m*****n
发帖数: 1631 | 1 这哥们是AMD粉
But, McConnell, who has a Sector Weight rating on the shares, writes “we
remain on the sidelines,” because AMD is going to take some share at
computer original equipment makers (OEM):
After shortages throughout Q2, supply chain conversations indicate that
supply of AMD’s Polaris GPUs (RX 480/470/460) in the channel significantly
improved last quarter. Despite the supply improvement, desktop graphics card
manufacturers noted very little shift in market share to Polaris (100 to
300 basis ... 阅读全帖 |
|
|
发帖数: 1 | 3 在计算机时代，有个著名的摩尔定律，就是说同样成本每隔18个月晶体管数量会翻倍，
反过来同样数量晶体管成本会减半，这个规律已经很好地吻合了最近50年的发展，并且
可以衍生到很多类似的领域：存储、功耗、带宽、像素。
两年前的这个月是摩尔定律发表50周年，整整半个世纪。当时IEEE Spectrum为了纪念
摩尔定律50周年，特地发表了四篇关于摩尔定律的文章。其中一篇的题目叫“Moore’s
Law is Dying (and That Could Be Good)”，说的是摩尔定律不可能继续下去了。
1965年微芯片上的元件数增加了1倍，Gordon Moore于是预言这一趋势近期内将继续。
1975年他修改为每两年翻一翻，后来又说是18个月，或者说按指数律增长，每年46%。
这就是摩尔定律。摩尔定律预言了半导体产业50年的发展。摩尔预言了一个光辉的未来
，改变了世界。
这样高速的增长在其他产业是见不到的。美国的主粮玉米从1950年以后平均产量每年增
长2%，蒸汽涡轮式发电机把热能转换为电能，其效率在20世纪年增长率为1.5%，而1881
-2014室内灯光有效性（流明每瓦）年平均增长2.... 阅读全帖 |
|
m*****n
发帖数: 1631 | 4 Growing AI City Partner Support
More than 50 NVIDIA AI city partners already help customers reveal insights
and take real-time action using deep learning on NVIDIA GPUs. Among them are
industry leaders such as Avigilon, Dahua, Hanwha Techwin, Hikvision and
Milestone. |
|
p**f
发帖数: 3549 | 5 快要看不懂了。。。
NVIDIA’s Deep Learning Accelerator (DLA)
With this context in mind, then, it makes sense for NVIDIA to build a fixed
function accelerator that acts as an efficient inference engine as part of a
larger solution. NVIDIA announced that its next generation DrivePX platform
for autonomous vehicles, the Xavier SOC, would consist of ARM CPU cores,
Volta GPU cores, and a fixed function Deep Learning Accelerator (DLA) for
inference. This approach, the company says, will result in higher
perform... 阅读全帖 |
|
l*******m
发帖数: 1096 | 6 AMD根本没有embedded GPUs的产品。 |
|
p**f
发帖数: 3549 | 7 https://blogs.nvidia.com/blog/2017/09/25/auto-startups-nvidia-drive/
还有，among the GTC China announcements:
Alibaba, Baidu, and Tencent are all deploying Tesla V100 GPU accelerators in
their cloud services.
Huawei, Inspur, and Lenovo have all adopted our HGX server architecture to
build accelerated data centers with Tesla V100 GPUs.
Alibaba, Tencent, Baidu, JD.com, and iFlyTech have all adopted our GPU
inferencing platform. |
|
m*****n
发帖数: 1631 | 8 AWS has released its most powerful and expensive instance aimed at AI
developers.
Liam Tung
By Liam Tung | October 26, 2017 -- 11:34 GMT (04:34 PDT) | Topic: Cloud
Amazon Web Services (AWS) has launched new P3 instances on its EC2 cloud
computing service which are powered by Nvidia's Tesla Volta architecture
V100 GPUs and promise to dramatically speed up the training of machine
learning models. |
|
发帖数: 1 | 9 I saw this last year. The amd turmoil was before the Ryzen.
http://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility
It's coming from the mouth of the guy started the intel/amd rumor a year ago.
