Загрузка видео...

Не удалось загрузить видео

На главную

Stuck waiting on training? Hit fast-forward. Amazon EC2 Trn3 UltraServers enable up to 4.4x higher performance, 3.9x higher memory bandwidth and 4x better performance/watt, meaning lower training costs. Built for agentic AI, MoE, and video models, with native PyTorch integration so you don’t change a single line of code.

awscloud's profile picture

Amazon Web Services

2,217,437 subscribers

1,281,551 просмотров • 6 месяцев назад •via X (Twitter)

ОбразованиеЗдоровье и велнесНаука и технологии
LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

When your AI models dream of big compute, they dream of this. 🚀 Introducing Amazon EC2 P6e-GB200 UltraServers, purpose-built for training & deploying frontier #AI models. Big dreams demand even bigger computing power. #AWS 👉
0:38

When your AI models dream of big compute, they dream of this. 🚀 Introducing Amazon EC2 P6e-GB200 UltraServers, purpose-built for training & deploying frontier #AI models. Big dreams demand even bigger computing power. #AWS 👉

Amazon Web Services

35,909 просмотров • 11 месяцев назад

The #IntelGaudi 3 #AI accelerator offers a highly competitive alternative to NVIDIA’s H100 with higher performance, increased scalability, and PyTorch integration. Explore more key product benefits.
0:29

The #IntelGaudi 3 #AI accelerator offers a highly competitive alternative to NVIDIA’s H100 with higher performance, increased scalability, and PyTorch integration. Explore more key product benefits.

Intel AI

71,832 просмотров • 2 лет назад

$HASHAI is Ahead of the Curve ✅ Today’s Amazon Web Services outage highlights the advantage of truly decentralised, reliable infrastructure. 🔹Hash AI’s Node network is purpose-built for real-world, high-performance computing, built on Ethereum. From training and deploying AI models to executing complex data processing and ETL workflows. 🔹Experience lower costs, higher reliability, and complete freedom from vendor lock-in with infrastructure engineered for scale, speed, and resilience. Rent now → Learn more →
0:41

$HASHAI is Ahead of the Curve ✅ Today’s Amazon Web Services outage highlights the advantage of truly decentralised, reliable infrastructure. 🔹Hash AI’s Node network is purpose-built for real-world, high-performance computing, built on Ethereum. From training and deploying AI models to executing complex data processing and ETL workflows. 🔹Experience lower costs, higher reliability, and complete freedom from vendor lock-in with infrastructure engineered for scale, speed, and resilience. Rent now → Learn more →

Hash AI

17,248 просмотров • 8 месяцев назад

Excited to be launching a series of short videos (5 min or less) to optimize the training performance of your PyTorch models! Covering - mixed-precision training - multi-GPU training strategies - Compiling PyTorch models - and finding good batch sizes 🔗
0:35

Excited to be launching a series of short videos (5 min or less) to optimize the training performance of your PyTorch models! Covering - mixed-precision training - multi-GPU training strategies - Compiling PyTorch models - and finding good batch sizes 🔗

Sebastian Raschka

120,786 просмотров • 3 лет назад

Amazon Redshift RG, powered by AWS Graviton, deliver better performance - running data warehouse and data lake workloads up to 2.4x as fast as previous generation RA3 at 30% lower cost per vCPU. Run SQL analytics across your data warehouse and data lake from a single engine — including on open formats such as Apache Iceberg and Apache Parquet. Built to meet the demands of today’s analytics and agentic AI workloads.
0:28

Amazon Redshift RG, powered by AWS Graviton, deliver better performance - running data warehouse and data lake workloads up to 2.4x as fast as previous generation RA3 at 30% lower cost per vCPU. Run SQL analytics across your data warehouse and data lake from a single engine — including on open formats such as Apache Iceberg and Apache Parquet. Built to meet the demands of today’s analytics and agentic AI workloads.

Amazon Web Services

13,444 просмотров • 1 месяц назад

Meet Amazon Nova 2 reasoning models Nova has introduced its next generation of foundation models that deliver frontier intelligence with industry-leading price performance 1/ Amazon Nova 2 Lite, is a fast, cost-effective model designed for everyday AI tasks. 2/ Amazon Nova 2 Pro (preview), Nova’s most intelligent model built for complex AI workloads. These models offer developers control to manage thinking budgets, native support for agentic workflows, and provide a one-million-token context window for richer interactions.
0:56

Meet Amazon Nova 2 reasoning models Nova has introduced its next generation of foundation models that deliver frontier intelligence with industry-leading price performance 1/ Amazon Nova 2 Lite, is a fast, cost-effective model designed for everyday AI tasks. 2/ Amazon Nova 2 Pro (preview), Nova’s most intelligent model built for complex AI workloads. These models offer developers control to manage thinking budgets, native support for agentic workflows, and provide a one-million-token context window for richer interactions.

