GPU-accelerated Apache Spark with Amazon EMR and NVIDIA RTX PRO 4500 on Amazon EC2 G7 instances runs up to 3.7x faster

Post Syndicated from McCall Peltier original https://aws.amazon.com/blogs/big-data/gpu-accelerated-apache-spark-with-amazon-emr-and-nvidia-rtx-pro-4500-on-amazon-ec2-g7-instances-runs-up-to-3-7x-faster/

For years, Apache Spark has been the backbone of large-scale data processing. However, as datasets grow and artificial intelligence and machine learning (AI/ML) pipelines become more complex, modern workloads demand more computational power. Feature engineering for machine learning models, large-scale extract, transform, and load (ETL) transformations, and real-time analytics workloads are computationally intensive by nature. GPU-accelerated instances improve performance and transform jobs that once took hours into minutes, so you can iterate on models faster and reduce operational costs. You can process larger datasets in single batches, make decisions in real time, and achieve strong performance without over-provisioning infrastructure.

We’re excited to share the benchmarking results on Amazon EMR with Amazon Elastic Compute Cloud (Amazon EC2) G7 instances, powered by NVIDIA RTX PRO 4500 Blackwell Server Edition GPUs. For data engineers and data scientists running Apache Spark workloads, this means faster pipelines, shorter iteration cycles, and more time spent on insights.

Amazon EMR on EKS natively supports the NVIDIA cuDF plugin for Apache Spark. This support is the result of joint engineering between AWS and NVIDIA to qualify the cuDF plugin for Amazon EMR, co-optimize Spark execution paths for RTX PRO 4500, and validate performance at scale through shared TPC-DS benchmarking on Amazon EC2 G7 instances. Now, Apache Spark workloads on Amazon EMR on EKS run up to 3.7x faster with Amazon EC2 G7 GPU instances than with comparable CPU instances, and require no changes to existing Spark code.

In the TPC-DS 3 TB benchmark, at the 64 GB memory tier, EC2 G7 instances with RTX PRO finished in 4.7 minutes. If you run large-scale data processing pipelines, you can cut job run times by more than two-thirds while maintaining full compatibility with the applications you already have in production.

The use cases that benefit most are those where speed directly unlocks business value. In AI/ML feature engineering, faster Spark jobs mean data science teams can iterate on features more quickly, reducing the time from raw data to trained model. In complex ETL pipelines, like financial transactions, clickstream aggregation, or supply chain data consolidation, GPU acceleration compresses multi-hour batch windows into near-real-time processing. For real-time analytics, teams running fraud detection, personalization engines, or operational dashboards can process larger volumes of data within tighter latency windows, without redesigning their architecture.

Beyond data analytics, the G7 instances will support a broad range of AI and graphics workloads, including conversational AI, content generation, recommender systems, and video streaming and rendering. Built on the AWS Nitro System, they deliver the security and resource efficiency that production AI, analytics, and graphics workloads demand.

The following sections walk through the cluster configuration, benchmark methodology, and performance results.

Cluster configuration

We benchmarked four instance types to measure the real-world performance of G7 GPU instances against comparable CPU instances for Spark SQL performance. The g7.4xlarge also provides 80 Gbps network bandwidth (compared to 15–17 Gbps on the CPU baselines) and uses RapidsShuffleManager. However, CPU runs showed no evidence of being network- or shuffle-bound at this cluster scale. All tests used Amazon EMR on EKS 7.12.0 with Apache Spark 3.5.6 and cuDF plugin 26.04.2, running the full TPC-DS benchmark at 3 TB scale across 103 queries. Each experiment ran 5 iterations. We report the median. Data was stored as Parquet on Amazon Simple Storage Service (Amazon S3) (same-region gateway endpoint). All instances were launched in a single Availability Zone.

Instance specifications

All four instance types share the same compute footprint of 16 vCPUs and 64 GB system RAM. The g7.4xlarge additionally includes an NVIDIA RTX PRO 4500 Blackwell GPU with 32 GB of dedicated video memory (VRAM), which the cuDF plugin uses to accelerate Spark SQL operations. The baseline for all speedup and cost comparisons is m9gd.4xlarge (Graviton), the lowest-cost CPU instance in the group.

. g7.4xlarge m9gd.4xlarge m8id.4xlarge m8a.4xlarge
Architecture x86_64 arm64 (Graviton) x86_64 x86_64
vCPU 16 16 16 16
RAM 64 GB 64 GB 64 GB 64 GB
GPU 1× RTX PRO 4500 Blackwell (32 GB VRAM)
NVMe 875 GB 950 GB 950 GB EBS only (GP3 16k IOPS and 2000 MB/s throughput to match NVMe
Network 80 Gbps Up to 17 Gbps Up to 15 Gbps Up to 15 Gbps

The g7.4xlarge uses the RTX PRO 4500 Blackwell Server Edition GPU. The CPU baselines cover all three major architectures: m8id.4xlarge (Intel x86), m8a.4xlarge (AMD x86), and m9gd.4xlarge (Graviton arm64).

Spark configuration

All instances used eight executor nodes with the following configuration:

Configuration GPU instances CPU instances
Amazon EMR release emr-7.12.0-spark-rapids-latest emr-7.12.0-latest
executor.cores 14 14
executor.instances 8 8
executor.memory 20G 20G
executor.memoryOverhead 30G 30G
spark.plugins com.nvidia.spark.SQLPlugin
rapids.memory.pinnedPool.size 8G
rapids.sql.concurrentGpuTasks 3
shuffle.manager RapidsShuffleManager default (sort)
sql.adaptive.enabled true true
io.compression.codec zstd zstd

CPU instances use the same 30 GB memoryOverhead as GPU to make sure that the memory comparison is apples-to-apples. This setting reserves off-heap memory for shuffle and caching on both sides.

For GPU instances, the cuDF plugin offloads Spark SQL operations to the GPU automatically. No code changes are required. The executor.memoryOverhead value is set higher on GPU instances to accommodate GPU memory management and the RAPIDS shuffle manager.

The cuDF plugin automatically falls back to CPU execution for unsupported operators and user-defined functions (UDFs). Your job still completes, but those stages run without GPU acceleration. To identify which operations run on GPU compared to CPU, set spark.rapids.sql.explain=NOT_ON_GPU in your Spark configuration. For a pre-migration assessment of your workloads, use the NVIDIA cuDF tool to estimate GPU acceleration potential before moving to G7 instances.

To tune settings like concurrentGpuTasks and pinnedPool.size, use the Spark History Server on Amazon EMR on EKS, which provides per-stage execution details to identify CPU fallback and shuffle bottlenecks.

Getting started

Reference the Using cuDF Accelerator for Apache Spark with Amazon EMR on EKS for detailed setup instructions.

