Re: [Computer-go] Google used TPUs for AlphaGo

[甲斐徳本]

Body of the Japanese text says "TPU is an ASIC developed for Deep Learning
with 'ten' times better performance per Watt compared to other technologies
such as GPU or FPGA, according to Pichai."

So may be it is easier to consider "10 times over Stratix(1) or
Vertex7/Xilinx(2)" and not 13 times over 'other' lesser hardware?

*1 http://research.microsoft.com/pubs/240715/CNN%20Whitepaper.pdf  (see
Table 1)
*2 http://cadlab.cs.ucla.edu/~cong/slides/fpga2015_chen.pdf

Tokumoto

On Sat, May 21, 2016 at 4:17 PM, Darren Cook  wrote:

> > http://itpro.nikkeibp.co.jp/atcl/column/15/061500148/051900060/
> > (in Japanese).  The performance/watt is about 13 times better,
> > a photo in the article shows.
>
> Has anyone found out exactly what the "Other" in the photo is? The
> Google blog was also rather vague on this.
>
> (If you didn't click through, the chart just say "Relative TPU
> Performance/Watt", with Other being between 0 and 4, and TPU being
> between 11 and 14.)
>
> Darren
>
> ___
> Computer-go mailing list
> Computer-go@computer-go.org
> http://computer-go.org/mailman/listinfo/computer-go
>
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Google used TPUs for AlphaGo

[Darren Cook]

> http://itpro.nikkeibp.co.jp/atcl/column/15/061500148/051900060/ 
> (in Japanese).  The performance/watt is about 13 times better, 
> a photo in the article shows.

Has anyone found out exactly what the "Other" in the photo is? The
Google blog was also rather vague on this.

(If you didn't click through, the chart just say "Relative TPU
Performance/Watt", with Other being between 0 and 4, and TPU being
between 11 and 14.)

Darren

___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Google used TPUs for AlphaGo

[Hideki Kato]

Some photos here.
http://itpro.nikkeibp.co.jp/atcl/column/15/061500148/051900060/ 
(in Japanese).  The performance/watt is about 13 times better, 
a photo in the article shows.

Hideki

Petr Baudis: <20160519105443.go22...@machine.or.cz>: 
>  Hi,
>
>  it seems that Google in fact used TPUs for AlphaGo rather than GPUs:
>
>   
>https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html
>
>It's be interesting to know what the speedup factor against, say,
>Tesla K40 is.  In the Nature paper, they talk about GPUs, but they
>could have switched to TPUs for training only after then Fan Hui matches.
>
>   Petr Baudis
>___
>Computer-go mailing list
>Computer-go@computer-go.org
>http://computer-go.org/mailman/listinfo/computer-go
-- 
Hideki Kato 
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

Re: [Computer-go] Google used TPUs for AlphaGo

[Hideki Kato]

DGX-1 looks not shipped yet.  TPUs has been used more than 1 
year.  This could be a big advantage for Google.

Hideki

Darren Cook: <573da3db.8000...@dcook.org>: 
>> It's be interesting to know what the speedup factor against, say,
>> Tesla K40 is.
>
>Or against the P100 chip [1], which claims the same "order of magnitude"
>speed-up on neural nets by doing the same thing (half-precision floating
>point).
>
>Darren
>
>[1]:
>http://nvidianews.nvidia.com/news/nvidia-delivers-massive-performance-leap-for-deep-learning-hpc-applications-with-nvidia-tesla-p100-accelerators
>
>___
>Computer-go mailing list
>Computer-go@computer-go.org
>http://computer-go.org/mailman/listinfo/computer-go
-- 
Hideki Kato 
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

[Computer-go] Google used TPUs for AlphaGo

[Petr Baudis]

  Hi,

  it seems that Google in fact used TPUs for AlphaGo rather than GPUs:


https://cloudplatform.googleblog.com/2016/05/Google-supercharges-machine-learning-tasks-with-custom-chip.html

It's be interesting to know what the speedup factor against, say,
Tesla K40 is.  In the Nature paper, they talk about GPUs, but they
could have switched to TPUs for training only after then Fan Hui matches.

Petr Baudis
___
Computer-go mailing list
Computer-go@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go