"What HardOCP knows from sources inside and outside the company, AMD has a
problem on its hands, as both these products have come up significantly
short of where these were supposed to land. But that is OK for AMD, it will
simply send Chris Hook out to fall on the sword and tell... 阅读全帖 |
|
y*********u
发帖数: 3664 | 10 “1. Recently begun OEM shipments of its Ryzen desktop CPUs (they gradually
launched over the course of 2017) in volume.
2. Launched its Vega high-end GPUs in Q3, and forecast they'll ramp with
OEMs in Q4.
3. Begun shipping its Epyc server CPUs -- like Ryzen parts, they're based on
AMD's new Zen CPU core architecture -- to cloud service providers, ahead of
major Q4 OEM launches.
4. Forecast its Ryzen Mobile notebook processors -- they pair a Ryzen CPU
with a Vega-architecture GPU -- would begin s... 阅读全帖 |
|
y*********u
发帖数: 3664 | 11 “1. Recently begun OEM shipments of its Ryzen desktop CPUs (they gradually
launched over the course of 2017) in volume.
2. Launched its Vega high-end GPUs in Q3, and forecast they'll ramp with
OEMs in Q4.
3. Begun shipping its Epyc server CPUs -- like Ryzen parts, they're based on
AMD's new Zen CPU core architecture -- to cloud service providers, ahead of
major Q4 OEM launches.
4. Forecast its Ryzen Mobile notebook processors -- they pair a Ryzen CPU
with a Vega-architecture GPU -- would begin s... 阅读全帖 |
|
|
G**Y
发帖数: 33224 | 13 I have some. More than most (I'm guessing) but a lot less than some of the
wealthy Litecoin holders out there. I want to think that litecoins are very
fairly distributed.
I had 2 CPUs mining from the start. I stopped mining when GPUs came on the
scene. You can probably figure out how many coins I mined. But I bought a
lot more on BTC-e. At one point, when the price was around $0.007, I
considered plopping down $7000 to buy a million litecoins just to have a
million. For some reason, I didn't. Oh... 阅读全帖 |
|
y*********u
发帖数: 3664 | 14 你这对比不对
Pegasus is built on two Xavier SoCs and two next-generation NVIDIA GPUs.
Customers will get the first samples of Pegasus with its 320 trillion
operations per second of processing performance in mid-2018.
应该是Pegasus v.s. “ 2 x EyeQ5 + 1 x ATOM"
TOPs |
|
b***d
发帖数: 288 | 15 我关心的是这2项， 女大 1000 不是梦阿
Datacenter
Announced that NVIDIA Tesla® V100 GPU accelerators are now available
through every major computer maker and have been chosen by every major cloud
to deliver AI and high performance computing.
Added a record 34 GPU-accelerated systems to the Top500 supercomputer list,
bringing the company’s total to 87.
Announced partnerships to further AI in key vertical industries, including
initiatives with GE Health and Nuance in medical imaging; Baker Hughes, a GE
compa... 阅读全帖 |
|
H*****l
发帖数: 445 | 16 看准时机建仓吧，上看260. 就看性能提升多少了。
如果还是GP102框架，估计250~260. 如果GP104+GDDR6，上300都有可能。 |
|
p**f
发帖数: 3549 | 17 麻痹Pascal架构早过时了。Volta都粗来一年了，还玩Pascal啊，能卖出去吗。最新消
息粗来了，2080/2070用的是Turing架构，可能是Volta小改马甲。不过Volta只针对
data center/ai市场。Turing针对gaming市场，估计把那些AI tensor flow相关的功能
挖掉了。Turing会阉割出各种核心，对应2060/2050/2030一堆中高低端消费卡。Ampere
才是volta/turing的真正下一代架构，目前还没影子。
不知道按摩店如何接招。目测NV要一桶浆糊了，虽然现在几乎已经是了。 |
|
d*****1
发帖数: 8618 | 18 按摩店真爱股东路过， 放弃治疗才是王道
按摩店接个屁招, 技术不是一个档次的，给女大提鞋都不配，放弃这一代游戏市场老老
实实把产量提上来多卖点矿卡才是正经，挣了钱投入在epyc上，再挣了钱的话下一代显
卡构架学
ryzen的彻底推倒从头再来，别放不下面子当了婊子还要牌坊
Ampere |
|
p**f
发帖数: 3549 | 19 美帝能源部被憋国刺激到，要大搞超算。新签了个单，放伯克利实验室。不过总预算也
才1.5个亿。亮点是用按摩店的EPYC+女大的GPU，没牙膏厂什么事。
之前的DOE两台也是IBM Power9+GPU，也没牙膏厂什么事。看来牙膏厂CPU要被超算放弃
了。。。
1.5亿不算多，但趋势是按摩店的EPYC被主流接受了。后面会雄起。。。
One day after unveiling the world's third fastest supercomputer at the
Lawrence Livermore National Laboratory, the US Department of Energy (DOE)
has announced yet another super-fast machine is on its way.
