Integrated Graphics vs Dedicated GPU: Do You Need One?

If you're speccing a new PC in 2026 and wondering whether you can skip the graphics card entirely, you're not being cheap, you're being sensible. Integrated graphics vs dedicated GPU used to be a short conversation: buy a card if you want to play games, don't if you don't. That's no longer true. AMD and Intel have both put serious graphics silicon inside their processors, dedicated GPU prices have climbed hard through 2026, and a lot of buyers are asking a genuinely open question: do I need a separate graphics card at all?

The honest answer is that it depends on what "gaming" means to you, and on a RAM trade-off that most guides skip entirely. Here's how to work out which side of the line you're on.

Why This Question Wasn't Worth Asking Two Years Ago

Integrated graphics (an iGPU built into the CPU die) used to mean one thing: enough grunt for a desktop, a browser and maybe a game of Solitaire. Anything with real 3D load, a modern game, video editing with GPU acceleration, needed a dedicated card with its own memory and power delivery. That gap has narrowed a lot faster than most people's mental model of "integrated graphics" has updated.

At the same time, as we've covered in our GPU prices UK 2026 piece, dedicated card prices have been anything but stable this year, and entry-level cards that used to be an easy £180 add-on now sit meaningfully higher. When the cheapest sensible dedicated GPU costs more than it did twelve months ago, "can the CPU's own graphics do this instead" stops being a niche question and starts being the first thing a lot of buyers should check.

What Modern Integrated Graphics Can Actually Do

The two chips worth knowing by name are AMD's Ryzen AI 300 series (codenamed Strix Point, and the beefier Strix Halo variant) with Radeon 800M-series graphics, and Intel's Core Ultra 200V (Lunar Lake) with Arc graphics. Both put considerably more silicon and, crucially, considerably more memory bandwidth behind the graphics portion of the chip than the integrated graphics of five years ago. Handheld gaming PCs are the clearest proof this isn't marketing: the current wave of Windows handhelds runs current-generation games at native 1080p using nothing but integrated graphics, because there's no room inside a handheld for a separate card.

Where it's genuinely fine

Web browsing, office work, video calls, 4K video playback and light creative work (photo editing, basic video cuts) run without complaint on any modern iGPU, including the much more modest graphics built into mainstream desktop CPUs from our CPU buying guide. Older or less demanding games, esports titles like Counter-Strike 2, Valorant and League of Legends, and most 2D or stylised indie games, run comfortably at 1080p on the better integrated chips, and reasonably on the average ones if you're willing to drop settings.

Where it starts to strain

Anything visually demanding released in the last two or three years, at 1080p or above, with ray tracing or high-refresh targets, is still a different conversation. Integrated graphics shares system memory with the CPU rather than having its own dedicated pool, and that bandwidth ceiling shows up exactly where you'd expect: frame time consistency in busy scenes, texture streaming in open-world games, and anything asking for more than about 4-6GB of usable graphics memory. If you've read our guide to how much VRAM you actually need, the same logic applies here, just with a much lower ceiling before things get uncomfortable.

The RAM Trade-off Nobody Mentions

This is the part that actually changes the buying decision, and it's specific to 2026's market. Integrated graphics performance is far more sensitive to your system memory than most buyers expect, because the iGPU draws its working memory from the same pool and the same channels as the CPU. Two consequences follow directly:

• Dual-channel isn't optional. A single stick of RAM roughly halves the memory bandwidth available to integrated graphics, which can cost you a large chunk of gaming performance you'd never notice with a dedicated card. If you're going the iGPU route, two matched sticks is the baseline, not a nice-to-have.
• Faster RAM matters more here than almost anywhere else in a build. Where a dedicated GPU shrugs off slower memory, an iGPU's frame rate scales visibly with memory speed. That pushes you toward faster kits from our RAM category, at exactly the point in 2026 when, as we've written about in our RAM pricing coverage, memory costs more than it used to.

So the saving from skipping a graphics card is real, but it's partly clawed back by needing a better memory kit than you'd otherwise buy. Budget for that honestly rather than assuming the whole GPU price disappears from your total.

When You Should Still Buy a Dedicated GPU

Skip the debate entirely and buy a card if any of these apply to you: you play current AAA games at 1080p or above and want stable frame rates on high settings, you want 1440p or 4K at all, you care about ray tracing, or you do GPU-accelerated creative work (serious video editing, 3D rendering) or local AI workloads that need dedicated VRAM. In all of these cases, entry and mid-range cards from our GPU category, think the current RTX 4060/5060-class and RX 7600/9060-class tiers, remain the sensible choice, and our GPU buying guide covers how to pick between them properly.

When Integrated Graphics Is the Right Call

The other side of the line: general-purpose desktops, home office and study machines, HTPCs, and light or older-game-focused builds where the money saved is better spent elsewhere in the build, a faster CPU, more storage, a better monitor. If that's your use case, put the budget you'd have spent on a card into the CPU itself and the RAM kit that feeds it, check our CPU buying guide and motherboard guide for platforms that pair a strong iGPU with good memory support, and revisit adding a dedicated card later. Most current motherboards make that upgrade trivial, so going without one now isn't a decision you're locked into.

Two Builds Worth Comparing Directly

It's easier to decide with two concrete builds in front of you rather than an abstract argument. An iGPU-only build puts its money into a stronger processor and a fast 32GB dual-channel kit, skips the graphics card line entirely, and comes out well ahead on price for general use, study, work, and older or lighter games. An entry dGPU build takes a more modest CPU, a standard-speed 32GB kit, and puts the difference into a budget card from our GPU category, and it's the one that actually holds up for current games at 1080p on medium-to-high settings.

The trap is comparing the iGPU build's price against the dGPU build's price without upgrading the RAM in the iGPU build to match what integrated graphics actually needs. Do that comparison properly, CPU plus the right memory kit versus CPU plus card plus standard memory, on our all products page, and the real gap between the two routes is usually smaller than either camp claims, and easier to justify either way.

The Bottom Line

Integrated graphics in 2026 has closed enough of the gap that it's a legitimate choice for a real slice of buyers, not just a fallback for people who can't afford better. But it's a choice with a condition attached: it only performs the way the marketing suggests if you pair it with proper dual-channel, reasonably fast memory, which costs more than it did a year ago. Work out honestly which games and workloads you actually run, price the RAM you'd need alongside the CPU on our all products page, and compare that total against a budget dedicated card before you decide. For a lot of general-use builds, skipping the graphics card is now a genuinely good call. For anyone who wants to game properly, it still isn't.