I am sorry for the thumbnail debacle. That being said, I was introduced to you for the first time via this podcast and am quickly going thru your course.
Same here, but I did printout about first 100 pages of AMD's GCN 5 "Vega" reference manual. There is some good stuff in there that is outlines. It's something that Sean Barret relied on when he made his voxel raycasting engine.
It might be naïve, but I believe a thumbnail featuring a guest prominently should be approved by them before going live. Unfortunately, common courtesy and respect seem to be fading away.
They said they would do that going forward, so hopefully I was the canary in the coalmine for this and they will get their thumbnails approved by guests in the future.
The GPU video is a very good idea. When at university, I remember a professor saying that a CPU is like a few fast sport cars and a GPU is a big number of slow motor cycles.
It is a very bad analogy. Says nothing about the memory model and sharing related behavior of GPUs compared to CPUs and does not say anything about the things which are better done with a GPU compared to CPUs.
Looking forward to the GPU video. However, I would love the full version if you have one. It would be great to go through every aspect of the GPU in the performance aware way.
Does the content moving to your own backend means I won't need to download from substack the videos of this course (painfully) ? If I understood correctly and that is the case, if you could add the possibility to search for content with the same granularity as on handmade hero website, it would be much appreciated (I understand the work it may imply).
After learning about how the cpu cores and memory works, it seems to me that I can sense waste in hardware too. If we could get rid of some of that, maybe a worse process node could give competetive performance and then we would not need billions to produce a good processor.
GA144 is an example of what can be done on a shoestring budget: 100 GOPS/s on 1/8 mm² of 180 nm silicon. (Only very limited operations, but still impressive to me.)
I am sorry for the thumbnail debacle. That being said, I was introduced to you for the first time via this podcast and am quickly going thru your course.
Casey appears on a podcast...
Podcast host: "Do I hear boss music?"
Grug simple guy. Grug see Casey Muratori, Grug click. That Grug clickbait.
I was just trying to figure out what a GPU core is a few weeks ago. There is nothing good out there.
I was specifically trying to understand the huge gap between apple silicon and Nvidia.
Looking forward to the video!
Same here, but I did printout about first 100 pages of AMD's GCN 5 "Vega" reference manual. There is some good stuff in there that is outlines. It's something that Sean Barret relied on when he made his voxel raycasting engine.
It might be naïve, but I believe a thumbnail featuring a guest prominently should be approved by them before going live. Unfortunately, common courtesy and respect seem to be fading away.
They said they would do that going forward, so hopefully I was the canary in the coalmine for this and they will get their thumbnails approved by guests in the future.
- Casey
The thumbnail seems to be fixed.
The GPU video is a very good idea. When at university, I remember a professor saying that a CPU is like a few fast sport cars and a GPU is a big number of slow motor cycles.
It is a very bad analogy. Says nothing about the memory model and sharing related behavior of GPUs compared to CPUs and does not say anything about the things which are better done with a GPU compared to CPUs.
I guess regular course content is on pause in the meantime?
(GPU video is very much welcome!)
Not really, I should just post the next video. I have it done, I just kept thinking I was "one more day" from having the GPU core video ready :(
- Casey
Not seeing the “You Suck” thumbnail on my end fyi
Looking forward to the GPU video. However, I would love the full version if you have one. It would be great to go through every aspect of the GPU in the performance aware way.
Does the content moving to your own backend means I won't need to download from substack the videos of this course (painfully) ? If I understood correctly and that is the case, if you could add the possibility to search for content with the same granularity as on handmade hero website, it would be much appreciated (I understand the work it may imply).
We've got a whole thing planned, I'll post more information when it's available. How well it all works will be another question, of course :)
- Casey
After learning about how the cpu cores and memory works, it seems to me that I can sense waste in hardware too. If we could get rid of some of that, maybe a worse process node could give competetive performance and then we would not need billions to produce a good processor.
GA144 is an example of what can be done on a shoestring budget: 100 GOPS/s on 1/8 mm² of 180 nm silicon. (Only very limited operations, but still impressive to me.)