HP Spectre Folio Review Check-In: Performance

Given the questions many still have about the performance of Intel’s Y-series processors, I thought it might be helpful to take a longer look at this ahead of my formal review.

As you may recall, I wrote a similar pre-review article about the new MacBook Air. In that case, I was pleasantly surprised by its real-world performance, and I was hoping that the HP Spectre Folio would follow a similar trajectory, given that it uses a very similar chipset.

My early experience with the Folio, however, is a bit more nuanced than that. And while I’m not sure yet whether to attribute the differences to the platforms which each device runs—macOS on the MacBook Air and Windows 10 on the Folio—that’s where I’m leaning at the moment.

As a bit of background, you may recall that Intel engaged in some shenanigans as it made the shift to its so-called 8th-generation Core processors, which first launched over a year ago. And among the more innocuous changes it made during this time frame was to rebrand (most of) its reviled Core M processors as Y-series Core processors. Which can be confusing.

Core M-based PCs were notorious for their performance issues. For example, in late 2015, I received the first-generation HP Spectre x2, which was an elegant-looking Surface Pro-like portable PC. The Core M chipset it utilized allowed HP to build something that was thin, light, and fanless. But it was hobbled by performance problems. In fact, one of the strange issues with using Windows with the Core M chipset was that initial tasks—like installing applications or even running them for the first time—would often be very slow. But the act of using them, over time, was fine.

(HP switched to Core i5 or better processors for the second-generation Spectre x2, which was a dramatically better PC as a result.)

When Apple announced the new MacBook Air last month, it said only that the processor was “an 8th-generation Intel Core” CPU. But we quickly learned that this was a dual-core Y-series chipset. And not a mainstream quad-core U-series chip as we see in virtually all laptops today.

This was initially disappointing. Like the Core M chips of old, Y-series chips run much cooler than their U-series counterparts, and they consume much less power: Usually 4.5-5 watts vs. 15 watts. The advantage is that their thermal requirements are less, so systems built on Y-series chipsets are generally thinner and lighter than those built on U-series chipsets, and they can provide better battery life while not requiring a fan. So they’re silent too.

The downside to Y-series chips is that they are not as powerful as their U-series counterparts. And that’s true during normal use as well as in “Turbo” mode when the processor can crank up the speed (and the power consumption, to 10 watts) to meet whatever need arises. Y-series chips tend to have smaller and shorter performance curves, basically.

The design of the new MacBook Air was initially confusing. Despite using a Y-series part, it does have a single fan, and it can run quite loudly under load. As it turns out, Apple is using a somewhat unique Y-series part that runs at 7 watts, not 5 or less. Theoretically, such a chipset should offer better performance across the board—both in normal usage and its bursts—and based on my real-world experience, it succeeds. Through some combination of the chipset and, I think, efficiencies built into macOS, I’ve never noticed any performance lulls or weirdnesses with the new MacBook Air.

HP announced the Spectre Folio in early October, and everything about this innovative new PC was interesting. But its use of a dual-core Y-series CPU was … disappointing: I’m still stung by the performance issues from the Core M era.

In speaking with HP at that event and since, however, I was told that Y-series processors have come a long way since Core M. And that HP itself has learned how to better utilize this chipset in its PC designs. In a recent briefing, I was shown some PCMark 10 benchmark comparisons that favorably compared the Spectre Folio to other modern Y-series and U-series PCs.

As you may know, I prefer real-world experience to benchmarks. But I also see the value in apples-to-apples comparisons, and I do mix in my own tests for performance and battery life. But I was curious about HP’s use of PCMark 10 for performance as opposed to, say, Cinebench or Geekbench.

HP told me that those tests were very CPU-bound—akin to using Handbrake to encode 4K video to 1080p, a performance test that I do perform on every PC I review—and thus weren’t necessarily indicative of real-world usage. This corresponds to my general feeling about benchmarks, so I was intrigued. PCMark 10 is more balanced overall, and it includes a productivity test that uses the chipset’s Turbo Boost feature to demonstrate how a PC can perform in standard productivity tasks. You know, the way people might actually use the thing.