The new Perlmutter system will be built for the National Energy Research
Scientific Computing (NERSC) center at the Lawrence Berke... 阅读全帖 |
|
t***t
发帖数: 7 | 20 Job Description :
The positions are part of the funded project “Scalable Algorithm Design for
Emerging Parallel Architectures”. The work will be carried out at the
School of Computer Engineering at NTU. The scope of the project is to
develop parallel algorithms and efficient implementations for solving large
sparse linear systems over GF(2) on GPUs using the CUDA programming model.
Solving such systems is an important task in the cryptanalysis of asymmetric
encryption schemes like RSA. For examp |
|
t***t
发帖数: 7 | 21 Job Description:
The positions are part of the funded project “Designing Scalable Algorithms
on GPU-accelerated Hybrid Architectures”. The work will be carried out at
Fraunhofer IDM Centre@NTUhttp://www3.ntu.edu.sg/fraunhofer/. The scope of the project is to design and develop new scalable parallel algorithms and efficient implementations for solving compute-intensive problems in computational biology, physics simulations, etc. We propose the development of these new algorithms for emerging ma... 阅读全帖 |
|
h****6
发帖数: 42 | 22 Linux C++ multithreading consultant
Duties:
Design and implement a C++ parallel programming framework to enable real
time image analysis and high performance computing in a hybrid platform with
multiple GPUs and CPUs.
Skills:
5-10 years C++ with multithreading in Linux/Unix.
Well versed in OO paradigms
Excellent written and verbal communication skills
Experience with cuda, Boost, and intel TBB is a big plus
Experience with statistical modeling and image processing is a plus
Rate: DOE
Term: 10 mo |
|
|
m********u
发帖数: 3942 | 24 【 以下文字转载自 JobHunting 讨论区 】
发信人: missingyou (miss), 信区: JobHunting
标 题: Sr Deep Learning Scientist
发信站: BBS 未名空间站 (Mon Aug 28 22:04:14 2017, 美东)
最好有手机开发经验，需要hands-on，如果背景够强可以consider to be leader role
for new R&D center
地点硅谷，站内或者邮件 [email protected]
- Deep Learning be must domain expertise
- PhD+2Y or MS+5Y
- Track record of designing architecture and developing algorithms to
accelerate computational demanding applications
- Experience in working with deep learning frameworks like Caffe, Te... 阅读全帖 |
|
g*****9
发帖数: 4125 | 25 I have built myself kind of super desktop because I does a lot of fluid
dynamics and finite element analysis.
Here is a little detail for you computer guru on this board.
OS Windows XP 64bit
Motherboard Tyan S7025
Tesla GPUs 4x Tesla C1060 (consume most of the power)
Power Supply Coolmax CUQ-1350B rated at 1350W
Case Lian Li PC-P80
Monitor 2 NEC MultiSync LCD3090WQXI rated at 150W each
Yes, it is a 960 core desktop and most of my analysis still takes hours to
days to run.
