3 ms·

Is your argument “if I remove the context and then negate the details, then the sentence is meaningless…”? Sure, I agree with that. There is plenty of informa

Is your argument “if I remove the context and then negate the details, then the sentence is meaningless…”? Sure, I agree with that. There is plenty of information in the statement: > Ryzen AI software lets developers take machine learning models […] and deploy them on laptops powered by Ryzen AI processors […]. If I have a passing understanding of computers, I know everything I need to decide if this is interesting to me or not. It is for machine learning developers. It helps them deploy their models to a specific AI processor.

roenxi 3y ago
The sentence is meaningful both ways. The argument is if [X] and [not X] are both true, X doesn't rule out any possible universes. And if it doesn't rule out any possible universes, it doesn't tell us anything about what they have done - because universes where they didn't do anything are still valid. That is a handy test of whether a sentence is marketing fluff or whether it contains data. > If I have a passing understanding of computers, I know everything I need to decide if this is interesting to me or not. It is for machine learning developers. It helps them deploy their models to a specific AI processor. I can already deploy models to my processor though, and it is an AMD processor that doesn't support Ryzen AI. I've used it to develop AI models too, since I have an AMD GPU and can't use that to develop AI models because it crashes the driver when I do. So while you can deploy models to a Ryzen AI processor, you can also deploy models to non-Ryzen AI processors. So, logically and practically, they haven't told us anything. The issue is with the limited understanding you've taken on for your argument, you didn't realise that all computers can already do what they promised Ryzen AI processors can do. Compare that to eightysixfour who articulates what the thing actually does - lots of TOPS, and a 150% improvement in TOPS on the previous generation. His sentence contains information and I doubt my CPU can match that TOPS performance. And I wasn't developing an INT8 model which is probably also relevant. Furthermore, it is that in 10 years or so models will be much bigger and more power hungry, possibly with different architectures. So there will be models that we won't be able to deploy on many CPUs even if they are powered by Ryzen AI. So not only is the documentation on Ryzen AI not telling me what they can do, even what the documentation promises can't be logically done. Which isn't a problem because that is more of a silly "gotcha!" style observation, but it really highlights how little useful information is in the docs intro material and marketing copy when it comes to identifying what it is this thing does.
- eightysixfour 3y ago
  First, I am eightysixfour. Second, if someone says “AI processor” the context is quite obvious that it is a dedicated AI accelerator, like a Coral TPU. Especially since the first sentence on the page defines it. > AMD Ryzen™ AI Software enables developers to take full advantage of AMD XDNA™ architecture integrated in select AMD Ryzen AI processors. Third, > you didn't realise that all computers can already do what they promised Ryzen AI processors can do In this text they didn’t promise Ryzen AI processors could do anything, they promised the Ryzen AI software would let you deploy models to Ryzen AI processors. > The issue is with the limited understanding you've taken on for your argument, you didn't realise that all computers can already do what they promised Ryzen AI processors can do. I have deployed machine learning models to the edge in production on accelerators and CPUs. I use Llamas on my home rig, both CPU and CUDA. I am quite aware of what they can do, which is why I already knew the performance of the Ryzen AI processor. You are the one whose limited understanding kept you from recognizing that “Ryzen AI processor” is a distinct thing from a CPU or GPU. Your complaint is the equivalent of complaining that the home page for CUDA’s documentation doesn’t explain what a GPU is and why it is different from a CPU, which can do the same computations. You also seem moderately confused about the use cases for these relatively low performance, but very low power ML accelerators. The first thing that shipped on Ryzen’s AI accelerator is a model from Microsoft that makes it look like you are making eye contact with a camera. Not a giant LLM. They don’t compete with GPUs.
  - roenxi 3y ago
    > Second, if someone says “AI processor” the context is quite obvious that it is a dedicated AI accelerator, like a Coral TPU. Especially since the first sentence on the page defines it. The first sentence doesn't define anything, the first sentence is "AMD Ryzen™ AI Software enables developers to take full advantage of AMD XDNA™ architecture integrated in select AMD Ryzen AI processors. Ryzen AI software intelligently optimizes AI tasks and workloads, freeing-up CPU and GPU resources, providing optimal performance at lower power." Which is at least better than meaningless but still doesn't explain what Ryzan AI is going to do. An architecture isn't an end goal. Intelligently optimizing AI tasks might be, but frankly AMD has been so behind in this space I don't believe the vapidity without an actual "we implement the such-and-such algorithm" or "the optimisation we refer to is ...". They're providing no evidence to believe they've optimised anything and their track record inspires little confidence. In my experience their AI software can't do BLAS while my GPU is running an X server and so "optimised" could just mean that I will be turning my computer off by choice after I buy one. Which would be a welcome improvement I admit, but they aren't exactly spelling out what the win is going to be. But communicating that there is a part of a processor that processes tasks is not informative. They might not even process "AI" for all I know, the AI I want to run on my GPU right now isn't represented as an ONNX model. Not a problem, but it continues to go to how low in information the headline copy is. "AI" isn't a real term, it is an umbrella. None of the AIs for the games I know of can be offloaded to Ryzen AI, for example. So there is some fairly specific set of things that Ryzan AI does which they aren't telling me. You are, but you obviously weren't involved in writing the documentation (or, I suppose, if you were the evidence is your editor murdered your work). Now if you already know what the chip does then sure, you don't need any of the copy or documentation. But there has to be an on ramp for people who don't know and they didn't include anything useful. > ... they promised the Ryzen AI software would let you deploy models to Ryzen AI processors. You can't name a processor that models can't be deployed to. They're big slightly-nonlinear operations. Every processor can do that. The sentence in the docs is literally information free. > I have deployed machine learning models to the edge in production on accelerators and CPUs. I use Llamas on my home rig, both CPU and CUDA. I am quite aware of what they can do, which is why I already knew the performance of the Ryzen AI processor. You are the one whose limited understanding kept you from recognizing that “Ryzen AI processor” is a distinct thing from a CPU or GPU. And you know what is interesting about this long list? None of it involves reference to the marketing copy or documentation that I'm talking about. Because as far as a cursory glance goes, nothing explains what "Ryzen AI" does. Just to circle back, you proffered a link to that documentation as a description of what these chips will do. Now I suspect you're about to argue that someone who isn't already in the field shouldn't bother with it. That says about as much as I have about the quality of the documentation - it isn't very good. im3w1l back at the thread root had a very valid question. > Your complaint is the equivalent of complaining that the home page for CUDA’s documentation doesn’t explain what a GPU is and why it is different from a CPU, which can do the same computations. THe CUDA homepage (https://developer.nvidia.com/cuda-toolkit https://developer.nvidia.com/cuda-toolkit), first paragraph, says "The toolkit includes GPU-accelerated libraries, debugging and optimization tools, a C/C++ compiler, and a runtime library." Then a bit further down the page has a link to the libraries they're talking about: https://developer.nvidia.com/gpu-accelerated-libraries https://developer.nvidia.com/gpu-accelerated-libraries. Which include, BLAS, FTT a random number generator, some other cool things. 10/10, I can tell what CUDA does. Does Ryzan AI do all this stuff? It allows me to deploy models and I'd need all that to deploy a models. I doubt it does. But the docs don't say what it does in the intro so... I dunno. It is probably in there somewhere. They defined ROCm as an umbrella term that included ROCm software once in their ROCm docs, so anything is possible. Being unable to explain what it is they are doing is a real pattern with AMD in the AI space.

