I don't mean to condescend, I just want to make sure we're on the same page.
Not at all, in fact, I really appreciate the effort.
TDP is a measurement in watts of the amount of thermal energy output when the CPU is under "heavy load" at base clock speed under normal use circumstances. So the absolute maximum could be a good bit higher than the actual TDP rating reported by the chip manufacturer under sustained, maximum load at maximum turbo boost frequencies. As I've said before, these frequencies are ambitious of Intel and the TDP that is stamped on their product is no indication that it is the TDP for anything other than base clock frequencies.
Precisely. In fact, I'd like to expand on this a bit. The notion of TDP, how it's used by Intel, is inseparable from the base frequency and it is first and foremost a marketing notion. It sets the expectation of the CPU performance. For Intel, its along the lines of "if you are running this CPU inside a chassis that is designed to provide/dissipate at least TDP of power in sustained multi-core operation, then our CPU will run at at least base frequency". To put it differently, that's what you are buying. If your CPU is consistently not hitting base, its underperforming and you have reasons to complain or ask for a refund. However, it also means that any operation beyond the base is not guaranteed — unless an OEM explicitly makes such a guarantee. The issue here is that performance and power is not linear. Again, as MSI Titan shows us, we need to increase the power consumption of the i9 by 2-2.t times to get 30% more performance, which is outside of any realm of reason (its crazy even by overlocker's standards).
As to your second point, I hope it's fine that I rearrange/pick some core statements from it so that I can reply to them consistently.
This is where I have a problem with Apple and their decision on the 2018 MacBook Pro 15". At heavy load, even at stock frequencies, the device has a very hard time staying cool when ambient temp is room temperature.
I have to admit, I don't understand this. I have tested out a bunch of 2018 15" MBPs and I have also read most external reviews and I don't see anything out of ordinary with its temps. Sure, it gets hot under heavy load, but so did every MBP before it, and also so does every other laptop on the market. These CPUs have a very wide range of operating modes, and the power usage of the i9 CPU in particular ranges from under 5 Watts to over 90 Watts (all values I've seen personally with my model). That said, the computer is absolutely cool and quiet under normal office operation (by which I mean the CPU being 90% in idle).
It was so bad after initial release, they had to address it in a firmware update (firmware updates aren't very common unless something really went wrong--how was an essential hardware controller like this missed in testing?). That firmware update was essentially a fan speed curve, among other tweaks we don't really know because it's an Apple product and update details are very non-transparent.
Thats where you are wrong. The original issue had nothing to do with cooling or thermal performance of the chassis. It was a bug in the power management firmware. The effect of this bug was that the system did not respect the power limits and was boosting beyond reasonable ranges. As the result of this, the CPU power system was overheating and sending out an emergency throttle signal. To use an ever popular car analogy, think of an misconfigured injection system that would inject too much fuel for the engine to handle, leading to detonation issues.
I don't know whether the fix included modification of fan speeds, but its main purpose was to fix the bug, so that the power system behaves properly.
firmware updates aren't very common unless something really went wrong--how was an essential hardware controller like this missed in testing?
Apple themselves claimed that the issue was a missing cryptographic key. It is very much possible that the issue occurred during the final packaging of the firmware software and did not show in the previous testing. (Which by the way is an additional reason why Apple should improve their software practices, but thats a different topic).
When I bought my MacBook Pro with the i7-8750H I knew I was getting roughly the same value as the i9 for sustained workloads because I knew they both share the same exact thermal headroom. What I didn't know is that the chassis is insufficient in cooling the chip's thermal output at stock frequencies under heavy load without the internal fans spinning at 5k RPM and the excessive amount of noise it generates.
I don't know why you find it surprising? As we discussed above, the chip is consuming/radiating around 45watts of power at this point, in addition to the power used by the surrounding circuitry. Thats quite a lot of heat. You can't dissipate this without running the fans at high speed.
My MacBook sounds like a helicopter about to take flight at heavy load with turbo boost disabled.
Yes, they are quite loud at load. Has been like this for years though. There are certainly quieter laptops out there. However, how is this an argument for chassis's thermal inadequacy? Noteboockcheck measured 33.9dB on the 2018 model when subjecting it to sustained Cinebench loop, the 2013 model under similar circumstances scored around 35 dB. Both laptops could go as loud 45-47db under stress test. The new thinner chassis is not necessarily any louder and it runs the CPUs within specs.
So why put them in there in the first place when a lower stock frequency and lower maximum turbo boost frequency chip could replace it, using the thermal headroom to achieve the same exact frequencies and get extremely close to the same exact performance (not counting cache size, cache speed, and other non-frequency performance enhancements)?
Ah, and now we get to the most interesting point. The thing is, I (partially) agree. Apple could (and maybe even should) have skipped the i9 option. However, here are my thoughts:
- There is more to performance than sustained multi-core operation. The true advantage of the i9 is the very high burst speed. This is where its outperforming the i7 models and this is there reason why I got it — the extra burst helps out quite a bit with code compilation and short scripts that I run all the time. Same goes for extra cache size.
- Skipping the i9 might have been even more of a PR disaster than using it. The "dissapointing" i9 performance (even though, I think its more appropriate to talk about i7's performance exceeding expectations) is
not exclusive to Apple, and its only the high-end gaming laptops that show systematic improvements in benchmarks with i9 vs i7. However, other laptops do include them, even if the performance is not what one wants to see. Imagine if Apple decided that offering an i9 option wouldn't make any sense to the customers. I have a strong suspicion that the discussions we would have right now on the forums would be something like this "Apple doesn't care about pro users, where is my i9? I am switching to Dell XPS, it has an i9? — But it runs barely faster than the i7.... — You are just an Apple fanboy ignoring the fact that Apple doesn't care about performance anymore"
So all in all, yeah, it's a mixed bag. Personally, as someone who has professionally worked and maintained close to a hundred of MacBook Pros for the last 10 years or so, I consider the new chassis to be a great little performer, which is in no way inferior to any other laptop chassis Apple ever made. And of course, if they increased the form factor/reduced the battery etc., they could most likely boost the performance somewhat. Would it make a big difference though? I doubt it. MBP's design is a compromise, as it has always been. I am very well aware of the fact that there are other laptops out there — and in the meantime they are almost as compact — which can outperform the MBP in certain areas. Still, when I look at these laptops in detail, the MBP always comes up me as a more balanced and user-friendly tool — at least in my perspective.
At any rate, we can hope that next CPU generation will improve the efficiency to the levels where this entire mess can be simplified.