That with a quite few 1 |
|
m*******I
发帖数: 1049 | 26 ☆─────────────────────────────────────☆
flora6866 (Flora) 于 (Mon Nov 2 16:29:31 2009, 美东) 提到:
我家俩口人，9月从外周搬来，现住在2bb的公寓，家里热水和cooking都是用电
近来收到了第一张电费账单，40天用了1480kwh （actual reading）, $250. 我们平时
用电节约，正常做饭，不开空调，很少用dryer干衣服，用的最多的就是电脑。
翻看以前的电费账单，同样的情况，每月消耗不到500kwh, 怎么到了这里用电能翻到3
翻？
早上让物业读了我的电表，又吓我一跳，10天用了1200kwh,照这速度，我这个月账单会
超过$500.到冬天开暖气不得1000刀！？ 太离谱了
我给PEPCO打了电话，他们说会派人调查，要4-6周。我担心他们要是查不出什么名堂，
我难道就只能认了吗？
大家的电费都是多少？有没有人跟PEPCO打过类似交到？
☆─────────────────────────────────────☆
gezhi89 (zhonda) 于 (Mo... 阅读全帖 |
|
O**l
发帖数: 12923 | 27
well
这个我当然知道 我们在讨论heuristic MCT是不是MCT的问题 和DCNN有什么关系， 我是
在强调MCT远远比heuristic重要 没MCT之前不管什么heuristic你1D都上不了 有了MCT
你能
上5D
pure
这是常识 DCNN拟合最多做做planning 你不search 简直开玩笑 你完全不理解DCNN MCT
的本质是啥 会不会下围棋也存疑
这是FB最新的paper里的话 你自己仔细读读再来说
http://arxiv.org/pdf/1511.06410v2.pdf
From the experiments, we clearly show that DCNN is tactically weak due to
the lack of search.
Search is a way to explore the solution space conditioned on the current
board situation, and build
a non-parametric local model for the game. The l... 阅读全帖 |
|
O**l
发帖数: 12923 | 28 这是FB论文里自己说的
From the experiments, we clearly show that DCNN is tactically weak due to
the lack of search.
Search is a way to explore the solution space conditioned on the current
board situation, and build
a non-parametric local model for the game. The local model is more flexible
than the global model
learned from massive training data and more adapted to the current situation
. The state-of-the-art
approach in computer Go is Monte-Carlo Tree Search (MCTS). Fig. 4 shows its
basic principle.
Com... 阅读全帖 |
|
G**********2
发帖数: 4138 | 29 Found the following on Wikipedia:
The distributed version in October 2015 was using 1,202 CPUs and 176 GPUs,[5
] and Google has not publicly explained what hardware and software changes
have improved its performance from October 2015 to March 2016, so the March
matches may well make use of significantly more hardware. |
|
b***k
发帖数: 1661 | 30 An early version of AlphaGo was tested on hardware with various numbers of
CPUs and GPUs, running in asynchronous or distributed mode. Two seconds of
thinking time was given to each move. The resulting Elo ratings are listed
below.[6] In the matches with more time per move higher ratings are achieved.
Configuration and performance
Configuration Search
threads No. of CPU No. of GPU Elo rating
Single[6] p. 10-11 40 48 1 2,151
Single 40 48 2 2,738
Single 40 ... 阅读全帖 |
|
i*******8
发帖数: 7955 | 31 算法最简单但是也最brute force的反锯齿就是渲染高于显示解析度的图然后down
sample到显示解析度。
Multisample anti-aliasing
The term generally refers to a special case of supersampling. Initial
implementations of full-scene anti-aliasing (FSAA) worked conceptually by
simply rendering a scene at a higher resolution, and then downsampling to a
lower-resolution output. Most modern GPUs are capable of this form of
antialiasing, but it greatly taxes resources such as texture bandwidth and
fillrate. (If a game is highly TCL-bound or CPU-bou... 阅读全帖 |
|
a***e
发帖数: 27968 | 32
you got to be kidding. The highest performing chip from tsmc are those gpus.
at 40nm, they max out at around 1ghz. Those arms are max out at 1.5ghz.
what is the max 45nm Intel chip could achieve? X86 chip is a lot more
complex than arm |
|
s*****e
发帖数: 21415 | 33 对，我想起来amd也不是去TSMC流片的，是在ibm支持下自建的fab
是牢
gpus. |
|
t******g
发帖数: 17520 | 34 Why AlphaGo Is Not AI
By Jean-Christophe Baillie
Posted 17 Mar 2016 | 20:50 GMT
This is a guest post. The views expressed here are solely those of the
author and do not represent positions of IEEE Spectrum or the IEEE.
What is AI and what is not AI is, to some extent, a matter of definition.
