So, why haven't you pre-ordered the Pascal-powered meme-learning box?
>>53884429
BECAUSE THEY LIED TO MEEEEE
>No new architecture
>Maxwell shrink with some add-ons
>NO ACTUAL CARD READY
Why did they do this... THERE WAS SO MUCH HYPE
>>53884429
I already have a computer
>>53884514
>he thought Jen-Hsun Huang was going to announce GTX 1080
Check out this cuck.
>>53884548
I at least wanted SOMETHING!
not fucking CARS!
FUCKING CAAAAAARS
$129,000
$129,000
$129,000
$129,000
$129,000
>>53884557
you fucked too much traps dude, relax.
>>53884429
Because AMD's current card is already superior
FirePro S9300 X2 = 13.9 TFLOPS Single Precision @ 28nm + HBM 1 & 300w TDP
NVIDIA Tesla Pascal = 10.6TFLOPS Single Precision @ 14nm + HBM 2 & 300w TDP
>>53884679
Too bad everything is written in CUDA, and CUDA doesn't run on AMD GPUs.
>>53884689
But anon it was just cracked a cross compiler was made
http://venturebeat.com/2016/03/09/otoy-breakthrough-lets-game-developers-run-the-best-graphics-software-across-platforms/
>>53884717
Have they re-implemented cuDNN to run on Radeons too?
Until that happens, machine learning users will stick with Nvidia.
>>53884752
I'm sure someone will at some point, or they'll just write some new open source stuff to do the job.
Wouldn't AMD cards be better for that anyways given their superior compute performance at pretty much all levels?
>>53884818
http://www.hpcwire.com/off-the-wire/amd-gpuopen-fuels-seismic-supercomputing-efforts-for-geoscience-leader-cgg/
it's already starting
>>53884818
>>53884910
TensorFlow only runs on CUDA.
So does every other major ML framework.
and I'm sure they're tired of it
>>53886478
meant for
>>53885025
>>53884429
> 3.2 kW in 5.25" tall box
If I really need to warm up that much, it would be cheaper and easier to light my house on fire without the $129k accessory.