roenxi 3y ago

The sentence is meaningful both ways. The argument is if [X] and [not X] are both true, X doesn't rule out any possible universes. And if it doesn't rule out any possible universes, it doesn't tell us anything about what they have done - because universes where they didn't do anything are still valid. That is a handy test of whether a sentence is marketing fluff or whether it contains data. > If I have a passing understanding of computers, I know everything I need to decide if this is interesting to me or not. It is for machine learning developers. It helps them deploy their models to a specific AI processor. I can already deploy models to my processor though, and it is an AMD processor that doesn't support Ryzen AI. I've used it to develop AI models too, since I have an AMD GPU and can't use that to develop AI models because it crashes the driver when I do. So while you can deploy models to a Ryzen AI processor, you can also deploy models to non-Ryzen AI processors. So, logically and practically, they haven't told us anything. The issue is with the limited understanding you've taken on for your argument, you didn't realise that all computers can already do what they promised Ryzen AI processors can do. Compare that to eightysixfour who articulates what the thing actually does - lots of TOPS, and a 150% improvement in TOPS on the previous generation. His sentence contains information and I doubt my CPU can match that TOPS performance. And I wasn't developing an INT8 model which is probably also relevant. Furthermore, it is that in 10 years or so models will be much bigger and more power hungry, possibly with different architectures. So there will be models that we won't be able to deploy on many CPUs even if they are powered by Ryzen AI. So not only is the documentation on Ryzen AI not telling me what they can do, even what the documentation promises can't be logically done. Which isn't a problem because that is more of a silly "gotcha!" style observation, but it really highlights how little useful information is in the docs intro material and marketing copy when it comes to identifying what it is this thing does.

