Intel can use up their excess capacities, making currently high-priced DRAM for profit, gain goodwill for rescuing the PC market, which in turn will sell more Intel CPUs as well. Sounds like a win to me. What do you think?

searches

It sounds like they are, in fact, looking into making DRAM. Just that they’re interested in making better DRAM for AI applications.

https://www.tomshardware.com/pc-components/dram/intel-and-softbank-collaborate-on-power-efficient-hbm-substitute-for-ai-data-centers-says-report

American chip giant Intel has partnered with Japanese tech and investment powerhouse SoftBank to build a stacked DRAM substitute for HBM. According to Nikkei Asia, the two industry behemoths set up Saimemory to build a prototype based on Intel technology and patents from Japanese academia, including the University of Tokyo. The company is targeting a completed prototype and mass production viability assessment by 2027, with an end goal of commercialization before the end of the decade.

Most AI processors use HBM or high-bandwidth memory chips, which are perfect for temporarily storing the massive amount of data that AI GPUs process. However, these ICs are complex to manufacture and are relatively expensive. Aside from that, they get hot pretty quickly and require relatively more power. The partnership aims to solve this by stacking DRAM chips and then figuring out a way to wire them more efficiently. By doing so, the stacked DRAM chip’s power consumption is halved versus a similar HBM chip.

If successful, SoftBank says that it wants to have priority for the supply of these chips. At the moment, only three companies produce the latest HBM chips: Samsung, SK hynix, and Micron. The insatiable demand for AI chips means that HBM supply can be hard to get by, so Saimemory aims to corner the market with its substitute, at least for Japanese data centers. This will also be the first time that Japan aims to become a major memory chip supplier in over 20 years.

I suppose if they can make better AI-oriented DRAM than the Big Three and make enough of it to satisfy AI demand at some point, that might make the Big Three redirect some of their output back to DIMMs.

“Partial workaround” wou’d probably be more accurate. As the article body points out, DDR5 SO-DIMM prices are also up, albeit not as much as DDR5 DIMM prices.

But it’s substantial enough of a price difference to be interesting, especially with larger-capacity SO-DIMMs.

EDIT: For those not familiar, SO-DIMMs are “laptop memory” and DIMMs are “desktop memory”.

While I agree that I don’t think that an LLM is going to do the heavy lifting of making full use of Rust’s type system, I assume that Rust has some way of overriding type-induced checks. If your goal is just to get to a mechanically-equivalent-to-C++ Rust version, rather than making full use of its type system to try to make the code as correct as possible, you could maybe do that. It could provide the benefit of a starting place to start using the type system to do additional checks.

Most of that is setting up third-party apt repos, which I don’t believe is necessary. Steam’s in the Debian trixie repo.

https://packages.debian.org/stable/steam

EDIT: I’d guess that the following would probably work on a Debian trixie system:

If you have your system set up for only 64-bit packages, you’d need this at some point prior to the install, to let your system use 32-bit packages, since Steam’s only available as a 32-bit binary:

$ sudo dpkg --add-architecture i386

I think that deciding whether to use both 64-bit and 32-bit packages or not is an option in the Debian installer, but I might be misremembering.

You can update your list of packages at this point, upgrade, all that, but that goes for any install operation; there’s nothing specific to Steam there. If you’ve just added 32-bit packages for the first time above, then you probably do want to update the list of packages, since your system won’t have a list of 32-bit packages yet.

$ sudo apt update

But then it’s just like any other installation of software.

$ sudo apt install steam

That actually just contains, as I recall, the Steam installer — enough to pull down and install the current Steam environment for a given user, which happens next time you run the Steam binary.

$ steam

EDIT2: I guess that assumes that you do have “contrib” enabled on the Debian repo, and I don’t know whether that’s enabled by default by the Debian installer or whether it’s an option during install or what. I do distinctly remember one point in time when “non-free-firmware” was not enabled by default, because I always had to turn it on to get support for <random hardware device with closed-source firmware blobs>, but I don’t know whether contrib is always enabled or not. I have main, contrib, non-free, and non-free-firmware enabled. From /etc/apt/sources.list.d/debian.sources:

Types: deb deb-src
URIs: http://mirror.i3d.net/debian/
Suites: trixie
Components: main contrib non-free non-free-firmware 
Signed-By: /usr/share/keyrings/debian-archive-keyring.gpg

“My goal is to eliminate every line of C and C++ from Microsoft by 2030,” Microsoft distinguished engineer Galen Hunt wrote in a recent LinkedIn post.

