They have onboard graphics to process tasks locally without relying on a remote server, improving speed and privacy.
They have onboard graphics to process tasks locally without relying on a remote server, improving speed and privacy.
I was considering my own configuration; a 4K main gaming screen and a few old 1920x1200 satellite monitors (for emails or whatever). Right now I’m using the satellite monitor onboard and the 4K on a 7900XTX. I think it doesn’t really change much in real life, but for deciding where to place the modest workload of running that system, I’d like options—between using the onboard, sharing the XTX, or even getting a cheap PCIe card. I doubt it’s a big deal unless I’m doing something extreme, like opening a Chrome tab. In practice, nothing much is happening while I’m actually gaming. Anyway, this got me thinking. The x3D versions of CPUs seem geared toward gamers unless the 3D cache really boosts other apps. Otherwise, why not opt for a simpler CPU on a budget, or go for a threadripper if you need top performance? Why would anyone outside my target audience need the onboard unit except maybe me? Isn’t it just taking up space on the die, adding complexity, increasing power draw? It feels a bit like adding a tow-hook to a racing car.
The Intel Core 14900K includes built-in graphics. The area it occupies is within the IO die, making it an affordable chip and ensuring clock and power gating for minimal energy consumption. AMD is only utilizing a 2CU GPU, which doesn’t significantly increase space on that die. It can still handle certain GPU workloads using far less power than sending data directly to the GPU. This setup lets you run the CPU without a dedicated GPU for systems that don’t require one, and it’s useful for debugging. x3D isn’t primarily designed for gamers—it’s just an unintended side benefit. For compute-heavy tasks involving large datasets, performance improves noticeably because you avoid cache misses when using structures larger than 20MB.
I’m leaning toward this scenario for those who make poor choices by switching vendors out of nowhere. Or maybe you’re building a PC and still missing a GPU. This chip is great for everyday tasks on a fast CPU, so it might work just fine for a while. Or for another reason entirely.
The 5000 series wasn't like that. The 7000 and 9000 models were different.
The latest IO die for Zen 4 and previously used in Zen 5 includes a 2 CU GPU. It isn't found in any Zen 3 models.
They're right; every engineering challenge has a price. The space and voltage loss through power gates aren't free—about a quarter of the die area. That means you could have produced roughly 25% more chips per wafer. Even with a wafer still over 10K, it's not insignificant.