Prerequisites

Before running GPU-accelerated Spark on Amazon EMR on EKS, make sure the following are in place:

  • Amazon EMR on EKS release version 6.9.0 or later (this post uses emr-7.12.0-spark-rapids-latest).

The -spark-rapids release variant ships the NVIDIA cuDF plugin pre-installed.

  • Amazon Elastic Kubernetes Service (Amazon EKS) cluster with a GPU-enabled node group using G7 instances.
  • Node AMI: AL2023_x86_64_NVIDIA (Amazon EKS optimized accelerated AMI).
  • NVIDIA device plugin installed in the cluster to expose GPUs to Kubernetes pods:
    kubectl apply -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v0.9.0/nvidia-device-plugin.yml

  • Amazon EMR on EKS virtual cluster registered to the EKS namespace.

To validate GPU availability on your nodes:

kubectl get nodes "-o=custom-columns=NAME:.metadata.name,GPU:.status.allocatable.nvidia\.com/gpu"

Note: Getting started with GPU-accelerated Spark on Amazon EMR is straightforward. To use the latest cuDF plugin, overlay the latest version (for example, 26.04.2 as of May 2026) onto the Amazon EMR RAPIDS image using an initContainer technique. This replaces the bundled cuDF JAR  with a newer version while preserving all other Amazon EMR dependencies. We recommend using the latest Amazon EMR release to get the most up-to-date cuDF plugin for better performance. In our benchmarks, upgrading from cuDF plugin 25.08.0 to 26.04.2 reduced runtime by 36–38 percent. Download the latest cuDF plugin JAR from the NVIDIA repository. AWS Support covers Amazon EMR. For issues specific to a cuDF JAR, file a GitHub issue or contact NVIDIA at .

Performance benchmarks and cost efficiency

We ran the full TPC-DS benchmark suite (103 queries) at 3 TB scale on 8-node clusters in us-east-1. The following table summarizes the results:

. GPU instances CPU instances
Cost per run $2.06 $2.93–$3.18
Total time (103 queries) 281s (4.7 min) 1,010–1,043s (16.8–17.4 min)
Speedup compared to CPU instances 3.7× baseline

Cost per run is the total cluster cost for the benchmark’s duration: Cluster $/hr × (median runtime ÷ 3,600). The hourly rate combines the EC2 On-Demand cost for all 8 nodes and the Amazon EMR on EKS charge for the vCPU and memory the Spark pods consume. Both are billed per second (one-minute minimum), so you pay only for what a job uses while it runs. All runs used Amazon EMR on EKS 7.12.0 in us-east-1, with 8 × 4xlarge nodes (128 vCPU) on both the GPU and CPU sides. The g7.4xlarge cluster runs at $26.35/hr (8 × $3.042 EC2 = $24.34, plus $2.01 for Amazon EMR on EKS) and finishes in 281 seconds, at $2.06 per run. The CPU clusters run at a lower hourly rate ($10.43–$10.99) but take 1,010–1,043 seconds, landing at $2.93–$3.18 per run. All prices reflect On-Demand pricing in us-east-1 as of May 2026. G7 instances are also eligible for EC2 Spot and Compute Savings Plans, which can further reduce costs for recurring batch workloads.

Cost-per-run calculations include EC2 and Amazon EMR charges only. They exclude the EKS control-plane fee, EBS volumes, S3 request and storage costs, and the driver pod.

Bar chart of total TPC-DS runtime by instance type, showing g7.4xlarge finishing far faster than the CPU instances

Figure 1: Total runtime by instance type for all 103 TPC-DS queries at 3 TB scale. The g7.4xlarge with GPU acceleration completed the benchmark in 4.7 minutes, 3.7× faster than CPU instances (16.8-17.4 minutes)

Bar chart of total cost per benchmark run by instance type, showing the g7.4xlarge GPU instance costing less than the CPU instances

Figure 2: Total cost per benchmark run, including both Amazon EC2 instance and Amazon EMR on EKS cost across all 8 nodes. Despite a ~2.5× higher hourly rate, the g7.4xlarge GPU instance costs up to 31% less per run than Graviton because it finishes the workload 3.7× faster

Where GPU acceleration excels

GPU acceleration completed the 103-query power run in 281s compared to 1,032s on CPU, an overall 3.7× speedup that saves 750 seconds per run. GPU was faster on 102 of 103 query executions.

GPU acceleration delivers the largest gains on the long-running, compute- and shuffle-heavy queries where kernel throughput outweighs launch overhead. The biggest absolute time savings:

Query CPU time GPU time Speedup Time saved
q24 (part 1+2) 81.6s 15.9s ~5.1× 65.7s
q23 (part 1+2) 79.4s 16.4s ~4.9× 63.1s
q93 63.7s 5.6s 11.4× 58.1s
q76 30.4s 3.4s 9.0× 27.0s
q64 35.4s 8.5s 4.2× 26.9s
q50 27.6s 3.5s 7.9× 24.0s

Speedup distribution across all 103 executions:

Speedup band Queries
≥5× 15
4–5× 13
3–4× 21
2–3× 26
1–2× 27
<1× (CPU faster) 1

Median per-query speedup 2.94× (geomean 2.84×). The heaviest wins (q50, q76, q93) are aggregation- and shuffle-join-intensive queries that convert cleanly to GpuHashAggregate and GpuBroadcastHashJoin.

Where CPU wins

With RAPIDS 26.04.2, the following query showcases a workload pattern where CPU was faster:

Query CPU time GPU time Ratio Root cause
q16 0.96s 1.44s CPU 1.5x faster Trivial/near-empty scan. Sub-second runtime where GPU kernel-launch overhead is not amortized

Choosing the right instance

Instance Best for Summary
g7.4xlarge (RTX PRO GPU) Fastest and most cost-effective Up to 3.7× faster than comparable CPU instances and up to 31% cheaper per run. Completes in 4.7 min compared to 17.2 min. Best choice for both speed and cost efficiency.
CPU instances (m8a / m8id / m9gd) Flexibility, availability, and always-on workloads Multiple architecture options deliver similar Spark SQL performance. Choose CPU when GPUs are unavailable, when clusters need to remain running continuously (for example, overnight jobs ready for next-day analysis), or when workloads cannot use GPU acceleration. CPU instances offer broad availability and predictable capacity without startup delays.

G7 instances require a G-instance vCPU service quota in your account (default is often 0 for GPU types). Request a quota increase through the Service Quotas console, or use On-Demand Capacity Reservations (ODCRs) to guarantee availability for recurring batch jobs.

Based on these benchmark results, consider evaluating GPU acceleration for your own Apache Spark workloads. Start by identifying compute-intensive operations in your current pipelines, particularly those involving large-scale aggregations, joins, or machine learning feature engineering that could benefit from the performance improvements demonstrated here.

