Why does an AMD GPU use more vram than an Nvidia one
>>107620995Unused Vram is wasted Vram, same question nigeroids on this board ask about RAM.
>>107620995For the same reason DX12 apps don't work well with Nwordia on Linux - optimization issues. "Huang first" is the modus operandi for game developers simply because there's more Nvidia users (and Mr. Jacket sends loads of money to those who suck his cock).
>>107620995because it's made of burned up jews
>>107620995tuesday
>>107620995Shitty drivers
>>107620995The Last of Us 1 is literally using 12GB of VRAM in 3K, imagine not having at least 16GB of VRAM in 2025.
>>107620995drivers
>>107620995Nvidia simply makes better product.
>>107621945>1440p>3k
>>107620995because it needs to uses more power too because it needs tooit even uses more system ram because it needs too
>>1076219451440p is 2.5k2k is FullHD and i dont give a fuck that everyone says differentIm right you are all wrong 1440p is 2.5kSimple as
I will enjoy 24gb of rdna3 vram for 4K/60+ AAAslop and emulation purposes up to 240hz. I defeated the market buy building this 2 years ago. My 32gb ram kit is 4x the price kek. Gaytracing lmao.
>>107620995Because Nvidia patented the superior math homework.CUDA contains the most efficient memory interlacing technique, which is to say it has fancy math that allows you to squeeze out more performance per byte in memory.AMD can’t use the CUDA formulas without getting sued even though everyone knows how it works. (Maybe the chinese will rip CUDA off eventually one day, who knows) AMD has to legally settle for almost as good memory management instead like what’s in ROCM.It’s not really _that_ much of an advantage though.HBM non-CUDA absolutely trounces GDDR CUDA.And APUs with in built memory giga rape everything which is why nvidia is trying so hard to get their ARM APUs operational.AMD is incompetent however and Lisa Su has only been coasting off of the Zen project, which was originally organized by Rory Read. She done almost fuck all with Zen’s massive success except pretend to be relevant in AI for shareholders and be Jensen’s cousin.
>>107623183AMD made HIP so its going to work, some performace loss happense with wrapper but not enough to offset AMDs superior raster performance100% Nvidia will try to postpone it as much as they can since they cant win a case against software that just circumvents CUDA monopoly
>>107621945Imagine playing the last of us
>>107623290This. Give me one good game that uses more than 12 gb vram. SOrry, honey, but you can't. Simple as.
>>107620995you paid for the hardware. you enabled hardware acceleration. let the software utilize it.
