5·
6 hours agoIf you set up a Kerberos server and use that for sign-in on both the Linux and MacOS systems, then you can have matching goetic demon accounts on both Linux and MacOS. :-)
Off-and-on trying out an account over at @[email protected] due to scraping bots bogging down lemmy.today to the point of near-unusability.
- 11 Posts
- 1.39K Comments
Joined 2 years ago
Cake day: October 4th, 2023
You are not logged in. If you use a Fediverse account that is able to follow users, you can follow this user.
linuxmemes
I believe that that’s a MacOS X system. One can only imagine what kind of madness goes on over there.
1·9 hours agoAh, thanks. Looks like they enabled zram in Fedora 33:
https://fedoraproject.org/wiki/Changes/SwapOnZRAM#Why_not_zswap?
I commented elsewhere in the thread that one option that can mitigate limited RAM for some users is to get a fast, dedicated NVMe swap device, stick a large pagefile/paging partition on it, and let the OS page out stuff that isn’t actively being used. Flash memory prices are up too, but are vastly cheaper than RAM.
My guess is that this generally isn’t the ideal solution for situations where one RAM-hungry game is what’s eating up all the memory, but for some things you mention (like wanting to leave a bunch of browser tabs open while going to play a game), I’d expect it to be pretty effective.
dev tasks, builds…etc
I don’t know how applicable it is to your use case, but there’s ccache to cache compiled binaries and distcc to do distributed C/C++ builds across multiple machines, if you can coral up some older machines.
It looks like Mozilla’s sccache does both caching and distributed builds, and supports Rust as well. I haven’t used it myself.
The big unknown that’s been a popular topic of discussion is whether Valve locked in a long-running contract for the hardware before the RAM price increases happened. If they did, then they can probably offer favorable prices, and they’re probably sitting pretty. If not, then they won’t.
My guess is that they didn’t, since:
-
They announced that they would hold off on announcing pricing due to still working on figuring out the hardware cost (which I suspect very likely includes the RAM situation).
-
I’d bet that they have a high degree of risk in the number of units that the Steam Machine 2.0 will sell. The Steam Deck was an unexpectedly large success. Steam Machine 1.0 kinda flopped. Steam Machine 2.0 could go down either route. They probably don’t want to contract to have a ton of units built and then have huge oversupply. Even major PC vendors like Dell and Lenovo got blindsided and were unprepared, and I suspect that they’re in a much less-risky position to commit to a given level of sales and doing long-running purchases than Valve is.
I’ve even seen some articles propose that the radical increase in RAM prices might cause Steam Machine 2.0’s release to be postponed, if Valve didn’t have long-running contracts in place and doesn’t think that it can succeed at a higher price point than they anticipated.
-
Honestly, a system with 64GB of memory is pretty well-provisioned compared to a typical prebuilt computer system from a major vendor.
I’ve felt that historically, PC vendors have always scrimped too far on RAM. In late 2025 with our RAM shortage, it’d be understandable, but in many prior years, it just looked like a false economy to me. Especially on systems with rotational drives — the OS is going to use any excess RAM for caching, and that’s usually a major performance gain if one has rotational drives sitting around.
EDIT: And battery. At least in 2025, a lot of people are using SSD storage, and caching that in RAM isn’t as huge a win as it is with rotational drives. But lithium batteries have gotten steadily cheaper over the years. The fact that smartphone, tablet, and laptop vendors aren’t jamming a ton of battery in their devices in 2025 is kinda crazy to me.
GPU prices
Outside of maybe integrated GPUs, I doubt it, because they need their own memory and are constrained by the same bottleneck — DRAM.
Or at least CPU prices?
I’ve read one article arguing that CPU prices will likely drop during the RAM shortage.
I don’t know if that’s actually true — I think that depends very much on the ability of CPU manufacturers to economically scale down their production to match demand, and I don’t know to what degree that is possible. If they need to commit to a given amount of production in advance, then yeah, probably.
Go back a couple years, and DRAM manufacturers — who are currently making a ton of money due to the massive surge in demand from AI — were losing a ton of money, because they couldn’t inexpensively rapidly scale production up and down to match demand. I don’t know what the economics are like for CPUs.
https://finance.yahoo.com/news/fear-dram-glut-stifling-micron-155958125.html
November 5, 2018
To be clear, the oversupply concerns that have plagued Micron Technology (NASDAQ:MU) shares for weeks now are completely valid. Micron stock has fallen as much as 40% just since June on this deteriorating dynamic.
In short, the world doesn’t need as many memory chips as Micron and rivals like Samsung (OTCMKTS:SSNLF) and SK Hynix are collectively making. The glut is forcing the price of DRAM (dynamic random access memory) modules so low that it’s increasingly tougher to make a buck in the business.
