Intel Nervana

[NumLock21]

Intel is planning to ship their first Intel Nervana NNP by the end of this year. NNP stands for Neural Network Processor and the Intel Nervana was specifically designed for deep learning. The NNP can be used in a wide range of application like,

Quote

weather prediction and autonomous vehicles to targeted advertising on social media.

Code name Lake Crest, Intel Nervana will have,

Quote

HBM2, FlexPoint, cacheless software memory management, and a high-speed interconnect. On the technical side of matters, the details remain the same from previous reports. Intel states that the NNP does not have a "standard cache hierarchy," however it does still have on-chip memory for performance reasons (I expect serving as registers and the like). Managing that memory is done by software, taking advantage of deep learning workloads where operations and memory accesses are mostly known before execution. Subsequently, the lack of cache controllers and coherency logic frees up die space. Otherwise for off-die memory, the processor has 32GB of HBM2 (4 8-Hi 1GB stacks) on the shared interposer, resulting in 8 terabits/s of access bandwidth.

It will be send to selected customers and one of them will be Facebook because they have actually collaborated with Intel on designing the Neural Network Processor. It will compete against Nvidia's deep learning hardware, their Volta based Tesla.

 

 

https://www.anandtech.com/show/11942/intel-shipping-nervana-neural-network-processor-first-silicon-before-year-end

[Guest]

Good to see more manufacturers adopting HBM even if its mostly just in the High end/research space.

[Drak3]

I prefer Kurt Cobain's Nirvana, TBH.

[Taf the Ghost]

5 hours ago, tjcater said:

Good to see more manufacturers adopting HBM even if its mostly just in the High end/research space.

Yup, but it also explains a lot of the HBM2 issues. 

[leadeater]

6 hours ago, NumLock21 said:

targeted advertising on social media.

No, bad Intel. Now go sit in the corner and think about what you did.

[Bcat00]

6 minutes ago, leadeater said:

No, bad Intel. Now go sit in the corner and think about what you did.

Would you rather they just shoot annoying advertisement like a decade ago?

The ads nowadays are actually pretty decent, not totally intrusive if the site you are visiting is legit. 

[leadeater]

12 minutes ago, Bcat00 said:

Would you rather they just shoot annoying advertisement like a decade ago?

The ads nowadays are actually pretty decent, not totally intrusive if the site you are visiting is legit. 

Quality of ads isn't the point or the issue. A great ad is a great ad, there are many I've seen that I like very much on TV and on the Internet. Using AI or other large computational resources to mass collect information and specifically target an individual is different, ads are not the issue.

[Bcat00]

How do you think good ads get into a viewer's eye?

just because one guy like car ads doesnt mean i like car ads, that's what this is going to fix.

 

send ads that people will actually look at.

17 minutes ago, leadeater said:

Quality of ads isn't the point or the issue. A great ad is a great ad, there are many I've seen that I like very much on TV and on the Internet. Using AI or other large computational resources to mass collect information and specifically target an individual is different, ads are not the issue.

eg: the annoying car ads on youtube. If any ad is to pop up for me then i want to see it food related.

[leadeater]

22 minutes ago, Bcat00 said:

How do you think good ads get into a viewer's eye?

just because one guy like car ads doesnt mean i like car ads, that's what this is going to fix.

 

send ads that people will actually look at.

eg: the annoying car ads on youtube. If any ad is to pop up for me then i want to see it food related.

There are other ways to achieve this without using AI, it's actually very simple. Target ads to the content so it's relevant to subject material and therefore likely to the viewer.

 

To me this isn't even mostly about privacy or data collection, something you should think about much more deeply than you appear to have been doing, but about wasted resources and funds. This whole endeavour of pumping money in to the advertising industry is more than a waste of money it's essentially stealing money out of the economy that could be used for much more productive things.

 

To be clear I support ad makers and people in advertising and marketing roles, those jobs are required and provide people with jobs and opportunities to branch out to other areas of their wider chosen field if they wish. Dumping ridiculous amounts money and resources in to computer science research to facilitate advertising I do not and will never support.

 

This is off topic anyway so if you want to discuss it more create a topic for it but to be frank refer to last sentence, my view on it isn't going to change. 

 

7 hours ago, NumLock21 said:

weather prediction and autonomous vehicles

Yes great, do this.

 

7 hours ago, NumLock21 said:

targeted advertising on social media.

This is bad, do not do this.

[Bcat00]

2 minutes ago, leadeater said:

There are other ways to achieve this without using AI, it's actually very simple. Target ads to the content so it's relevant to subject material and therefore likely to the viewer.

 

To me this isn't even mostly about privacy or data collection, something you should think about much more deeply than you appear to have been doing, but about wasted resources and funds. This whole endeavour of pumping money in to the advertising industry is more than a waste of money it's essentially stealing money out of the economy that could be used for much more productive things.

 

well then i guess we have very different opinions on what is considered wasted resource so ill end it here. 

[Sniperfox47]

Quote

Tensor based architecture

Seems everyone is moving in this direction. Wonder how long before tensor optimized units take over for training too, or is that what Nirvana is supposed to be for?

[dizmo]

Compete against. We'd advance so much faster if companies would just work together. Granted that will never happen, but one can dream.