15-Line JavaScript Formula Estimates VRAM Needed to Run LLMs Locally

·1 views

A technical guide published by EU hardware retailer Mineshop.eu explains how to estimate GPU VRAM requirements for running large language models locally using a compact JavaScript function. The calculation accounts for three key components: model weights (parameter count multiplied by bits-per-parameter), the KV cache (which stores key-value vectors per layer and token), and a runtime headroom reserve of around 2 GiB. For example, running a 4-bit quantised 8B-parameter model at an 8,192-token context requires approximately 7.28 GiB of VRAM in total. Larger models and longer contexts scale the requirement significantly — a 70B model at 32K context needs nearly 50 GiB. The post includes worked examples and test assertions to validate the formula's accuracy across different model sizes and cache configurations.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Developer builds 175+ browser-based tools, shares hard lessons on Google indexing

A developer built PlainToolbox, a free collection of over 175 browser-only utilities including calculators, converters, and text tools, all running without a backend or user sign-up. The project uses a shared data-driven engine in Next.js, where each tool is defined as a single object that automatically generates its page, schema, sitemap entry, and FAQ content. Despite initial crawling, Google dropped nearly all tool pages, flagging them as 'Crawled, currently not indexed,' prompting the developer to remove generic tools, add answer-first content blocks, and avoid fake freshness timestamps. The developer noted that on a new domain, technical SEO only establishes eligibility to rank, while backlinks and real-world mentions carry far greater weight. Notably, the site's earliest traceable traffic came from AI assistants rather than traditional search, suggesting niche single-purpose pages may find an audience through that channel first.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Shopify to Support Multiple Barcodes Per Variant, But API Still Returns One

Shopify is preparing to expand its barcode field to hold up to 20 values per product variant, a significant change from the current single-string limitation. The update matters because GS1 standards require distinct GTINs for each packaging level — individual units use EAN-13 barcodes, while cases and cartons require separate ITF-14 identifiers. A purchase order for 240 units, for example, correctly generates 240 item labels, 24 inner labels, and 2 carton labels — three different outputs from one input figure. Until the API is updated to return multiple values, developers are storing inner and carton GTINs in Shopify metafields, which hold the data but enforce none of the packaging hierarchy logic. Using a fallback GTIN across packaging levels is flagged as a critical error, since a mislabeled carton scanning as a single unit can cause downstream warehouse mistakes.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Base Bridge Rated 8/10 Risk: Critical Bugs and Governance Flaws Flagged in $3.15B Protocol

A security assessment of Base Bridge, the primary asset-transfer gateway between Ethereum mainnet and the Base L2 rollup, has assigned it a cumulative risk score of 8 out of 10, classifying it as high risk. The bridge holds approximately $3.15 billion in total value locked, making it a high-value target for potential attackers. Researchers identified critical smart contract vulnerabilities including a re-entrancy flaw in the withdrawal function, missing input validation, and improper nonce handling that could enable replay attacks and double-spending. Governance weaknesses were also flagged, notably an upgradeable proxy contract lacking a multi-signature timelock and an admin key controlled by a single externally owned account. The assessment, prepared by a senior DeFi security researcher and dated September 30, 2026, also highlighted economic design gaps and insufficient monitoring as contributing factors to the bridge's overall systemic risk.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Actually Clarify System Design Requirements in Technical Interviews

A software engineering guide published on DEV Community breaks down how to effectively clarify requirements during system design interviews, going beyond the generic advice of simply 'asking for requirements.' The piece uses a URL shortener as a running example to show how targeted questions — such as whether users are authenticated or what the read-to-write traffic ratio is — directly shape architectural decisions. The author distinguishes between functional requirements, which define what users can do, and non-functional requirements, which set performance, availability, and latency constraints. Rather than treating these as abstract checklists, the guide frames each question as a way to uncover consequences that drive design choices, such as separating analytics from the critical redirect path. The core takeaway is that good requirement clarification follows a chain of reasoning: requirement leads to consequence, which leads to architecture.

0 comments Read more at DEV Community