the same battery life and whatever I was working on sitting there waiting. Apple has set a standard that I think in the year 2021 is totally reasonable to expect from a laptop. The concept of power management isn't one, I, as an end user, think I should have to worry about any more.
With the speed of modern nvme drives I would much rather hibernate than sleep. The ability to know you can come back in a week to the same amount of battery is way better than saving a couple seconds off resuming from sleep.
Coming from the Apple side of things: I don't care what you call it, I just want to close my laptop, then open it and get on with working, and Apple seems to do this really well.
> The ability to know you can come back in a week to the same amount of battery

That's pretty much the modern MacBook experience. And I don't care if it suspends, hibernates or serruptitiously crawls over to a power outlet in the middle of the night to recharge, I as an end user should never need to think about such things.

Close lid. Wait hours or days. Open Lid. Work.

I like to keep things simple too.

I'm on a MBP where I pretty much only reboot for system and security updates and my workflow is the same as yours but two steps shorter. I basically never need to power on or power off. From my perspective it seems to me that you're adding steps and complexity. What does power cycling provide that I'm missing?
People work in different ways. I, for example, can't fathom why you would not want to have your computer and work quickly accessible.

One of my favorite M1 features is that it finally recovers from sleep fast enough that it doesn't interfere with my train of thoughts.

Does anyone see a product to buy on this page? As I click through the models, all I see are "pricing unavailable" for the laptop kits.
With enough clicking, I got to this page [1] which lists a 5-pack of the laptop elements for $674.10 - I am not an expert, but that seems like you get 5 for $674.10 and not $674.10 each. Of course, you have to add your compute element and the SSD. Am I reading this right?

[1] https://www.intel.com/content/www/us/en/products/sku/212650/...

Looks like it — the compute element will probably be another $300-600, right?
It says it's for "channel partners", so they're probably looking to sell in bulk to resellers, maybe vendors like System76 who don't make the laptops they sell.
The guy's experience with shitty drivers reflects what people have been saying in reviews of NUCs for years. They appear to suck.
I love my NUC Extreme compute element for my gaming rig. It performs extremely well, and my whole rig is the size of two full size graphics cards side by side, so it'd fit into a shoe box.

The board is also very customizable -- I can scale down to ~3.0 ghz x8 for nearly silent operation, or up to ~4.5 Ghz before thermal throttling takes over. Plenty of timing, voltage, and thermal options. On a more spacious case with better circulation, I'm sure it'd go higher.

I got this case for it: https://www.newegg.com/white-cooler-master-mastercase-nc100-... but my experience with coolermaster customer support was not great.

This sounds very interesting. I've been looking for a reasonably powerful, yet compact and silent gaming rig for a while (not too far from a mac mini size-wise) . I had ignored the nucs assuming they'd be too compact and therefore very noisy.

What kind of gpu did you put in it?

I've scraped by with an older GPU for now. I'll probably get a geforce 3xxx when hell freezes over.
I'm not sure I like this. It's a laptop, but if 2 years from now I want to upgrade its CPU, will there be a compute module that fits? It's matches the expandability of a MacBook with the beauty of a Dell Vostro - not a hugely attractive proposition.

Good you can make it faster, if Intel decides to make a compatible module in the future.

Is the memory limit of max 16GB artificial? To make sure it's not used in server environments or something? Or does some component(s) get significantly cheaper with that limitation? Seems low.
It's to artificially limit value. other NUCs have 8GB soldered RAM with no expansion slots.