Conclusion

Amazon EMR on EKS with NVIDIA RTX PRO 4500 together provide a meaningful step forward for teams running data-intensive Spark workloads at scale. Whether you’re building ML pipelines that demand rapid feature iteration, running complex ETL transformations across massive datasets, or powering real-time analytics that can’t afford to wait on slow batch jobs, GPU-accelerated Spark on G7 delivers the performance and speed to do more. As data and AI workloads continue to evolve, GPU-accelerated analytics on Amazon EMR is becoming the foundation for data teams. Get started with GPU-accelerated Spark on Amazon EMR on EKS today by visiting Amazon EMR documentation to launch your first G7-powered cluster and see the performance gains for yourself.


About the authors

McCall Peltier

McCall Peltier

McCall is a Senior Product Marketing Manager at AWS focused on data processing services, including Amazon EMR. She leads messaging and launches that support customers building modern data platforms on AWS, collaborating across product and field teams to drive adoption and customer impact.

Karthik Prabhakar

Karthik Prabhakar

Karthik is a Data Processing Engines Architect for Amazon EMR at Amazon Web Services (AWS). He specializes in distributed systems architecture and query optimization, working with customers to solve complex performance challenges in large-scale data processing workloads. His focus spans engine internals, cost-optimization strategies, and architectural patterns that enable customers to run petabyte-scale analytics efficiently

Kshitija Dound

Kshitija Dound

Kshitija is a Specialist Solutions Architect at AWS based in New York City, focusing on data and AI. She collaborates with customers to transform their ideas into cloud solutions, using AWS Big Data and AI services. She also engages in public speaking opportunities, sharing her expertise on cloud technologies, industry trends, and career in the cloud. In her spare time, Kshitija enjoys exploring museums, indulging in art, and embracing NYC’s outdoor scene.

Kinshuk Paharae

Kinshuk Pahare

Kinshuk is head of product for data processing, leading product teams for AWS Glue, Amazon EMR, and Amazon Athena. He has been with AWS for over 6 years.

Happy 20th Birthday, Amazon EC2

Post Syndicated from Channy Yun (윤석찬) original https://aws.amazon.com/blogs/aws/happy-20th-birthday-amazon-ec2/

Twenty years ago today, Jeff Barr wrote a blog post that launched the Amazon EC2 Beta. That single post introduced resizable Linux virtual servers in the cloud, billed by the hour, with one instance type (m1.small) in one Region (US East). It was minimal yet useful, and it changed how the world thinks about computing infrastructure.

In 2021, Jeff covered the fifteen years of EC2 with the backstory and memorable EC2 launches. Over the last five years, AWS has continued to push the boundaries of what cloud computing can deliver, building custom silicon for general-purpose and AI workloads and expanding EC2 into new form factors and deployment models that our customers in 2006 could not have imagined.

The 20 years in brief
In his 15th anniversary post, Jeff chose important milestones of EC2 that established the foundational building blocks that customers still rely on today. Amazon Elastic Block Store (2008) provided persistent block storage. Elastic Load Balancing, Auto Scaling, and Amazon CloudWatch (2009) made applications scalable and highly available. Amazon Virtual Private Cloud (2009) gave customers logically isolated networks. AWS Nitro System (2017) enabled faster innovation and enhanced security. AWS Graviton processors (2018) were designed for cost-sensitive scale-out workloads.

Over 20 years, EC2 grew from one to over 1,200 instance types to meet customer needs across general-purpose, compute-, memory-, and storage-optimized, accelerated computing, and high-performance computing families. These instances expanded from one AWS Region to 39 Regions globally. AWS also extended EC2 beyond the Region boundary with AWS Outposts (2018) running EC2 instances locally, AWS Local Zones (2019) place globally, and AWS Wavelength (2019) inside global 5G telecommunications carrier networks.

While I hate to play favorites, I want to choose some of my favorite EC2 launches of the past five years:

  • AWS Inferentia for ML inference at scale (2019): We introduced purpose-built ML inference instances (inf1) with AWS Inferentia chips. Amazon EC2 Inf2 instances became generally available in April 2023 for large-scale generative AI inference workloads. Together with the Inferentia family, AWS Trainium instances now give customers a full stack of AWS-designed silicon optimized for every phase of the AI lifecycle both inference and training.
  • EC2 Mac instances (2020): The first Mac instances (mac1) were built on Apple Mac mini with Intel Core i7 (Coffee Lake) on the AWS Nitro System. Mac M1 (mac2) instances launched in July 2022 as the first Arm-based macOS instances on EC2. M2 Pro Mac instances followed in 2023, M4 and M4 Pro Mac instances in 2025, M3 Ultra Mac instances and M4 Max Mac instances in 2026, giving Apple developers a complete range of cloud-based build and test environments for macOS, iOS, iPadOS, tvOS, watchOS, and visionOS apps.
  • AWS Trainium for full-stack AI workloads at scale (2021): In November 2021, we previewed Trn1 instances with AWS Trainium accelerators optimized for high-performance deep learning training. In December 2024, Trn2 instances powered by AWS Trainium2 launched, with Trn2 UltraServers linking 64 Trainium2 accelerators via NeuronLink for training trillion-parameter foundation models. At AWS re:Invent 2025, Trn3 UltraServers powered by AWS Trainium3 deliver the best token economics for next-generation agentic, reasoning, and video generation applications. A single Trn3 UltraServer interconnects up to 144 Trainium3 chips to train and serve the largest frontier models. Now, AWS Trainium3 delivers the leading price-performance for high-performance AI training and inference at scale.
  • EC2 Capacity Blocks for ML (2023): This new EC2 usage model further democratizes ML democratizes ML by making it easy to access GPU instances to train and deploy ML and generative AI models. You reserve the GPU capacity you need (initially P5 instances) for a future date and only for the duration you require. In November 2024, EC2 Capacity Blocks for ML added supported for provisioning in a matter of minutes and extending up to six months. Now, EC2 Capacity Blocks for ML supports P6-B300, P6-B200, P5e, P5en, P4d, P4de, Trn1, Trn2, and Trn3 instances in addition to P5.
  • AWS Graviton5 (2025): Building on eight years of Graviton innovation since 2018, we previewed Graviton5 chips in AWS re:Invent 2025 and launched M9g and M9gd instances powered by Graviton5 and built on the sixth-generation AWS Nitro System. C9g and C9gd followed in June 2026. Now, Graviton5 features 192 cores, a 5x larger cache, and up to 33% lower inter-core latency, making it well suited for the growing demands of agentic AI workloads such as real-time reasoning, code generation, and multi-step task orchestration that require continuous, high-throughput CPU compute at scale.
  • AWS Nitro Isolation Engine (2026): Customers wanted to see, not just hear from us, proof of workload isolation in the Nitro Hypervisor. The Nitro Isolation Engine is a purpose-built component inside the Nitro Hypervisor, harnessing formal verification to provide mathematical assurance that customer workloads are isolated from each other and AWS operators, pioneering a new standard for mathematically proven cloud security. This feature is also based on the sixth-generation AWS Nitro System which has continued to evolve since our introduction in 2017.

