Intel Tries To Conquer Mobile Market By Investing In Chinese Chip Makers

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

milkod2001

Distinguished
Apr 20, 2010
231
0
18,710
14
@amk-aka-Phantom

''Good point but thing is: There's no need to run x86 desktop app on smartphone/tablet. Nobody is asking for it.


What nonsense! If by "nobody" you mean "you and people you know", maybe yes, but my next purchase is an x86 tablet, maybe a Surface 4 Pro (3 can GTFO with Haswell, Broadwell is almost here). I'll dual-boot it with Ubuntu (yes, it's possible) and use Windows at home for entertainment and Ubuntu at work. It will be a huge step up from Android tablets - in terms of entertainment I'll be able to play many Windows strategies (great on touch screen bla bla''

lol,



by nobody i mean market, majority of customers.

1) tablets and phones are for consummation of content(audio, video, internet, Facebook, Twitter and other crap). For all that Android/iOS ARM based tables/phones are more then good enough. That's the majority of customers.

2) then for creation of content, coding, calculations etc for all that there is Windows and Intel based laptops,desktop etc. x86 is here to stay for next 10 years at least.

3) i believe there's no market for mostly overpriced Intel chips / bloated Windows based mobile
devices. Intel + MS combo does not work. Both company are just too greedy, slow and ignorant.

4) you are going to dual boot your non existing Surface PRO 4 and play windows strategies with touch? haha good one.





 

milkod2001

Distinguished
Apr 20, 2010
231
0
18,710
14
@amk-aka-Phantom

''Good point but thing is: There's no need to run x86 desktop app on smartphone/tablet. Nobody is asking for it.


What nonsense! If by "nobody" you mean "you and people you know", maybe yes, but my next purchase is an x86 tablet, maybe a Surface 4 Pro (3 can GTFO with Haswell, Broadwell is almost here). I'll dual-boot it with Ubuntu (yes, it's possible) and use Windows at home for entertainment and Ubuntu at work. It will be a huge step up from Android tablets - in terms of entertainment I'll be able to play many Windows strategies (great on touch screen bla bla''

lol,



by nobody i mean market, majority of customers.

1) tablets and phones are for consummation of content(audio, video, internet, Facebook, Twitter and other crap). For all that Android/iOS ARM based tables/phones are more then good enough. That's the majority of customers.

2) then for creation of content, coding, calculations etc for all that there is Windows and Intel based laptops,desktop etc. x86 is here to stay for next 10 years at least.

3) i believe there's no market for mostly overpriced Intel chips / bloated Windows based mobile
devices. Intel + MS combo does not work. Both company are just too greedy, slow and ignorant.

4) you are going to dual boot your non existing Surface PRO 4 and play windows strategies with touch? haha good one.





 

milkod2001

Distinguished
Apr 20, 2010
231
0
18,710
14
@amk-aka-Phantom

''Good point but thing is: There's no need to run x86 desktop app on smartphone/tablet. Nobody is asking for it.


What nonsense! If by "nobody" you mean "you and people you know", maybe yes, but my next purchase is an x86 tablet, maybe a Surface 4 Pro (3 can GTFO with Haswell, Broadwell is almost here). I'll dual-boot it with Ubuntu (yes, it's possible) and use Windows at home for entertainment and Ubuntu at work. It will be a huge step up from Android tablets - in terms of entertainment I'll be able to play many Windows strategies (great on touch screen bla bla''

lol,



by nobody i mean market, majority of customers.

1) tablets and phones are for consummation of content(audio, video, internet, Facebook, Twitter and other crap). For all that Android/iOS ARM based tables/phones are more then good enough. That's the majority of customers.

2) then for creation of content, coding, calculations etc for all that there is Windows and Intel based laptops,desktop etc. x86 is here to stay for next 10 years at least.

3) i believe there's no market for mostly overpriced Intel chips / bloated Windows based mobile
devices. Intel + MS combo does not work. Both company are just too greedy, slow and ignorant.

4) you are going to dual boot your non existing Surface PRO 4 and play windows strategies with touch? haha good one.





 

amk-aka-Phantom

Distinguished
Mar 10, 2011
3,004
0
20,860
31
How good are those Intel benchmarks running ARM code?

