新div
liked
There are also images of this news, so I think it's better to read the source 👇👇👇
NVIDIA Blackwell Is Up & Running In Data Centers: NVLINK Delivers To 1.4 Tb/s, More GPU Details, First-Ever FP4 GenAI Image
Hassan Mujtaba • August 23, 2024 at 11:00 AM EST
NVIDIA Blackwell Is Up & Running In Data Centers: NVLINK Delivers To 1.4 Tb/s, More GPU Details, First-Ever FP4 GenAI Image
Hassan Mujtaba • August 23, 2024 at 11:00 AM EST
Translated
12
新div
liked
NVIDIA Announces Launch of Blackwell Servers, Dismisses Delay Rumors, and Reveals Many Technological Innovations | XenoSpectrum
2024/8/24 (full text reprint) It's a long sentence 👇
Nvidia announced the commencement of operation of the next-generation AI platform Blackwell,I dismissed the rumor about the delay. Ahead of the Hot Chips 2024 conference, the company disclosed the state of the installation and configuration of the Blackwell server, and also clarified details of advanced technologies that will shape the future of AI and high-performance computing, such as innovative liquid cooling technology and new quantization systems.
Nvidia's BlackWell: A comprehensive platform that opens up a new era of AI
Nvidia's BlackWell is designed as a comprehensive ecosystem that goes beyond the boundaries of a simple GPU. Dave Salvator (NVIDIA's Accelerated Computing Product Division Director) said, “NVIDIA Blackwell is a platform, and GPUs are the beginning...
2024/8/24 (full text reprint) It's a long sentence 👇
Nvidia announced the commencement of operation of the next-generation AI platform Blackwell,I dismissed the rumor about the delay. Ahead of the Hot Chips 2024 conference, the company disclosed the state of the installation and configuration of the Blackwell server, and also clarified details of advanced technologies that will shape the future of AI and high-performance computing, such as innovative liquid cooling technology and new quantization systems.
Nvidia's BlackWell: A comprehensive platform that opens up a new era of AI
Nvidia's BlackWell is designed as a comprehensive ecosystem that goes beyond the boundaries of a simple GPU. Dave Salvator (NVIDIA's Accelerated Computing Product Division Director) said, “NVIDIA Blackwell is a platform, and GPUs are the beginning...
Translated
18
2
新div
liked
$NVIDIA (NVDA.US)$
Nvidia is scheduled to announce details of the next-generation GPU architecture “Blackwell” at the Hot Chips meeting next week. In this post, we will analyze Blackwell's technological innovations and their impact on the AI market based on published information.
1 ️ ECT Blackwell Architecture Innovations
a) Transistor count and manufacturing process:
Blackwell is equipped with 208 billion transistors and is expected to be manufactured using TSMC's 4nm process. This is a significant increase from the current Hopper (H100) 80 billion transistor, and a dramatic improvement in computational capability with a single chip is expected.
b) Innovations in chip-to-chip communication:
Due to the evolution of NVSwitch technology, the bandwidth between GPU pairs will double from 900 GB/sec to 1.8 TB/sec. This is an important evolution that prevents data transfer between chips from becoming a bottleneck in operations that require parallel processing, such as large-scale language models (LLM)...
Nvidia is scheduled to announce details of the next-generation GPU architecture “Blackwell” at the Hot Chips meeting next week. In this post, we will analyze Blackwell's technological innovations and their impact on the AI market based on published information.
1 ️ ECT Blackwell Architecture Innovations
a) Transistor count and manufacturing process:
Blackwell is equipped with 208 billion transistors and is expected to be manufactured using TSMC's 4nm process. This is a significant increase from the current Hopper (H100) 80 billion transistor, and a dramatic improvement in computational capability with a single chip is expected.
b) Innovations in chip-to-chip communication:
Due to the evolution of NVSwitch technology, the bandwidth between GPU pairs will double from 900 GB/sec to 1.8 TB/sec. This is an important evolution that prevents data transfer between chips from becoming a bottleneck in operations that require parallel processing, such as large-scale language models (LLM)...
Translated
![Nvidia $NVDA Blackwell Architecture: Technological Innovations and Implications for AI Platforms](https://sgsnsimg.moomoo.com/sns_client_feed/182331612/20240824/93786b48f9a7e87d0c59ba99037b36ad.jpg?area=105&is_public=true)
![Nvidia $NVDA Blackwell Architecture: Technological Innovations and Implications for AI Platforms](https://sgsnsimg.moomoo.com/sns_client_feed/182331612/20240824/5d817dd5a997ebe9a5fd2d60230fd79b.jpg?area=105&is_public=true)
10
新div
liked
$Direxion Daily Semiconductor Bull 3x Shares ETF (SOXL.US)$
I'm really looking forward to N's financial results announcement~![]()
![]()
![]()
![]()
NVIDIA Announces Launch of Blackwell Servers, Dismisses Delay Rumors, and Reveals Many Technological Innovations | XenoSpectrum
NVIDIA Blackwell Is Up & Running In Data Centers: NVLINK Delivers To 1.4 Tb/s, More GPU Details, First-Ever FP4 GenAI Image
I'm predicting 140 or more from Monday to Tuesday, and 150 to 160 when announcing financial results,
I also feel like I'm going any further than that![]()
![]()
![]()
Next week is N. Settlement Festival![]()
Maybe I'll eat Chateaubriand if I make a profit![]()
I'm really looking forward to N's financial results announcement~
NVIDIA Announces Launch of Blackwell Servers, Dismisses Delay Rumors, and Reveals Many Technological Innovations | XenoSpectrum
NVIDIA Blackwell Is Up & Running In Data Centers: NVLINK Delivers To 1.4 Tb/s, More GPU Details, First-Ever FP4 GenAI Image
I'm predicting 140 or more from Monday to Tuesday, and 150 to 160 when announcing financial results,
I also feel like I'm going any further than that
Next week is N. Settlement Festival
Translated
28
2
新div
liked
$NVIDIA (NVDA.US)$
How long will this sales growth continue?![]()
Since I owned it in May last year, I haven't been able to get it out of my head ever since.