“Our strategy is to combine AI and Algorithms to rewrite Microsoft’s largest codebases,” he added. “Our North Star is ‘1 engineer, 1 month, 1 million lines of code.’”

Well, I expect it’ll be exciting, one way or another.

I remember banging on updating controller firmware using Windows VMs (in my case, with XBox controllers) and didn’t get it working after some time, though theoretically it should be possible.

Just using an Internet-connected console is all it takes, so if you know anyone that has one, that’s probably a more convenient route.

I kind of wish places like GameStop would offer this as a service or let people do it, since they have demo consoles sitting there anyway (or did last time I was in one).

It’d be nice if Sony and Microsoft went out of their way to support fwupd, but I suppose in Microsoft’s case it’s a direct competitor (with Steam on Linux) and in Sony’s case, probably niche enough that they don’t see much point. Sony’s trying to make money on selling access to make games for their console, not selling controllers.

Killing DIMM production.

That’s of the three major companies that make RAM chips.

There are other companies that make DIMMs. They just buy chips from the RAM chip manufacturers to do it. PNY or Kingston, say.

Micron was just doing a vertically-integrated thing where they did both the chips and DIMMs.

EDIT: Looking back at the article, it does say that.

I mean, efficient in terms of memory utilization, like. Obviously there are gonna be associated costs and drawbacks with having remote compute.

Just that if the world has only N GB of RAM, you can probably get more out of it on some system running a bunch of containers, where any inactive memory gets used by some other container.

Apparently there are m.2 NVMe drives with DRAM caches.

I don’t know if anyone makes a pure DRAM NVMe drive — it’d forget its contents every boot — but if so, on Linux, you could make the block device a swap partition.

https://en.wikipedia.org/wiki/Zram

One of the mechanisms for compressing memory in Linux. Trades CPU time for effectively having more RAM Recent versions of Fedora apparently have it on by default.

I’ve read that zswap, another mechanism, is preferable on newer systems with NVMe/SSD, where paging isn’t as painful; that only compresses pages going to swap, but requires that you actually have some swap. I haven’t used either.

Probably someone should try benchmarking them for various workloads if systems are going to be running on much less memory for a while. Was more of an edge case thing that not many people cared about, but if operating with less memory is suddenly more important, might have broader interest.

On Linux, also possible to opt for lighter-on-memory versions of a lot of software that you’re kinda committing to using the Microsoft-provided version of on Windows. File browser, compositor, etc.

https://obsolescence.wixsite.com/obsolescence/cpm-internals

CP/M requires a minimum of 20K RAM, although realistically, 48K is the bare minimum. Most systems have the maximum 64K.

Sounds like it can’t address > 2¹⁶ bytes.

Honestly, it’ll be more efficient to have memory in a datacenter in that hardware in a datacenter will see higher average capacity utilization, but it’s gonna drive up datacenter prices too.

I don’t think that the NVMe shortage is that big of a deal in terms of using it for swap. It’s much cheaper than DRAM per GB. You don’t need that much.

Windows 11 can run on 4GB. That’s the minimum for the listed requirements, and the other day, I saw Best Buy selling a 4GB model, and I see some systems for sale online. I would imagine that it’s not ideal.

https://www.microsoft.com/en-us/windows/windows-11-specifications

Minimum system requirements for Copilot+ PCs

RAM: 16 GB DDR5/LPDDR5

I think that OpenAI has probably kind of bashed a hole in the bottom of Microsoft’s boat on the local AI stuff, if 8GB is going to be midrange.

mid-range laptops to 8GB

My not-terribly-new phone has 12GB of memory, and I’m pretty sure that Android is a lot lighter on memory than the Windows 11 that I suspect a lot of these are going to be running.

Biometrics are irrevocable. If you’re worried about stolen personal data, they are not what I would be moving to.

DDR4 RAM is presently cheaper than DDR5, but it has also increased dramatically in price recently.

https://pcpartpicker.com/trends/price/memory/

DDR4:

https://lemmy.today/pictrs/image/ed889201-f9e6-46ec-81a8-832f6bfc63ed.jpeg

DDR5:

https://lemmy.today/pictrs/image/35d03746-8d9c-443f-808f-8c88f2914b73.jpeg

I have never seen a DIMM fail, and don’t believe that failure is remotely on the order of something like hard drives or fans.

I’d imagine, based on the fact that modern DIMMs often have heat spreaders, that there’s probably some level of thermal stress that can cause problems. But it’d be pretty low on my concern list as computer components go.