Dell Precision 7560 Review: 16 GB RTX A5000 and 128 GB RAM, Off-Lease
Who is this laptop for?
The Precision 7560 is Dell’s 2021 full-thickness 15.6” mobile workstation, now flowing off corporate leases in serious volume. In its RTX A5000 configuration it is the direct rival to the ThinkPad P15 Gen 2: 16 GB of CUDA VRAM, four RAM slots reaching 128 GB, and cooling built for week-long sustained loads rather than gaming bursts.
🎓 Students (Budget: £850–£1,250)
Almost certainly more machine — and more money — than you need. A Legion 5 Gen 6 covers learning-grade Stable Diffusion and LLM work for £550–£750. The 7560 makes student sense only for thesis-level work: training runs that go overnight, or models that genuinely need 16 GB of VRAM and 64 GB+ of system RAM.
👨💻 ML Engineers & Data Scientists
This is the target buyer. QLoRA fine-tuning of 7B–13B models, FLUX.1 in fp8, GPU-resident 13B inference, and — thanks to the four SO-DIMM slots — CPU+RAM offloading of 70B Q4 models if you fit 64–128 GB. ECC options on both VRAM (standard on the A5000) and system RAM (with Xeon configs) matter if you run multi-hour jobs where a single bit-flip wastes a night.
🏢 Small Teams & Startups
The strongest business case in our 16 GB tier. These are ex-fleet machines: expect Dell service-tag history, refurbisher stock with 12-month warranties and VAT invoices, and boring, repairable, socketed internals. Buying two or three identical off-lease 7560s is a realistic way to equip a small ML team without cloud bills.
What can it actually run?
| Task | Works? | Notes |
|---|---|---|
| GitHub Copilot / Cursor AI | ✅ Yes | API-based, any laptop works |
| Whisper transcription (local) | ✅ Yes | large-v3 at ~7–10× realtime on GPU (estimated) |
| Ollama 7B (Llama 3, Mistral) | ✅ Yes | ~40–55 tok/s, fully in VRAM (estimated) |
| Ollama 13B | ✅ Yes | ~22–30 tok/s, fully in VRAM |
| Stable Diffusion 1.5 | ✅ Yes | ~6–9 s per 512×512 image at 20 steps |
| Stable Diffusion XL | ✅ Yes | ~16–24 s per 1024×1024 image at 20 steps |
| ComfyUI / FLUX.1 | ✅ Yes | FLUX.1 dev fp8 fits in 16 GB; ~2–3.5 min per image |
| LoRA fine-tuning (small model) | ✅ Yes | 7B QLoRA comfortable; 13B QLoRA feasible; ECC helps long runs |
Key:
- ✅ Yes — works well
- ⚠️ Possible but slow — usable with patience
- ❌ No — hardware limitation prevents this
Full Specifications
| Component | Specification |
|---|---|
| CPU | Intel Core i7-11850H (8C/16T) — i9-11950H and Xeon W-11955M variants exist |
| CPU Generation | Intel 11th Gen (Tiger Lake H) |
| RAM | 32 GB DDR4-3200 typical (4× SO-DIMM, up to 128 GB; ECC with Xeon) |
| Storage | 512 GB NVMe (primary slot Gen 4; up to three M.2 slots depending on config) |
| GPU | NVIDIA RTX A5000 Laptop GPU, ~110 W sustained in this chassis |
| VRAM | 16 GB GDDR6 with ECC (256-bit, ~448 GB/s) |
| Display | 15.6” FHD 500-nit or UHD options |
| Battery | 95 Wh |
| Weight | from ~2.5 kg |
| TDP | 45 W CPU + ~110 W GPU (sustained) |
| AI Score | 86/100 |
AI Performance in Practice
The mobile RTX A5000 is GA104 silicon with a workstation badge: the same 16 GB / 256-bit / ~448 GB/s memory system as the 16 GB RTX 3080 Laptop GPU in the Legion 7 Gen 6. The Legion runs its GPU at up to 165 W and wins raw speed by an estimated 15–25%; the Precision runs ~110 W, quieter and cooler, and holds that number indefinitely. For inference the difference is modest — token generation is memory-bandwidth-bound, and both cards have the same bandwidth.
Estimated from comparable GA104 results (we have not benchmarked this exact unit): Llama 3.1 8B Q4 at 40–55 tok/s; 13B Q4 fully GPU-resident at 22–30 tok/s; SDXL at 16–24 s per 20-step image. Where the 7560 pulls ahead of every gaming laptop is system memory: with 64–128 GB of DDR4 you can run llama.cpp with a 70B Q4 model split across GPU and RAM — slow (roughly 1.5–3 tok/s, estimated), but possible, which it simply isn’t on 32 GB-max machines.
