Hifihedgehog
(Hifihedgehog - Waiting for Surface Pro 10!)
224
That’s the hitch. VPN drivers have to be recompiled by each of the providers for ARM, and there is currently no emulation layer available. However, in theory, it should be quite feasible since it operates within the OS space only and does not directly interface and control hardware like a typical driver would. There is still tepid developer resolve towards Windows on ARM, so it would be a huge boon if Microsoft flexed their software muscles for once and offered an emulation layer for the VPN component.
The several major drivers that are currently missing from the mix:
Apple Mobile Device USB Driver
Many external webcam and video capture devices
Custom audio interfaces, which are actually increasingly common thanks to the increased popularity of hi-fi dongles
Unified keyboard and mouse dongle control drivers (clarification: mouse and keyboard input itself will still work but custom macros flashing, scroll speed adjustment, lighting, and other features that require a custom driver to interface with the device controllers are widely unsupported; further, pairing and sync functions that require initiation from the host computer and cannot occur via a hardware button on the dongle also do not work)
External hubs and docks may or may not have working wired Ethernet depending on if the Ethernet chipsets on them have ARM64 drivers. I can at least happily report that Microsoft’s Surface Docks work well.
This is a mere smattering of issues, but the current driver situation is an absolute mess due to reluctance from hardware makers after getting burned in the past. From all the performance indications I see, there is good reason for hardware companies to eventually compile for ARM64, but they will be playing it safe and cautiously waiting first for customer bases to form before compiling, validating, and publishing.
I’m failing on the point of my message - I just don’t see how you maintain 100% compatibility between Intel and ARM within the same OS environment. Aren’t you asking developers to write to two different architectures for that “native app” performance?
Eltos is right and it’s also about getting devs to write to Windows APIS and frameworks which are common now and have been since the second annual cumulative update to Windows 11, way back when.
One of the ongoing legacies of Windows generally has been the ability/tendency for developers to “roll their own” versus using standard windows libraries and function calls.
There was justifiable reasons to do this in the XP and even early Windows 7 days as technology often was advancing faster than Windows updates were being released.
But in the last few years I’ve been part of meetings where this or that dev is complaining about this or that bug or lack of functionality in Windows only to have MS respond with “are you aware of API or framework xxx that does exactly what you are asking?”
Even a company as niche as ours has gotten a handful of functionality adds (mostly around power management) to the Windows Libraries and APIs.
I mean after all, the wild west that was Windows development was the original core reason behind the albatross that is the registry.
It’s a failing that Apple long ago left behind, out of necessity when they switched from PowerPC and Intel and which benefited them greatly with the switch to ARM as well.
PS: There is still a “registry” in the newest WOA dev builds, but it’s read only and has to be explicitly unlocked by a developer to make use of it though MS strongly encourages devs to use their standard and now cross architecture resources instead.
PPS: To reference something that @Hifihedgehog and I were talking about above, that is also one of the sources of contention for both VPNs and printer drivers as well where devs want to use what they want to use instead of the existing, provided tools.
Hifihedgehog
(Hifihedgehog - Waiting for Surface Pro 10!)
230
Yes, per Qualcomm, TDP can be tuned down to as little as a few watts, much like Core M/Y-class brought to the table. The advantage is it would be significantly more performant than either AMD or Intel can pull off given its far higher efficiency across its power/clock speed curve.
I have yet to see anything from Intel or AMD that can pull off a ~1000 multicore score in Cinebench 2024 at 23 watts (note the highlighted 997 under “Thing and Light-Observed” below). The only thing from those two that comes in a distant third is AMD’s Ryzen 7 8700G, which has a 65W TDP. Qualcomm Snapdragon X’s sole competitor on that front is Apple with the 12-core M3 Pro, which has a 27-watt TDP. So Snapdragon X Elite efficiency is basically one-to-one with Apple’s latest M3 chips, and on an older 4nm process compared to Apple M3 series’ more efficient 3nm process, no less.
@Hifihedgehog get us an 11" Surface Pro 10 TOMORROW - in those most recent benchmarks the QC was 2.75x single core and 4.1x multi-core over my SP8 i5. Even matched my M1 Max toe for toe!
Seriously - if QC/MS could push out an iPad Pro 11 size Surface Pro in the 1500/10000 range and day long battery, Windows would sweep the field…
Wouldn’t be a clean sweep unless they trade out MPP for Wacom. I’ll stick with my IPP until that happens. Or until Samsung makes their own Windows Surface like clone with Wacom.
Yeah, something like that. I’d love an 11" version. I’d even use the dumb kickstand for Wacom with all day battery.
2 Likes
Hifihedgehog
(Hifihedgehog - Waiting for Surface Pro 10!)
235
That is why I am thinking they are playing the game of sitting and waiting while the technology simmers into a next-gen homerun. Unlike the R&D slug that is Apple’s chip team these days, like the rest of the industry, Snapdragon X Elite’s next-gen improvement will likely have >10% IPC gains over first-gen Snapdragon X series, and then it will not just be them telling them “well, we beat Apple M3, and we match them more or less in efficiency.” When Nuvia was formed, there was a serious case of brain drain from Apple’s microarchitectural development staff. What they want is to say, we do not just meet, but beat Apple all across the board in every single metric, top to bottom. The efficiency should be so good that they will totally trounce them and can say confidently, “If you like performance, don’t be stupid and overpay. Don’t buy Apple. Buy Snapdragon X series.”
I wonder if they will put these in Android phones? Wouldn’t that make them as performant as iPhones?
Hifihedgehog
(Hifihedgehog - Waiting for Surface Pro 10!)
237
Funny you should ask! They are promising optimized Oyron cores, now at 3nm, in Snapdragon 8 Gen 4. Today’s big rumor drop (posted some hours ago) is 3500 single core score in Geekbench 6, along with an unspecified sizeable lead in multicore and graphics performance over Apple’s next-gen A18. The revenge of Android is on the cusp of reality.
Hifihedgehog
(Hifihedgehog - Waiting for Surface Pro 10!)
240
I’ll agree there wholeheartedly, but give developers an audience and a stage worthy of an illustrious performance and processor chops will set the stage for a show worth the price of admission.
It has never been about performance for Android apps. Even with the outrageous fees and restrictions, developers make way more money (or at least believe they do) on their apps on IOS than on Android. There are lots of examples of apps on Apple that cost $3-4 or more when they are completely free (and add supported) on Android. I’ve known plenty of new developers over the years who only wanted to learn Swift for mobile apps, and had no desire to learn Java or Kotlin simply because they thought they would make more money with Apple. It may just be perception at this point, but it’s where developers have been for awhile. All processors have been able to realistically outpace what we actually do with mobile apps for quite some time. It’s the question on how to get developers engaged with Android.
