Apple MacBook Pro 13 and Intel Core i5-1038NG7 shown to offer comparable performance with 15 W Ice Lake processors despite 28 W TDP; significant improvements expected for Tiger Lake-U too

[Redaktion]

The Core i5-1038NG7 and Core i7-1068NG7 have started appearing on multiple benchmarking websites following the launch of the MacBook Pro 13, and they seem to be incapable of outperforming processors like the Core i7-1065G7 despite their 28 W TDPs. The cooling system of the MacBook Pro 13 could be at play, with the 28 W Core i5-1038NG7 also falling well short of early examples of Tiger Lake-U processors.

https://www.notebookcheck.net/Apple-MacBook-Pro-13-and-Intel-Core-i5-1038NG7-shown-to-offer-comparable-performance-with-15-W-Ice-Lake-processors-despite-28-W-TDP-significant-improvements-expected-for-Tiger-Lake-U-too.465627.0.html

[Valantar]

Not hitting peak boost whatsoever is a serious issue - I would say grounds for a recall, frankly. Even AMD's early Ryzen 3000 with the shitty BIOS managed to hit those boost levels even for tiny tiny amounts of time - and of course the subsequent BIOS fix removed the issue altogether.

On the other hand I wonder how sustained performance plays out - the higher TDP (and hopefully matching cooling system) ought to make for less throttling over time than 15W SKUs. Fire Strike is by no means a long enough test to demonstrate this, so I guess we'll see once NBC gets these laptops in for proper testing.

[Tov]

I understand they can't use Renoir because upcoming ARM power chip won't surpass it. Keeping Intel with low power making future ARM look better at how many times CPU and GPU they will be stronger.

[Spunjji]

Looking forward to a fair and balanced meltdown from the popular tech press on these chips (and Comet Lake) failing to reach their boost clocks.

Any day now.

[william blake]

@rogame asserts that top-tier Tiger Lake-U processors may offer up to a 40% CPU and GPU performance uplift over the Core i5-1038NG7

oh really? same power, same node, +40% performance?

[Mission]

True. www.MissionQuest.org

[JMac]

You realise those other ice lake CPUs are also running at around 25 W too right?

[Turbocharge]

JMac: you mean it's 15W running at base clock and 25W with tBoost?

[_MT_]

What did you expect? We all know that the chips are not running at TDP. And that they tend to be thermally limited. Higher TDP simply means higher base frequency. But nobody is benchmarking them at base frequencies. At best, Apple might be getting a high bin silicon (perhaps the highest bin). Laptop's design is the important variable. This applies in general to processors in the same line with the same core count. At best, the more expensive model is higher bin (putting aside cache size and such). Meaning more efficient. And that's the only difference that matters if you're thermally limited. With given efficiency, frequency (and therefore performance) is a function of cooling. Up to a power limit. Higher power limit only matters if the cooling can take you there. And the performance is going to come at the cost of more heat and therefore potentially more noise. Cooling really is perhaps the biggest feature in a laptop. When it comes to performance.

You might infer that usage of a higher TDP chip means better cooling. As TDP should guide cooler selection. But that just might be a wishful thinking. And you don't need a higher TDP chip to make use of cooling headroom. That's what boost is about. And the 28 W version has only 100 MHz higher boost IIRC. Negligible.

[_MT_]

Not hitting peak boost whatsoever is a serious issue - I would say grounds for a recall, frankly.

As long as it's only because of insufficient cooling (or because the manufacturer chose to impose a stricter power limit), it's not Intel's fault. And good luck trying to get the manufacturer to recall the laptop. I generally ignore advertising so I don't know what promises they make. But I wouldn't hold my breath.

[S.Yu]

I understand they can't use Renoir because upcoming ARM power chip won't surpass it. Keeping Intel with low power making future ARM look better at how many times CPU and GPU they will be stronger.

Seriously? They're messing up this gen to sell next gen? I honestly don't think I've ever come across a case in which a company plays its cards like this.

You realise those other ice lake CPUs are also running at around 25 W too right?

Some are 25W(like my XPS13 2-in-1, one of the reasons I bought it is a stable boost, far better than for example the SP with the same chip, which also happens to be more expensive), others are really 15W. Anyhow, there should still be a slight difference between 28W and 25W if properly implemented, especially since Apple's likely getting a good bin.

[Valantar]

What did you expect? We all know that the chips are not running at TDP. And that they tend to be thermally limited. Higher TDP simply means higher base frequency. But nobody is benchmarking them at base frequencies. At best, Apple might be getting a high bin silicon (perhaps the highest bin). Laptop's design is the important variable. This applies in general to processors in the same line with the same core count. At best, the more expensive model is higher bin (putting aside cache size and such). Meaning more efficient. And that's the only difference that matters if you're thermally limited. With given efficiency, frequency (and therefore performance) is a function of cooling. Up to a power limit. Higher power limit only matters if the cooling can take you there. And the performance is going to come at the cost of more heat and therefore potentially more noise. Cooling really is perhaps the biggest feature in a laptop. When it comes to performance.

You might infer that usage of a higher TDP chip means better cooling. As TDP should guide cooler selection. But that just might be a wishful thinking. And you don't need a higher TDP chip to make use of cooling headroom. That's what boost is about. And the 28 W version has only 100 MHz higher boost IIRC. Negligible.

You're absolutely right, but in previous generations when chips haven't been pushed quite as far, the difference between, say, an i5 and an i7 in the same thermal envelope has tended to be that the i7 is slightly higher clocked, slightly more efficient, but also allowed to pull slightly more power. Of course this relies on the cooling system being more than simply adequate, which is increasingly rare with the ongoing quest for making laptops so thin they become transparent (at least it seems like that's the goal). Now, instead of this system - which actually worked! - we have chips boosting to many times TDP for short periods (which is a good thing overall, but stresses cooling severely), increasingly underbuilt cooling, and far more leeway for OEMs to configure chips thermally (often to the detriment of the laptop). Thus we end up with i5s outperforming i7s as the latter enters a thermal throttling loop with the former stabilizing at a higher average performance level, and even "28W" chips not beating "15W" chips (that might be configured to 25W, though nobody will tell you) as cooling is limiting the chip from actually consuming its TDP in steady-state. It's all quite frustrating really.

[Philip]

The verge has MacBook pro 13 outperforming Dell XPS 2-in-1 and surface laptop 3: www.theverge.com/2020/5/7/21249966/apple-13-inch-macbook-pro-magic-keyboard-performance-video-premiere-affinity

[KPOM]

This is clickbait. Apple's 28W chips have long been slightly souped up versions of the 15W chips (which really top out at 25W). But even MaxTech (which started the whole "overheating" controversy) noted that in its tests the Ice Lake i5 MacBook Pro outperformed the Dell XPS 13 i7-1065G7.

[S.Yu]

The verge has MacBook pro 13 outperforming Dell XPS 2-in-1 and surface laptop 3: www.theverge.com/2020/5/7/21249966/apple-13-inch-macbook-pro-magic-keyboard-performance-video-premiere-affinity

By all accounts Dell's XPS13 2-in-1 does allow a stable 25W, so the reported difference is far greater than reasonable. Maybe there's a difference in the way Premiere is compiled between Windows and MacOS, anyhow a single app that's not even a benchmark is not a reliable indicator.