Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Moore’s Law is dead – Long live the chiplet

semiwiki.com

Post reply on HN
Print view View original

91–100 of 123 posts

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

#91

Post by adwn »

pjmlp wrote: post #88
miohtama wrote: post #87

Earlier quoted context omitted.

I thought most generic computation workloads are ill-suited for GPUs. A normal web SaaS application is full of if branches and JMP instructions. Running this on GPU would slow it down, not speed it up.

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.

The point stands, the vast majority of workloads are unsuited for GPUs, either because they are full of divergent branches, or because the data transfer and synchronization overheads would cancel any performance gains.

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

#92

Post by zarzavat »

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.

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?

Here the advantage of GPUs (performance) is also the biggest disadvantage: a gigantic range of performance profiles. At least with CPUs the oldest CPU is only going to be a small integer factor slower than a new one in single thread.

What unites gamers and machine learning is an expectation that the user has a reasonably recent and capable GPU. But these are small, self-selecting populations.

On the server side the issue is cost. GPUs are expensive, and usually not necessary, so nobody is going to write code that requires one without a good reason.

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

#94

Post by TomVDB »

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.

That seems like a stretch? Isn’t the more obvious explanation that laptops with a real GPU are much more expensive and that the weaker, integrated GPUs are more than good enough for the vast majority of business use?

Today’s iGPUs are fast enough comfortable run plenty of games.

I have work provided high-end POS Dell Precision engineering laptop. It has an Nvidia discrete GPU, but I don’t think I’ve ever actually needed its power, and I’d gladly trade it for a laptop without…

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

#95

Post by WastingMyTime89 »

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.

That’s clearly untrue. Employers source computers from a few selected vendors and generally issue computers with average specifications because they can buy them in bulk and they are good enough. You can get a laptop with a more powerful GPU at most place if you actually need one.

No one is scared of employees gaming. Employees can’t install applications themselves on their laptops at most place.

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

#96

Post by galangalalgol »

pjmlp wrote: post #88
miohtama wrote: post #87

Earlier quoted context omitted.

I thought most generic computation workloads are ill-suited for GPUs. A normal web SaaS application is full of if branches and JMP instructions. Running this on GPU would slow it down, not speed it up.

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.

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

#97

Post by mjburgess »

mildmotive wrote: post #47
Ma8ee wrote: post #36

Earlier quoted context omitted.

But the inefficiencies you sometimes see today can’t even be explained with any bad choice of language. You can create more than fast enough programs with interpreted languages with garbage collection. But then of course you need to know at least a little bit about data structures and not doing dozens of REST calls anytime anyone taps the screen.

I only partially agree, an interpreted language can act as a hinderance when doing optimisations, for various reasons. But how we write the code also matters of course. Furthermore, competency also costs money. I’m not saying this to be mean to people abusing REST calls.

Most code is so bad that "optimization" isnt the issue, at least in no sense that a language/runtime would help with.

It's just designed with O(N^3) architectures, "servers on servers" and "callbacks on callbacks", etc. because, seemingly, new features are made by very jnr developers -- who glue these idiot-proof frameworks together ---- trading needed-knowhow for polynomial behaviour.

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

#98

Post by psykus »

dagw wrote: post #37
michaelt wrote: post #27

Earlier quoted context omitted.

> Have clock speeds really plateaued? Pentium 4 HT 3.8F, November 2004, 3.8GHz, 115W TDP Core i9-13900KF, October 2022, 3.0GHz, 125W TDP Of course, the latter does give you 8 performance cores and 16 efficiency cores so performance-per-watt has clearly improved; and it has 'turbo boost'. But in terms of sustained single-core performance? It's clear Intel's attention has been elsewhere. Such as on the laptop market, w…

CPUMark single threaded score for the P4: ~620 CPUMark single threaded score for the 13900K ~4800 So obviously a lot has changed and improved.

I'm curious if single threaded* games/applications that were CPU limited when the P4 originally came out run better on the 13900k with the same code.

That was my impression from games at the time, that they were coded with an expectation that clock speeds would keep going up in the future. But they didn't and the games probably run just as bad now as they did before.

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

#99

Post by brundolf »

ordu wrote: post #38
brundolf wrote: post #17

Lotta people here not reading the article: > However, as it has in the past, the semiconductor ecosystem is adapting and as Chiplet technology builds traction, we will very likely see a period of accelerating innovation and new market opportunities opening as we move forward. The whole premise is that chip innovation (and overall computing power) is continuing to accelerate, even though "Moore's Law as we've known it…

I'd say that the premise of the article is captured in the title perfectly. One needn't to read the article to get it. So, we can strengthen your assertion and to say that "lotta people here are not reading the title." People react to the first part of the title. Seems they are really exasperated by all these repeated deaths of Moore's law.

In the title it's a little subtle due to the wordplay and terminology, but yeah, it is present. Definitely a lot of knee-jerking going on

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

#100

Post by PaulHoule »

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.

That said people are always discovering algorithms that get better performance than you'd expect from new hardware.

For instance a frightening amount of CPU is spent in financial messaging systems on validating UTF-8, parsing XML and JSON, converting numbers written in decimal digits to binary and things like that. You'd think these are "embarrassingly serial" problems but with clever coding and advanced SIMD instructions such as AVX-512 they can be accelerated for throughput, latency, and economy.

The benefits of the GPU are great enough that you might do more "work" but get the job done faster because it can be done in parallel.

For instance the algorithms used by the old A.I. ("expert systems") parallelize better than you might think (though not as well as the Japanese hoped they would in the 1980s) despite being super-branchy. Currently fashionable neural networks (called "connectionist" back in the day) require only predicated branching (which side of the ReLU are you on?) but spend a lot of calculations on parts of the network which might not be meaningful for the current inference. It depends on the details, but you might be better doing many more operations if you can do them in parallel.

Given that GPUs are out there and that so many people are working on them I think the range of what you can do with them is going to increase, though I think few people will be writing application logic on them directly, but they will increasingly use libraries and frameworks. For instance, see

https://arxiv.org/pdf/1709.02520.pdf

Post reply on HN