Intel's 14th-generation Raptor Lake Refresh is here and to nobody's surprise, it was basically the same thing as 13th-generation CPUs. All we really got were higher price tags and slightly higher clock speeds, and although the Core i7-14700K does offer a new core configuration of 8 P-cores and 12 E-cores, it's surprisingly power hungry and only offers a decent improvement to multi-threaded performance. It's natural to be disappointed with Intel, and it's easy to draw parallels to Intel's similarly poor 7th-generation Kaby Lake CPUs, and then get depressed about the state of things.
But I really don't mind these refreshed CPUs. I definitely don't think they're worth buying (and I personally expressed my disappointment in my own reviews), but all the doom and gloom surrounding 14th Generation is unearned. Meteor Lake desktop CPUs clearly weren't going to work out, and whether we got a refresh or Meteor Lake on the desktop, we wouldn't really care when a much more powerful generation is already on the horizon.
Meteor Lake is a good laptop chip, but looks like a poor replacement for Raptor Lake on the desktop
We've known for a while that Meteor Lake's desktop version was axed some time ago, but it's not confirmed exactly why. In my opinion, it's largely down to two problems: Meteor Lake can't hit the frequencies Raptor Lake can, and it would be too expensive even if it could. Of course, Intel hasn't officially confirmed either of these things, but based on what we know today it seems likely this is what's going on.
Obviously, Meteor Lake hasn't come out yet and it's only due in late December, but a steady stream of leaks is painting a fairly clear picture that clock speeds are significantly down. The highest-end SKU to have leaked thus far is the Core Ultra 9 185H, which is supposedly rated for 5.1GHz according to Chinese leaker Golden Pig Upgrade in September. That clock speed seems to be locked in as the first leaked Geekbench result of the 185H was spotted just a few days ago, and it hit just 5.1GHz in the benchmark.
5.1GHz isn't slow by any means, but for a flagship laptop processor it's quite slow. The Core i9-13980HX hits up to 5.6GHz, which is about 10% higher on what appears to be a flagship in the same vein as the 185H. If Meteor Lake is this limited in laptops, it would only be worse on desktops, where Raptor Lake has more room to stretch its legs. I also want to note that Meteor Lake only 6 P-cores and 16 E-cores, 2 P-cores behind Raptor Lake, and that further reduces the potential of Meteor Lake's performance. Even a tie in multi-threaded performance seems like the best case for Meteor Lake.
The key reason why Meteor Lake has a frequency regression is certainly down to the process its made on, Intel 4. This is a brand-new node and like many new, cutting-edge nodes, it can't hit the high clock speeds of more mature nodes like Intel 7, used for Raptor Lake. But new nodes also tend to be more expensive nodes, not just to pay off the development costs, but also because defects tend to be more numerous when a node first launches, and the more defects there are, the fewer CPUs that can be made per silicon wafer.
The best case scenario for Meteor Lake on the desktop would probably be a tie against Raptor Lake and higher prices, and in that light I think we can all understand why we got this boring refresh instead of something new. Of course, it's not great that Meteor Lake wasn't good enough for the desktop, but it is what it is.
Intel isn't desperate, the next generation is already on the way
Though I'm not impressed with 14th-generation CPUs, I'm not really worried that it's a sign of a relapsed Intel that can't get products out on time. It's true that Raptor Lake probably wouldn't have existed if Meteor Lake came out on time, but that was really happenstance, and we're probably not looking at the first of many refreshes or chips designed on the Intel 7 node like we saw with 14nm from 2015 to 2021. Things are different this time.
Intel has really turned itself around when it comes to its fabs, which had formerly struggled to get its Intel 7 node (known back then as 10nm) in an acceptable state up until 12th-generation Alder Lake came out in 2021. Intel has bounced back remarkably from these troubles and is now on track to launch four more nodes by the end of 2024, by making incremental gains with every new node rather than putting all its eggs in one basket and trying to make huge gains with a single node that takes years to develop.
When it comes to the desktop, the node to keep an eye on is 20A as it's the node that upcoming Arrow Lake desktop CPUs will be fabbed on. Intel will start manufacturing 20A chips in the first half of next year, and the assumption is that Arrow Lake will be the first. That doesn't necessarily mean we'll get next-generation CPUs in just a few short months, but that Arrow Lake should have plenty of time to come out in 2024.
Of course, being the first product to be made on a new node could mean that Arrow Lake also shares Meteor Lake's frequency issue, but that should be offset with a much better architecture that will likely offer significantly improved IPC and more cache. Plus, Arrow Lake should be much more efficient than Raptor Lake, which has a poor relationship with power consumption. Plus, Arrow Lake should be fairly cheap for Intel to make, as it will likely use two or three of Meteor Lake's four tiles, and that means saving money on both development and production. In this sense, Meteor Lake is walking so Arrow Lake can run.
I also expect Arrow Lake will have enough performance to go toe-to-toe with Ryzen 8000 (or whatever Zen 5 desktop CPUs are coming out). While AMD has a great track record with its Ryzen CPUs, it's going to have to contend with the fact that TSMC's 3nm node really isn't all that great. 3nm's upgrade over 5nm is far worse than 5nm's improvement over 7nm all across the board: gains in density, clock speed, and efficiency are all lower. It's hard to imagine that Arrow Lake won't be at least a pretty decent chip.
14th-generation CPUs inadvertently highlighted Intel's weaknesses
It's easy to get short-sighted when it comes to tech. 14th-generation CPUs didn't introduce any new problems but just put all of 13th-generation's existing problems at the forefront of our minds. That's pretty ironic considering 14th-generation CPUs were supposed to regenerate interest in Intel's product stack, not make users worried that the company is backsliding into failure.
The thing is, lots of great CPUs are preceded by really boring or even awful ones. AMD's legendary Ryzen 3000 series was preceded by Ryzen 2000, which was basically a refresh of Ryzen 1000; Intel's 12th-generation Alder Lake came just months after critically panned 11th-generation Rocket Lake chips. So, I'm looking forward to Arrow Lake, and I'm betting we'll all forget about this refresh in a year's time.