Amazon Web Services

3,958,154 просмотров • 6 месяцев назад

Warm up to improved workload performance with the #AWS Cloud. ☁️💻☀️ Powered by next gen AWS Graviton3E processors, Amazon EC2 Hpc7g instances deliver up to 60% better performance for compute-intensive workloads over the previous gen. #EC2 #Compute 👉
0:30

Warm up to improved workload performance with the #AWS Cloud. ☁️💻☀️ Powered by next gen AWS Graviton3E processors, Amazon EC2 Hpc7g instances deliver up to 60% better performance for compute-intensive workloads over the previous gen. #EC2 #Compute 👉

Amazon Web Services

27,858 просмотров • 3 лет назад

$AMD CEO Lisa Su said Helios has industry-leading specs and is on track to launch this year. Built with $META, Helios targets large-scale AI training & inference aiming for up to 10x performance where performance per dollar matters.
1:17

$AMD CEO Lisa Su said Helios has industry-leading specs and is on track to launch this year. Built with $META, Helios targets large-scale AI training & inference aiming for up to 10x performance where performance per dollar matters.

Shay Boloor

117,981 просмотров • 5 месяцев назад

Does your business have the performance it needs to support your changing AI infrastructure? AMD and Microsoft Azure are built for next-level speed and efficiency. From processing transactions to handling complex workloads, leadership performance per watt helps you take charge of change.
0:43

Does your business have the performance it needs to support your changing AI infrastructure? AMD and Microsoft Azure are built for next-level speed and efficiency. From processing transactions to handling complex workloads, leadership performance per watt helps you take charge of change.

AMD

13,458 просмотров • 1 год назад

This is my favorite clip of the new Elon pod. He opens up saying xAI struggles with memory usage/bandwidth and CUDA kernel optimization (matmul, attention, MoE, etc). If you are good kernel or performance engineering in general, you should apply. Steer the world in a better direction.
0:37

This is my favorite clip of the new Elon pod. He opens up saying xAI struggles with memory usage/bandwidth and CUDA kernel optimization (matmul, attention, MoE, etc). If you are good kernel or performance engineering in general, you should apply. Steer the world in a better direction.

Elliot Arledge

163,922 просмотров • 5 месяцев назад

Today, we’re launching the Parallel Search API, the most accurate web search for AI agents, built using our proprietary web index and retrieval infrastructure. Traditional search ranks URLs for humans to click. AI search needs something different: the right tokens in their context window. Parallel’s Search API is built from the ground up for AI agents, resulting in higher accuracy, lower costs, and lower latency for end-to-end agentic workflows.
1:20

Today, we’re launching the Parallel Search API, the most accurate web search for AI agents, built using our proprietary web index and retrieval infrastructure. Traditional search ranks URLs for humans to click. AI search needs something different: the right tokens in their context window. Parallel’s Search API is built from the ground up for AI agents, resulting in higher accuracy, lower costs, and lower latency for end-to-end agentic workflows.

Parallel Web Systems

669,541 просмотров • 7 месяцев назад

Video of SAFTI City – Army’s main training ground for urban ops. Instrumented with sensors for video playback and interactive targets that can also shoot back, it brings training to a much higher level of realism.
0:42

Video of SAFTI City – Army’s main training ground for urban ops. Instrumented with sensors for video playback and interactive targets that can also shoot back, it brings training to a much higher level of realism.

Ng Eng Hen

46,846 просмотров • 1 год назад

Another high intent Friday, mic'd up on the High Performance training floor. Training hard, moving fast.
1:43

Another high intent Friday, mic'd up on the High Performance training floor. Training hard, moving fast.

