Graphics Cards Price Comparison
Compare 1,353 graphics cards from ASUS, GIGABYTE, MSI and more — find the best price on RTX 5070, RX 9070 and beyond.
Graphics Cards price comparison UK
Few components define a PC's capability quite as sharply as the graphics card. Whether you're pushing frames in the latest AAA titles, running AI inference workloads, or driving a professional CAD suite, the GPU you choose sets a hard ceiling on what's possible. Our catalogue currently spans 1,353 products — from a budget GT 1030 at 56 £ to workstation-class cards nudging 1,391 £ — which tells you everything about how wide this market really is.
ASUS leads the field in sheer product volume with 220 listings, though its average price sits noticeably higher than rivals like Zotac or Palit. That gap isn't arbitrary: ASUS's ROG Strix and TUF Gaming lines carry premium cooling solutions and factory overclocks, whereas Palit and Inno3D tend to offer more restrained designs at keener prices. If you're after the same GPU die for less, those brands are worth a serious look — the underlying silicon is identical.
The current generation tells a clear story. NVIDIA's Blackwell architecture (RTX 50-series) now dominates the mid-to-high end, with the RTX 5070 emerging as the most-contested card in the catalogue — multiple board partners including MSI, PNY, Palit, and GIGABYTE are all competing around the same price point. AMD's RDNA 4 response, the Radeon RX 9070, offers 16 GB of GDDR6 and represents a compelling alternative for those who don't need DLSS. Intel's Arc B580, meanwhile, has quietly become the most interesting budget option — 12 GB of GDDR6 for well under 450 £ is genuinely hard to argue with at that price tier.
One thing our data makes plain: the professional segment (PNY's RTX A-series, HP workstation cards) skews the average price dramatically upward. These cards aren't for gaming — they carry certified drivers for Autodesk, Solidworks, and similar software, and their pricing reflects that. If you're building a gaming rig, you can safely ignore anything labelled "workstation" and focus on the GeForce RTX or Radeon RX lines instead.
Pairing your GPU with the right motherboard and sufficient memory is just as important as the card itself — a bottlenecked CPU or starved RAM will waste whatever you spend on the GPU. Use MagicPrices to compare offers across Currys, Amazon.co.uk, and specialist retailers before committing; prices on new-generation cards can shift significantly week to week.
How to Choose the Right Graphics Card
With RTX 5070s from four different board partners all competing at nearly the same price, the GPU market in 2026 is simultaneously more exciting and more confusing than it's been in years. The right card depends less on brand loyalty and more on three hard questions: how much VRAM do you actually need, what's your PSU rated for, and are you gaming or working?
VRAM: the number that matters most right now
8 GB was the mainstream standard for years — it's now a liability. Modern titles like Alan Wake 2 and Cyberpunk 2077 at 4K can exceed 10 GB of VRAM, and once a card runs out, performance doesn't degrade gracefully; it falls off a cliff. Our firm recommendation: 12 GB is the minimum for a card you expect to use for three or more years. The RTX 5060 Ti in 8 GB form is a notable trap — the 16 GB variant costs more but avoids a ceiling you'll hit sooner than you think. For professional workloads (3D rendering, large AI models), 24 GB or more is the sensible floor.
Architecture generation and driver longevity
Buying a Turing or Ampere card today means you're already one or two generations behind on driver optimisation and feature support. NVIDIA's Blackwell (RTX 50-series) and AMD's RDNA 4 (RX 9000-series) are the current architectures — they'll receive active driver updates for the longest period. Ada Lovelace (RTX 40-series) cards may appear at attractive prices as stock clears, and they remain excellent, but factor in that DLSS 4 and Multi Frame Generation are Blackwell-exclusive. Older architectures like Turing (RTX 20-series) should only be considered if the price is exceptionally low.
Power delivery: know your PSU before you buy
This is the most commonly overlooked factor. An RTX 5070 draws around 250W; an RTX 5070 Ti pushes past 300W. If your PSU is a 550W unit from a budget build three years ago, you may have a problem. Check the 12VHPWR connector requirement on high-end cards — some older PSUs need an adapter, and a poor-quality adapter is a genuine fire risk. As a rule of thumb: budget 100W headroom above the card's TDP, and stick to PSUs from reputable brands (Seasonic, Corsair, be quiet!). If you're upgrading from a low-TDP card like a GT 1030, a PSU upgrade may be unavoidable.