The foundation underneath it all
Despite two decades of innovation, the fundamental value proposition of Amazon EC2 has not changed. Customers use it to get secure, resizable compute capacity in minutes, pay only for what they consume, and scale on demand without long-term commitments. That same flexibility now extends to AI workloads at a scale no one anticipated in 2006.

EC2 remains the foundational compute layer of AWS. Amazon ECS, Amazon EKS, AWS Lambda, AWS Fargate, AWS Batch, Amazon EMR, Amazon SageMaker AI, and Amazon Bedrock ultimately run on EC2 capacity. Every architectural pattern customers have built over the past twenty years, from simple web servers to trillion-parameter foundation model training clusters, starts with a decision to launch an instance.

We made strong foundational decisions in 2006, and we left room for the service to grow. Twenty years later, that strategy of creating services that are minimal-yet-useful, launching quickly, and iterating rapidly in response to your feedback continues to guide how we build. The next twenty years of cloud computing will demand capabilities we have not yet imagined. Amazon EC2 will continue to be the foundation where your workloads run.

To learn more about Amazon EC2, visit the Amazon EC2 product page or check out what’s new with EC2.

Channy

Why Your GPU Is Sitting Idle: The Data Pipeline Problem No One Talks About

Post Syndicated from Maddie Presland original https://www.backblaze.com/blog/why-your-gpu-is-sitting-idle-the-data-pipeline-problem-no-one-talks-about/

A decorative image showing cloud storage and AI icons.

At NVIDIA’s GTC conference, Adobe’s CTO Ely Greenfield walked the audience through the company’s three-year journey building frontier generative AI models from scratch. And the AI training data pipeline required to keep thousands of GPUs productive.

Along the way, he showed a profiler readout from their early training runs—a visualization of exactly how much time each GPU was spending on actual computation versus sitting idle. It revealed that roughly two-thirds of GPU time was spent simply waiting for data. And that idle time had a price tag. 

“If we were putting a million dollars into training,” he told the room, “that was $600,000 we were burning away on GPUs sitting and doing nothing.” In other words, roughly sixty cents of every dollar spent on GPU compute was being wasted.

Greenfield’s team traced the waste to two culprits, neither of which is unique to Adobe: whether data can reach your GPUs fast enough, and whether the work it represents is distributed evenly once it arrives.

One culprit is a storage problem. The other isn’t, but you won’t be able to fix it until you’ve solved the first one.

Culprit #1: Storage and retrieval speed

Adobe’s training data lived in petabytes of distributed cloud storage and had to be shipped out to thousands of GPUs constantly, over standard Ethernet. The dataset itself was enormous and varied—images and video, low-res and high-res, simple formats and expensive codecs, all moving at once. Standard networking, which was built for retrieving individual files on request, wasn’t designed for that kind of sustained, parallel, petabyte-scale movement. It became a massive bottleneck.

Checkpointing compounded the problem. As a safeguard, the training run would periodically write a complete copy of the model back to storage. Most of those checkpoints were never needed again. But writing and reading them still consumed real GPU time, which cost a lot of money whether they were used or not. 

How Adobe fixed it 

Adobe’s solution had two parts. 

First, they replaced standard Ethernet with a high-performance networking fabric designed for the petabyte-scale traffic distributed AI training generates. Now, data could finally move at the pace their GPUs needed. 

Second, they changed how they saved checkpoints. Instead of writing one giant file containing the entire model, they began breaking the model into smaller pieces and saving pipeline fragments to many places at once. Saving and loading checkpoints now takes significantly less time than it used to. If a checkpoint ever needs to be reassembled, that’s slightly slower, but it’s a rare event. The savings on every other write happen continuously, across thousands of GPUs, around the clock.

The underlying lesson is that for model training, parallel access and sustained high throughput are baseline requirements. Training data has to be immediately accessible at the pace your GPUs consume it, not tucked away in a storage tier that takes minutes or hours to retrieve.

Culprit #2: The data loader problem

A balanced data loader can only do its job if the data it’s balancing is actually available the moment it’s needed. That makes fast storage a precondition for everything else in the pipeline. But even after fixing the storage and retrieval speed culprit, Adobe still had a problem: the way training data was divided across the cluster meant some GPUs were doing far more work than others.

Their pipeline used a straightforward data-parallelism approach: slice the training data into equal-sized chunks and assign one chunk to each GPU in the cluster. But equal-sized chunks weren’t equal work. Some GPUs got simple, low-resolution assets that processed in seconds; others got large, complex files that took minutes. The fast GPUs finished early and sat idle waiting for the slow ones to catch up. And then all of them waited again while their results were merged into one updated model before the next round could begin. 

How Adobe fixed it

Adobe stopped treating all data as equivalent. They custom-built a balanced data loader that understood the processing cost of each asset and distributed work so every GPU finished at roughly the same time. Then they restructured how computation was divided across the cluster to make the merging step dramatically cheaper.

A perfectly balanced data loader still idles if the data it’s waiting on hasn’t arrived yet. This is why storage can’t be an afterthought. The loader optimizes what happens once data is there, but storage determines whether it’s there at all. 

And building something like Adobe’s balanced data loader takes real engineering investment, including time spent profiling workloads, testing distribution strategies, and tuning until every GPU finishes at roughly the same pace. Teams still fighting storage bottlenecks rarely get to that work. Their engineers are busy figuring out why GPUs are idle in the first place, not optimizing how work gets distributed once data arrives. Removing the storage bottleneck frees up the engineering time needed to tackle the data loader problem properly.

From 40 to 80 cents

After addressing both culprits, Adobe’s GPUs ran at roughly 80% utilization. After accounting for the coordination overhead inherent to running thousands of machines together, 80% is close to the practical ceiling. That’s the difference between a GPU cluster that’s mostly waiting and one that’s mostly working.

The Storage Side of the Solution

Backblaze B2 Overdrive addresses the storage side of exactly the problem Adobe ran into. B2 is always-hot object storage, with no tiering, no retrieval delays, and no waiting for data to be promoted from a cold tier before training can begin. Training data stays immediately accessible whether it was written an hour ago or six months ago.

But availability alone isn’t enough if data can’t move fast enough to keep up with the cluster. B2 Overdrive adds the throughput layer: at up to 1Tbps, it’s designed to keep petabyte-scale GPU clusters fed continuously, not just handle occasional bursts of traffic. Both layers are S3-compatible, so they drop into existing PyTorch or TensorFlow pipelines without a rewrite. And because egress is free, moving training data between storage and GPU compute—across regions or providers—doesn’t add a cost penalty on top of a performance one.

