Live data from Hacker News

RAM now represents 35 percent of bill of materials for HP PCs

arstechnica.com

201–210 of 355 posts

Re: RAM now represents 35 percent of bill of materials for HP PCs

#201
post #39

Earlier quoted context omitted.

The worry is that these high prices aren't going to last long. And by the time you spend years building the capacity, the prices plummet making your facility uneconomical to run. Ram will always be in some demand, but that doesn't mean it's viable for everyone to start building production.

There's a few things to note here: 1) Prices aren't returning to "normal". The only way they will is if the hyperscalers and AI companies start to implode -- which will kill a huge portion of the US economy and lead to global recession, so, cheap RAM but nobody can afford it 2) By building up capacity you influence the outcome. If someone else enters the DRAM space, the duopoly has to actually start thinking about co…

> 1) Prices aren't returning to "normal".

> The only way they will is if the hyperscalers and AI companies start to implode -- which will kill a huge portion of the US economy and lead to global recession, so, cheap RAM but nobody can afford it

RAM isn’t some commodity that gets mined at a fixed rate and therefore costs more when people want large amounts of it. It’s a manufactured good, made from raw materials that are available in huge quantity, that was produced and sold at a profit at 2024 prices, even accounting for the capex needed to produce it.

Two things have changed. First, demand increased quickly. Second, big buyers sort of demonstrated that they’re willing to pay current prices, at least temporarily, so maybe the demand price elasticity has changed, or at least people’s perception of it has changed.

None of prevents the price from going back down. The high prices have made it economical for new manufacturers to invest more to compete — look at CXMT. And CXMT doesn’t have EUV machines, which doesn’t appear to be a showstopper for them.

Re: RAM now represents 35 percent of bill of materials for HP PCs

#202

I had my formative years in programming when memory usage was something you still worried about as a programmer. And then memory expanded so much that all kinds of “optimal” patterns for programming just become nearly irrelevant. Will we start to actually consider this in software solutions again as a result?

I just heard in a podcast, they talked about how powerful our devices are today but do not feel faster than they did 15 years ago and that it's because of what you write here.

A lot of that is on the OS vendors (and security requirements drive some inefficiencies that didn't used to be needed either).

Re: RAM now represents 35 percent of bill of materials for HP PCs

#203

If you are on Linux you can 'download' some RAM. Enable zram, configure sysctl variables to make good use of it. Note that it won't help you if your workload makes use of all your RAM at once. If you have a bunch of stuff running in the background it will help a lot. I get 2 to 3 compression factor at all times with zstd. I calculated the utility to be as if I had 20GB extra RAM for what I do.

That’s not particularly unique, Mac OS and Windows have had compressed memory for years. No fiddling with setting needed either.

It's not just that it's compressed - the OS is also intelligently handling which memory should be on hardware vs. virtual. Effectively a lot of the memory concerns have been offloaded to the OS and the VM where one exists.

Re: RAM now represents 35 percent of bill of materials for HP PCs

#204

Maybe this RAMmageddon will trigger a wave of optimized softwares that don't need GBs of memory for anything and everything.

Let's see. A) Programmers will get their shit together and start shipping lean software. OR B) New laptops will become neutered thin clients, and all the heavy lifting will be done by cloud service providers. Which one seems more likely?

B. But those neutered thin clients do not have enough ram to run software anyway even if all the real computation is in the cloud.

Re: RAM now represents 35 percent of bill of materials for HP PCs

#206
post #9

I think Europe should invest into manufacturing RAM. RAM isn't going anywhere, all of modern compute uses it. This would be an opportunity to create domestic supply of it.

> Europe should invest into manufacturing RAM

It should. And it should enact the political reforms they would make large capital projects like fabs possible. The current confederacy is proving just as much a stepping stone for Europe as it was for America. I’m not saying a full united Europe should emerge. But a system of vetoes is barely a system at all.

Re: RAM now represents 35 percent of bill of materials for HP PCs

#207

Earlier quoted context omitted.

>Modern RAM is made in fabs, which are ridiculously expensive to manufacture. Modern EUV lithography machines cost around 500M each. They're manufactured by hand. Only one company in the world knows how to manufacture them right now. You're wrong here. You don't need the most cutting edge ASML EUV machines to make RAM. Most RAM fabs still use standard DUV.

> You're wrong here. You don't need the most cutting edge ASML EUV machines to make RAM. Most RAM fabs still use standard DUV. Ah. Please check that. Which types of DRAM can be made in a DUV fab? Obviously the older ones, but are those obsolete for new computers. This really matters.

CXMT’s entire portfolio is made without EUV, and CXMT claims to have acceptable yield and performance comparable to other producers.

Keep in mind that the high bandwidths of modern RAM modules aren’t really a property of the RAM cells so much as a property of the read and write circuitry and the DDR or HBM transceivers, and those are a large part of the IP but a small part of the die. There is no such thing as “double data rate” or “high bandwidth” DRAM cells. Even DRAM cells from the 1990s could be read in microseconds. Reading and streaming your fancy AI model weights is an embarrassingly parallel problem and even 1 TB/sec does not even come close to stressing the ability of the raw cells to be read. This in contrast to, say, modern tensor processors where the actual ALUs set a hard cap on throughput and everyone works hard to come closer to the cap.

Take a look at what makes a modern computer with good RAM performance work: it’s the interconnect between the RAM and processor.

Re: RAM now represents 35 percent of bill of materials for HP PCs

#208
post #203

Earlier quoted context omitted.

That’s not particularly unique, Mac OS and Windows have had compressed memory for years. No fiddling with setting needed either.

It's not just that it's compressed - the OS is also intelligently handling which memory should be on hardware vs. virtual. Effectively a lot of the memory concerns have been offloaded to the OS and the VM where one exists.

Isn't that literally every modern OS, always, unless you tell it to act differently?

Re: RAM now represents 35 percent of bill of materials for HP PCs

#209
post #94

Earlier quoted context omitted.

AI demand isn't going away. It will just move from the data center to the local machine. On device AI is much better for the customer than it being in the cloud. Expecting people to stick with a few dozen gb of hbm is going to be the 'no one needs more than 640kb' of the 2030s.

> AI demand isn't going away. I'm not sure about that. When was the last time you have used Copilot prompt in Run dialog or Notepad?

About 10 minutes ago in Emacs.

Re: RAM now represents 35 percent of bill of materials for HP PCs

#210

Earlier quoted context omitted.

> And then memory expanded so much that all kinds of “optimal” patterns for programming just become nearly irrelevant. I don't think that ever happened. Using relatively sparse amount of memory turns into better cache management which in turn usually improves performance drastically. And in embedded stuff being good with memory management can make the difference between 'works' and 'fail'.

The need to use optimal patterns didn't go away, but the techniques certainly did. Just as a quick example, it's usually a bad idea now to use lookup tables to accelerate small math workloads. The lookup table creates memory pressure on the cache, which ends up degrading performance on modern systems. Back in the 1980s, lookup tables were by far the dominant technique because math was *slow.*

The way to approach this is to benchmark and then pick the best solution.
Post reply on HN