Intel's Future Laptops Will Have Memory Sticks Again (theverge.com) 47
Intel is rolling back one of the biggest changes to its laptop chips in years. The Verge: Remember how this fall's Lunar Lake laptops ditched the idea of memory sticks, putting a fixed amount of RAM on the processor package instead? Intel CEO Pat Gelsinger now says that turned out to be a financial mistake, and Intel won't do it again. Oh, and he may be axing desktop GPUs, too. Future Intel generations of chips, including Panther Lake and Nova Lake, won't have baked-on memory. "It's not a good way to run the business, so it really is for us a one-off with Lunar Lake," said Gelsinger on Intel's Q3 2024 earnings call, as spotted by VideoCardz.
I love to have 64GB on my laptop... (Score:4, Interesting)
I am a memory-lover - I love to have 64GB on my laptop and twice that on desktop.
The fact that on macbook it was not possible and event 48GB means a huge cost increase is what is keeping me on x86....
Re: (Score:2)
Re: (Score:2, Funny)
And yet, my two year old MacBook Pro has 64GB - so it has been possible for at least a few years...
And yet, you only have another seven years left on that MacBook mortgage.
Re: (Score:1, Informative)
The fact that on macbook it was not possible and event 48GB means a huge cost increase
You can now order MacBook Pros with 128GB of memory. I think you missed a few updates.
It's a little expensive, yes, but also very fast and reliable memory.
Re: I love to have 64GB on my laptop... (Score:2)
Fast and reliable isn't something attributable to apple hardware. Expensive, yes. Though I'm sure that in your case, it will be a cold day in hell before somebody rips your butterfly keyboard from your cold dead hands. Assuming you weren't holding it wrong, anyways.
The most profitable company on earth (Score:5, Interesting)
Re:The most profitable company on earth (Score:5, Informative)
They're running short on money. Fab spending is cleaning them out, and their nodes are not healthy (yet). It costs money to develop dGPUs. Plus Arc hasn't sold well yet, so it can't carry its own weight. It doesn't help that Ponte Vecchio failed massively and Rialto Bridge was scrapped.
That being said, we need an alternative to Nvidia. AMD has pretty much noped out of the high-end gamer dGPU market as of RDNA4, and it's not clear when (if ever) they may return to it. Competition from Battlemage and RDNA4 in the mid-to-low range would be very nice.
Re: (Score:1)
They just got billions [reuters.com] of socialist money [cnbc.com] to build a plant(s).
Re: (Score:3)
Those grants are tollgated based on meeting milestones.
Intel isn't meeting the milestones, so they're not getting the money. They have to spend up front, and then they get reimbursed under the CHIPS Act if they actually build shit that works.
Otherwise known as "the way it should be done, if it must be done."
Re: (Score:2)
$10 billion isn't enough. Their foundry business is bleeding them dry. Look at their Q3 financials:
https://www.intc.com/news-even... [intc.com]
Pure pain.
Re: (Score:3)
Unfortunately for PC graphics enthusiasts
Shouldn't that say fortunately? As in "fortunately Intel will now stop trying to push its mediocre graphics subsystem on people and leave it to a company that knows what it's doing"?
Re: (Score:3)
No, they don't make video cards. In fact, video cards is a minority of their sales (depending on who you ask, anywhere from 10-20% of their sales are GPUs)..
Intel's GPUs have a history of poor performance. Sure, their Iris Pro and Arc graphics do better, but they're stuck in the midrange market and noncompetitive given the amount of midrange GPUs from better known names.
Intel's trying, but the market isn't responding very positi
Nvidia's "AI accelerators" (Score:1)
Re: (Score:2)
That's because their GPUs are shit. Why do you think people go and buy Nvidia products if their system already has an Intel GPU in it?
I absolutely hate it, but... (Score:4, Interesting)
... the future is apparently ARM cpu's and Nvidia gpu's, with AMD as the distant second backup player, and Intel a once dominant company that goes under. In a decade, Intel will probably be bought up for their IP rights. Even if they lose money initially, Intel should have doubled down on the powerful desktop gpu sector.
Problem is they need money (Score:2)
If we had strong antitrust law enforcement this wouldn't be a problem. But we did away with that in the 1980s.
Re: (Score:3)
Intel should have doubled down on the powerful desktop gpu sector.
The problem is that everyone thinks another market player is a competitor.
Instead of just a player.
