When I was in computer science in college back in the 1900s, they told us that the USSR was responsible for some of the best algorithms still in use at the time (both for RAM and CPU usage) because they just had less to work with.
Does it count as optimization? For supported hardware, it’s certainly cool. But it’s not making inefficient software better, it’s making better use of the hardware.
So using the back of a knife to cut something is not more efficient than using the sharp side? I would say making better use of hardware is also an efficiency gain. You use the same hardware to do more, while not driving up cost significantly. That is efficiency in my book.
It’s not using the same hardware, this seems to use a special compression chip.
To me this feels more like the benefits of a CPU or GPU upgrade. Things run faster, but not due to software improvements. When people talk about things getting optimized due to the high RAM prices, I think most people are referring to software improvements.
In another comment, I posted a link to the CRAM presentation that mentioned a chip.
This article also mentions some hardware:
Because CRAM is stored in RAM and treated as RAM, with full cacheline/byte access, it can be accessed in a read-only fashion with little delay; just the cost of hardware-offloaded compression.
I don’t think it would say that if they meant the CPU was doing the compression.
Whenever big tech companies like Meta do useful things like support Overture Maps Foundation (which is also supported by Microsoft and Amazon) they are doing it to blunt some competitive advantage another tech company has, such as google in this case. They support open source insofar as they can’t monetize whatever they are open sourcing.
So there are some benefits to the ram crisis? People actually optimize things because of limited resources.
This is still just a bandaid over running 10 copies of chromium for each electron app.
When I was in computer science in college back in the 1900s, they told us that the USSR was responsible for some of the best algorithms still in use at the time (both for RAM and CPU usage) because they just had less to work with.
Guess we’re all in that boat now.
That certainly seems to be a genre of article right now.
I feel like they might want to start with something a little simpler, like landing the first humans on Jupiter or making Windows 11 fly on 1024GB RAM.
Does it count as optimization? For supported hardware, it’s certainly cool. But it’s not making inefficient software better, it’s making better use of the hardware.
So using the back of a knife to cut something is not more efficient than using the sharp side? I would say making better use of hardware is also an efficiency gain. You use the same hardware to do more, while not driving up cost significantly. That is efficiency in my book.
It’s not using the same hardware, this seems to use a special compression chip.
To me this feels more like the benefits of a CPU or GPU upgrade. Things run faster, but not due to software improvements. When people talk about things getting optimized due to the high RAM prices, I think most people are referring to software improvements.
There is no mention of any extra hardware here though. The article even suggests this to benefit existing hardware like the steam deck.
In another comment, I posted a link to the CRAM presentation that mentioned a chip.
This article also mentions some hardware:
I don’t think it would say that if they meant the CPU was doing the compression.
Sounds like optimisation.
You know it’s getting weird out there when Meta seems to be doing something that’s actually useful.
they ain’t doing it for us, they’re doing it for themselves.
Whenever big tech companies like Meta do useful things like support Overture Maps Foundation (which is also supported by Microsoft and Amazon) they are doing it to blunt some competitive advantage another tech company has, such as google in this case. They support open source insofar as they can’t monetize whatever they are open sourcing.
ZSTD and BYRFS, among many other smaller things, hello?
They’ve been doing useful engineering things since the start, it’s just that it pales in comparison to all the bad things
They only did a few useful things to be more efficient in doing the bad things
Small performance improvements translate into large savings when you got millions of servers.
Necessity is the mother of invention.
This is very true, cheap RAM and storage just lead to inefficient, bloated code.