Suddenly not so good, huh?
What "ARM code"? Applications for Android are built in Java which is cross-platform and Intel chips run them just fine, if not better than most ARM counterparts. Go read a Notebookcheck review of any such tablet and skip to benchmarks. The only apps that don't work are games that have parts of them coded in native C specifically for ARM, but most smart devs now simply push the correct version of the game to your device to avoid that. By the way, they are forced to do that even with different ARM chips, because the differences are apparently significant between Mali, PowerVR and Adreno graphics - when was the last time we had to do that on x86 systems? DirectX took care of such BS.

by nobody i mean market, majority of customers.

1) tablets and phones are for consummation of content(audio, video, internet, Facebook, Twitter and other crap). For all that Android/iOS ARM based tables/phones are more then good enough. That's the majority of customers.

2) then for creation of content, coding, calculations etc for all that there is Windows and Intel based laptops,desktop etc. x86 is here to stay for next 10 years at least.

3) i believe there's no market for mostly overpriced Intel chips / bloated Windows based mobile
devices. Intel + MS combo does not work. Both company are just too greedy, slow and ignorant.

4) you are going to dual boot your non existing Surface PRO 4 and play windows strategies with touch? haha good one.
Majority of customers, yes. Any idea how many people live in Asia? More than in the rest of the countries combined. That's "majority of customers". And this is where Intel's stuff with Windows 8 is aiming for. Tons of people are already buying x86 tabs, pleased with how much better they are than most Android ones even for media consumption. On Android and iOS, things like USB and HDMI are still exotics that require special adapters and support... on Windows it's natural. Plus you can run all the Windows applications. I only see advantages. I don't really care that Intel or MS are greedy and neither do most customers. Best offer wins, idealism loses. If anyone's slow here, it's ARM, which is only now hurriedly pumping out 64-bit chips because Google wants Android to be 64-bit to match iOS - where were they before?
For the rest:

1) I agree that ARM is "good enough" but that doesn't mean cheap x86 won't be, either. See why above.

3) You believe != reality. See why above.

4) And why not? Surface 3 Pro dual-boots Ubuntu out of the box. I played Heroes of Might and Magic 5 on my friend's Windows tablet and it was awesome, great entertainment on the road instead of these pathetic ports of shooters controlled with thumbs that riddle Android "gaming". Civilization 5 even has a special touch-optimized version - I own it on Steam, first game I'll load on the new tablet, whatever it'll be. And there surely will be a Surface 4 Pro - Broadwell is going to bring so many power consumption improvements that it's probably already in the works, MS won't pass that up. All this combined, I don't really see what's so funny. It's something that none of ARM devices can do and I and many others will enjoy it a lot.
 

milkod2001

Distinguished
Apr 20, 2010
231
0
18,710
14
@amk-aka-Phantom

i don't know where this idea 'people living in Asia want Windows tablets' comes from but im not buying it.

Think about price and customers thinking when buying products:

For 300 bucks or even much less you can get Android based tablet with some nice 1080p IPS panel.

For the same money you CAN NOT get Windows based tablet with same 1080p screen. What you 'll get is some tablet with crap 720p TN screen and brutally under powered Intel CPU and crappy GPU. Add bloated Windows into that, your storage will become very limited as Windows takes quite a good bit of that plus Windows experience on such weak hardware is not exactly great experience. I don't think even for consumption(video, audio) windows tablet can offer some advantages if any.

Now what device you think most customers will get?

Then if you want more snappy Windows/Intel device you might get something with Intel Core M. Do you have any idea how much is Intel planning to charge for 1? Around 300 bucks a piece(CPU only), that's the price of entire Android tablet. Again even if Core M perform much better then any ARM chip, in order to be used in mobile devices(fanless) its frequency is very low and in Windows environment it might not feel as the snappiest device ever.

I read somewhere that if Surface Pro will be upgraded with Intel Core M it actually will be downgrade from CPU standpoint. then again Surface Pro(SP) is not exactly something MS is hitting records in sales with. im not saying it's a bad device but its price is out of reach of majority(especially in ASIA).

Don't get me wrong, i do like idea of Windows tablets but like i said MS + Intel combo does not make it to work plus alternative from Android + ARM just runs circles around it from price and user experience stand point.
 