Thermal behaviour
Workstation cooling with headroom: dual fans, thick fin stacks and a chassis that doesn’t care about being thin. Multi-hour QLoRA runs and overnight batch generation hold clocks flat, with fan noise in the “audible hum” range rather than the Legion’s hair-dryer register. This machine was ISV-certified to render for days; a fine-tuning weekend doesn’t scare it.
Battery life under AI load
95 Wh sounds generous, but the maths doesn’t change: expect roughly 60–75 minutes of sustained GPU inference on battery, with the GPU power-capped. The 240 W barrel adapter is required for full performance — USB-C alone won’t feed an A5000 at 110 W. Confirm the correct adapter is included; replacements run £40–£60.
What to Check Before Buying (Used)
The GPU config lottery — verify before paying. The 7560 shipped with everything from a T1200 (4 GB) through RTX A2000 (4 GB), A3000 (6 GB) and A4000 (8 GB) to the A5000 (16 GB). Listings love to say “NVIDIA RTX Workstation Graphics”. Demand the Dell service tag (look up the factory config on Dell’s support site) or a GPU-Z screenshot showing 16384 MB. An A4000 8 GB unit is still a good machine — see our ZBook Fury G8 review for what 8 GB buys — but it’s a different price bracket.
Battery health
powercfg /batteryreport as always; fleet machines often show high cycle counts. Below 70% of the 95 Wh design capacity, negotiate — replacements are ~£60–£90 and DIY-friendly.
GPU throttling test FurMark 15 minutes with HWiNFO: a healthy A5000 in this chassis holds ~100–110 W and 1,200–1,400 MHz under ~80 °C. Sudden drops to double-digit wattage indicate paste or a clogged fin stack — both fixable, both negotiating points.
Storage health CrystalDiskInfo per drive. Fleet 7560s often ran encrypted 24/7 — Power On Hours above 20,000 is common; insist on 0 reallocated sectors. Spare M.2 slots make adding a model-library drive trivial.
RAM slots Four slots is the 7560’s superpower — test that all four work (boot with sticks rotated, or check HWiNFO reports all channels). Mixed-speed fleet RAM is common and silently costs CPU-inference performance.
Model-specific issues to watch for Check BIOS for a lingering admin password and confirm Computrace/Absolute is deactivated — both occasionally survive careless fleet offboarding. Verify the display config matches the listing (FHD 500-nit vs dim base panels). Hinges and USB-C ports take fleet abuse first; test video out on both Thunderbolt ports.
Where to Buy in the UK
The best places to find a used Precision 7560 in the UK:
Back Market UK — steady workstation stock from professional refurbishers, typically £800–£1,100 for A4000/A5000 configs with 12-month warranty. Spec sheets are usually accurate, but confirm the exact GPU before ordering.
Dell Refurbished / Dell Financial Services outlet — official off-lease channel; pricier, but the config is exactly what the service tag says and business invoicing is painless.
eBay UK — the widest choice of A5000/64 GB+ configs, often from IT disposal companies with 30-day returns. Fine if the seller has 100% feedback and provides the service tag up front. Avoid listings without photos of the actual unit.
What to avoid: “RTX graphics” listings with no model number, prices that look like A5000 money for what the service tag reveals as an A2000, and the visually identical Precision 7550 (2020, older 10th-gen CPUs) sold as a 7560.
Verdict
AI Score: 86/100 — Pro AI
The Precision 7560 with an RTX A5000 is what serious, boring AI capability looks like on the used market: 16 GB of ECC VRAM, expandability no gaming laptop matches, and thermals designed for sustained work. It gives up 15–25% raw GPU speed to the Legion 7 Gen 6 and weighs enough that you’ll think twice about carrying it daily — but it will still be calmly fine-tuning models long after flashier machines have throttled.
Buy if: you fine-tune or batch-generate for hours at a time, want 64–128 GB RAM headroom for large-model offloading, or need an invoice, a warranty and ECC for business use.
Don’t buy if: you commute with your laptop, want maximum tokens per pound (the Legion 7 is faster for less), or your workloads fit happily in 8 GB VRAM.
Better alternatives: the ThinkPad P15 Gen 2 (£900–£1,300) is the same recipe with a better keyboard; the Lenovo Legion 7 Gen 6 (£800–£1,150) trades pro build for outright speed.