For example, would an Intel i5 processor for (e.g.) the Microsoft Surface Pro 4 be different from a similarly-named processor for a standard laptop?
Yes. Very much so.
To understand why, read on.
You need to look at specific model numbers and part numbers, not just generic brands or series such as "i5" in order to really understand the difference.
The simplest way is to google around until you can find a spec sheet detailed enough to tell you the exact model number of the CPU. For example "5th generation Core i5" is not specific; "Core i5-6400" is specific.
Then go to cpubenchmark.net and go to the "Searchable CPU List". Then compare the benchmark figures for the CPU and another known CPU (either a desktop CPU of the same generation, or a larger scale laptop CPU).
Intel had less variety and flexibility in their CPU lineups for laptops and smaller scale devices in the past, but the current lineup of desktop/laptop/converged device processors consists of the following product categories, in decreasing order of CPU throughput and performance (and GPU performance, if a GPU is included):
Skylake-E: Not yet released, and won't be for a while, but the "-E" moniker means this is for Enthusiasts, and is the largest-scale, highest power consumption and highest performance processor you can install in desktop motherboards. The latest ones currently released are Haswell-E, several microarchitectures old, but still faster than mainstream Skylake. No GPU.
Skylake-S: Can further be subdivided into "Performance" and "Mainstream", these are full size desktop processors that come with a GPU. They're significantly cheaper than Enthusiast processors, and plenty fast enough to play high-end games. They use significantly less power than -E processors and have fewer CPU cores. Some laptop manufacturers use these processors in very large (17"+) gaming laptops.
Skylake-H: The largest, highest performance Intel CPUs you can buy for laptops. These usually go into laptops at least in the 14-15" size class at the minimum, if not larger, because they require a larger battery to power them for an appreciable length of time.
Skylake-U: These are used in ultra light-weight laptops, 2-in-1s, and some "full-fat" Windows 10 64-bit tablets (with or without a detachable keyboard).
Skylake-Y: These are the tiniest, lowest performance, and lowest power consumption CPUs available, while still having many of the architectural features of the other Skylake CPUs. They're just scaled down significantly to allow them to physically fit in very small, battery-powered computers, like 7 and 8" tablets with a battery weighing only a few hundred grams.
Within each series there is a price range from cheapo (lower clocks, fewer CPU cores, and slower graphics chipsets) to high-end (highest clocks, most features, most capable), while generally requiring the same amount of power, on average, within the series.
Within this Skylake generation, you can see a progressive decrease in Thermal Design Power (TDP) and a proportional decrease in performance (benchmark scores) as you go from Skylake-S, to H, to U, to Y. You can commit this to memory using the following facts:
- A processor with a "K" in its model number, or without any letters at all in the 4-digit model number, is a Skylake-S.
- A processor with an "H" is a Skylake-H.
- A processor with a "U" in the model number is a Skylake-U.
- A processor with a "Y" in the model number is a Skylake-Y.
- ESHUY -- the five series sorted in descending performance/size/power requirements order. Or just SHUY for short, since the -E series is not out yet for Skylake (doesn't stop people from buying older-generation -E series processors, though).
To take an example from your question, the Surface Pro 4 uses the "U" series processor, so, objectively speaking, they ARE, in fact, going to be significantly slower (and use significantly less energy, and generate significantly less heat) than an -H, -S, or -E processor.
Let's get quantitative, shall we?
Compare one of the CPUs from the Surface Pro 4 (several CPU options are available, but all of them are Skylake-U), the i7-6650U.
Compare the i7-6650U to the i7-6770HQ. One is a Skylake-U, one is a Skylake-H. You'd think they'd be similar, on the surface, because they're both "Skylake i7 CPUs designed for laptops", but they're really quite different.
CPU Benchmark link here. (They don't have the i7-6770HQ benchmarked yet, but the i7-6700HQ is close.)
GPU Benchmark link here. (They don't have the Iris Pro 580 benchmarked yet because it's so new, so I'm linking a fairly unimpressive Nvidia GeForce that I expect is close to what Iris Pro 580 will be able to do.)
Feature || i7-6650U || i7-6770HQ
Cores || 2 || 4
Cache || 4 MB || 6 MB
TDP || 15 W || 45 W
Max RAM || 32 GB || 64 GB
GPU || Iris 540 || Iris Pro 580
Passmark Score || 4918 || A little more than 7998
GPU Score || 1446 || Approx. 2100-2200 (speculative)
Atom is a distinct microarchitecture from Intel's "main" CPU lineup. Although the latest Atom processors borrow heavily from the design of architectures like Skylake and Broadwell, they are considered to be a distinct lineup with different design goals.
I like to think of the mainline CPU lineup as being optimized to sit roughly in the middle of the power/performance/cost triangle on average, with the most expensive CPUs pushed a little more away from the cost side, and "budget" CPUs (Pentium, Celeron, i3) pushed a little more toward the cost side.
Atom, on the other hand, is always extremely cost-conscious, because it's designed to be put in devices whose entire unit cost is pretty low (or as part of a much larger product where there isn't a big budget for electronics, like an affordable car's infotainment system).
They also don't make Atom processors at a large enough scale to compete with their main line of processors, so we don't really know what the Atom lineup would look like if they took those same design decisions and tried to produce laptop or desktop parts. So, generally speaking, Atom processors focus on value, extremely low power, excellent power efficiency (good performance for the amount of watts it's drawing), and integration of an entire system's worth of microelectronics on one chip (a so-called "System on Chip" (SoC)).
If you see a smartphone or small tablet with an Intel CPU in it, it probably has Atom, not Skylake or a similar mainline CPU in it. One notable exception is the Intel Next Unit of Computing (NUC), which, while it looks like it's a tiny little box that would only have room for Atom, actually comes with a "-U" branded processor, like Skylake-U. Since the NUC is not a battery-powered device, power draw was less of a concern, so all they needed to do is make sure they can stick a heatsink/fan solution in there that would keep the processor cool.
And that's what a lot of the decision of which series to use comes down to: batteries (and cost). The energy density and voltage output of modern batteries is somewhat less than what we'd like it to be, so if you can imagine a world where our batteries are significantly better than they are now, Intel could conceivably put a higher-performance series of processors into ever-smaller form factors (I think we're a little better-off with cooling; they could probably figure out a clever way to keep these cool if they had enough energy to run the thing on batteries).