Jaroslav Jandek

Honorable
Jan 13, 2014
103
0
10,680
0

The cheap Atom tablets have IPS screens (T100, Venue Pro, etc.) and in case of T100, you get a pretty nice keyboard too (that's what I am writing this post on). The CPU is anything but underpowered - higher benchmark scores than any current ARM SoC (see anand, engadget, etc. - slightly faster than both Tegra 4 and Snap 8xx). The GPU is slightly weaker, not sure if crappy is the right word, though (e.g. my T100 runs Starcraft 2 @30+ fps).


Really not an issue. What annoys me the most is actually the storage subsystem's performance - you are better off using a network drive anyway.


Let me guess, you don't own a Windows tablet (or 2-in-1)? I have 2 droid tablets that are now mostly catching dust because of the T100. Borrow one from a friend or something...

I have also done some (imprecise) power measurements on my ex-phone Nexus 5 (S800) vs. T100 (Z3740) - both on 1080p screens running some games: Nexus 5 lasted about an hour consuming ~11W and T100 slightly over 7 hours consuming ~6.5W (the whole device). I admit these amateur benchmarks were anything but precise, but you get the picture with a large margin of error (and you can try it yourself too).
 

somebodyspecial

Honorable
Sep 20, 2012
1,459
0
11,310
20


Let me know when the INTEL mobile division stops selling 1.15B mobile chip for 50mil in revenue (IE, losing 1.1Billion per quarter, that is REVENUE, not 50mil profit...). You are GIVING your chips away here. If I have to give away every chip I make for mobile just to get someone to use it, I call that losing the race. I'm not an Intel hater, I want them to buy NV and take over...LOL. Either that or ARM will take more share and weaken one of our best tech companies. If NV is a big winner in that, maybe it doesn't matter much, but if it ends up being Samsung gaining in the end, that isn't a USA company. So to me, Intel buying NV is the winning move here. Nobody on Arm's side could take out an Intel/NV combo punch. You don't seem to read financial reports/quarterly reports much. You also don't seem to understand Intel is now losing 4Billion+ per year if the current rate keeps up and it looks like it will (probably approach a 4.5Billion loss on mobile this year).

Explain to me how 1.1B loss per quarter on your mobile division is good please...LOL. That's not hate for Intel, it's just seeing reality for exactly what it is. Thanks for playing.
 

somebodyspecial

Honorable
Sep 20, 2012
1,459
0
11,310
20


What atom tablet has a faster score than K1 in gpu? Atom isn't facing T4, it's facing (and getting killed by) K1. See anandtech, toms etc review of K1 tablet. T100 got killed.
http://anandtech.com/show/8296/the-nvidia-shield-tablet-review/5
1/3rd the score of K1 for T100 and even worse in most benchmarks (IE, 1/4 of K1 in TrexHD offscreen). Atom sucks ;) Intel's next chip will be facing a 20nm version probably in H1 2015 and it will have a Denver core or better in it too. Those will start to take some low end desktops no doubt, just as they have taken 21% of the ENTIRE notebook market.

http://anandtech.com/show/8296/the-nvidia-shield-tablet-review/4
In cpu perf, it only tied (557 to 558) once benchmark and lost all others by far vs. K1. So even the cpu side sucks for now. Also note that webprxt benchmark is intel work, so it's no surprise they tie that one and lose all others (lost kracken, sunspider & Octane by far). Not sure why you're comparing to T4 here when the competition is now K1 and there are many devices on the way (next month a new nexus) etc. K1 lasts 11-13hrs in a chromebook so I don't see Intel as power sipping vs. NV here or that couldn't be done (that 11-13hrs is more than Intel versions by the same maker - Acer).
 

anthony8989

Distinguished
Feb 2, 2013
653
2
19,165
57


Broadwell can't save Intel in the mobile market. I feel like you're overestimating its capabilities and perhaps ignoring its premium price .

