Intel claims new i9 is faster than M1 Max

This feels like Intel blowing smoke. Comparing an HK processor to the M1 Pro/Max feels disingenuous.

AnandTech compared the 11th generation HK to the M1 Max and found the M1 Max to use a lot less power (https://www.anandtech.com/show/17024/apple-m1-max-performanc...). The HK used 2.7x, 2.1x, 2.3x, 2.0x, 3.2x, 1.9x, 1.3x, 1.3x, and then the gaming ones with discrete GPU where the combo used 3.2x and 2.4x. The MSI laptop with that HK chip in it is a 6.4 pound laptop with a 280W charger.

What's so great about my M1 Pro MacBook is that I never hear fan noise. It doesn't get warm. I don't have to worry about running my battery down with IntelliJ. I don't have to worry about covering the vents with a blanket. I just get to use it.

Intel is also announcing new 28W P-series processors and 9-15W U-series processors. The M1 (non-Pro/Max) fits into a fan-less MacBook Air. While my MacBook Pro has a fan, it basically never comes on (which definitely wasn't the case with my 10th generation Intel MacBook Pro). Comparing a 45W HK part to an M1 Pro/Max that seems to be well below 28W just doesn't seem like a real comparison.

We'll see when AnandTech and others get their hands on it, but if it's achieving marginal performance gains while using double the power, that's not impressive.

It kinda leaves me with the question: Are the HK processors a desperate attempt for Intel to get positive PR? When Apple showed off the M1 Max, it was showing off a chip that would fit in a gorgeous, convenient laptop. All of a sudden, you could have all that power without compromise. The HK processors feel like Intel saying, "see, you can have that power with us too," except that it doesn't come without the compromises. It comes with a 280W power adapter, a large and heavy form factor, heat, and low battery life. Or maybe Intel has pulled it off...but it seems like a desperate attempt to say "we're #1...if you don't look closely."

Plus, one of the big advantages of the M1 is the unified memory. When Affinity Photo was benchmarking the M1 Max, it really showed the importance of the unified memory: https://twitter.com/andysomerfield/status/145262392072144896....

Maybe Intel has pulled off something amazing. It seems more likely that they've produced a low-volume 45W part that will let them claim a benchmark crown while their 28W, 15W, and 9W parts are still behind. The fact that they're trumpeting a 45W part immediately makes me think that they're making a bad faith comparison.

I hope Intel does well because competition is good for me as a consumer. This comparison doesn't make me think Intel has caught Apple.

Well Intel is also claiming lower energy usage at equivalent performance (see other posts in this thread). I take that with a grain of salt, but that's the claim. If true then you could power-limit it and use less energy to do the same work as the M1 (although it's not clear even from Intel's marketing that theirs scales down quite so well as M1 so as a general purpose low power / low heat / long battery life laptop that you don't run flat out all the time it's possible it would still be better overall even if Intel's claims hold up).

Couple things that come to mind in response to that chart:

Its impressive, but:

Generally, take Intel's marketing material with a grain of salt. They're notorious for exaggerating, cherry-picking, and out-right lying in marketing materials.

Related to that: While both the M1 Max & 12900HK have on-board iGPUs, the M1 Max has an especially powerful GPU for its class; one of the best iGPUs ever produced. That's going to contribute massively to package power draw, which is what they're measuring on the X-axis; if the Xe G7 can't compete, then these graphs are disingenuous to the overall experience some customers would be looking for (alternatively, other customers will be going dGPU anyway, so it doesn't matter).

Related to that: Its unclear from the chart whether this is per-core relative perf or entire package and multithreaded perf. I'm guessing the latter, due to the size of that x-axis arc and because its roughly where we'd expect a 12th gen 14/20 configuration to land; that's an obscene amount of threads for mobile, far outpacing the 10/10 in M1 Max. In fact, I'd argue: if this is a multithread test, a 40% increase in cores w/ a 100% increase in SMT threads only arriving at a miniscule (~5%?) jump in perf at similar power draw is average, not good. The single-core perf, even on their high-draw cores under load, may be rather poor; so its end-of-day effectiveness may be very workload dependent.

Overall, I'm hopeful the real-world performance matches the expectations they've set, and I'm bullish on Intel's strategy. Chip manufacturing is hard, but TSMC is becoming a bottleneck for both Apple & AMD. Hyper-optimizing large processes may work out for them; but the proof will be in the pudding.

> that makes it difficult to manage your compilers and libs on macOS

A failure of the C/C++ toolchain, mostly.

Every time they zig or zag on the architecture, though, it presents a risk to their third-party ecosystem. Yeah, there will be a few years where a backwards-compatibility penalty box is offered, but eventually someone's x86-64 only tools or extensions will break.

They also threw out the ability to say "If your company/tasks don't work with MacOS, you can always dual-boot full-fat Windows."

They have to be betting that their platform is still big and compelling enough to keep bringing in users and developers to replace anything they lose to the x86-64 abandonment.

I tend to think that there's no real magic secret to ARM for low power, it's more that we've spent the last 20 years optimizing them for low-power devices while x86 was traditionally performance-at-all-cost.