Checkpointing benefits from the same foundation. Adobe’s solution of breaking the model into fragments and saving them to many places at once only works if the underlying storage is fast and parallel enough to make it pay off. Always-hot, high-throughput storage makes checkpoint writes and recoveries faster across the board, so they cost less GPU time whether they happen rarely or often.

Backblaze can’t write your data loader for you, but it can make sure that once you’ve built one, it isn’t waiting on storage to do its job.

Ready to remove the storage bottleneck from your training pipeline? Learn more about Backblaze B2 Overdrive.

The post Why Your GPU Is Sitting Idle: The Data Pipeline Problem No One Talks About appeared first on Backblaze Blog | Cloud Storage & Cloud Backup

[$] Old-school calendaring at the command line with Remind

Post Syndicated from jzb original https://lwn.net/Articles/1090376/

Remind is a
command-line calendar and alarm program, with an optional Tk-based graphical
interface, for Linux and Unix-like operating systems. It has its own scripting
language that allows users to create reminders that are difficult (if not
impossible) to specify in other calendaring programs. It is wholly unsuitable for use in
corporate environments that require calendar sharing and exchanging meeting
invitations; however, it may be precisely the calendaring tool for users who
prefer the command line and fast, flexible tools that can help keep track of
messy schedules.

Озеленяването при нови строежи – планове, документи и как се крият

Post Syndicated from Боян Юруков original https://yurukov.net/blog/2026/ozelenyavane-1/

Всички знаем, че в строителството има нередности. Не са нужни дори пет минути в който и да е град в България без да се сблъскаш с пресни примери разпознаващи се с просто око. Всички знаем, че има изисквания за озеленяване при строежи, както и знаем, че повсеместно не се спазват. На места липсва съвсем озеленяване, а на повечето от останалите е бутафорно или вече заменено с паркоместа и кафенета.

Озеленяването не е просто желание или за красота. Има критична роля в чистотата на въздуха, намаляване на шумовото замърсяване, защитата от наводнения и биоразнообразието в града, което от своя страна помага за намаление на вредители. Специално за наводненията видяхме ярък пример наскоро, където наистина сериозно количество дъжд веднага потече по улиците вместо да се задържи от огромното количество сгради построени наскоро. Запечатването на повърхностния слой е сериозен проблеми прави озеленяването все по-важно.

Да започнем от началото

Писал съм и преди за озеленяването, изискванията и защо не се спазва. Описах и колко е мъчно да се накара районна администация да направи проверка и да защити дори няколко квадрата трева и едно дърво. Писах какво обществото има предвид когато говори за презастрояване и че зелените площи са съществен компонент от това. Последната статия, впрочем, беше подбудена от среща с тогавашния в.и. главен архитект. Това той сподели личното си мнение, че изискванията за озеленяване били неадекватни и затова той и други не следели за тях много-много. Възможно е и да има връзка с оценката, която други негови колеги дадоха, че при строго спазване на дори старите изисквания интензивността на новото застрояване на София би намаляла с 20 до 30%, а доста сгради построени в последните пет години биха загубили я паркоместата си, я цял вход или крило. Всъщност, изискванията за озеленяването са дори твърде занижени и дават предпоставки за злоупотреби.

От доста време се опитвам да разбера този проблем от нормативна, административна, практическа и дори корупционна гледна точка. Затова съм пускал десетки искания за достъп до обществена информация до различни институции. В повечето случаи получавам входящи номера или части от документи, които хем нарочно не ми дават никаква информация, хем номинално се водят за отговор. Затова подобрявах исканията ми по ЗДОИ и питах отново. В други случаи получавах отговор, че такава информация или документи липсват, което само по себе си е полезно, защото знам, че лъжат и ми помага в следващото искане. В трети случаи се налагаше да изчаквам шест месеца да пусна ново искане с надеждата, че ще забравят с какви извинения са отказали преди или какво са скрили с справките. Последното работи учудващо добре.

След четири години игра на котка и мишка се фокусирах върху четири сгради като примери за подобни практики. Не защото са специални, а просто защото често минавам покрай тях и виждам какво се случва. Отделно са добри примери за наглост в контраст със заявките за плувнали в зеленина и устойчивост сгради. В крайна сметка едва тази година получих достъп до документите на проектите им и мога да говоря с числа и конкретни примери за нарушенията.

Ще ги събера във втората част на тази статия. Първо трябва да обсъдим защо беше толкова труден процесът. Подобно на промените в имотния регистър ще покаже отново лобистки текстове в закона и нарочна липса на прозрачност целяща опазването на именно такива схеми. Надявам се, че ще внесе и повече разбиране какво виждате в снимките в следващата част.

Преди това обаче важно уточнение: знам, че далеч не всички термини, които използвам са точни и обясненията изчерпателни. Целта ми тук е да внеса яснота на разговорен език, а не да цитирам наредби и лекции по архитектура. Ще се радвам на всяко допълнение и корекция и ще го обсъдим в коментарите.

Какво точно търсех?

Обичам да казвам, че отворените данни ни помагат да задаваме по-добри въпроси. Същото може да се каже и тук. Всички гледаме в разрешението за строеж, а то е само последната фаза от дълъг процес понякога продължаващ над десетилетие. Преди него има виза за проектиране, често промяна на ПУП, серия от градоустройствени заповеди, обсъждания в специализирани комисии и всичко това на база инвестиционни проекти. Нерядко има и промени и заповеди след разрешението за строеж като намаляване броя на паркоместата, промени в разпределението, а в един случай, на който попаднах тия дни – цялостна промяна на предназначението, конструкцията, мястото и всичко по сградата независимо, че оригиналното разрешение вече няма правно действие.

Всички тези стъпки носят риск за нарушения, пропуски, умишлено забавяне и корупция, особено предвид, че процесът упорито не се дигитализира, за да е лесно проследим. Част от документите като искания за промени на ПУП, издадените визи и заповеди са публични. Строителните книжа заедно с всички тези административни документи следва да се качват в регистъра по ЗУТ, който обаче все още не е видял бял свят – мина обсъждане и от началото на мандата чака само подписа на Шишков.

Час от тези строителни книжа и инвестиционното намерение са и плановете за озеленяване. Това са няколко скици, които в детайли показват къде ще са зелените площи, колко ще е дълбок почвения слой, къде и колко дървета и храсти ще има и какъв процент от изискванията ще бъде постигнат.

Защо е важно?

