Hype train chooo chooo za laku noć ili dobro jutro
Citiraj:
If this new dynamic L1 clustering scheme is the infinity cache included in RDNA2, so we can actually say that RDNA2 is a complete monster. Also, this method is able to boost performance in #DeepLearning applications by 2.3x on average (up to 3.9x).
Really Impressive, AMD!
|
Citiraj:
|
This is pretty crazy if its implemented into RDNA2. It sounds like it will scale up as more CU's are added. I guess we wait until AMD announces but Nvidia may have some real reasons to be concerned.
|