There is no denying that AlphaGo, the Go-playing artificial intelligence
designed by Google DeepMind that recently beat world champion Lee Sedol, and
similar deep learning approaches have managed to solve qu... 阅读全帖 |
|
s****m
发帖数: 6 | 35 Dr. Wei Zhang, a professor in the Electrical and Computer Engineering
Department at Virginia Commonwealth University, is currently looking for
highly-motivated graduate students to perform research in the fields of real
-time and embedded systems, compiler, computer architecture, and graphics
processing units (GPU) computing.
Applicants should have a M.S or B.S degree in Computer Science, Computer
Engineering or a closely related discipline. Excellent programming skill in
C/C++ (or Java) is a mu... 阅读全帖 |
|
s****m
发帖数: 6 | 36 Dr. Wei Zhang, a professor in the Electrical and Computer Engineering
Department at Virginia Commonwealth University, is currently looking for
highly-motivated graduate students to perform research in the fields of real
-time and embedded systems, compiler, computer architecture, and graphics
processing units (GPU) computing.
Applicants should have a M.S or B.S degree in Computer Science, Computer
Engineering or a closely related discipline. Excellent programming skill in
C/C++ (or Java) is a mu... 阅读全帖 |
|
r******y
发帖数: 3838 | 37 http://www.macrumors.com/2011/01/14/ios-4-3-beta-hints-at-openc
devices/
When it comes to the graphics performance on Apple's iOS devices, the
company has been sticking with the
powerful POWERVR SGX535 GPU from Imagination Technologies for quite some
time, introducing it on the
iPhone 3GS and carrying it through into the company's own A4 system-on-chip
designs for the brains of the
iPad and iPhone 4. But hints included in the new iOS 4.3 beta seeded to
developers earlier this week indicate
that ... 阅读全帖 |
|
r******y
发帖数: 3838 | 38 转一篇，不过我觉得apple的ambition不在此。
Should Apple Buy nVidia?
I've read comments from people elsewhere that have made the case that it
makes sense for Apple. nVidia's
market cap is under 4B of which nearly 50% is cash. They could pay a 100%
premium and the net
acquisition cost would still be around $6B. Apple has lots and lots of cash
. Since Apple has no overlap with
nVidia there should not be a show stopper problem from the regulators -- at
most I think they might have
to agree to continue making e... 阅读全帖 |
|
w********1
发帖数: 3492 | 39 Fri, 29 Jun 2012 11:15:55 PDT
Just after the launch of the Retina MacBook Pro earlier this month,
AnandTech provided a first glimpse of the machine's display performance,
noting the various resolution options available to users and examining how
its color and contrast compares to other notebooks.
After having more time to analyze the new machine, AnandTech last week
published its full review of the Retina MacBook Pro, bringing its thorough
and technically-detailed perspective to the report. Whi... 阅读全帖 |
|
a****a
发帖数: 5763 | 40 http://bbs.weiphone.com/read-htm-tid-556999.html
上篇讲到OpenCL的好处, 但是如果要让Mac用户的日常应用切实的体验到这些优势和影
响, 就需要程序员把OpenCL技术加入到他们的程序中去. 而历来, GPGPU的编程语音一
直没有在传统的桌面程序中得到广泛的应用. 这当中有几个原因.
最初, 针对GPU的编程需要用汇编语言, 而汇编语言跟传统用于桌面程序开发的语言的
差别很大. 在桌面程序的开发上, 广泛应用GUI, API等工具. 即使后来出现的类似C语
音的一些编程语言, 保留了一些针对图形的特性. 但大部分的开发者对于这部分东西仍
然感到很头痛, 像一门他们不太懂的外语一样避而远之.
其次, 即使使用GPU能在很大程度上提高速度, 但依赖于GPU来进行通用计算还是一个问
题, 就是会让程序本身的兼容性变得比较狭窄. 很多老款的GPU, 特别是笔记本上的,
无法运行像CUDA这样的语言.