"Broadwell parts in general will certainly be faster/better than the current Haswell parts – Intel doesn't typically "go backwards" on processor updates – but you shouldn't expect twice the performance at the same power. Instead, Broadwell-Y should offer better performance than Haswell-Y using much less power, but if you reduce total power use by 2X you could increase performance by 5% and still claim a doubling of performance per Watt. And that's basically what Intel is doing here. Intel estimates the core Broadwell architecture to be around 5% faster than Haswell at the same clocks; specifically, IPC (Instructions Per Cycle) are up ~5% on average. Similarly, changes and improvements to the graphics portion of the processor should deliver more performance at a lower power draw. Add in slightly higher clock speeds and you get a faster part than last generation that uses less power. These are all good improvements, but ultimately it comes down to the final user experience and the cost." http://www.anandtech.com/show/8358/short-bytes-intels-core-m-and-broadwell-y-soc
 

Jaroslav Jandek

Honorable
Jan 13, 2014
103
0
10,680
0

Everything got killed by K1 in GPU performance :) and I totally forgot to consider K1, good point. btw. I said the GPU is slower...


Nope. FYI WebXPRT is from ex-eTesting guys. And it is actually measuring real user-visible performance (HTML5+Javascript). It isn't "just" a javascript benchmark like Octane and the others. Although the performance of K1 in Octane is surprising - I suspect they stealth-added (or I missed it) compute support to their LLVM jitter (which would explain the performance boost over the other ARMs, since it shouldn't actually be much faster by itself).


My T100 lasts 11 hours (after disabling some services like SQL Server and setting power saving to the max.) for Office / browsing / small Visual Studio projects.
My point with the power measurements was gaming, though. The Shield lasts 2 hours, the T100 lasts 7 (not that surprising given the performance, but still, I prefer endurance over performance in a mobile device).
Although I am considering buying the Shield tablet for kids (if only to mess around with it myself).

Also can't wait for K1 and Broadwell-Y comparison, even though those are in very different price ranges.
 

Sakkura

Illustrious

Haswell is already far ahead of any mobile CPU architecture, and they're talking about Broadwell being slightly faster - while driving power consumption way down. That is exactly what will save Intel when it comes to high-performance mobile devices. Cooperating with these Chinese chip makers is apparently an important part of their plan for the budget segment.
 

anthony8989

Distinguished
Feb 2, 2013
653
2
19,165
57


Well you've lost me now. This Chinese deal has almost nothing to do with Broadwell-Y , I'm not sure why it was brought it up in this thread. Intel is trying to push its Atom processors onto the low-end Chinese SoC market. And the strategy here is that Chinese manufacturers can bring Atom equipped units into a lower price point by cheapening out the other components.

The reason it's a bad strategy is ARM is getting ready to ship out 20nm chips at roughly the same time, and already has a very competitive lineup of products representing all price points as it is.
 

anthony8989

Distinguished
Feb 2, 2013
653
2
19,165
57
Mhm Broadwell-Y notwithstanding, Atoms are inherently inferior products competing in an already saturated market dominated by ARM SoCs. I don't see this strategy paying off in the long or short term. May be I'm just cynical.
 

somebodyspecial

Honorable
Sep 20, 2012
1,459
0
11,310
20


Shield doesn't last two hours. It only does that when you MAX the gpu at 100% using a benchmark. When limiting the games to 30fps (which is still faster than the rest of the field) it nets much better battery life as anandtech's 2nd follow up revealed.
http://anandtech.com/show/8329/revisiting-shield-tablet-gaming-ux-and-battery-life
You can clearly see it doubles the life when used like you would in a game today. 4.35hrs vs. 5.3hrs for regular shield from last year.

"By capping T-Rex to 30 FPS, the SHIELD Tablet actually comes quite close to the battery life delivered by SHIELD Portable with significantly more performance. The SHIELD Portable also needed a larger 28.8 WHr battery and a smaller, lower power 5" display in order to achieve its extra runtime. It's clear that the new Kepler GPU architecture, improved CPU, and 28HPm process are enabling much better experiences compared to what we see on SHIELD Portable with Tegra 4."

So give the tablet the same battery as shield handheld and it would be BEATING that unit in battery. You're not understanding what is happening here ;) Larger battery+smaller display by far, and barely beating Shield Tablet, while LOSING in perf. K1 is a lot better than you're realizing.