Gaming vs professional: two completely different markets
Consumer GeForce RTX and Radeon RX cards are optimised for gaming performance per pound. Professional cards — PNY's RTX A-series, HP's workstation GPUs — carry ISV-certified drivers for CAD, simulation, and media production software, and command a significant price premium for that certification. Buying a professional card for gaming is wasteful; buying a gaming card for Solidworks or Catia is unsupported and may produce incorrect results. Be clear about your primary use case before filtering by price.
Cooling solution and physical fit
Triple-fan coolers on flagship cards can exceed 330mm in length — measure your case before ordering. Beyond length, consider noise: dual-fan designs on mid-range cards like the MSI Ventus 2X are compact and quiet enough for most builds, but sustained workloads (rendering, mining) will push them harder than a triple-fan alternative. Liquid-cooled variants exist but add complexity. For small form-factor builds, look specifically for SFF (short) variants — GIGABYTE's Windforce OC SFF is a good example of a full-performance card in a compact footprint.
Board partner differences: more than just aesthetics
All RTX 5070 cards use the same NVIDIA die, but board partners differ in meaningful ways. Factory overclocks (the "OC" suffix) typically add 2–5% performance. Cooling quality varies: ASUS's TUF and ROG Strix coolers are among the best, Palit's Infinity 3 is competent but noisier under load. Warranty terms matter too — check whether the retailer or the manufacturer handles claims, and whether John Lewis's extended guarantee adds value over the standard two-year UK statutory right. Zotac and Palit often undercut ASUS and MSI by a meaningful margin for near-identical performance.
- Entry-level and legacy (From 56 £ to 290 £) : Mostly older-generation cards (GT 1030, GTX 1650) and professional entry-level options like the RTX A400. Suitable for basic display output, light office use, or upgrading a system with no dedicated GPU. Not recommended for modern gaming — VRAM and architecture limitations will frustrate quickly. Intel's Arc B580 sits just above this band and is a far better choice if your budget stretches.
- The mainstream sweet spot (From 290 £ to 450 £) : Where most buyers should be looking. The RTX 5060 Ti (8 GB and 16 GB variants) and Intel Arc B580 live here, alongside clearance RTX 40-series mid-range cards. The 16 GB RTX 5060 Ti at the top of this band is arguably the best value in the entire catalogue right now — enough VRAM to stay relevant, current architecture, and competitive pricing from Currys and Amazon.
- High-performance gaming (From 450 £ to 735 £) : The RTX 5070 dominates this segment, with multiple board partners (ASUS Prime, MSI Ventus, PNY OC, Palit Infinity 3, GIGABYTE Windforce) all clustering tightly. The AMD RX 9070 with 16 GB GDDR6 is the main alternative and worth serious consideration if you're not invested in DLSS. Meaningful 4K gaming becomes realistic here, and ray tracing stops being a slideshow.
- Enthusiast and workstation (Over 735 £) : RTX 5070 Ti, 5080, and 5090 territory, plus professional workstation cards from PNY and HP. For most gamers, diminishing returns set in sharply — the jump from RTX 5070 to 5080 costs far more than the performance gain justifies. This band makes sense for content creators needing 24 GB+ VRAM, or professionals requiring ISV-certified drivers. Watch for Boxing Day and Black Friday deals if you're targeting this segment.
Top products
- ASUS Prime -RTX5070-O12G NVIDIA GeForce RTX 5070 12 GB GDDR7 (ASUS) : The most-listed RTX 5070 in the catalogue and a solid all-rounder — ASUS's Prime cooler is quieter than the Ventus but the card is longer than average, so check your case clearance before ordering.
- MSI GeForce RTX 5070 12G VENTUS 2X OC NVIDIA 12 GB GDDR7 (MSI) : The compact choice among RTX 5070 options — the dual-fan Ventus 2X design keeps dimensions manageable and pricing sharp. Not the quietest under sustained load, but excellent value for mid-tower builds.
- Palit GeForce RTX 5070 Infinity 3 OC NVIDIA 12 GB GDDR7 (Palit) : The outsider pick — Palit undercuts ASUS and MSI noticeably for the same Blackwell die with a factory overclock included. Build quality is a step below the premium brands, but for pure performance-per-pound this is hard to ignore.