If my core business is CPUs, I do not need to jump up and invent my own GPUs to compete with one who is already good at GPUs.
American companies always want/try to destroy each other. It seems not plausible to them that cooperative markets like in Japan in the 1980s, simply work better and make more companies happy and make more people happy and
scaling smaller (Score:1)
There are probably things I don't understand about making chips on silicone. As things scale smaller it seems that tighter integration of components is a good thing. Have to wonder if this is the right direction in 2024.
Re: (Score:3)
Re: (Score:2)
Yeah, probably.
Re: scaling smaller (Score:2, Interesting)
Really depends on the component. For DRAM, the only component that benefits from tighter integration is the memory controller, and that was moved on to the CPU die about 20 years ago, first by AMD, then by Intel. CAS latency is high enough that you're not actually limited by the speed of light, instead you're limited by how long it takes for the controller to synchronize with the RAM chips. Being on-die does you no favors. SRAM is a different story, but we haven't had that off-die in ages anyways.
For GPU, y
Re: (Score:2)
Chips on silicone
= snack food crumbs on a Vegas hooker.
Re: (Score:1)
Everyone is very impressed with your pedantry.
https://hackaday.com/2022/01/2... [hackaday.com]
Maybe it's just Intel (Score:2)
Soldering RAM on the MB seems to have worked fine for Apple. Maybe Intel's fan-base just aren't Loyal like Apple's since there there are alternatives, regardless of whether the RAM is upgradeable or not.
Re:Maybe it's just Intel (Score:4, Interesting)
So soldering memory onto the motherboard is not what Intel did. What Intel did was attach the memory into the CPU package. The difference is that Intel now has to source the memory and get memory margin on the whole processor package rather than have the OEM source memory and assemble it into the system. This leads to a massive increase in inventory requirements and a lowering of margins that is unsustainable for Intel's business model.
Intel OEMs have been soldering down memory on motherboards for a few years making laptop upgrades untenable which is a huge change. The first thing you should do when you get a new laptop is get the largest Memory upgrade you can get off of Amazon and put that memory stick into the laptop.
Re: (Score:2, Interesting)
So soldering memory onto the motherboard is not what Intel did. What Intel did was attach the memory into the CPU package.
That's also what Apple does.
The difference is that Intel sells CPUs to OEMs, and Apple sells complete systems to end users.
Since Apple needs to deal with memory inventory either way, it makes sense to integrate it for better performance.
Re: (Score:2)
So soldering memory onto the motherboard is not what Intel did. What Intel did was attach the memory into the CPU package.
It was a weak attempt at creating a vertical monopoly. Some third rate VP probably got a promotion based on the idea.
Re: (Score:2, Interesting)
The memory on Apple silicon is on-package, not soldered to the motherboard. Trying to move all of the different stages of the memory hierarchy closer to the CPU is how computer engineering has been solving the performance disparity for the past 30 years. Without a major revolution in memory performance, the future will likely have memory on-die.
Re: (Score:2)
The "problem" for Intel is that those pesky PC and laptop producers keep building products with upgradable memory and storage capacity, so they can't get away with charging $800 for a 2 TB SSD upgrade like Apple can.
I love modularity (Score:5, Insightful)
I would rather have a slightly thicker laptop than one I can't upgrade or repair.
In my opinion, a laptop should be a chassis and backplane into which all the other components can be mounted without tools. Sliding in a new graphics card would be a wonderful option.
Re: (Score:3)
I would rather have a slightly thicker laptop than one I can't upgrade or repair.
Thicker, upgradeable laptops are available, but they're a niche market, so you'll likely pay more for the ability to upgrade than if you'd just bought a better-spec machine in the first place.
Re: (Score:1)
Nothing's stopping you. They still sell luggable computers that are fully moduler, save the screen and keyboard. Not quite as big as a Compaq Portable, think a small briefcase. Unfasten the cover to reveal the screen, unstow the keyboard and you ha
Re: (Score:3)
Re: I love modularity (Score:2, Insightful)
Good (Score:3)
Our current products suck! (Score:2)
Re: (Score:2)
Lunar Lake is about the best thing in Intel's entire product portfolio (for now). Unless Arrow Lake-H makes a big splash vs Strix Point.
Re: (Score:2)
He's basically telling everyone that their current laptops stink.
No. He's saying that current laptops are less profitable.
Sounded good to me. (Score:2)
Re: Sounded good to me. (Score:2)
I thought that DRAM Was Reliable (Score:2)