Moore’s Law is dead – Long live the chiplet • Hacker News Board

Re: Moore’s Law is dead – Long live the chiplet

#101

Post by briffle » Fri, Oct 07, 2022, 4:20 PM UTC

pjmlp wrote: post #85

PaulHoule wrote: post #70

Moore’s law is alive but the benefits are diminishing. Until 2005 or so, shrinking transistors automatically increased speed and reduced power consumption. When that ran out of steam, the industry went to multi core and massive parallelism with GPUs. Until recently each shrink also lowered the cost per transistor, but that seems to have run out also and has something to do with why Intel was stuck at 14nm for so long…

The problem is that GPUs are mostly underutilized outside games and machine learning, because the industry still hasn't moved away from the concept only a few selected group of developers can enjoy tooling to program them. So everyone that works in other domains, without access to libraries written by the GPU druids, largely ignores their existence.

Can you blame them? I just built a nice custom PC for my Son, with an 6 core cpu (with graphics, a ryzen -g class) 32GB of ram, 1TB NVMe hard drive, nice case, etc.

That cost about the same as a single mid-range video card. (Nvidia RTX 3070)

Why on earth would you add a requirement to your software/workflow that doubles your cost, and is just about impossible to find in stock?

Re: Moore’s Law is dead – Long live the chiplet

#102

Post by irrational » Fri, Oct 07, 2022, 4:40 PM UTC

briffle wrote: post #101

pjmlp wrote: post #85

Earlier quoted context omitted.

The problem is that GPUs are mostly underutilized outside games and machine learning, because the industry still hasn't moved away from the concept only a few selected group of developers can enjoy tooling to program them. So everyone that works in other domains, without access to libraries written by the GPU druids, largely ignores their existence.

Can you blame them? I just built a nice custom PC for my Son, with an 6 core cpu (with graphics, a ryzen -g class) 32GB of ram, 1TB NVMe hard drive, nice case, etc. That cost about the same as a single mid-range video card. (Nvidia RTX 3070) Why on earth would you add a requirement to your software/workflow that doubles your cost, and is just about impossible to find in stock?

I remember building my own PCs back in the 90s and early 2000s. In 2022 I don’t think any of my kids has ever even seen a desktop computer. It is all laptops, tablets, and phones.

Re: Moore’s Law is dead – Long live the chiplet

#103

Post by pjmlp » Fri, Oct 07, 2022, 4:43 PM UTC

zarzavat wrote: post #92

pjmlp wrote: post #85

Earlier quoted context omitted.

The problem is that GPUs are mostly underutilized outside games and machine learning, because the industry still hasn't moved away from the concept only a few selected group of developers can enjoy tooling to program them. So everyone that works in other domains, without access to libraries written by the GPU druids, largely ignores their existence.

You can have compute shaders in WebGL2, and WebGPU is around the corner. GPU power is available but then you run into the thorny issue of specs... Consumer machines vary wildly in their GPU capabilities, especially VRAM. So how do you know that your nice accelerated algorithm is going to work if the user has an old GPU? And what do you do if it doesn’t work? Run on the CPU? Tell the user their machine is too weak? He…

Good luck letting the average JavaScript coder take advantage of them.

This is the problem, GPUs are still a very specialised skill.

Re: Moore’s Law is dead – Long live the chiplet

#104

Post by duped » Fri, Oct 07, 2022, 4:52 PM UTC

PaulHoule wrote: post #70

Moore’s law is alive but the benefits are diminishing. Until 2005 or so, shrinking transistors automatically increased speed and reduced power consumption. When that ran out of steam, the industry went to multi core and massive parallelism with GPUs. Until recently each shrink also lowered the cost per transistor, but that seems to have run out also and has something to do with why Intel was stuck at 14nm for so long…

> If manufacturing overhead is low, two chiplets give you twice the transistors at twice the cost

That doesn't take into account yields. One chip with twice the transistors is physically larger than two chips with half as many, and more likely to have a defect during production.

Re: Moore’s Law is dead – Long live the chiplet

#105

Post by warmwaffles » Fri, Oct 07, 2022, 4:57 PM UTC

froggertoaster wrote: post #5

Moore's Law is dead - can anyone think of a more cliche article title in the tech world?

