Huge NVIDIA Hopper H100 GPU, 144-Core CPU, & DGX H100 Systems
video description
Date: 2022-03-26
Related videos
Comments and reviews: 9
Llewln
Working in ML, most experimentation is done on GPUs such as 1080/3090/Tesla series with large VRAM buffers. When running ML in the real world (ie on a computer or phone) the move is currently towards ml specific hardware (TPUs) as a single TPU can equal a GPUs power while using less energy and being a lot cheaper.
The Nvidia stuff presented will be more aimed at massive companies training huge models (think BERT) or larger institutions in general. I hope Nvidia comes to the TPU world or that competitors reach the market soond as they'll be the real game changers!
reply
Working in ML, most experimentation is done on GPUs such as 1080/3090/Tesla series with large VRAM buffers. When running ML in the real world (ie on a computer or phone) the move is currently towards ml specific hardware (TPUs) as a single TPU can equal a GPUs power while using less energy and being a lot cheaper.
The Nvidia stuff presented will be more aimed at massive companies training huge models (think BERT) or larger institutions in general. I hope Nvidia comes to the TPU world or that competitors reach the market soond as they'll be the real game changers!
reply
darkSorceror
What they don't show with the live translation is that different grammars need to be translated with different word order. To have any chance of accuracy, it could only translate one sentence at a time. You can't break it down into smaller phrases unless there are strict rules on what languages are inter-compatible. This would introduce a noticeable latency, which is not presented in the short clip of Ms Uncannie McValley.
reply
What they don't show with the live translation is that different grammars need to be translated with different word order. To have any chance of accuracy, it could only translate one sentence at a time. You can't break it down into smaller phrases unless there are strict rules on what languages are inter-compatible. This would introduce a noticeable latency, which is not presented in the short clip of Ms Uncannie McValley.
reply
Jason
Psh, after working in the entertainment industry you don't even need to memorize a script and have an earpiece feeding you the lines (for some reason you need to have extensive training and in some cases be certified by a media coach).
reply
Psh, after working in the entertainment industry you don't even need to memorize a script and have an earpiece feeding you the lines (for some reason you need to have extensive training and in some cases be certified by a media coach).
reply
soundcloudcom
were interested in this subject of matter, would you please share your thougts on how its available, how it would work out for the consumer, what pricerange it is and how you would preferably use it?
reply
were interested in this subject of matter, would you please share your thougts on how its available, how it would work out for the consumer, what pricerange it is and how you would preferably use it?
reply
Silver
As far as gaming goes, all I need to do is stay ahead of the consoles with my PC. They are not going to go very far ahead of the hottest console. They will make sure it will play on the consoles.
reply
As far as gaming goes, all I need to do is stay ahead of the consoles with my PC. They are not going to go very far ahead of the hottest console. They will make sure it will play on the consoles.
reply
ArtistaFrustrado
hmm... we could use a 10 bucks teleprompter, or we could invest in an overengineered AI worth millions of dollars that edits the eyes on the video
yeah i can't choose either
reply
hmm... we could use a 10 bucks teleprompter, or we could invest in an overengineered AI worth millions of dollars that edits the eyes on the video
yeah i can't choose either
reply
DD
It took EIGHT 3090's to render the Star War RT demo in real time two years ago... flawless Realtime raytracing I wonder how close we'll get this time w/ Lovelace. .Here's hope(er)!
reply
It took EIGHT 3090's to render the Star War RT demo in real time two years ago... flawless Realtime raytracing I wonder how close we'll get this time w/ Lovelace. .Here's hope(er)!
reply
Star
That NVIDIA chip gives me hope for a Nintendo console on the future following the sucess of the SWITCH which would have an SoC lightyears better than the terrible Tegra X1.
reply
That NVIDIA chip gives me hope for a Nintendo console on the future following the sucess of the SWITCH which would have an SoC lightyears better than the terrible Tegra X1.
reply
Zzzz
AMD been going this sort of thing COMBINE CPU and GPU together and you call it APU that is used for GAME SYSTEMS PLAYSTATION 4 ,XBOX ONE , PLAYSTATION 5 and XBOX SEIRES X/S
reply
AMD been going this sort of thing COMBINE CPU and GPU together and you call it APU that is used for GAME SYSTEMS PLAYSTATION 4 ,XBOX ONE , PLAYSTATION 5 and XBOX SEIRES X/S
reply
Add a review, comment
Other channel videos