eightysixfour 3y ago
First, I am eightysixfour. Second, if someone says “AI processor” the context is quite obvious that it is a dedicated AI accelerator, like a Coral TPU. Especially since the first sentence on the page defines it. > AMD Ryzen™ AI Software enables developers to take full advantage of AMD XDNA™ architecture integrated in select AMD Ryzen AI processors. Third, > you didn't realise that all computers can already do what they promised Ryzen AI processors can do In this text they didn’t promise Ryzen AI processors could do anything, they promised the Ryzen AI software would let you deploy models to Ryzen AI processors. > The issue is with the limited understanding you've taken on for your argument, you didn't realise that all computers can already do what they promised Ryzen AI processors can do. I have deployed machine learning models to the edge in production on accelerators and CPUs. I use Llamas on my home rig, both CPU and CUDA. I am quite aware of what they can do, which is why I already knew the performance of the Ryzen AI processor. You are the one whose limited understanding kept you from recognizing that “Ryzen AI processor” is a distinct thing from a CPU or GPU. Your complaint is the equivalent of complaining that the home page for CUDA’s documentation doesn’t explain what a GPU is and why it is different from a CPU, which can do the same computations. You also seem moderately confused about the use cases for these relatively low performance, but very low power ML accelerators. The first thing that shipped on Ryzen’s AI accelerator is a model from Microsoft that makes it look like you are making eye contact with a camera. Not a giant LLM. They don’t compete with GPUs.
- roenxi 3y ago
  > Second, if someone says “AI processor” the context is quite obvious that it is a dedicated AI accelerator, like a Coral TPU. Especially since the first sentence on the page defines it. The first sentence doesn't define anything, the first sentence is "AMD Ryzen™ AI Software enables developers to take full advantage of AMD XDNA™ architecture integrated in select AMD Ryzen AI processors. Ryzen AI software intelligently optimizes AI tasks and workloads, freeing-up CPU and GPU resources, providing optimal performance at lower power." Which is at least better than meaningless but still doesn't explain what Ryzan AI is going to do. An architecture isn't an end goal. Intelligently optimizing AI tasks might be, but frankly AMD has been so behind in this space I don't believe the vapidity without an actual "we implement the such-and-such algorithm" or "the optimisation we refer to is ...". They're providing no evidence to believe they've optimised anything and their track record inspires little confidence. In my experience their AI software can't do BLAS while my GPU is running an X server and so "optimised" could just mean that I will be turning my computer off by choice after I buy one. Which would be a welcome improvement I admit, but they aren't exactly spelling out what the win is going to be. But communicating that there is a part of a processor that processes tasks is not informative. They might not even process "AI" for all I know, the AI I want to run on my GPU right now isn't represented as an ONNX model. Not a problem, but it continues to go to how low in information the headline copy is. "AI" isn't a real term, it is an umbrella. None of the AIs for the games I know of can be offloaded to Ryzen AI, for example. So there is some fairly specific set of things that Ryzan AI does which they aren't telling me. You are, but you obviously weren't involved in writing the documentation (or, I suppose, if you were the evidence is your editor murdered your work). Now if you already know what the chip does then sure, you don't need any of the copy or documentation. But there has to be an on ramp for people who don't know and they didn't include anything useful. > ... they promised the Ryzen AI software would let you deploy models to Ryzen AI processors. You can't name a processor that models can't be deployed to. They're big slightly-nonlinear operations. Every processor can do that. The sentence in the docs is literally information free. > I have deployed machine learning models to the edge in production on accelerators and CPUs. I use Llamas on my home rig, both CPU and CUDA. I am quite aware of what they can do, which is why I already knew the performance of the Ryzen AI processor. You are the one whose limited understanding kept you from recognizing that “Ryzen AI processor” is a distinct thing from a CPU or GPU. And you know what is interesting about this long list? None of it involves reference to the marketing copy or documentation that I'm talking about. Because as far as a cursory glance goes, nothing explains what "Ryzen AI" does. Just to circle back, you proffered a link to that documentation as a description of what these chips will do. Now I suspect you're about to argue that someone who isn't already in the field shouldn't bother with it. That says about as much as I have about the quality of the documentation - it isn't very good. im3w1l back at the thread root had a very valid question. > Your complaint is the equivalent of complaining that the home page for CUDA’s documentation doesn’t explain what a GPU is and why it is different from a CPU, which can do the same computations. THe CUDA homepage (https://developer.nvidia.com/cuda-toolkit https://developer.nvidia.com/cuda-toolkit), first paragraph, says "The toolkit includes GPU-accelerated libraries, debugging and optimization tools, a C/C++ compiler, and a runtime library." Then a bit further down the page has a link to the libraries they're talking about: https://developer.nvidia.com/gpu-accelerated-libraries https://developer.nvidia.com/gpu-accelerated-libraries. Which include, BLAS, FTT a random number generator, some other cool things. 10/10, I can tell what CUDA does. Does Ryzan AI do all this stuff? It allows me to deploy models and I'd need all that to deploy a models. I doubt it does. But the docs don't say what it does in the intro so... I dunno. It is probably in there somewhere. They defined ROCm as an umbrella term that included ROCm software once in their ROCm docs, so anything is possible. Being unable to explain what it is they are doing is a real pattern with AMD in the AI space.