Linked by Thom Holwerda on Wed 4th Jul 2018 22:30 UTC
Windows

A couple of months ago, it was reported that Microsoft will be launching a cheaper Surface tablet. According to the original report, it was going to include an Intel Core M processor, also known as the Y-series. As we noted at the time, this didn't make sense, given the $281 price point for a Core m3 and the fact that it's supposed to go into a $399 tablet. It would probably be the most inexpensive Core M device ever.

But according to a report from WinFuture, the $399 tablet will include Intel's Pentium CPUs, and that makes a lot more sense. The base model will have a Pentium Silver N5000, which is a quad-core, 32-bit 'Gemini Lake' processor that's clocked at 1.1GHz.

I find this absolutely puzzling. My Surface Pro 4 with its Core i5 processor isn't exactly a speedy computer, and going down to mere Pentium processors surely makes these new rumoured Surface devices even slower. On top of that, didn't Microsoft just make a whole big deal out of Windows on ARM, which would surely be a far better fit for such a cheaper Surface tablet? Or would ARM processor at these price points be even slower? Surely this device will have to be locked into using Microsoft Store applications, since classic Win32 applications will have a lot of trouble functioning properly on such processors.

If this rumour is true, these cheap Surfaces are going to deliver a terrible user experience.

Thread beginning with comment 659548
To read all comments associated with this story, please click here.
I suspect that's basically it...
by bhtooefr on Thu 5th Jul 2018 01:43 UTC
bhtooefr
Member since:
2009-02-19

...that an ARM would be much slower at *EMULATING* x86, than this thing would be at running native x86: http://browser.geekbench.com/v4/cpu/compare/8271545?baseline=873063...

That performance from the N5000 is, on single-threaded, on par with Core 2-era stuff, which is still serviceable for light use (really, disk is more important, the device needs a good SSD, instead of cheap eMMC).

And, store apps aren't the problem for ARM, most of them already exist in ARM versions AFAIK. It's Win32 apps, though, and the problem is that most people aren't buying Windows to get the Windows brand, they're buying Windows to get a Win32 runtime that just works (as opposed to WINE and ReactOS, which don't just work).

Reply Score: 3

kuiash Member since:
2018-05-21

Hi. I've done a little work in this direction recently.

Turns out, yes, you are right! Emulating x86 on ARM is slow(ish). But that only turns out to be true of benchmark/test code.

Example - An application draws some $stuff on the screen.

Depending on where the emulation kicks in this may, or may not, be slow. If the UI/OS libraries are not emulated (maybe a thunking layer) then the libraries run at "full speed". The actual drawing is handled by a device driver that is running native code and the GPU is doing the actual graphics inner loops.

Some stuff does run terribly badly - notably SIMD optimised code. Filtering and any graphics in software.

But a surprising amount doesn't even need emulating.

Conclusion - slow but not as slow as one might expect.

Reply Parent Score: 2