而苹果在设计OpenCL的时候,有一个很关键的决定就是要让OpenCL不仅能运行在GPU上,
而且还要运行在CPU上. 一个OpenCL程序... 阅读全帖 |
|
m*********t
发帖数: 527 | 41 我回答你提出的问题关于命令行的，你又跑题跑到了并行计算。。。而且，关于并行和
gpu 你说的基本不对。自己去看文档再来
http://www.apple.com/science/solutions/clustercomputingresource
另外
PGI and NVIDIA have worked in cooperation to develop CUDA Fortran. CUDA
Fortran includes a Fortran 2003 compiler and tool chain for programming
NVIDIA GPUs using Fortran. Available in PGI 2010 and later release, CUDA
Fortran is supported on Linux, Mac OS X and Windows |
|
s*********y
发帖数: 6151 | 42 Shipping times for Apple's 15" MacBook Pro with discrete GPU slip to 2-3
weeks, point to refresh
By AppleInsider Staff
Friday, April 24, 2015, 12:39 am PT (03:39 am ET)
Days after apparent supply constraints slowed ship times for certain Apple's
15-inch MacBook Pro models to 1-2 weeks, the Online Apple Store on Friday
extended those estimates to 2-3 weeks.
On Wednesday, AppleInsider reported that 15-inch MacBook Pro with Retina
display inventory has been fluctuating over the past month. With dis... 阅读全帖 |
|
r******y
发帖数: 3838 | 43 【 以下文字转载自 Apple 讨论区 】
发信人: rhapsody ( input type=\"button\" value=), 信区: Apple
标 题: iOS 4.3 Beta Hints at OpenCL-Capable SGX543 GPU in Future Devices
发信站: BBS 未名空间站 (Tue Mar 8 19:46:52 2011, 美东)
http://www.macrumors.com/2011/01/14/ios-4-3-beta-hints-at-openc
devices/
When it comes to the graphics performance on Apple's iOS devices, the
company has been sticking with the
powerful POWERVR SGX535 GPU from Imagination Technologies for quite some
time, introducing it on the
iPhone 3GS and carrying ... 阅读全帖 |
|
k**********g
发帖数: 989 | 44 Try CPU companies
Intel, AMD, also mobile CPU companies e.g. Qualcomm, etc.
Might also take a look at Raspberry Pi, high performance-to-power ratio GPUs
. |
|
w*******r
发帖数: 128 | 45
功耗要求不高means the application does not have stringent requirement on
power consumption. Thus you can use devices like GeForce GPUs which can
consume lots of power. |
|
w***n
发帖数: 1084 | 46 I used cuda a year ago. There was some limitations, especially when you have
multiple GPUs. Not sure if they fixed it or not.
I haven't used opencl yet. |
|
c*******3
发帖数: 21 | 47 Job Title: CUDA Compute Developer Technology Engineer
Location: Beijing
Job Description/Qualifications:
To spearhead its breakthroughs in High Performance Computing and Visual
Computing, NVIDIA is building a team with a unique and critical position
within the company and is looking for passionate world-class software
engineers.
In this role, you will open up new domains of application for GPUs by
researching and developing GPU computing algorithms, driving their adoption
with key application dev... 阅读全帖 |
|
s****m
发帖数: 6 | 48 Dr. Wei Zhang, a professor in the Electrical and Computer Engineering
Department at Virginia Commonwealth University, is currently looking for
highly-motivated graduate students to perform research in the fields of real
-time and embedded systems, compiler, computer architecture, and graphics
processing units (GPU) computing.
Applicants should have a M.S or B.S degree in Computer Science, Computer
Engineering or a closely related discipline. Excellent programming skill in
C/C++ (or Java) is a mu... 阅读全帖 |
|
a***a
发帖数: 34 | 49 The company I work for is looking for a research software engineer with
strong C++, graphics and/or GUI development experience.
Contact me if you are interested.
Some requirements are listed here:
● BS degree or higher in computer/biomedical science or engineering, or
a related field.
● Proficient in object-oriented software development using C/C++
● Good knowledge on building cross-platform code (e.g. using CMAKE,
Visual Studio)
● Understanding and prior experience with OpenGL or 3D... 阅读全帖 |
|
a***a
发帖数: 34 | 50 The company I work for is looking for a research software engineer with
strong C++, graphics and/or GUI development experience.
Contact me if you are interested.
Some requirements are listed here:
● BS degree or higher in computer/biomedical science or engineering, or
a related field.
● Proficient in object-oriented software development using C/C++
● Good knowledge on building cross-platform code (e.g. using CMAKE,
Visual Studio)
● Understanding and prior experience with OpenGL or 3D... 阅读全帖 |
|