- ASUS TUF Gaming TUF-RTX5060TI-O16G-GAMING NVIDIA GeForce RTX 5060 Ti 16 GB GDDR7 (ASUS) : The 16 GB variant is the one to buy — the extra VRAM over the 8 GB model is worth every penny given where game requirements are heading. Best-in-class cooling from ASUS's TUF line makes this a near-silent performer under typical gaming loads.
- Intel Arc B580 12 GB GDDR6 (Intel) : The budget wildcard that earns its place — 12 GB GDDR6 at this price is genuinely exceptional, and modern game performance punches above its weight. Avoid if your library includes many pre-2016 titles; driver maturity for legacy games still lags behind NVIDIA and AMD.
Related categories
Frequently Asked Questions
How much VRAM do I actually need for gaming in 2026?
12 GB is the practical minimum for a future-proof gaming card in 2026. At 1080p you can get away with 8 GB today, but several recent titles already exceed that at higher settings, and a card bought now should last three or more years. If you're gaming at 1440p or 4K, 16 GB gives meaningful headroom. The 8 GB RTX 5060 Ti is a specific card to approach cautiously — the 16 GB variant costs more but avoids a VRAM ceiling you'll likely hit within 18 months.
Is the RTX 5070 worth buying over the RTX 4070 Super at a lower price?
Yes, in most cases — the Blackwell architecture advantage is real, not just marketing. DLSS 4 with Multi Frame Generation is exclusive to RTX 50-series and provides a substantial frame-rate uplift in supported titles. The RTX 5070 also ships with GDDR7 memory offering significantly higher bandwidth. If you find an RTX 4070 Super at a steep discount (clearance pricing), it remains a capable card, but for a new build at similar prices, the RTX 5070 is the stronger long-term choice.
What's the difference between a consumer GeForce card and a professional RTX A-series card?
Professional cards like the RTX A400 carry ISV-certified drivers validated for CAD and simulation software — consumer cards do not. For gaming, a professional card offers no advantage and costs considerably more. For Autodesk, Solidworks, or Catia, a certified driver prevents rendering errors and ensures software support. Don't buy a workstation card for gaming, and don't buy a GeForce card expecting it to behave like a Quadro in professional software.
Does it matter which board partner I choose for the RTX 5070?
The performance difference between board partners is small — typically 2–5% — but cooling quality, noise, and warranty handling vary more meaningfully. ASUS TUF and ROG Strix coolers are consistently among the best for thermal performance. Palit and Zotac tend to undercut ASUS and MSI on price for comparable performance. Check the physical dimensions carefully — some triple-fan designs exceed 330mm and won't fit in mid-tower cases without checking clearance first.
Should I avoid buying a graphics card with only 8 GB of VRAM?
For a new purchase in 2026, yes — 8 GB VRAM is increasingly a liability rather than a safe choice. Several current AAA titles already push past 8 GB at high settings, and the situation will only worsen. The only exception is if you're buying a legacy card (GT 1030, GTX 1650) for a specific low-demand use case like basic display output or light office work, where VRAM capacity is irrelevant. For any gaming-focused purchase, stretch to 12 GB minimum.
Will an RTX 5070 work in my existing PC, or do I need to upgrade other components?
The RTX 5070 uses a PCIe x16 slot and is compatible with most motherboards from the last decade, but your PSU is the critical check. The card draws around 250W and requires a 12VHPWR connector (or an adapter from dual 8-pin). A 650W PSU is the recommended minimum for a full system. Also verify your case has sufficient clearance — most RTX 5070 cards are between 280mm and 320mm long. If your current PSU is below 550W or from an unknown brand, budget for a replacement alongside the card.
Is the Intel Arc B580 a genuine alternative, or is it a compromise?
The Intel Arc B580 is a genuine contender at its price point, not just a budget fallback. 12 GB of GDDR6 for well under the category median is exceptional value, and rasterisation performance is competitive with cards costing significantly more. The caveats are real though: driver maturity still lags behind NVIDIA and AMD, older DirectX 9 titles can exhibit issues, and there's no equivalent to DLSS or FSR 4. For a primarily modern-game library on a tight budget, it's hard to beat. For older or more varied game libraries, stick with NVIDIA or AMD.