Integrated graphics (also known as integrated GPUs or iGPUs for short) have been AMD’s bread and butter for years. They power AMD’s so-called Accelerated Processing Units (APU), which are just CPUs with Radeon graphics. These APUs have been found in places like the Xbox and PlayStation since 2013, and they also power handheld gaming PCs like the Steam Deck and the Asus ROG Ally. Integrated graphics have opened lots of doors for AMD, doors that are at the same time closed to Intel — its iGPUs just can’t compete.
However, Intel won’t always be in the rearview mirror. In fact, the upcoming 14th-generation Meteor Lake chips have the potential to put Intel on parity with AMD or even outright take the crown. Beating AMD at its own game could open up some desperately needed opportunities for Intel as the company is going through financial turmoil.
Although I see lots of potential for Intel with Meteor Lake, I’m not totally convinced it’s a slam dunk either.
Why integrated graphics are so important for AMD and Intel
Source: AMD
Integrated graphics are a bit of a joke, at least in the PC gaming and content creator spaces. They’re way slower than regular graphics cards (or discrete GPUs, dGPUs for short) and are often recommended only to people who can’t afford a real graphics card. I see people criticizing Intel and AMD for dedicating so much silicon area to integrated graphics on desktop chips since gamers just use graphics cards anyway. What people tend to miss is that Intel and AMD aren’t just selling CPUs to desktop gamers.
Firstly, a CPU with integrated graphics doesn’t require a graphics card for video output and other functions, which not only reduces the price of a computer but also cuts other GPU manufacturers out of the picture; this is what killed Nvidia’s chipsets with integrated graphics. Additionally, a CPU with an iGPU is also suitable for other markets like laptops and embedded systems. It’s for these reasons that AMD suddenly added integrated graphics to its enthusiast-class CPUs in Ryzen 7000.
Integrated graphics don’t have to be just for basic video output though, and AMD has been proving that for years with some great CPUs. Unlike Intel’s integrated graphics, APUs have been able to deliver respectable gaming performance at 720p and 1080p for over a decade, and AMD’s latest Ryzen 7000 APUs using the Phoenix chip are apparently so fast that they’re perfectly suitable for 1080p in AAA games. AMD claims its fastest APU is up to two times faster than Intel’s fastest P-class laptop CPU, which is not great for Intel.
Losing the performance crown by this much means Intel’s chips will almost certainly never be considered for consoles, handheld gaming PCs, and other devices where graphical horsepower matters. Financially, Intel is in a similar situation AMD was in back in the early 2010s, and APUs were AMD’s lifeline. That avenue isn’t open to Intel.
Meteor Lake could propel Intel to first place in performance
Although I’ve criticized Meteor Lake quite a bit for its quirky design, I do think it has some strengths, especially when it comes to graphics. Meteor Lake is Intel’s first consumer CPU to use “tiles,” which are basically individual silicon dies that have specific purposes. They’re similar in concept to AMD’s chiplets, but are far more specialized and custom. One of the strengths of this approach is that it enables Intel to make powerful integrated GPUs for select Meteor Lake chips.
Intel has already confirmed that it will make at least two different GPUs for its Meteor Lake chips using TSMC’s 5nm node, which is on par with AMD’s RX 7000 series. A standalone GPU tile allows Intel to maximize performance if it so desires since integrated graphics on a monolithic CPU (which means no chiplets or tiles) have to be limited in size in order to be practical. Because AMD prefers to just make one consumer APU rather than multiple ones with different iGPUs, it might be leaving performance on the table. Intel would have no such problem with this tile-based approach.
In the best-case scenario, Intel could compete with AMD everywhere, from business laptops and desktops to gaming consoles both at home and handheld. Even coming within striking distance of AMD’s integrated graphics in performance and efficiency could bring in some new customers for Intel. There’s even the possibility of beating AMD at its own semi-custom game. This is the segment that has AMD making custom chips for various clients like Sony, Microsoft, and Valve. A tile-based CPU would be even more customizable and potentially a major competitor.
Producing Meteor Lake cheaply might be a big problem for Intel
Source: Intel
But Meteor Lake has some significant problems in one important area: price. For starters, designing multiple pieces of silicon is expensive (especially the way Intel is doing it), which is why AMD almost never makes more than two APUs per generation. Putting all these tiles together is also yet another step in the manufacturing process, which means increased costs and potential for manufacturing defects. The cost of making tile-based CPUs could be a big problem for Intel here, and that really limits Meteor Lake’s potential.
One of the key reasons why AMD has been so successful in selling APUs across so many segments is down to how cheap AMD can make them. Home consoles and handheld gaming PCs (or at least the Steam Deck) target relatively low price points, and AMD APUs using older, cheaper nodes are undoubtedly more capable of making those price points feasible. I can’t see Meteor Lake doing the same thing with its state-of-the-art hardware until years down the line when TSMC 5nm and Intel 4 are no longer super expensive.
Additionally, Intel has a very poor track record when it comes to high-performance integrated graphics. In 2018, Intel launched Kaby Lake G, a quad-core Intel CPU with AMD graphics and integrated HBM2 memory, very similar to Meteor Lake’s multiple tiles. However, it was a commercial failure and was featured only in the HP Spectre, Dell XPS, and Intel’s own NUCs. Using cutting-edge packaging and HBM2 undoubtedly made Kaby Lake G expensive to produce and was likely a key factor in its poor sales.
There’s also Tiger Lake-U from 2020, which came with a big boost in integrated graphics for the express purpose of beating AMD’s APUs and Nvidia’s MX-class GPUs. While it sold far better than Kaby Lake G, it still wasn’t enough. Lots of Tiger Lake-U laptops were paired with Nvidia GPUs anyway, and driver issues prevented Tiger Lake’s iGPU from being consistent. Furthermore, Intel hasn’t upgraded its integrated graphics since, which is why Ryzen 7000 is so much faster than Alder Lake and Raptor Lake.
Cutting-edge technology could be a double-edged sword for Meteor Lake, at least for now
AMD, the company that pioneered chiplets, still hasn’t replaced monolithic APUs, and I think there are lots of good reasons why. Since chips with powerful integrated graphics only need to cover a relatively slim part of the market, it’s easier to just make one or two new designs every generation and cut those new designs down as needed. Additionally, AMD relies on older 7nm APUs from the Ryzen 4000, 5000, and 6000 series for devices that need a more cost-effective solution. It’s a strategy in stark contrast to Meteor Lake.
Meteor Lake’s usage of brand-new nodes and architectures will certainly help in the fight against AMD’s APUs when it comes to performance, but I expect Intel to struggle with pricing and value. If Meteor Lake is significantly more expensive than AMD APUs, then someone has to foot the bill. It’ll probably be Intel, which is less than ideal since the company desperately needs to start making more money.
On the other hand, if Intel can master its tile technology with Meteor Lake, then it might set up the company for a better future. Manufacturing Meteor Lake and its successors will hopefully become cheaper over time, which should even the playing field and could allow Intel to break into AMD’s niche in semi-custom. A sudden foray into one of AMD’s most historically important segments would be good for both Intel’s business and its reputation. Although I think it’s less likely things will turn out so well for Meteor Lake, at least Intel has a shot of making some magic happen.