←back to thread

335 points aspenmayer | 2 comments | | HN request time: 0.626s | source
Show context
GeekyBear ◴[] No.45008439[source]
Didn't we already cross this particular Rubicon during the auto bailout a decade ago?

Other examples:

> Since the 1950s, the federal government has stepped in as a backstop for railroads, farm credit, airlines (twice), automotive companies, savings and loan companies, banks, and farmers.

Every situation has its own idiosyncrasies, but in each, the federal government intervened to stabilize a critical industry, avoiding systemic collapse that surely would have left the average taxpayer much worse off. In some instances, the treasury guaranteed loans, meaning that creditors would not suffer if the relevant industry could not generate sufficient revenue to pay back the loans, leading to less onerous interest rates.

A second option was that the government would provide loans at relatively low interest rates to ensure that industries remained solvent.

In a third option, the United States Treasury would take an ownership stake in some of these companies in what amounts to an “at-the-market” offering, in which the companies involved issue more shares at their current market price to the government in exchange for cash to continue business operations.

https://chicagopolicyreview.org/2022/08/23/piece-of-the-acti...

replies(16): >>45008458 #>>45008466 #>>45008546 #>>45008710 #>>45008838 #>>45009730 #>>45009928 #>>45010439 #>>45010649 #>>45011116 #>>45011228 #>>45011332 #>>45012047 #>>45012973 #>>45014055 #>>45018837 #
JKCalhoun ◴[] No.45008710[source]
What happened to Intel? Did they need a bailout?
replies(5): >>45008786 #>>45009769 #>>45009869 #>>45010341 #>>45012781 #
downrightmike ◴[] No.45008786[source]
My intel powered workstations from 2008 have chips that are more than decent enough for modern computing.

But in reality, they gave up on R&D, Missed Apple mobile chips and then put a lot of MBAs in top leadership positions to make their financial schemes work. Those schemes did not work and they were left without technical leadership. When they got the tech leadership, the board just gave up and fired him and brought in a chop shop CEO to part intel out.

For some reason, the US admin thinks this is a good buy. You probably would too, if you bankrupted multiple casinos.

Intel is going the way of SunBeam, Sears and Toys r Us. The board failed to stop that. And failure attracts more failure.

replies(6): >>45009339 #>>45009848 #>>45010867 #>>45011514 #>>45012095 #>>45012513 #
pixelpoet ◴[] No.45009848[source]
Depends what you consider "modern computing"; I wouldn't like to try Rust development on a 2008 machine, let alone things like path tracing.
replies(2): >>45011910 #>>45011928 #
torginus ◴[] No.45011910[source]
Watch this video:

https://www.youtube.com/watch?v=m7PVZixO35c

New computers aren't as much faster as benchmarks would lead you to believe. If your code relies on unpredictable branches, dependept memory accesses in a tight loop, or very tight feedback loops inside computations, new CPUs won't have that much of an edge.

replies(1): >>45014049 #
1. pixelpoet ◴[] No.45014049[source]
I'm writing rendering engines, so quite up to speed on how fast CPUs and GPUs are, where the bottlenecks are etc :) In fact I worked on the latest Cinebench as part of Redshift dev.
replies(1): >>45017244 #
2. torginus ◴[] No.45017244[source]
Ah sorry, didn't mean to sound so haughty, I really just quickly type these comments off before I start my workday, really should give more thought to sounding a bit more tactful.

It's really cool they you get to meet people like you here in the comments :)

However I think the video raises an interesting point - I imagine you're an expert on modern CPU optimization, and spend quite a bit of time thinking about cache access patterns, exposing instruction-level parallelism to the CPU and other such gory bits without which your code wouldn't be running as fast on modern CPUs.

Most of us regular folks however don't really write code like that, and I think for that sort of code, I don't think modern CPUs run as much faster than an ancient Core 2 Duo, as benchmarks would suggest, but we can expect results somewhat more like in the video I linked.

Congrats on your work getting used in Cinebench, but I can't help but ask you - what's your thoughts on your work getting used to basically benchmark every CPU under the sun by tech reviewers, who then make recommendations on whether to buy that CPU or another one. Looping back to the previous point, I think your code has quite likely very different performance characteristics than the sort of generic stuff most people end up running, and this sort of benchmarking creates an incentive for manufacturers to crank out CPUs that are good at running hyper-optimized path tracing code, while ignoring real-world performance. What are your thoughts on that?