Dran

Those types of attacks are called “side-channel” attacks. If you wanted to look up stories of more. Fascinating history there.

Since nobody seemed to actually answer your question: the answer is that ram is actually really simple electrically. Modern DDR5 is very difficult and expensive to manufacture at scale, but is very simple to design.

If someone were to try and poison a memory package, it would be massively obvious by virtue of the package being larger, being very electrically noisy, or by sucking an order of magnitude more power to function.

DRAM sockets are not generic pci busses, and cannot be used on a typical motherboard to load arbitrary hardware the way USB or PCIE can.

Also, the way ram works, you really couldn’t do much more than read contents and relay via an on-dir radio, which would have to be super short range. Even something as “simple” as Bluetooth or wifi would be too big, too slow and take too much power to still function as a memory die.

You should be way more scared of cloud services, appliances, or iot devices than a stick of DDR5.

Tl/DR: it’d be prohibitively expensive and itd have nowhere to go, if it could even work at all without Corsair noticing.

For music enthusiasts plexamp is also basically unbeatable. I welcome the day open source catches up.

Basic output? probably. Gaming? Probably not. Again same problem of some cards missing hardware components that are required for a rendering pipeline, regardless of physical or virtual outputs

The 6000 Blackwell has all of this and would totally work, the H200 does not, etc

You could definitely do that for rasterization, but if any part of the graphics API required something the card didn’t have (like ray tracing) you couldn’t synchronize the work between it and another RT capable card fast enough.

See also: why they killed crossfire and sli in the gaming space, and why the derivative works of that technology only exists in the data center for non latency-intensive workloads.

Common misconception, display head =/= display socket.

It’s not just missing the physical plug, it’s missing the entire logic circuit dedicated to rendering and encoding a display output. Even if you wired an hdmi or dp to the board, it would have nothing to drive and time the signal.

That circuitry isn’t a separate chip it’s physically part of the GPU die architecture and you can’t add it on later via a mod.

AI GPUs either just use the same standard driver packages that gaming GPUs use, or they use slightly cut down ones because they physically lack internal hardware like display heads or rt cores. With those it’s not a matter of drivers, it’s a physical lack of hardware that a game engine or display would expect.

If you had an rtx 6000 Blackwell in your PC right now you’d just install the same drivers as the 5090, 5050, or 3060 (its the same driver across all three of these)

People shit on it but there’s a lot of good open-source tooling that supports it.

There are nist l1 profiles

Tutorials and guides for everything

etc

Part of being a good sysadmin is knowing when not to reinvent the wheel. Ubuntu has a lot of options for vetted, hardened, “other people’s wheels.”

Also, for posterity, the competent ones are running the headless, server version of Ubuntu. (As opposed to the bloated mess that is Ubuntu Desktop). The server version catches a lot of flack it doesn’t deserve.

Code rewrites are always going to have growing pains. Rewriting gnu-corrutils in rust is a noble effort.

Microvms or containers could give you external control of the networking. Then you would put whatever you want behind warp inside the warp container/vm.

+1 to this observation. I run zfs arrays at both home and work and it’s way more likely that your controller is flaking than you have that many simultaneous drive failures.

The unfortunate reality though is that you can’t trust the current copy of this data, even the snapshots, unless the restore passes a scrub post-restore.

I use discord quite heavily; the TOS violation and risk of ban has me a little risk averse when it comes to discord.

At 22mb it must reverse-engineer the protocol rather than just being a web wrapper.

Bring back the 3070ti and 3090 tbh

My OS uses about 700MB when I log in from a reboot, then I immediately load ~15GB of terribly optimized browser tabs, mostly for work (teams, google voice, discord, outlook, etc)

Inference is dirt cheap in comparison. Hundreds to thousands of concurrent users can be served by hardware costing in the high-thousands to low-ten-thousands.

Training those same foundational models is weeks to months of time on tens to hundreds of millions worth of hardware.

The simpler the arbitrary string/blob parsing logic the less this happens

https://app.opencve.io/cve/?product=grub2&vendor=gnu

I agree with you that it’d be nice if the cuts were a little shallower and allowed for an encrypted boot partition, but you could still have the system reasonably secure by encrypting the data partitions and signing the entire boot process to detect and abort decryption if the boot partition doesn’t match signatures. You already have to do this with the efi partition if you’re particularly paranoid about that attack vector, so this really isn’t a new one.

If caddy is acting as a proxy for anything, you should not need to forward that port externally. Local host firewalls allowing traffic on your local network is sufficient.

Depending on your physical host layout you may be looking at an issue with nat reflection.

You have not given us enough about your topology to assist in troubleshooting.

The tl/dw is that a retrospective of history often allows us to draw parallels to the present and predict the future.

There are server chips like the E7-8891 v3 which lived in a weird middle ground of supporting both ddr3 and ddr4. On paper, it’s about on par with a ryzen 5 5500 and they’re about $20 on US eBay. I’ve been toying with the idea of buying an aftermarket/used server board to see if it holds up the way it appears to on paper. $20 for a CPU (could even slot 2), $80 for a board, $40 for 32gb of ddr3 in quad chanel. ~$160 for a set of core components doesn’t seem that bad in modern times, especially if you can use quad/oct channel to offset the bandwidth difference between ddr3 and ddr4.

I think finding a cooler and a case would be the hardest part

+1 for Niagra. It takes a few days to get used to but it’s the launcher every power user didn’t know they wanted. Lifetime purchase options and a very responsive/passionate dev