imaginary_num6er@alien.topB to Hardware@hardware.watchEnglish · 1 year agoNvidia Blackwell GB202 GPU Rumored to Feature 384-bit GDDR7www.tomshardware.comexternal-linkmessage-square47fedilinkarrow-up11arrow-down10
arrow-up11arrow-down1external-linkNvidia Blackwell GB202 GPU Rumored to Feature 384-bit GDDR7www.tomshardware.comimaginary_num6er@alien.topB to Hardware@hardware.watchEnglish · 1 year agomessage-square47fedilink
minus-squareBaalii@alien.topBlinkfedilinkEnglisharrow-up1·1 year agoYou should be looking at transistor amount if anything at all, “cuda cores” is only somewhat useful when looking at different products within the same generation.
minus-squareResponsibleJudge3172@alien.topBlinkfedilinkEnglisharrow-up1·1 year agoStill very accurate if you know what to look for. For example, the reason why Ampere vs Turing CUDA cores scale different will let you predict how an Ampere GPU scales vs Turing GPU. It’s also why we knew how Ada would scale linearly except with 4090 that was nerfed to be more efficient
minus-squareResponsibleJudge3172@alien.topBlinkfedilinkEnglisharrow-up1·1 year agoI guess people don’t dig into white papers to learn about how and why the architectures perform as they do
You should be looking at transistor amount if anything at all, “cuda cores” is only somewhat useful when looking at different products within the same generation.
Still very accurate if you know what to look for.
For example, the reason why Ampere vs Turing CUDA cores scale different will let you predict how an Ampere GPU scales vs Turing GPU.
It’s also why we knew how Ada would scale linearly except with 4090 that was nerfed to be more efficient
I guess people don’t dig into white papers to learn about how and why the architectures perform as they do