SUPPORT OUR WORK: Use code PUCKGAMERSNEXUS for 10% off anything on the store, like our MICROSLOP T-Shirt https://store.gamersnexus.net/products/microslop-par...
I’m just using modular to describe like… a series of connected independent things that is not inside of one box.
I don’t know that there’s an actual term for this, and I get that yes standard PCs have modular component sets and configurations, but I mean more like a motorcycle and a side car, than … a car model with different trim levels.
Uh anyway, how far away are we from CPU integrated gfx being equivalent to modern GPUs?
Probably infinity amount of time.
Because CPUs and GPUs are basically defined by the different kinds of math they are optimized to do.
When you shove both of those things into one chip, that’s an APU, like what the Steam Deck or some consoles have.
When you jam CPU and basically LLM optimized chiplets together, thats apparently called an NPU.
All those AI Max’s and what not.
They can do GPU type things, but not as well as a similar caliber of APU… LLM math is similar to but not quite the same as GPU math… GPUs need a whole bunch of shit to handle unpredictable/rapidly changing ‘problems’ they are solving, whereas LLM get gigantic problems and then chew through them one at a time.
I’m just using modular to describe like… a series of connected independent things that is not inside of one box.
I don’t know that there’s an actual term for this, and I get that yes standard PCs have modular component sets and configurations, but I mean more like a motorcycle and a side car, than … a car model with different trim levels.
Uh anyway, how far away are we from CPU integrated gfx being equivalent to modern GPUs?
Probably infinity amount of time.
Because CPUs and GPUs are basically defined by the different kinds of math they are optimized to do.
When you shove both of those things into one chip, that’s an APU, like what the Steam Deck or some consoles have.
When you jam CPU and basically LLM optimized chiplets together, thats apparently called an NPU.
All those AI Max’s and what not.
They can do GPU type things, but not as well as a similar caliber of APU… LLM math is similar to but not quite the same as GPU math… GPUs need a whole bunch of shit to handle unpredictable/rapidly changing ‘problems’ they are solving, whereas LLM get gigantic problems and then chew through them one at a time.