Quantcast
[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology


View post   

[ Toggle deleted replies ]
File: 528 KB, 746x685, intel.png [View same] [iqdb] [saucenao] [google] [report]
62953568 No.62953568 [Reply] [Original] [archived.moe] [rbt]

It's happening:

>Intel aims to conquer AI with the Nervana processor
>It will release its first all-purpose AI chips by the end of 2017.

INTEL WILL BE THE FIRST TRILLION DOLLAR MARKET CAP COMPANY

SCREENCAP THIS

Nvidiots and Googlebots BTFO

>> No.62953581

>>62953568
Yet
Another
Blunder

>> No.62953610
File: 495 KB, 1070x601, 1495358050273.png [View same] [iqdb] [saucenao] [google] [report]
62953610

>>62953568
>intel fags starting to go after nvidia
nvidia and amd take turns rapeing intel

>> No.62953636

>>62953568
Dead in 3...2...1...
Intel will eventually surpass google in dead projects, but google's dead projects cost less.

>> No.62953655

>>62953636
google's dead projects are just employees fucking around in their spare time

>> No.62953667

>>62953568
Uses EMIB and what looks like an Omni-Path?
>>62953610
That's Nervana and not Intel though.

>> No.62953801

>>62953568

>All-purpose AI processor

What were they thinking? Doesn't Google have purpose-built AI systems that they sell? Wouldn't those work way better?

>> No.62953832

>>62953801
TPU/TPU2 are mainly for inference.
NNP is training-focused as far as I can guess.

>> No.62953864

Volta available now

Lake Crest, nowhere out in the field until 2018 at best

>> No.62953876

>>62953667
Its intel you mongoloid

>> No.62953893

>>62953864
>Volta available now
>815mm^2
>available
>now
Not so sure about that, mr. Jacketman.
>>62953876
It's Nervana. Intel simply bought them much the same way they bought Altera.

>> No.62953929

>>62953893
Volta available now, shipping now, in customers hands now

You can stay mad forever, Nervana shill, Volta can be fabbed with good yields at TSMC unlike the constant failure of Intel not even being able to ship their bread and butter Core CPUs with 10nm process

>> No.62953936

Nerve + Nirvana (Eastern enlightenment) = a very good name. Great job, Intel. I will buy more Intel products.

>> No.62953966

Let me guess, little to no software support, minimum 6 months late, and available only to "enterprise" customers with minimum order size of $1M locking out start-ups who would have built OSS library ecosystem?

>> No.62953969

>>62953929
>Volta available now
Where?
>shipping now
Where? All I see is "agreements" and zero GPUs shipped.
>in customers hands now
Sorta. Chinks at Baidu reported success using meme cores for training in the lab.

>> No.62954036

>>62953893
>It's Nervana. Intel simply bought them much the same way they bought Altera.
Doesn't matter it's intel now :^)

>> No.62954101
File: 361 KB, 2001x1502, google ia.jpg [View same] [iqdb] [saucenao] [google] [report]
62954101

>>62953568

ok

>> No.62954119

>>62953568
>Intel making care bears

>> No.62954196

>>62953568
Did they rename Larrabee again? kek

>> No.62954207

>>62954196
No, Knights Mill was before that.
This is difference but almost certainly DOA because no CUDA & libs.

>> No.62954223

>>62954207
>no cuda
https://github.com/RadeonOpenCompute/hcc/wiki

>> No.62954240

>>62954223
How the fuck CL is related to fucking ROCm?
ROCm is Radeon-specific.

>> No.62954283

>>62954240
>HIP provides a set of tools and API for converting CUDA applications into a portable C++ API.

>> No.62954295

>>62954283
And then you'll have to hand-optimize it for fucking CL.
The main benefit of CUDA is not CUDA but libs, they are good and plenty.

>> No.62954328 [DELETED] 

its happening

>> No.62954385

>>62953568
>Nervana processor
Useless

>> No.62954528

>>62953568


well its probably good if you delid it.

and only in certain types of AI that need fast IPC.

>> No.62954688
File: 54 KB, 507x464, top ebin.png [View same] [iqdb] [saucenao] [google] [report]
62954688

>>62953568
Make more SKUs!

>> No.62954713

Retard here.
What would make these chips better than any other chip for ai training? This seems like marketing hype

>> No.62954729

>INTEL WILL BE THE FIRST TRILLION DOLLAR MARKET CAP COMPANY

>AAPL: $828.86 BN
>INTC: $186.97 BN

Statically speaking that would be Apple. Intel isn't even close.

>> No.62954744
File: 155 KB, 804x978, AMDvIntel.jpg [View same] [iqdb] [saucenao] [google] [report]
62954744

>>62954688

>> No.62954749

>>62954713
Nothing, it's just another Itanium-level blunder.

>> No.62956313

>>62953568
Even the name is pajeet.

>> No.62957267

>>62953568
>AI
I hate how everyone is abusing this. THERE IS NO REAL AI, IT'S IMPOSSIBLE

>> No.62957459
File: 113 KB, 480x360, file_0.png [View same] [iqdb] [saucenao] [google] [report]
62957459

>>62953568
>nervana
>nerv
Uh oh

>> No.62957517

>>62957267
There's no real AI now. That's why everyone is researching it. What point is there in investing this much time and energy into something we already have? Top-down AIs are still pretty far off, but I wouldn't be surprised to see some pretty good bottom-up AIs in the next decade or two.

>> No.62957982

>>62953568
>INTEL WILL BE THE FIRST TRILLION DOLLAR MARKET CAP COMPANY

All the more reason to invest into competition

>> No.62960057

>>62953969
http://www.thinkmate.com/system/gpx-xt4-24s1-4nvlink

Scroll down to Co-Processors, the V100 is $8000.

>> No.62960240 [DELETED] 

>>62953568

>> No.62961288

>>62953568
remember xpoint memory

>> No.62962254

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

>Presumably, the NNP will be fabricated on the TSMC 28nm process that Lake Crest was intended for

>28nm
>Not even a FINFET process

INTO THE TRASH IT GOES

>>
Name (leave empty)
Comment (leave empty)
Name
E-mail
Subject
Comment
Password [?]Password used for file deletion.
Captcha
Action