We had a glut of DRAM as late as early this year:
https://evertiq.com/news/56996
Weak Demand and Inventory Backlogs
Both the DRAM and the NAND markets are still in a state of oversupply, with excess inventory leading to significant price declines through Q4 2024 and Q1 2025. This is driven by multiple factors such as weak consumer demand.
Memory manufacturers ramped up production during previous periods of strong demand, but the market failed to meet these forecasts. This has resulted in inventory backlogs that now weigh on prices.
That’s fair, but my understanding is that VRChat, despite the name, isn’t a VR-only thing.
Honestly, I kinda wish that Bethesda would do a new release of Skyrim that aims at playing well with massive mod sets. Like, slash load time for huge mod counts via defaulting to lazy-loading a lot more stuff. Help avoid or resolve mod conflicts. Let the game intelligently deal with texture resolutions; have mods just provide a single high-resolution image and let the game and scale down and apply GPU texture compression appropriate to a given system, rather than having the developers do tweaking at creation time. Improve multicore support (Starfield has already done that, so they’ve already done the technical work).
I haven’t used it, but my understanding is that it’s vaguely like Second Life, popular with folks creating adult-content-oriented-worlds.
From a technical standpoint, that might actually be a pretty good example of a game that would benefit from cloud gaming, since I assume that it’s not all that latency-critical, not the way an FPS would be.
I guess that there would potentially be privacy issues with adult content stuff that would argue against cloud hosting, but in the case of VRChat, the service itself is already living in the cloud, so…shrugs
If it’s a leak in a mod and some pages just aren’t being accessed at all, then I’d think that the OS might be able to just page them out.
It might be possible to crank up the amount of swap you have and put that swap on a relatively-fast storage device. Preferably NVMe, or maybe SATA-attached SSD. I mean, yeah, SSD prices are up too, but you don’t need all that much space to just store swap, and it’s vastly cheaper than DRAM.
If you have a spare NVMe slot on your system or a free spot to mount a 2.5 inch SATA drive and SATA plug, should be good.
If you have a free PCIe slot, doing a quick Amazon search, looks like a PCIe card with a beefy heatsink to provide an M.2 slot to mount a single stick of NVMe can be had for $14:
https://www.amazon.com/Sabrent-NVMe-PCIe-Aluminum-EC-PCIE/dp/B084GDY2PW
And a 128GB M.2 stick of NVMe for $20:
https://www.amazon.com/GALIMU-128GB-XP2000-Gen4x4-XP2000F128GInternal/dp/B0FY4CQRYF
I have no idea the degree to which “lots of cheap, fast swap” helps. It will probably depend a lot on a particular use case. In some cases, probably about as good as having the memory. My guess is that in general, it’ll tend to be more helpful on systems running lots of programs than on systems running one large game (though a leak might change that up), but hard to say without actual testing.
If a flash storage device is really heavily used, I imagine that it’ll probably eat through its lifetime write cycles relatively quickly, but if nothing else lives on the device, no biggie if it fails (well, not in terms of data loss for stored stuff), and I don’t expect it being 5 or 10 years until DRAM prices come back down, so it doesn’t need to last forever.
Probably be interesting to see some gaming sites benchmark some of these approaches.
I assume this:
The tech giant previously announced that users can pay for Windows 10 Extended Security Updates to get patches for another year, but this week it revealed additional enrollment options, including free alternatives for individual users.
Specifically, consumers can pay roughly $30 per PC (depending on location) to enroll in the ESU program and receive security updates for one year after Windows 10 reaches EOS.
If they don’t want to spend money, they can simply start using Windows Backup to sync their settings to the cloud. It’s worth noting that Microsoft recommends Windows Backup for backing up files and settings before switching to Windows 11.
Another ESU option that does not involve spending actual money is to enroll for 1,000 Microsoft Rewards points, which users earn for engaging with Microsoft products and services, such as Bing, Xbox and Microsoft Store.
“ESU coverage for personal devices runs from Oct. 15, 2025, through Oct. 13, 2026,” Microsoft’s Yusuf Mehdi explained.
So you can get one extra year, but you need to tie the PC’s Administrator account to a Microsoft account, and either need to pay a $30 subscription fee, spend their Microsoft Rewards points, or set the PC to sync to their cloud service.
What is the feasibility of getting a prebuilt gaming PC and using it for the parts I need/want and selling the rest of it?
I’m sure that you could do that, but I think part of the problem there is that everyone else is going to be in the same boat, short of RAM, and I’m not sure what demand there is for a gaming PC stripped of its RAM.
If there isn’t much demand, you might have trouble recouping what you spent on the parts you don’t want.
I read one article that CPU prices may drop, because the increased RAM prices will drive up PC prices, price some people out of the market, and so there will be less demand for CPUs.