Плановете за озеленяване са неразделна част от проекта и на тяхна база се издава разрешение за строеж. В последствие приемателната комисия сравнява тези и други документи с това, което на практика виждат на място и решават дали да издадат акт 16. Това е поне на теория. Конкретно за София и озеленяването подписът е на служител на районната община. По принцип трябва да е ландшафт архитект, но в практиката често самият районен кмет или трето лице, на което е делегирал, се подписва. Длъжни са обаче и потвърждават с подписва си, че довършеният имот отговаря на описаното в плана.

В статията си за бутафорното озеленяване описах някои от изискванията в наредбата за зелената система в София. Някои ключови аспекти са:

  • Почвеният слой трябва да е минимум 120 см. дълбочина при дърветата, а ако са в кашпа – поне 1 м3 обем на почвата
  • За храсти почвения слой трябва да е поне 60 см, а за трева – поне 40 см.
  • При строежи над 2 декара – задължителен резервоар за дъждовна вода и използването ѝ за поливане на зелените площи
  • Отстояние на дървета от сгради – 1.5 м. или и 3 м., ако дървото стига 5 метра.
  • Отстояние от бордюри – 70 см.
  • Отстояние от пътни платна – 2 м.
  • Отстояние от откоси и тераси – 1м.
  • Отстояние от стълбове – 4м.
Пример за липса на отстояние и достатъчно открита почва – само 30 см. от бордюра и под 50 см. страна. Дървото никога няма да се развие и има нужда от напояване, за да е умре съвсем.

Тези изисквания не значат, например, че не може да се засади дърво точно до бордюра на пътя без достатъчно почва. Безсмислено би било, защото ще изсъхне, но също така не следва да се брои при смятането на озеленяването. Това е важно, защото освен дялът земя като зелена площ има изискване и какъв процент от нея е във висока дървесна растителност. За широколистни дървета с височина между 3 и 5 м. на 12 годишна възраст броят 12 кв.м. площ за това изчисление. Над 5 м. – 20. При иглолистните е съответно 5 и 7 кв.м.

Ключов аспект тук е, че доста от тези изисквания влизат в сила през юли 2023, т.е. за разрешения за строеж издадени преди това важат старите разпоредби за отстояния и почвен слой. През 2019 г., например, изискват да има почвен слой от поне 60 см., а когато е не по-малко от 30 см. площта важи с индекс 50% към общия коефициент на озеленяване. Отстоянията от бордюри, стени и тераси обаче са били същите.

Трябва да се разбере също, че плановете не са абсолютни. Това е просто илюстрация как си представят, че ще стане озеленяването, но на терен може да се окаже друго. Аналогично скиците къде ще бъдат сградите почти никога не отговарят в мащаб – отстоянията не са същите, височината не е тази и прочие. Затова ключовия момент е при приемане на обект на груб строеж да не се гледа по документи, а да се измерват реалната височина и отстояния. Аналогично при премателната комисия не се гледа дали точно в това пространство има дърво, а колко са на брой, дали са същия вид (има значение колко големи ще станат) и дали почвения слой и отстоянията са правилните.

Защо не виждаме тези планове?

Най-честата причина за отказ при поискване на тези скици, е че се засягали правата на трети лица. Това са инвеститорските компании и архитектите. Те от своя страна изрично са искали да не получавам плановете за озеленяване. Доколкото не изясняват никога в отговорите си точно какви права биха били засегнати и по какъв начин, може само да гадаем, че всъщност замесените се притесняват от разкриване на нарушения.

Подобни откази са доста спорни, защото става въпрос за документи неразделна част от строителните книжа. ЗУТ изрично посочва, че поне част от тях трябва да се публикуват в регистъра (когато види бял свят), т.е. са публични по дефиниция. Отделно свободно може да се използват в рамките на административен процес и производство, част от който спокойно да изиска публичността им какъвто вече в случая с исканията за промяна на ПУП и други аналогични скици от същите инвестиционни проекти. Няма основателна причина да се прилагат тези правила за публичност за едни чертежи, а не за други. Най-вече законът за достъп до обществена информация предвижда, че дори да има съмнение за засегнати права, при надделяващ обществен интерес може да се предостави достъп при минимизиране на предполагаемата щета. Т.е. могат да редактират имена и елементи, които ги притесняват и пак да предоставят скиците.

Експлоатиране на авторски права за прикриване

Попаднах и на друг аргумент – авторски права. Законът третира архитектурните планове и всичко свързано с градоустройството по любопитен начин. През последните 10 години са вкарани множество стратегически промени, които силно ограничават разпространението на архитектурни планове освен, ако изрично не се изисква от друг закон – например както ЗУТ дава възможност това да се определели с наредба. Тук ограничението сериозно надвишава опазване на имуществените и неимуществените права. Забранили са притежанието на копия от архитектурни планове дори за лична употреба без търговска цел.

Както се сещате, в тази хипотеза попадат и плановете за озеленяване като част от архитектурния проект. Те съдържат лични данни като кой ги е изготвил и подписал. Тези данни стандартно се заличават по GDPR. Това от своя страна създава интересна хипотеза – ако все пак получа тези скици със скрити имена и подписи, мога ли да спазя Закона за авторските права и да опиша както съм задължен източникът и авторът му, когато не го знам.

Голяма част от възможностите за т.е. „безкористно използване“ на авторски материали като журналистически материали, критика, сатира и прочие нарочно и изрично изключват тези свързани с архитектурата и градоустройството. Трудно ми е да определя тези точки като лобистки опити да се скрие ключова за обществото информация злоупотребявайки с принципите на авторското право. Аналогичен случай имах преди години, когато направих карта на разрешителните за сеч в България и няколко лесовъда ме заплашиха със съд, тъй като имали авторски права върху актовете и никой нямал право да ги използва без тяхно съгласие.

Административни документи в квантова суперпозиция

Всичко описаното до тук не важи за официалните актове и документи издадени от различните институции като част от административния процес. По принцип всички те са по подразбиране публични и специално тези по ЗУТ трябва вече да се публикуват от общини, ДНСК и министерства. Някои, които засягат вътрешни процедури и актове стават публични и обект на поискване по ЗДОИ две години след издаването им. Дори за тях никой държавен служител няма право да откаже предоставяне на вътрешните номера на такива документи, както и да признае съществуването им, освен ако не са засекретени. Тук няма значение дали същите се отнасят или засягат трети лица.

Това значи, че документ като протокола от държавната приемателна комисия, разрешението за ползване и всички документи издадени от общината и ДНСК следва да са публични и то най-късно две години след датата. Отделно копия от протоколите се дават на подписалите ги страни и членове на комисията.

Именно заради тези важни аспекти обжалвах два отказа на ДНСК да ми предоставят точно такива актове издадени от тях, както и протоколи по приемане. В тях се съдържа ключова информация какво са видели, какви забележки са имали и кой се е подписал, че всичко е наред. Делата са насрочени в рамките на следващия месец и се надявам до края на годината да имаме резултат.