"After all, Tegra K1 delivers immense amounts of performance when necessary, but manages to sustain low temperatures and long battery life when it it isn't. More importantly, it's important to keep in mind that the Kepler GPU in Tegra K1 was designed for desktop and laptop use first. The Maxwell GPU in NVIDIA's Erista SoC is the first to be designed to target mobile devices first. That's when things get really interesting."

That said, I can't wait for M1...LOL. A 20nm M1 actually made for mobile first, should be interesting indeed.

RE: Webxprt
What I meant was Intel had a hand in optimizing webxprt. It is why they use it in their marketing material.
http://www.intel.com/newsroom/kits/idf/2014_spring/pdfs/2014_idf_prc-eul_english.pdf
They don't do that for nothing ;)

http://vr-zone.com/articles/nothing-redeeming-intel-mwc-2014/72692.html
"Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Intel is a sponsor and member of the BenchmarkXPRT Development Community, and was the major developer of the XPRT family of benchmarks."

That is INTEL's fineprint. INTEL was the major DEVELOPER (and they sponsor it...LOL) of this benchmark, hence my comment. I throw out results in my brain when reading reviews with this benchmark in them. :) VRZone is pointing it out for a reason too.

"But as discussed before on VR-Zone, WebXPRT 2013 and the benchmarking methodology used by Intel is problematic and shouldn’t be considered entirely objective. In fact, in the fine print of the press release has the following clause"

You can click the link to read about their impressions of the test also in that first link.
http://vr-zone.com/articles/intel-bay-trail-vs-arm/62551.html?utm_source=rss
Here's the link...Discussing how Intel wants people to use it because it's friendly to THEIR chips. ;) Perhaps you just missed all this. Not hard to do considering the size of the web. None of us can read it all ;)
 

Jaroslav Jandek

Honorable
Jan 13, 2014
103
0
10,680
0

That's much better. And if you can reach at least 6 hours by lowering quality, that would be fine for me (which is something I will test if I buy one). Ideally, I want something that can reach 8 hours of light gaming and 14 hours of normal use on battery (not even the T100 can do that, though - using a mobile charger for that, which sucks), preferably on Windows (for non-gaming use) and at least the performance of Z3740.


On PCs, yes (obviously).
WebXPRT is a browser-based benchmark, it depends on optimizations the browser makers have done (eg. Microsoft, Google, Mozilla, ...). It is practically impossible to optimize that benchmark for Intel CPUs :)
Same goes for the Android XPRT suite (it is optimized for ARM).

If this was a test of PC CPUs (AMD vs. Intel), yeah, that would be bad, but really, some journalists need to start using their brains if they claim this affects Android or Web...
 

somebodyspecial

Honorable
Sep 20, 2012
1,459
0
11,310
20


You do realize I quoted INTEL's own fine print right?:
"Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Intel is a sponsor and member of the BenchmarkXPRT Development Community, and was the major developer of the XPRT family of benchmarks."

They are quoting this for android devices, not just for PC's. They use this in the atom tablet etc benchmarks. IF you don't have a problem with the last part (they are the major developer of the benchmark), well, I don't even know what to say. ;) You should also be aware Intel (heck everyone probably does this) works closely with all of these guys on opmizations. IE w/google on ChromeOS opmtizations:
http://www.theregister.co.uk/2013/09/11/untel_reveal_haswell_based_chrome_os_devices/
"Fisher said that Intel is working closely with Google on Chrome OS optimizations for Intel chippery. "We're optimizing the kernel," he said. "We're optimizing drivers. We're working in WebKit and [Google's WebKit fork] Blink to optimize that experience – the browser. All aspects of the platform we're optimizing, to ensure that you get the best performance on Intel."

Just an example. It took a full 10sec to google that, and probably the same to find data on chrome browser etc. I think optimizing for it is not only possible (when you wrote most of the code even easier), it's a foregone conclusion it's getting done IMHO. But I guess we'll agree to disagree here. :) Why else is this the only CPU benchmark they won vs. K1? Odd, or just simply obvious what is happening? I say obvious :)
 

Jaroslav Jandek

Honorable
Jan 13, 2014
103
0
10,680
0
Yes.


The quote "Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors." is used in all documents where Intel shows benchmarks of anything - even server stuff...

Actually, from the same quote, they say they WERE (past tense) a major developer - now they contribute and sponsor - likely same as everyone else.