"I use arch btw"

Re: Moore’s Law is dead – Long live the chiplet

#106

Post by hedgehog » Fri, Oct 07, 2022, 5:15 PM UTC

nordsieck wrote: post #57

aargh_aargh wrote: post #19

For dummies like me who didn't know what a chiplet is: https://en.wikipedia.org/wiki/Chiplet This seems to be about the third reason listed: > Known good die (KGD): chiplets can be tested before assembly, improving the yield of the final device Problem: > In general, a killer defect is defined as a defect that is 20% the size of the fabrication node. For > example, a defect that is less than 9nm may be acceptable for…

The issue that you seem to be skirting around but not mentioning is, the chance a chip has a defect increases with die area since defects are randomly distributed across the surface of the wafer. Chiplets are a way for manufacturers to practically increase die area while keeping yields high.

In case it's not obvious the implication is that cost per good part starts to go up quickly. The wafer size and cost is essentially fixed so bigger dies mean both a lower % of parts are good and also fewer fit on the wafer in the first place. Wrong kind of hockey stick chart.

Re: Moore’s Law is dead – Long live the chiplet

#107

Post by JonChesterfield » Fri, Oct 07, 2022, 6:09 PM UTC

pjmlp wrote: post #85

PaulHoule wrote: post #70

Moore’s law is alive but the benefits are diminishing. Until 2005 or so, shrinking transistors automatically increased speed and reduced power consumption. When that ran out of steam, the industry went to multi core and massive parallelism with GPUs. Until recently each shrink also lowered the cost per transistor, but that seems to have run out also and has something to do with why Intel was stuck at 14nm for so long…

The problem is that GPUs are mostly underutilized outside games and machine learning, because the industry still hasn't moved away from the concept only a few selected group of developers can enjoy tooling to program them. So everyone that works in other domains, without access to libraries written by the GPU druids, largely ignores their existence.

The tooling is getting better. Debuggers are a thing now. You can program them in freestanding C++ with a little determination. Openmp target regions are friendlier syntax. Julia and a bunch of python machine learning things have GPU backends. They're still niche but slowly we make progress.

Re: Moore’s Law is dead – Long live the chiplet

#108

Post by 93po » Fri, Oct 07, 2022, 6:30 PM UTC

froggertoaster wrote: post #5

Moore's Law is dead - can anyone think of a more cliche article title in the tech world?

Breakthrough battery tech only 8 years away

Re: Moore’s Law is dead – Long live the chiplet

#109

Post by pjmlp » Fri, Oct 07, 2022, 7:05 PM UTC

PaulHoule wrote: post #89

pjmlp wrote: post #85

Earlier quoted context omitted.

The problem is that GPUs are mostly underutilized outside games and machine learning, because the industry still hasn't moved away from the concept only a few selected group of developers can enjoy tooling to program them. So everyone that works in other domains, without access to libraries written by the GPU druids, largely ignores their existence.

For one thing, most employers will refuse to issue a laptop with a real GPU to developers and other employees because they are afraid they will get used for games.

The integrated one would already be quite good, if there was a more mainstream way to make use of it.

Re: Moore’s Law is dead – Long live the chiplet

#110

Post by pjmlp » Fri, Oct 07, 2022, 7:38 PM UTC

galangalalgol wrote: post #96

pjmlp wrote: post #88

Earlier quoted context omitted.

Exposing GPU programming to anyone besides C, C++ and Fortran developers would already help, even if that would take a speed bump, as proven by the few attempts targeting PTX. I wasn't talking about Web apps.

Pytorch isn't just for ML, it can do normal signal processing or physics too. The julia libraries for cuda and roc and oneapi also are general enough for those uses and approachable. Both can fall back to cpu without much modification to the rest of your code. If you aren't doing signal processing, physics or something that would benefit from simd, then the gp is correct, a gpu won't do much for you.

Pytorch belongs to "... libraries written by the GPU druids..." on my comment.

And still requires specific skills to use, and is constrained to Python, C++ and Java based languages.

GPUs need to be exposed like SIMD, something that the language runtime takes care of, even if not perfect, better than not using them at all.