Компромисният път

Авторските права в България са доста строги и подобни тенденциозни ограничения специално за архитектурата съвсем естествено притесняват всеки. Дори когато няма зла умисъл и опит за прикриване, това механично води до отказ от действие „за всеки случай“. Както споменах, ЗДОИ дава възможност да се търсят алтернативи при надделяващ обществен интерес. В случая несъмнено има такъв предвид изложеното в началото на този текст. Затова на няколко поредни искания за достъп до обществена информация от Столична община ми дадоха възможност да видя все пак плановете за озеленяване. Условието бе нямам дигитално копие, да ги чета само физически на място и да не правя снимки.

Това ми позволи да видя истинските параметри, разпределени на озеленяването, колко дървета, вертикално озеленяване и отстояния са обещали, както и как тези планове са се променяли с времето. Това е важно, защото често не знаем какво търсим докато не го видим. Не стига просто да се искат няколко числа от документите. Пробвал съм и дори такива са ми били отказвани с аргумента, че не им била работа да ровят по дигиталните архиви. Тази седмица получих такъв отказ по свързана тема, за която ще пиша също скоро. Нямах възможност да видя също протоколите от приемателната комисия, защото те са били при ДНСК, а районната администрация, където са разглежданите сгради и е била страна, отрича да има копие.

Затова в следващата част на тази статия няма да мога да ви покажа самите планове. Първо, защото години наред получавах откази от вече бивши главни архитекти. Второ заради ограниченията в закона за авторското право целящи именно избягване на подобна прозрачност. Трето, защото все още нямаме публичен регистър по ЗУТ, който да се надгради да включва документи защитаващи обществения интерес заедно с други подобни като транспортния анализ и становищата за пожарна безопасност.

Този начин на предоставяне е силно ограничаващ, най-малкото защото мога само да преразкажа какво съм видял на база няколко числа в наредбата. Ландшафт архитект би хванал с един погледа всички нередности, ако документите бяха публични. Все пак, Столична община направи една крачка повече от всички други институции и предишни администрации и ми даде поне някакъв достъп. Следващата стъпка би била довършването на регистъра по ЗУТ от Шишков и изискване да се качват този тип документи там.

В близките дни ще публикувам следващата част от темата, че която ще покажа с конкретни примери какво видях, какво липсва и защо е толкова нужна прозрачност и безкомпромисност.

Security updates for Tuesday

Post Syndicated from jzb original https://lwn.net/Articles/1090524/

Security updates have been issued by AlmaLinux (cups-filters, gstreamer1-plugins-base, gstreamer1-plugins-good, kernel, mrtg, NetworkManager, nginx, nginx:1.24, nodejs24, perl-Date-Manip, python-pyasn1, python-urwid, python3.12, python3.14, and qemu-kvm), Debian (erlang, thunderbird, webkit2gtk, and zfs-linux), Fedora (calibre, chromium, freeipa, java-21-openjdk, java-21-openjdk-portable, java-25-openjdk, java-latest-openjdk, jfrog-cli, kernel, libxls, nextcloud, perl-URI, and samba), Gentoo (Incus), Mageia (kernel and kernel-linus), Oracle (ansible-core, cups-filters, curl, firefox, kernel, libcupsfilters, libreoffice, mrtg, NetworkManager, perl-Date-Manip, php:8.2, php:8.3, python-urwid, python3.14, qemu-kvm, and sqlite), Red Hat (assertj-core, httpd, and osbuild-composer), SUSE (buildah, comfyui, dracut, erlang, erlang27, grafana, kernel, libssh2_org, openvswitch, perl-Dancer2-Plugin-Auth-Extensible, postgresql17, python-cryptography, python-sqlparse, python311, python313-hpack, rpm, suseconnect-ng, thunderbird, and vim), and Ubuntu (async-http-client, curl, and ffmpeg).

Black Hat State of Security Vendors

Post Syndicated from Bruce Schneier original https://www.schneier.com/blog/archives/2026/08/black-hat-state-of-security-vendors.html

Andy Ellis has a roundup of the security vendors at Black Hat this year.

Key Takeaways: We have entered into an AI world. While nearly half of booths didn’t directly mention AI or agents in their taglines, the effects of AI are everywhere. Multiple spaces (Identity, SaaS, AppSec, Data) have almost every vendor leading with AI; existing unsolved problem areas just got worse.

At the same time, there’s a clear trichotomy in the market: tools that tell you how bad things are; tools that stop adversaries, and tools that prevent problems from occurring. While you’d suspect that the tools that fix things would dominate, the tools that merely tell you how bad things are seem to be frustratingly plentiful.

Supporting a technical AI-focused qualification for young people in England

Post Syndicated from Diane Dowling original https://www.raspberrypi.org/blog/support-for-ai-focused-qualification-young-people-england/

Young people see and use artificial intelligence (AI) systems frequently, and hear a lot about how they are applied across many areas of life. That makes them understandably eager to find out more about how AI works. For some, this interest extends to wanting to experiment with the technology to investigate their own questions and build something themselves. How can schools support learners who want to take their interest further and gain recognition for the knowledge and skills they develop along the way?

Computer science students at a desktop computer in a classroom.

For learners in the UK, there is currently no established qualification that develops technical AI skills. This can make it difficult to create space in the curriculum for dedicated AI learning, especially for learners from age 14, for whom preparation for high-stakes exams is prioritised over non-examined learning.

We have been working over the past year to develop a research-informed curriculum framework for data science, which lays out many of the foundational skills and knowledge needed to understand and develop AI models. Alongside this, we have been making the case in England for a Level 3 qualification in data science and AI. However, establishing a new qualification can take years.

In the meantime, we are starting to support students in England who choose an established route for extending their skills and getting that effort recognised: the Extended Project Qualification (EPQ).

An established qualification with room to explore AI

The EPQ is a well-established qualification for 16- to 19-year-olds offered in England, Northern Ireland, and Wales by several exam boards (including AQA, Pearson Edexcel, OCR, Eduqas/WJEC, and City & Guilds). It is completed by around 1 in 10 learners at this stage of education.

Learners pursuing an EPQ undertake an extended, self-directed investigation or project. They choose an area of interest where they want to develop their own project and take responsibility for planning, completing, and evaluating the project work.

A teenager learning computer science.

Its self-directed nature makes the EPQ particularly well suited to learners who want to explore emerging areas, such as artificial intelligence and data science. For students fascinated by AI, an EPQ offers an opportunity to do more than simply learn about the technology. With the right foundations and support, they can investigate a question of their own, build a machine learning model, and use that work to gain a recognised qualification.

We are now working to offer those foundations through a new introductory ‘Data Science and AI’ course that will enable learners to build a machine learning model as their EPQ artefact.

