TRANSPARENCY & SPECIFICATION STANDARDS

Hardware Badges & Verification Guide

Every badge on FindBestRig is earned mathematically through automated benchmark formulas, verified retailer telemetry, and physical hardware architecture. We ban subjective marketing fluff like “lowest price ever” or inflated “was/now” discounts.

Section 1 // Hardware Hierarchy

Hardware Tiers & Form Factors

How we categorize complete systems versus standalone upgrade components.

TIER 1 // STARTER (£800–£1,500)

Tier 1: Starter Machines

Budget-conscious prebuilts with 12GB to 24GB of fast memory. Ideal for everyday chat, personal coding assistance, and image generation using 8B to 14B models.

TIER 2 // SWEET SPOT (£1,800–£3,500)

Tier 2: Sweet Spot Workstations

The professional developer bracket featuring 24GB to 64GB memory. Built for multi-file coding agents, autonomous pipelines, and 32B parameter reasoning models (Qwen 2.5 32B, DeepSeek-R1 32B).

TIER 3 // MAXIMUM POWER (£3,200+)

Tier 3: Maximum Power Systems

High-memory workstations with 96GB to 256GB+ unified memory or 32GB CUDA setups. Runs full 70B models with long context windows and 120B+ MoE checkpoints.

Discrete GPU CardSWEET SPOT GPU

Standalone GPU Component Cards

Indicates a discrete graphics card sold on its own (such as an RTX 5080 or RTX 5090) rather than a complete computer. Clearly labeled to prevent mistaking a £700 upgrade card for a turnkey PC.

Section 2 // Hardware Architecture

Memory Architecture & Compute Badges

The physical mechanism your computer uses to store model weights and run parallel tensor calculations.

Discrete GPU

Dedicated Graphics (dGPU)

The machine features a separate graphics card (NVIDIA RTX or AMD Radeon) with its own high-speed GDDR6/GDDR7 video memory. Delivers the fastest token generation speeds and native CUDA software support.

Unified Memory

Unified System Architecture

Found in Apple Silicon (Mac Studio, Mac Mini) and AMD Ryzen AI Max systems. The CPU, GPU, and Neural Engine share a single high-bandwidth memory pool, allowing you to allocate up to 75%–80% of total system RAM to large AI models at low power draw.

Multi-GPU Cluster

Dual-GPU & Rack Configurations

Systems with 2 or more physical graphics cards (e.g. 2x RTX 3090 or dual RTX 5090s) connected over PCIe slots. Splitting model layers across two cards pools 48GB–64GB of fast GDDR VRAM.

~12.4 tok/s (70B)

Measured 70B Generation Speed

The benchmarked or estimated output rate when generating text with a 70-billion parameter model quantized to 4 bits (Q4_K_M). Human reading speed is ~4 to 6 tokens/sec; high-speed autonomous coding agents typically benefit from 15+ tokens/sec.

Section 3 // Sizing & Competitive Value

Model Fit & Value Badges

Calculated automatically from memory capacity, thermal limits, and index-wide price rankings.

70B Q4 Ready

48GB+ Fast Memory Milestone

Awarded only to machines with at least 48GB of accessible fast memory. A 70-billion parameter model compressed to 4-bit quantization requires ~40GB to load weights, plus 4GB–8GB for KV context cache. Machines with this badge run 70B models completely in fast memory without offloading to slow storage.

Fits 32B Unquantized

32GB Fast Memory Milestone

Awarded to 32GB–47GB hardware (e.g. RTX 5090 or Mac Mini M4 Pro 48GB). Confirms that 32B reasoning models (Qwen 2.5 32B, DeepSeek-R1 32B) fit cleanly at Q8 quantization or full precision with ample context buffer.

£29.47/GBBest £/GB in Tier 2

Cost-Per-Gigabyte Metric

Calculated by dividing the lowest verified retailer price by usable fast memory in gigabytes. Rigs that achieve the lowest number within their hardware tier or VRAM bracket earn competitive value honors.

Silent / 35W TDP

Low Power & Thermal Profile

Awarded to hardware pulling 150 watts or less under full local inference. These machines run quietly in home offices and can run 24/7 background agent pipelines with minimal impact on electricity bills.

Editor Pick

Editorial Recommendation

Designates a standout benchmark reference machine chosen by our editorial team for exceptional engineering balance, verified thermal stability, and unmatched value in its class.

Section 4 // Price Tracking & Compliance

Verified Deals & Freshness Badges

Automated proof of substantive price drops derived from append-only historical database tracking.

90-Day Low30-Day Low£200 below RRP

Age-Gated Deal Badges

Deal badges are mathematically restricted based on continuous logging history. A “30-Day Low” badge requires at least 30 days of logged prices before it can trigger; “90-Day Low” requires 90+ days. If an offer does not beat all previous recorded prices, no deal badge is displayed.

Verified 2h ago

Verification Freshness

Every retailer URL is checked periodically by our sync scrapers to confirm stock availability and live pricing. Each badge is backed by a machine-readable ISO-8601 timestamp in the page HTML so search engines and users can verify data freshness.

Section 5 // Wizard Matching

Rig Finder Recommendation Badges

Labels assigned by our 3-Question Rig Finder when matching your workload, budget, and form factor.

Top Recommendation

Top Recommendation

The highest-scoring system that satisfies your exact workload target (e.g. 70B reasoning or daily coding) while remaining comfortably within your chosen budget bracket and form factor.

Strong Alternative

Strong Alternative

The runner-up recommendation that also satisfies your workload, providing a different memory technology (e.g. discrete CUDA vs. unified memory) or slightly different cost-to-performance ratio.

Next-Tier Alternative

Closest Fallback Alternative

Shown when an exact filter combination has limited options (e.g. seeking an Upgradeable Tower under £1,500 for 70B models). Clarifies the nearest high-performing hardware that achieves your goal.

Rent Cloud GPU Power

Not Ready to Buy a Full Rig?

Don't want to spend £3,500+ upfront? Rent the GPU power instead.

Buying a dedicated AI workstation is a serious investment—and if you aren't ready to spend that much right now, you don't have to. Renting cloud GPUs gives you instant access to high-VRAM hardware like RTX 4090s and H100s on demand. Spin up an instance for fine-tuning or running massive 70B+ models, pay strictly for the minutes you use, and avoid the upfront hardware cost, electricity bills, and maintenance.

Vast.ai Global GPU Marketplace

Lowest $/hr
Unbeatable $/hr rates on community-hosted RTX 4090 & 3090 rigs
Deploy Pod →

RunPod Community & Secure Cloud

Community & Secure Cloud
RTX 4090 from $0.34/hr • H100 SXM from $1.99/hr (Billed in USD)
Deploy Pod →

FindBestRig independently benchmarks cloud GPU providers to compare rental economics against physical hardware.