Seventh-generation server hardware at Dropbox: our most efficient and capable architecture yet
Read Full ArticleSummary
Dropbox has unveiled its seventh-generation server hardware, marking a significant evolution in its infrastructure to support its growing product and user base. This new architecture incorporates advanced technologies such as Crush, Dexter, and Sonic for compute, database, and storage workloads, alongside new GPU tiers named Gumby and Godzilla. Key improvements include increased storage bandwidth, enhanced rack power, and a next-gen storage chassis designed to minimize vibration and heat. The design process emphasized collaboration with suppliers and a product-first mindset, ensuring that the hardware meets the unique demands of Dropbox's AI-driven services while maintaining efficiency and scalability.
Key Learnings
- 1The importance of supplier collaboration in hardware development to access emerging technologies and optimize designs for specific workloads.
- 2How the shift to higher-core CPUs necessitated a rethinking of thermal and power management strategies to accommodate increased power demands.
- 3The role of co-designing hardware and software to enhance performance and efficiency, particularly for AI workloads.
- 4The significance of balancing cooling and acoustic requirements in high-density storage systems to maintain performance and reliability.
- 5The strategic decision to integrate GPU capabilities into the architecture to support advanced AI functionalities and improve computational efficiency.
Who Should Read This
Senior Hardware Engineers designing high-performance server architectures for scalable cloud infrastructure
Test Your Knowledge
What trade-offs did Dropbox consider when selecting the new CPU architecture for their seventh-generation servers?
How does the integration of GPUs in the new hardware impact the overall performance of AI workloads?
What specific challenges did Dropbox face regarding thermal management, and how were they addressed in the new design?
In what ways did supplier partnerships influence the design and capabilities of the new server hardware?
How did the shift from dual-socket to single-socket design affect database performance and latency?
What lessons can be learned from Dropbox's approach to balancing power consumption and performance in their hardware architecture?
Topics
More articles about Gpus
Explore Gpus engineering →Expanding our Agentic Inference Cloud: Introducing GPU Droplets Powered by AMD Instinct™ MI350X GPUs
DigitalOcean has announced the launch of GPU Droplets powered by AMD Instinct™ MI350X GPUs, aimed at enhancing the capabilities of their Agentic Inference Cloud. These GPUs, built on the AMD CDNA™ 4...
Hack Week 2025: How these engineers liquid-cooled a GPU server
The article details a project undertaken during Hack Week 2025 at Dropbox, where engineers developed a custom liquid cooling system for GPU servers to address the increasing thermal demands of AI...
Sharks of DigitalOcean: Archana Kamath, Senior Director, IaaS
In this article, Archana Kamath, Senior Director of Compute and Network at DigitalOcean, discusses her experiences and insights regarding the company's approach to cloud computing and innovation. She...
Powered by DigitalOcean Hatch: Why Uxify’s Founders Always Choose DigitalOcean
The article discusses DigitalOcean's Hatch program, which supports startups by providing cloud infrastructure credits, personalized guidance, and technical support. Co-founders Georgi Petrov and...
Helping Startups Build Faster with an AI Startup Ecosystem
DigitalOcean is enhancing its support for AI startups through the launch of the AI Startup Ecosystem, which provides tailored cloud solutions, discounted infrastructure, and access to technical...
More from Dropbox Engineering
View Dropbox engineering blogs →Using LLMs to amplify human labeling and improve Dash search relevance
The article outlines how Dropbox Dash utilizes a retrieval-augmented generation (RAG) approach to enhance search relevance by integrating large language models (LLMs) with human labeling. It explains...
How low-bit inference enables efficient AI
The article discusses the advancements in large machine learning models and the challenges associated with their deployment, particularly focusing on low-bit inference techniques that enhance...
Insights from our executive roundtable on AI and engineering productivity
The article provides insights into Dropbox's approach to enhancing engineering productivity through the adoption of AI tools. It highlights the importance of aligning AI initiatives with business...
Engineering VP Josh Clemm on how we use knowledge graphs, MCP, and DSPy in Dash
In this article, Josh Clemm discusses the technical architecture behind Dropbox Dash, focusing on the integration of knowledge graphs, retrieval methods, and the use of large language models (LLMs)....
Inside the feature store powering real-time AI in Dropbox Dash
The article delves into the implementation of a feature store that powers the AI-driven Dropbox Dash, focusing on how it manages and delivers data signals for effective ranking and retrieval of...