Building the foundations for an independent AI investigation

Creating a machine learning model involves much more than choosing a statistical technique and using easily available training data. It involves understanding the problem being addressed, deciding what data to use, exploring and preparing that data, building and evaluating a model, and interpreting the model’s outputs.

Learners who want to do this work for their EPQ need not only technical knowledge and skills, but also skills for critical thinking, research, decision-making, evaluation, and reflection — skills that are valuable for any further study and future career.

To help learners develop these foundations, we are creating an introductory ‘Data Science and AI’ course, based on our curriculum framework for data science and designed to be completed before learners begin their AI-focused EPQ.

The course will guide students through a structured investigation based on a recognised data science lifecycle. By working through it, students will:

  • Learn about each stage of the data science lifecycle
  • Develop the technical understanding they need to make independent decisions during an AI investigation 
  • Gain practical experience of working with data and building a machine learning model
  • Learn about key concepts behind the techniques they use, through theoretical content that supports them to understand not just what to do, but why

The course will consist of 10 units, each involving 2 to 3 hours of independent study for learners. We are designing the activities with a no-code approach to allow students to focus on understanding data science concepts and processes.

Supporting independent project work

Importantly, the work learners produce during this course won’t be the EPQ investigation itself. Working through ‘Data Science and AI’ will give learners the knowledge and skills they need to undertake an AI-focused project with greater independence and confidence.

A young person in a university computing classroom.

After they have completed the course, when they begin their EPQ project pursuing a question that interests them, students can apply their new skills and knowledge to make decisions about their AI investigation and document their progress.

We are looking forward to supporting learners in England to turn their interest in one of today’s most significant technologies into the opportunity to develop independent research and project skills through a recognised qualification.

What’s next?

The ‘Data Science and AI’ course will be available to selected schools in England from September 2026, where students and teachers will test the materials and we will learn from their feedback to improve the course materials. We plan to make the course freely available in 2027. 

The post Supporting a technical AI-focused qualification for young people in England appeared first on Raspberry Pi Foundation.

Supporting an AI-focused qualification for young people in England

Post Syndicated from Diane Dowling original https://www.raspberrypi.org/blog/support-for-ai-focused-qualification-young-people-england/

Young people see and use artificial intelligence (AI) systems frequently, and hear a lot about how they are applied across many areas of life. That makes them understandably eager to find out more about how AI works. For some, this interest extends to wanting to experiment with the technology to investigate their own questions and build something themselves. How can schools support students who want to take their interest further and gain recognition for the knowledge and skills they develop along the way?

Computer science students at a desktop computer in a classroom.

Currently, there is no widely available subject qualification in AI for students in the UK. This can make it difficult to create space in the curriculum for dedicated AI learning, especially for students from age 14, for whom preparation for high-stakes exams is prioritised over non-examined learning.

We have been working over the past year to develop a research-informed curriculum framework for data science, which lays out many of the foundational skills and knowledge needed to understand and develop AI models. Alongside this, we have been making the case in England for a Level 3 qualification in data science and AI. However, establishing a new qualification can take years.

In the meantime, we are starting to support students in England who choose an established route for extending their skills and getting that effort recognised: the Extended Project Qualification (EPQ).

An established qualification with room to explore AI

The EPQ is a well-established qualification for 16- to 19-year-olds offered in England, Northern Ireland, and Wales by several exam boards (including AQA, Pearson Edexcel, OCR, Eduqas/WJEC, and City & Guilds). It is completed by around 1 in 10 learners at this stage of education.

Learners pursuing an EPQ undertake an extended, self-directed investigation or project. They choose an area of interest where they want to develop their own project and take responsibility for planning, completing, and evaluating the project work.

A teenager learning computer science.

Its self-directed nature makes the EPQ particularly well suited to learners who want to explore emerging areas, such as artificial intelligence and data science. For students fascinated by AI, an EPQ offers an opportunity to do more than simply learn about the technology. With the right foundations and support, they can investigate a question of their own, build a machine learning model, and use that work to gain a recognised qualification.

We are now working to offer those foundations through a new introductory ‘Data Science and AI’ course that will enable learners to build a machine learning (ML) model as their EPQ artefact.

Building the foundations for an independent AI investigation

Creating an ML model involves much more than choosing a statistical technique and using easily available training data. It involves understanding the problem being addressed, deciding what data to use, exploring and preparing that data, building and evaluating a model, and interpreting the model’s outputs.

Learners who want to do this work for their EPQ need not only technical knowledge and skills, but also skills for critical thinking, research, decision-making, evaluation, and reflection — skills that are valuable for any further study and future career.

To help learners develop these foundations, we are creating an introductory ‘Data Science and AI’ course, based on our curriculum framework for data science and designed to be completed before learners begin their AI-focused EPQ.

The course will guide students through a structured investigation based on a recognised data science lifecycle. By working through it, students will:

  • Learn about each stage of the data science lifecycle
  • Develop the technical understanding they need to make independent decisions during an AI investigation 
  • Gain practical experience of working with data and building a machine learning model
  • Learn about key concepts behind the techniques they use, through theoretical content that supports them to understand not just what to do, but why

The course will consist of 10 units, each involving 2 to 3 hours of independent study for learners. We are designing the activities with a no-code approach to allow students to focus on understanding data science concepts and processes.

Supporting independent project work

Importantly, the work learners produce during this course won’t be the EPQ investigation itself. Working through ‘Data Science and AI’ will give learners the knowledge and skills they need to undertake an AI-focused project with greater independence and confidence.

A young person in a university computing classroom.

After they have completed the course, when they begin their EPQ project pursuing a question that interests them, students can apply their new skills and knowledge to make decisions about their AI investigation and document their progress.

We are looking forward to supporting learners in England to turn their interest in one of today’s most significant technologies into the opportunity to develop independent research and project skills through a recognised qualification.

What’s next?

The ‘Data Science and AI’ course will be available to selected schools in England from September 2026, where students and teachers will test the materials and we will learn from their feedback to improve the course materials. We plan to make the course freely available in 2027. 

The post Supporting an AI-focused qualification for young people in England appeared first on Raspberry Pi Foundation.

Intel Crescent Island 160GB to 480GB LPDDR5X AI GPU at Hot Chips 2026

Post Syndicated from Patrick Kennedy original https://www.servethehome.com/intel-crescent-island-160gb-to-480gb-lpddr5x-ai-gpu-at-hot-chips-2026/

We learned more about the new 160GB to 480GB LPDDR5X Intel Crescent Island GPU focused on memory capacity at Hot Chips 2026

The post Intel Crescent Island 160GB to 480GB LPDDR5X AI GPU at Hot Chips 2026 appeared first on ServeTheHome.

The collective thoughts of the interwebz