When the Android benchmarks were concieved, there were no Bay Trail CPUs to run it. So, clearly, it was not optimized for Bay Trail, it couldn't be. You can bet it will be optimized now, when Bay Trail tablets start showing up - that I can agree on (but it will also be optimized for ARM, as it is pretty much the only platform it was running one before this).

This is also irrelevant, because there are no atom tablets running Android in the tests you provided (T100 is a Windows tablet).

The javascript has a standard called ECMA script. This is what browsers conform to - no low level optimization that actually leverage one CPU over the other - like GL extensions for OpenGL - like using Array.fastSortIntel - nothing like that exists, you only have Array.sort, which is, for example, implemented as a MergeSort by Mozilla (there are some non-standard browser-specific functions - nothing to do with the CPU, though). So no, you CANNOT optimize javascript benchmark for one CPU over the other. Unless you test like one javascript function that runs exceptionally well on Intel in a specific browser - which WebXPRT does not do - because it would be impossible to implement this way...
TL;DR;: you can trust WebXPRT, even if you are skeptical about the other benchmarks in the same family.


No, you are quite simply wrong. It would really help if you actually knew what you are talking about. By the way you can Google me just as easily and see who I am, that should clear up a few things...

You may want to re-read my previous posts. I have explained why. Apparently it is not that obvious, after all :)
 

somebodyspecial

Honorable
Sep 20, 2012
1,459
0
11,310
20
Rather than re-quote the whole post. I have a problem with them being the original devs, no matter what they're doing with it now (reminds me of sysmark oh so many years ago on toms, with Van Smith getting booted over reporting it). Oh and back then they did do things repeatedly (excel operations repeated etc) to show their strengths vs. AMD. So forgive me if I have my doubts about their innocence here when they DEVELOPED the code and they don't win any other benchmarks.

Also, apps are optimized in many cases LONG before the chips exist in a buy-able product. IE, AndroidL has been made on Denver even though you can't buy one until Nov in the form of a Nexus 9. I guess you've never had an EE sample. These chips are seeded to major devs (usually slower than release versions, maybe a few features off etc) so they can start work on new stuff all the time. Nothing new here.

Who you are (and I know who you are, so what you can google me also, here, anandtech as thejian, nobodyspecial on stock sites etc) doesn't change the fact that Intel wrote the code to begin with. I would have the same problem with AMD or Nvidia being the major developer of a benchmark (even claiming they are "NOW" just contributors like everyone else). It doesn't change the fact that they were talking about ANDROID devices, regardless of the T100 example and what it runs. If you say it's a browser based benchmark what is the point of telling me T100 ran windows?

I'm not alone in thinking Intel is likely being shady here as shown by link before. Call VRZone etc and let them all know you (think) are right and they are wrong ;)
 

Jaroslav Jandek

Honorable
Jan 13, 2014
103
0
10,680
0

You still don't get my point. As I repeatedly stated, browser-based benchmarks CANNOT be optimized for a specific CPU. The only thing that can be optimized is the browser itself - it still has to provide more-or-less the same javascript functionality as all the other browsers, though.
What you also don't get and I have mentioned is that WebXPRT is actually the only benchmark that does not measure pure javascript performance - that is what makes it a realistic benchmark, since web applications are rarely bottlenecked by javascript performance (unless they are doing something crazy, majority of them doesn't).
For example, we are developing a single-page Angular application (with ng-views - very javascript-heavy) and if you run it with a bit over-the-board settings, after data is fetched, processing takes a few milliseconds and then, the system waits for the browser's rendering engine to actually finish (you won't notice on a PC, but on a mobile SoC...)...

Again, irrelevant. Maybe you don't know the difference between a regular application and a browser-based one. Regular applications can be easily optimized (unless they are running inside a VM, in that case it is the VM's job to optimize for specific CPUs/instruction sets).