Driveline Baseball

13,382 просмотров • 10 месяцев назад

Tesla’s computer vision delivers unmatched performance at $1,000-$1,200, a fraction of competitors’ costs—3-4x higher for inferior solutions. Vertical integration of AI silicon, a decade-long bet by Musk, mirrors Apple’s edge over Android’s fragmentation. This is Tesla’s moat. Can rivals close the gap? Full episode with Jeff Lutz 🔋 in the replies 👇
1:05

Tesla’s computer vision delivers unmatched performance at $1,000-$1,200, a fraction of competitors’ costs—3-4x higher for inferior solutions. Vertical integration of AI silicon, a decade-long bet by Musk, mirrors Apple’s edge over Android’s fragmentation. This is Tesla’s moat. Can rivals close the gap? Full episode with Jeff Lutz 🔋 in the replies 👇

Farzad Content

10,522 просмотров • 10 месяцев назад

Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devices. Here’s what Gemma 4 unlocks for developers: — Intelligence-per-parameter: Our 31B (Dense) and 26B (MoE) models deliver state-of-the-art performance for their size, outcompeting models 20x their size on Arena.ai — Commercial flexibility: Released under a permissive Apache 2.0 license for complete developer flexibility and digital sovereignty — Agentic workflows: Native support for function-calling and structured JSON output allows you to build reliable, autonomous agents — Multimodal edge AI: The E2B and E4B models bring native vision, audio, and low latency to mobile and IoT devices — Long-context reasoning: Up to 256K context windows allow you to process entire repositories or large documents in a single prompt Whether you're building global applications in 140+ languages or local-first AI code assistants, Gemma 4 is built to be your foundation. Explore in Google AI Studio or download the weights on Hugging Face, Kaggle, and ollama.
2:51

Today, we’re launching Gemma 4, our most intelligent open models to date. Built with the same breakthrough technology as Gemini 3, Gemma 4 brings advanced reasoning to your personal hardware and devices. Here’s what Gemma 4 unlocks for developers: — Intelligence-per-parameter: Our 31B (Dense) and 26B (MoE) models deliver state-of-the-art performance for their size, outcompeting models 20x their size on Arena.ai — Commercial flexibility: Released under a permissive Apache 2.0 license for complete developer flexibility and digital sovereignty — Agentic workflows: Native support for function-calling and structured JSON output allows you to build reliable, autonomous agents — Multimodal edge AI: The E2B and E4B models bring native vision, audio, and low latency to mobile and IoT devices — Long-context reasoning: Up to 256K context windows allow you to process entire repositories or large documents in a single prompt Whether you're building global applications in 140+ languages or local-first AI code assistants, Gemma 4 is built to be your foundation. Explore in Google AI Studio or download the weights on Hugging Face, Kaggle, and ollama.

Google AI

1,668,807 просмотров • 2 месяцев назад

⚡️Training RL on world models used to be a headache. Not anymore. ⚡️ Meet *WorldCompass*, a major update to *HY World 1.5*: 🏃‍♂️ Fast rollouts 🏗️ stable train ✨ Higher quality paper: code: page:
0:26

⚡️Training RL on world models used to be a headache. Not anymore. ⚡️ Meet *WorldCompass*, a major update to *HY World 1.5*: 🏃‍♂️ Fast rollouts 🏗️ stable train ✨ Higher quality paper: code: page:

Tengfei Wang

15,385 просмотров • 4 месяцев назад

GPU tradeoff series: A100 is not much more powerful than 4090 🫠 GPU Perf and Price: - 4090: 330 fp16 TFLOPs, $1,749 - A100 (80GB): 312 fp16 TFLOPs, $20,000 > A100 is 11.4X more pricy Training speed for GPT-2(124M) with llm.c: - 4090: 153K tokens/s - A100 (80GB): 195K tokens/s > A100 is only 1.3X faster (both trained using a single card, A100 llm.c training is shown in the video, 4090 video is in the quoted tweet) Conclusion: 4090 has a much better cost vs performance ratio Why: As in the H100 vs. 4090 comparison, the biggest difference between A100 and 4090 is their GPU memory size/bandwidth and cross-GPU communication bandwidth, which does not matter too much if your model can fit into a single 4090. Specs: 4090: - GPU memory size: 24GB - memory bandwidth: 1 TB/s - communication bandwidth: 64 GB/s A100: - GPU memory size: 80GB - memory bandwidth: 2 TB/s - communication bandwidth: 900 GB/s Nvidia killed off NVLink (a high-speed communication link that connects GPUs) on 4090. (Jensen Huang smiling face) If multiple 4090s could be interconnected via NVLink, their performance would be closer to datacenter-grade A100 GPUs, even for training larger models. Additionally, 4090 isn't allowed in datacenters, that's how Nvidia makes 💰💰💰
0:32