I thought the later, the better
Well, usually that is true.
Not a hardware fix, but there’s memory compression. It sounds like Windows 11 defaults to having memory compression on:
Linux has zswap and zram to do memory compression, which I’ve mentioned here recently. I don’t know of any distros that turn it on by default. It sounds from recent reading like for modern systems with SSD swap, zswap is probably preferable to zram.
As (relatively) old as they are, midrange Core i5 chips from Intel’s 12th-, 13th-, and 14th-generation Core CPU lineups are still solid choices for budget-to-midrange PC builds.
I would be hesitant about obtaining secondhand 13th or 14th gen desktop Intel CPUs, since those are the ones that destroy themselves over time. There is no way to know whether they’ve been run on non-updated BIOSes and damaged themselves. I burned through an i9-13900 and an i9-14900 myself. Started with occasional errors and gradually got worse until they couldn’t even get through boot. I am sure that there are lots of people trying to unload damaged processors (knowingly or unknowingly) that have only seen the early stages of damage.
12th-gen CPUs are safe.
Consider pre-built systems. A quick glance at Dell’s Alienware lineup and Lenovo’s Legion lineup makes it clear that these towers still aren’t particularly price-competitive with similarly specced self-built PCs. This was true before there was a RAM shortage, and it’s true now. But for certain kinds of PCs, particularly budget PCs, it can still make more sense to buy than to build.
I just picked up two Alienware PCs for relatives to take advantage of this window, but it was only something like a two-week window, where Dell announced at the beginning of December that they were doing price increases to reflect the RAM shortage mid-December. I believe that that window is closed now (or, well, it might still be cheaper to get DIMMs with a PC than separate, but not to get memory that way at pre-memory-shortage prices any more).
EDIT: From memory, Lenovo announced that they were doing their RAM-induced price increases at the beginning of January, so for Lenovo, it might still work for another week-and-a-half or so.
EDIT2: 15th gen Intel CPUs are also safe WRT damage, but like AMD’s AM5-socket processors, they can’t use DDR4 memory, which is what the author is trying to find a route to do.
Bind mounts aren’t specific to Docker. You’re asking specifically about bind mounts as used by Docker?
12·4 days agoartificial scarcity
From the article:
This implies that Nvidia cannot get enough GDDR7 memory to produce GPUs at its current rate. Alternatively, it implies that Nvidia expects significantly reduced GPU sales in 2026, possibly due to rising NAND and DRAM costs and their impact on PC prices.
That seems a lot more plausible to me than Nvidia trying to intentionally constrain GPU supplies. They just need to expect that the memory shortage will affect their gaming sales. We know that there is a massive RAM shortage. I mean, you don’t need to introduce other factors to explain that move.
On a similar note, I’ve seen articles talking about how PC vendors expect revenue to drop drastically in the near term (since they’ll have to hike prices because of the RAM constraints driving up their RAM costs, which in turn will price some consumers out of the market, which will reduce their sales until more supply of RAM relative to demand becomes available). I haven’t seen anything about them cutting production, but I would expect them to do so; otherwise they’d just build up a bunch of computers that won’t sell.
And I’d bet that there are other companies that will be impacted, all kinds of things that one might not think of as computer vendors that use memory in their devices and suddenly need to hike prices and expect to see sales impacted.
This isn’t a huge increase. It’s not on the order of what just happened with RAM.
According to a report from Digitimes Asia (quoting Nikkei), HDD contract prices jumped roughly 4% quarter over quarter in Q4 2025.
I don’t use Kali Linux, but it sounds like it’s based on Debian’s testing release. Debian hasn’t packaged Blackwell drivers yet, so I wouldn’t be surprised if Kali doesn’t have them packaged either. You can download Blackwell drivers from Nvidia, but the Debian guys won’t have made sure that things don’t break with them.
https://wiki.debian.org/NvidiaGraphicsDrivers
https://www.nvidia.com/en-us/drivers/details/259042/
I don’t know why you wouldn’t be able to install the driver with the graphics card inserted.
The initrd contains drivers that aren’t directly built into the kernel.
Typically, the way this works on Debian with third-party drivers is that you have the proper linux-headers package matching your current kernel installed. Then a third-party package registers a DKMS module with the driver source, and when you install a new kernel, the driver gets recompiled for that kernel. That driver gets dropped into the initrd, the ramdisk with the out-of-kernel stuff required to boot.
I don’t use Nvidia hardware, so I can’t tell you if that’s what’s supposed to happen, but I would guess so.
If you’re not booting with it, my guess is that something isn’t working as part of that process. Either the Nvidia script didn’t register the module or it didn’t get rebuilt or the installed driver has some issue and isn’t working when you try to load it.
You can probably run
sudo dkms statusand it’ll show DKMS modules and their current status. That might be a starting point.