Given this and my subsequent testing, I’ve decided to add PCMark 10 (the overall score and the productivity score) to my testing criteria for all PC reviews going forward. But since I hadn’t been performing this test, I also decided to retroactively run the PCMark 10 benchmarks on several PCs I’ve already reviewed—or will soon review—to see how they stack up. I spent much of the weekend running those tests across several PCs.

These results are not final: Far too often, PCMark 10 tells me (after the lengthy benchmarks conclude, which is irritating) that it could not identify the GPU (or, in one case the CPU), which probably nullifies the results. But that said, I still find the results I’ve seen so far interesting. And they can perhaps help put the Spectre Folio in perspective.

So far, the best results I’ve seen came from the Lenovo ThinkPad X1 Extreme. Powered by an Intel Core i7-8750H chipset and NVIDIA GeForce GTX 1050 Ti graphics, this beast of a laptop scored 4551 overall and 6562 in the productivity benchmark. The new Intel NUC came in just behind, overall, with an overall score 4477. But it also beat the X1 Extreme in productivity, with a score of 7315.

The HP Spectre Folio scored the lowest of the five PCs I’ve tested so far, but that’s a little unfair since most of the other PCs are quad-core laptops or desktop PCs. (Actually, the X1 Extreme has a six-core processor, but whatever.) It netted 3002 overall and 5769 in productivity. But I also tested last year’s dual-core i5-based Surface Laptop, which scored 3034/5174. And that is interesting.

That is very interesting.

The Folio was only barely beat out by Surface Laptop; in fact, one might fairly rate the overall scores a tie. But it actually beat Surface Laptop in the productivity benchmark by a healthy margin.

(By the way, you might be wondering where the MacBook Air falls in these scores. Since PCMark 10 is not available for the Mac, there’s no way to know. I did run the benchmarks under Windows 10 using the Mac’s Boot Camp-based dual-boot environment. But you should ignore these scores since Apple’s Windows drivers and configuration are notably non-optimized. It scored 2201 overall and 3768 for productivity.)

Theoretically then, the Folio should perform about as well as the first-generation Surface Laptop. And to be clear, that PC is wonderful. And while I appreciate the addition of quad-core processors in the more recent Surface Laptop 2, the truth is that most users—doing “everyday” tasks like email, web browsing, word processing, and so on—would never notice any difference between the two generations. So the level of performance we see in the HP Spectre Folio should be equally good.

In my real-world usage over the past several days, however, I have in fact seen some of that weird initial performance lag that I used to experience with Core M-based PCs. It’s not as pronounced, for sure. But where the MacBook Air—again, possibly because of the efficiencies of macOS—experienced no such issues, the HP did. A little bit.

HP was upfront about this. It noted that some processor-intensive initial tasks, like syncing a large Outlook PST file for the first time, could take longer than is the case on a U-series CPU. And that if a user is interested in things like video rendering, using complex Photoshop filters, or other similar heavy-duty tasks, a U-series CPU-based PC would be the better choice.

Of course, such a PC would be thicker and heavier. And it would have a fan. The HP does not: So while the MacBook Air hissed like a pissed-off cobra during my Handbrake test and while running the PCMark 10 benchmarks, the Folio went about its business silently. When completed, it was just a little bit warm.

So what we come back to here is the same conversation about trade-offs that we would have had three years ago with the Core M-based Spectre x2. The difference is that the trade-offs have shifted into a less awkward area.

Yes, a U-series chipset does perform better. But when you combine the epic battery life, the always-on connectivity, and the stunning design of the HP Spectre Folio with what I feel is perfectly acceptable performance—in the everyday tasks that most users undertake—you arrive at something that might just make sense.

(Compare this to Surface Go, which seems unfair given the cost differential. But the comparison makes a bit more sense when you realize that both companies were attempting to solve the same compromises of performance, heat, silence, and portability. Surface Go offers anemic day-to-day performance in a too-small form factor and it gets only 4-5 hours of battery life. We can’t make these complaints about the Spectre Folio. Which, yes, is twice as expensive.)

So here’s where I’m at. Where the MacBook Air was a slam dunk from a performance perspective, the HP Spectre Folio requires a bit of time to get up to speed when you initially configure the device. But once your applications are all installed, OneDrive has synced, and everything is the way you want it, the Folio provides what can only be described as the normal, expected performance. And that’s a good thing.

I’ll make sure that remains the case during the review process.