It’s not that they don’t work. It’s how businesses handle hardware.
I worked at a few data centers on and off in my career. I got lots of hardware for free or on the cheap simply because the hardware was considered “EOL” after about 3 years, often when support contracts with the vendor ends.
There are a few things to consider.
Hardware that ages produce more errors, and those errors cost, one way or another.
Rack space is limited. A perfectly fine machine that consumes 2x the power for half the output cost. It’s cheaper to upgrade a perfectly fine working system simply because it performs better per watt in the same space.
Lastly. There are tax implications in buying new hardware that can often favor replacement.
I agree that there is hyperbole thrown around a lot here and its possible to still use some hardware for a long time or to sell it and recover some cost but my experience in planning compute at large companies is that spending money on hardware and upgrading can often result in saving money long term.
Even assuming your compute demands stay fixed, its possible that a future generation of accelerator will be sufficiently more power/cooling efficient for your workload that it is a positive return on investment to upgrade, more so when you take into account you can start depreciating them again.
If your compute demands aren't fixed you have to work around limited floor space/electricity/cooling capacity/network capacity/backup generators/etc and so moving to the next generation is required to meet demand without extremely expensive (and often slow) infrastructure projects.
Sure, but I don't think most people here are objecting to the obvious "3 years is enough for enterprise GPUs to become totally obsolete for cutting-edge workloads" point. They're just objecting to the rather bizarre notion that the hardware itself might physically break in that timeframe. Now, it would be one thing if that notion was supported by actual reliability studies drawn from that same environment - like we see for the Backblaze HDD lifecycle analyses. But instead we're just getting these weird rumors.
I agree that is a strange notion that would require some evidence and I see it in some other threads but looking at the parent comments going up it seems people are discussing economic usefulness so that is what I'm responding to.
A toy example: NeoCloud Inc builds a new datacenter full of the new H800 GPUs. It rents out a rack of them for $10/minute while paying $6/minute for electricity, interest, loan repayment, rent and staff.
Two years later, H900 is released for a similar price but it performs twice as many TFlOps/Watt. Now any datacenter using H900 can offer the same performance as NeoCloud Inc at $5/month, taking all their customers.
It's because they run 24/7 in a challenging environment. They will start dying at some point and if you aren't replacing them you will have a big problem when they all die en masse at the same time.
These things are like cars, they don't last forever and break down with usage. Yes, they can last 7 years in your home computer when you run it 1% of the time. They won't last that long in a data center where they are running 90% of the time.
A makeshift cryptomining rig is absolutely a "challenging environment" and most GPUs by far that went through that are just fine. The idea that the hardware might just die after 3 years' usage is bonkers.
Crypto miners undervote for efficiency GPUs and in general crypto mining is extremely light weight on GPUs compared to AI training or inference at scale
With good enough cooling they can run indefinitely!!!!! The vast majority of failures are either at the beginning due to defects or at the end due to cooling! It’s like the idea that no moving parts (except the HVAC) is somehow unreliable is coming out of thin air!
There’s plenty on eBay? But at the end of your comment you say “a rate cheaper than new” so maybe you mean you’d love to buy a discounted one. But they do seem to be available used.
Do you know how support contract lengths are determined? Seems like a path to force hardware refreshes with boilerplate failure data carried over from who knows when.
> The exporter doesn't give a shit because it doesn't affect them at all.
Well that’s not true. Otherwise you are going to have to explain why so many outside the USA were upset with tariffs, and why there were retaliatory ones applied on the inverse.
I make no other claim that your quoted assertion is wrong.
Maybe you don't use the mouse because it just doesn't work as expected? ;)
> So I am wondering, are people fighting using a Mac in the most effective way simply because of old patterns and habit
"Most effective" doesn't mean "most intuitive". I don't want to learn keyboard shortcuts just to move or resize a window. That's the entire premise of graphical user interfaces.
What if your needs aren't as simple and you want to increase the size just a bit to fit more text than the tile permits and you don't want to waste the whole screen for that?
I really hate this type of blog. It pollutes the world with this attitude of “I messed up, how I have to frame the problem in a way, and write a blog that lets my ego stay intact”, which results in blogs like this showing in decision making process as “why you should not use go”. And mostly people never look past the title.
The fact is go is portable, it provides the ability to cross compile out of the box and reasonably executed on other platforms it supports. But in this case, a decision that had little to do with go, the desire to use c code, a non go project, with their go project made things harder.
These are not “just a set of constraints you only notice once you trip over them”, this is trivializing the mistake.
Entire blog can be simplified to the following.
We were ignorant, and then had to do a bunch of work because we were ignorant. It’s a common story in software. I don’t expect everybody to get it right the first time. But what we don’t need is sensational titled blogs full of fluff to try to reason readers out of concluding the obvious. Somebody in charge made decisions uninformed and as a result the project became more complicated and probably took longer.
What is happening to this person sucks for sure. But one thing I have learned in this industry is the thing being replaced is always better than the thing replacing it when you talk to the team who built and run the thing being replaced.
They might try. But nobody is going to pay for a sub par experience.
Some executives will get a big bonus for pushing this out, and later another for reverting back.
I won’t buy a car that does not have CarPlay. And I know many who are the same. My phone is a centerpiece of my life (gahh I hate saying that), my car however is not.
> AI will be the worst thing to happen to society in a very long time.
Maybe. Keep going back and forth.
On one hand I might loose my job. On then other hand everybody might loose their job.
Ai is tricky. If we have a singularity event maybe one or two combines might take all the jobs overnight. Fine. But economies are weird. Once those jobs are automated and nobody has a job we probably won’t even need the jobs that were replaced.
Like today. We have jobs because some other thing came along and “made something easy”. Think about how many jobs we have simply because we as humans write bad software. If this goes away it’s not even about automation taking jobs, it’s about simply not needing huge swaths of jobs at all.
So I think about this and ponder. I’d all Job are basically worthless, then the “rich” people like to complain about, won’t be rich. They won’t have anything either. Simply put, nobody will have any money to buy things and thus the “rich” won’t have anybody to buy things to keep them rich.
So I think more. It’s really not an advantage for the rich and powerful to basically destroy what makes them rich.
For people to be rich they have to have a bunch of people to extract small amounts of money from. A starving and angry population is not going to be a fun place to live for anybody.
I actually think the last point isn’t exactly true. I mean certainly in aggregate, but not at the margins. If the new tech billionaire elite just want to get to space, then they just need enough people to mine a some minerals and metal and make their space ships. If you control all the AI bots you can make money that way.
Kinda the point being, a small number of companies could control all the resources and a few of the people and be rich that way. Yes you’re right, maybe the Walton’s and the other families who made their money on people having money won’t go away, but in theory you could have a group of super rich people just giving money to each other to build space ships and nuclear power plants and the like.
Finally. It’s not a cut and dry as one side or the other. People have lost their minds. It’s case by case for every product and every consumer.
Some companies might chose to loose the margin (few but still passable ). Some might try to pass some or all to the sale price (which creates all sorts another dynamics) and finally the customer does not have to buy that product. There are many note breakdowns that all adjust who pays and when they pay.
I worked at a few data centers on and off in my career. I got lots of hardware for free or on the cheap simply because the hardware was considered “EOL” after about 3 years, often when support contracts with the vendor ends.
There are a few things to consider.
Hardware that ages produce more errors, and those errors cost, one way or another.
Rack space is limited. A perfectly fine machine that consumes 2x the power for half the output cost. It’s cheaper to upgrade a perfectly fine working system simply because it performs better per watt in the same space.
Lastly. There are tax implications in buying new hardware that can often favor replacement.