We rely on GAFAM for much of our sales, but can we maintain growth when the number of orders from them saturates or declines?
Will there be customers who can replace them?
I can't forget the horror I felt when I saw the Netflix stock price drop by one-third in 2022. 😰
How long will this sales growth continue?
Since I owned it in May last year, I haven't been able to get it out of my head ever since.
We rely on GAFAM for much of our sales, but can we maintain growth when the number of orders from them saturates or declines?
I can't forget the horror I felt when I saw the Netflix stock price drop by one-third in 2022. 😰
Translated
8
新div
voted
新div
liked and commented on
well...
it's about which exit you're aiming for![]()
I'm at a loss as to whether to go with 3 baskets separately or put them all together and concentrate on 1 point![]()
it's about which exit you're aiming for
I'm at a loss as to whether to go with 3 baskets separately or put them all together and concentrate on 1 point
Translated
7
7
新div
liked and commented on
Buy stocks when they are cheap and sell when they are expensive! I have no choice but to focus on this, but it's not easy
Since they didn't sell during unrealized gains, they ended up with unrealized losses over and over
It's frustrating but I haven't been able to learn![]()
Buying and selling are both difficult to limit prices![]()
Buying and selling are both difficult to limit prices
Translated
15
18
新div
liked
Will the new Tesla “Model Y” be drastically increased in power, or the interior and exterior will also be revamped, and the texture will be improved | AUTOCAR JAPAN
Published: 2024.08.23 06:25 (some excerpts)
Tesla is planning to improve the Model Y. There is also a possibility that cruising range will increase due to improvements in aerodynamics and control systems. Photos of the prototype have been released online, and it has been suggested that the announcement is not far off.
Focus on aerodynamics, etc.
The Tesla Model Y, the best-selling car in the world, will receive drastic improvements in 2025, and it is expected that the interior and exterior will be revamped and performance improved.
Tesla's sales volume has not increased month-on-month since the end of 2023. The aim is to strengthen the product power of the Model Y in order to break away from this slump.
Published: 2024.08.23 06:25 (some excerpts)
Tesla is planning to improve the Model Y. There is also a possibility that cruising range will increase due to improvements in aerodynamics and control systems. Photos of the prototype have been released online, and it has been suggested that the announcement is not far off.
Focus on aerodynamics, etc.
The Tesla Model Y, the best-selling car in the world, will receive drastic improvements in 2025, and it is expected that the interior and exterior will be revamped and performance improved.
Tesla's sales volume has not increased month-on-month since the end of 2023. The aim is to strengthen the product power of the Model Y in order to break away from this slump.
Translated
8
新div
liked
$NVIDIA (NVDA.US)$
1. Getting started
With the rapid development of artificial intelligence (AI), computational resources required for large-scale language model (LLM) training are growing exponentially. This trend has brought new challenges to the network architecture of AI clusters. Conventional Clos networks are suitable for distributed processing of applications across data centers, but there is a possibility that specifications are excessive for AI-only clusters.
In this post, I will analyze in detail the innovative “Rail-Only” network architecture proposed by researchers at the Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence (CSAIL) and Meta Platforms. This new approach makes it possible to maintain performance while dramatically reducing the network cost of AI training systems.
2. Issues with traditional network architectures
2.1 C...
1. Getting started
With the rapid development of artificial intelligence (AI), computational resources required for large-scale language model (LLM) training are growing exponentially. This trend has brought new challenges to the network architecture of AI clusters. Conventional Clos networks are suitable for distributed processing of applications across data centers, but there is a possibility that specifications are excessive for AI-only clusters.
In this post, I will analyze in detail the innovative “Rail-Only” network architecture proposed by researchers at the Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence (CSAIL) and Meta Platforms. This new approach makes it possible to maintain performance while dramatically reducing the network cost of AI training systems.
2. Issues with traditional network architectures
2.1 C...
Translated
![Network architecture innovation in AI training systems: $NVDA enables Rail-only architectures with NVDA NVSwitch](https://sgsnsimg.moomoo.com/sns_client_feed/182331612/20240824/65bf584cec1e895c980e82b83eb147d1.jpg/thumb?area=105&is_public=true)
![Network architecture innovation in AI training systems: $NVDA enables Rail-only architectures with NVDA NVSwitch](https://sgsnsimg.moomoo.com/sns_client_feed/182331612/20240824/bc0879ba019ec4076fc3812612fa76a5.jpg/thumb?area=105&is_public=true)
![Network architecture innovation in AI training systems: $NVDA enables Rail-only architectures with NVDA NVSwitch](https://sgsnsimg.moomoo.com/sns_client_feed/182331612/20240824/3ded2f1d15d4294986c16930ed4827f1.jpg/thumb?area=105&is_public=true)
+1
11
1