Because it looked like you were claiming the benchmarks were run on the XPRT Android, which T100 cannot even run. If I misunderstood, you can safely ignore my response :)

Oh, Intel is definitely shady, just as much as everyone else :). Pretty much everyone shows only the benchmarks they perform well at. Last ARM vs. Intel bench (from ARM) I saw ran a fully ARM-optimized native game (in NDK) vs. the Atom running under VM (Dalvik) and IIRC I was not the only who pointed that out.
Also, I read the VRZone article. Sam Reynolds talks about XPRT platform being friendly to Intel - I don't disagree with that (thanks to you for providing the XPRT link). Funny thing is that he mentions that it is a browser-based benchmark which makes it inherently not capable of optimizations - you can easily confirm this by asking any experienced developer (I am sure you have some friends like that?) - obviously not me, since even though you know who I am (over 2 decades of experience as a developer), you still argue about something you clearly don't understand.
But he has a good point with saying that there’s the possibility of wide variances between browsers, which is of course true and maybe also one of the reasons why Intel was doing well in that benchmark - which would just mean that IE?+Windows+Intel > Chrome+Android+ARM with regards to actual browsing performance. But again, this has nothing to do with optimizations for a specific CPU.
Anyway, no point in correcting him on a year-old article - that's why I like you can PM the article authors directly here on Tom's.

Edit: looks like the boost for T100 in WebXPRT might be caused by crappy support for rendering HW acceleration in mobile WebView (poked around in Android - WebView's canvas is still not hardware-accelerated, at least not by default). By the way, Intel has also enabled HW acceleration with its Android kernel and fork (experimental). If all goes well, we'll see these changes with upcoming Intel-based devices (and Android L), which means with regards to browsing, those will be ahead of ARM-based ones.
 

somebodyspecial

Honorable
Sep 20, 2012
1,459
0
11,310
20


Google controls both the OS & the Browser (same for others, MS with Win/IE, apple). Intel can work with them to make these two points better optimized for the cpu, thus affecting the results with their fav benchmark (made easier by knowing exactly what goes on in said benchmark). Reduced to two sentences so you get what I mean. That was the point of showing Google and Intel's love-fest (as it was called in an article). They can work together to improve results of Intel's cpus (anyone can do this of course). This probably isn't as nefarious as bapco crap, but still...

I wasn't knocking your dev skills, I just don't think you are getting that no matter which point it is you say has to be optimized here (OS or browser or both, to affect results), I'm saying INTEL can get that done (with google, apple or MS, they all own the OS and their respective browser that runs on them). Intel even claimed they have (IIRC) 700 engineers working on optimizations here. How they spread that out is anyone's guess, but you get the point. It's aimed at the browsers and their respective OS's. Everyone will do this, but I'd have to think there is in inherent advantage to knowing exactly what is in the benchmark's code, and where an OS or Browser optimization should be done to affect the results. That is my point. We seem to be arguing semantics here. I would optimize the OS/Browser to affect the benchmark just like everyone else, but I have an advantage if I wrote the code at the center. Which you can take any way you want, but to me it means I have an easier time than the competition figuring out how to win in said benchmark. I don't have to be a 20yr dev or a rocket scientist to realize that. That isn't a comment on you, but rather a comment on how many ways there are to skin a cat while making most think I didn't skin it at all (because I'm just a contributor now) ;)

RE Android L: I'm sure with Google putting out Nexus 9 on AndroidL, they will be working hard with NV (qcom etc) to get the same hardware acceleration done if it isn't already done (we'll know in a few weeks I guess). I was kind of surprised to see a 64bit K1 in Nexus 9 in time for a Nov launch, so won't be surprised by an updated kernel for ARM. At worst we're talking a software patch here for most probably regarding new socs launching from here on I'd guess.
 

Jaroslav Jandek

Honorable
Jan 13, 2014
103
0
10,680
0
Not really. There are several layers of abstraction between the benchmark and native code. The web framework (javascript/css) is fairly simple and WebXPRT is a broad benchmark.
And to say someone would optimize the whole VM for a single benchmark application does not really make sense - not to mention that would also increase performance for every application - which is, among other things, the advantage of VMs.

Or in another words: when you optimize the VM, all applications (running on it) instantly benefit. In contrast - with native applications - you have to optimize yourself (assembly, platform-specific libraries, ... - aside from the compiler's general optimizations).

FYI: everyone can see the code of the benchmark, you can even debug and profile the code...



An unofficial ARMv8-A Android kernel and emulator has been available for months, so no surprise there.
 
Status
Not open for further replies.

ASK THE COMMUNITY