GPU tradeoff series: A100 is not much more powerful than 4090 🫠 GPU Perf and Price: - 4090: 330 fp16 TFLOPs, $1,749 - A100 (80GB): 312 fp16 TFLOPs, $20,000 > A100 is 11.4X more pricy Training speed for GPT-2(124M) with llm.c: - 4090: 153K tokens/s - A100 (80GB): 195K tokens/s > A100 is only 1.3X faster (both trained using a single card, A100 llm.c training is shown in the video, 4090 video is in the quoted tweet) Conclusion: 4090 has a much better cost vs performance ratio Why: As in the H100 vs. 4090 comparison, the biggest difference between A100 and 4090 is their GPU memory size/bandwidth and cross-GPU communication bandwidth, which does not matter too much if your model can fit into a single 4090. Specs: 4090: - GPU memory size: 24GB - memory bandwidth: 1 TB/s - communication bandwidth: 64 GB/s A100: - GPU memory size: 80GB - memory bandwidth: 2 TB/s - communication bandwidth: 900 GB/s Nvidia killed off NVLink (a high-speed communication link that connects GPUs) on 4090. (Jensen Huang smiling face) If multiple 4090s could be interconnected via NVLink, their performance would be closer to datacenter-grade A100 GPUs, even for training larger models. Additionally, 4090 isn't allowed in datacenters, that's how Nvidia makes 💰💰💰

Yuchen Jin

234,951 просмотров • 1 год назад

Here's what Donald Trump's executive orders and immigration actions will mean for you: Higher housing & food costs when construction and farm workers are deported. Higher costs across the board when consumers pay for Trump tariffs. So much for his promise of lower prices.
0:55

Here's what Donald Trump's executive orders and immigration actions will mean for you: Higher housing & food costs when construction and farm workers are deported. Higher costs across the board when consumers pay for Trump tariffs. So much for his promise of lower prices.

Adam Schiff

292,105 просмотров • 1 год назад

My startup, Positron AI just raised $51.6M Series A to rebuild the infrastructure powering AI inference and bring Superintelligence to everyone. Led by Valor Equity Partners (@valorep) , Atreides Management (Atreides Management, LP) , and DFJ Growth (DFJ Growth) —the same teams behind SpaceX, Tesla, xAI, and other companies pushing compute to its physical and economic limits. Why inference? Because as AI moves from research into production, the real bottleneck isn’t training—it’s deploying transformer models at scale. GPUs were built for flexibility, not predictable workloads. That mismatch means wasted bandwidth, poor density, and massive power costs. So at Positron, we built silicon specifically optimized for inference: •>90% memory bandwidth utilization (GPUs typically ~30%) •Multi-model hosting per card (higher density, lower power) •Zero code changes — drop in compatability with the HuggingFace Transformers ecosystem. U.S.-made silicon, stable supply chain, geopolitically resilient We’re already shipping and running in production environments today. Here’s what we’ve built, and link to our WSJ coverage and press release in the follow up reply ↓
6:20

My startup, Positron AI just raised $51.6M Series A to rebuild the infrastructure powering AI inference and bring Superintelligence to everyone. Led by Valor Equity Partners (@valorep) , Atreides Management (Atreides Management, LP) , and DFJ Growth (DFJ Growth) —the same teams behind SpaceX, Tesla, xAI, and other companies pushing compute to its physical and economic limits. Why inference? Because as AI moves from research into production, the real bottleneck isn’t training—it’s deploying transformer models at scale. GPUs were built for flexibility, not predictable workloads. That mismatch means wasted bandwidth, poor density, and massive power costs. So at Positron, we built silicon specifically optimized for inference: •>90% memory bandwidth utilization (GPUs typically ~30%) •Multi-model hosting per card (higher density, lower power) •Zero code changes — drop in compatability with the HuggingFace Transformers ecosystem. U.S.-made silicon, stable supply chain, geopolitically resilient We’re already shipping and running in production environments today. Here’s what we’ve built, and link to our WSJ coverage and press release in the follow up reply ↓

Thomas Sohmers

26,324 просмотров • 11 месяцев назад

Introducing HyperTune A performance engine designed to get more out of the hardware you already own Higher frames. Tighter responsiveness. Better stability. All in one click. Built for pros. Available to everyone.
1:49

Introducing HyperTune A performance engine designed to get more out of the hardware you already own Higher frames. Tighter responsiveness. Better stability. All in one click. Built for pros. Available to everyone.

HYPERTUNE

119,776 просмотров • 7 месяцев назад