Nvidia Volta Speculation Thread

Infinisearch said:
Double precision is fully programmable, while tensor cores are fixed function units aimed at AI. There is likely zero chance they'll be in consumer cards or even quadro's for that matter.
Click to expand...
They do go beyond AI though, it is just that Nvidia engineers are still learning what they can do with Tensor Cores and have stated in the past they will open further functions/libraries when they can.
In fact one can access the Tensor Cores themselves with C++ but yeah if Nvidia engineers are still working out what they can do with these beyond current DL related libraries then not sure what those outside of Nvidia would be able to do themselves.
Case in point it is probable Tensor Cores in the future could be used in part for simulation modelling with GEMM computation; part of this comes back to what one of the senior engineers has said in the past and the latest article from Nvidia.

That said I cannot see these anytime soon being anything but flagship mixed-precision GPU and the 'half-size' dedicated Tensor GPU Nvidia card (if they still intend to sell it).
Cheers
 
Just clicking on her name and reading the comments on her page tells it all! :rolleyes:
 
pharma said:
Just clicking on her name and reading the comments on her page tells it all! :rolleyes:
Click to expand...

What? Into the comment there is other people working into the industry responding like a guy from unity and saying he like the explanation. It is a compiler Dev... Not only a cosplayer ...

Edit: reason don't want to talk about the fact she is a woman
 
Last edited:
pharma said:
Just clicking on her name and reading the comments on her page tells it all! :rolleyes:
Click to expand...
what you're a misogynist and a shill?

Just so you know, she has posted lots of interesting things over quite a long time. So how about point out the issues with the content, because just like the first line of this post you don't have to be very clever to attack the player.
 
CSI PC said:
They do go beyond AI though, it is just that Nvidia engineers are still learning what they can do with Tensor Cores and have stated in the past they will open further functions/libraries when they can.
Click to expand...
Why the wait? Why not simply unblock and release full documentation on the cores and let end users find uses for the hardware themselves?
 
Grall said:
Why the wait? Why not simply unblock and release full documentation on the cores and let end users find uses for the hardware themselves?
Click to expand...
They’ve already done that.

I don’t understand the comment about “trying to figure out what they can do”. It’s a matrix multiplication. What is there to figure out???
 
Bondrewd said:
Is there anything anything wrong with weaboo cosplayers?
Click to expand...

Yeah, they're stupid and disgusting.

As for Volta: Will it support DX Feature Level 12_1 Full tier levels a'la Vega, and will it even be important for gaming since not even the console hardware supports these features fully?
 
Pinstripe said:
Yeah, they're stupid and disgusting.

As for Volta: Will it support DX Feature Level 12_1 Full tier levels a'la Vega, and will it even be important for gaming since not even the console hardware supports these features fully?
Click to expand...

This will be good, once it don't hold down the software anymore. By 2018 timeframe all games can be DX12 then, finally inviting Game Devs to exploit all power that DX12/Vulkan can offer. This may help the utilization of six and eight core CPUs(once they are not well utilized in barely any game), finally.
 
I have no idea about technical details but even if AI/DNN would go the way of Dodo bird there is life left in volta in HPC/supercomputers/simulations. She misses the point of volta which to me looks like to be a monster chip for everything. You install volta to cloud and can sell time for all kinds of computing loads. There is really decent amount of fp64 flops and api's/flexibility to harness that power. I find it hard to believe ai would disappear as a fad however I can easily believe the algorithms for ai would be refined/reinvented and ultimately rolled into specific ai accelerator training/inference chips.

https://www.nextplatform.com/2017/09/19/power9-rollout-begins-summit-sierra/
 
chris1515 said:
What? Into the comment there is other people working into the industry responding like a guy from unity and saying he like the explanation. It is a compiler Dev... Not on'y a cosplayer ...

Edit: reason don't want to talk about the fact she is a woman
Click to expand...
Maybe you could link it? Comments are tweets after all and without being logged into Twitter, navigating and reading Tweets is nothing short of annoying.
 
You must log in or register to reply here.

Similar threads

F
RDNA5 Speculation
Replies
2
Views
1K
Erinyes
E
D
Speculation and Rumors: Nvidia Blackwell ...
46 47 48
Replies
942
Views
105K
arandomguy
A
D
Show us your gaming set-up - PC Edition!
2 3
Replies
59
Views
6K
Davros
D
Jawed
  • Locked
NVidia Ada Speculation, Rumours and Discussion
104 105 106
Replies
2K
Views
247K
Malo
Malo
Rys
Reopening Architecture and Products + rules and guidelines changes
3 4 5
Replies
90
Views
16K
Shifty Geezer
Shifty Geezer
Back
Top