wonderfield
Supreme [H]ardness
- Joined
- Dec 11, 2011
- Messages
- 7,396
Building an ARM OS was a wise decision. After all, having an architecture-agnostic kernel is The Right Thing. Productizing what they had wasn't a wise decision.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
It's dead, Jim: "Microsoft Surface RT price drop continues"
- Thread starter pxc
- Start date
I wouldn't lump those two together for tablet-centric apps.
I don't know the current Google situation, but from the beginning Google didn't make a distinct tablet section and relied on stretched phone apps, which has definitely hampered the creation of tablet-centric apps for Android.
OTOH Apple has had a distinct App store section for the iPad since it was released. The last official count I saw for iPad was 275 000 iPad Apps. About half the number it has for the iPhone, but still quite a lot of iPad Centric apps.
The lack of great tablet-specific apps in the Play Store is still hampering adoption of Android tablets. The way that Google has allowed the hardware ecosystem to be constructed really means there's three styles of app layouts: phone, tablet and "phablet" (devices like the Nexus 7 and the Note 2 that are too small for high-res tablet apps, but large enough to make stretched phone apps ugly and simple-looking). Google either needs to do a better job setting hardware requirements for Play-certified devices, or they need to rethink Android's presentation framework and make it easier for developers to build apps that have great usability across many screen sizes and resolutions (easier said than done).
Though both of these are better than not having any tablet apps at all, which is basically how the Surface RT launched.
The vast majority of those engineers are simply doing chip layouts of standard ARM cores, that does nothing to move ARM state of the art. Even the few who do unique architecture enhancements are simply re-inventing the wheel. Apple and Qualcomm both did chips slightly less powerful than the A15 to use a bit less battery. That was just about the biggest thing that came out of all those extra engineers. A design tweak that doesn't feed back into main ARM line. Now ARM has released their A12 design doing much the same. They didn't have common effort driving ARM forward, they had three teams re-invent the same wheel three times over.
So there are not 4 times as many engineers pushing ARM forward, there is basically just ARM engineers, with a bunch of wheel spinning outside of ARM. I am sure Intel has more engineer working on x86 architecture than ARM does on ARM architecture.
This is demonstrably false. Intel development cycle (Tick-Tock) blithely ignores the much slower development cycle of their competitors like AMD. The have rapidly iterated way beyond AMD to the point that it is almost considered impolite to compare an AMD CPU to Intel CPUs that are so far ahead.
Get a little informed. Intel showed 48 core x86 back in 2009 on 65nm, they could easily do 64 core at 22nm if they could find a use for it.
http://www.pcper.com/reviews/Proces...-Processor-Single-chip-Cloud-Computer?aid=825
Exchange Multi-Core with MegaHertz and the 1980s will call for their Myth back. This is the same kind of nonsense all over again. Multi-core scaling starts running into overhead problems very early, even when you find appropriate software.
http://scalibq.wordpress.com/2012/06/01/multi-core-and-multi-threading/
If you have 2 cores with IPC of 4, and 8 cores with IPC of 1, the total amount of theoretical power is the same, but the 2 cores will ALWAYS be faster.
Intel is at reasonable parity TODAY, which is why there are so many Intel design wins showing up at Computex. 2 years is the timeframe, for Intel to be firmly out in front. ARM may be relegated to very low end.
DeathFromBelow
Supreme [H]ardness
- Joined
- Jul 15, 2005
- Messages
- 7,315
True, I shouldn't have said that. The point I was trying to make was that Apple and Google's tablet OS's 'grew' out of their phone OS and have been successful while Microsoft seems to have shot themselves in the foot by making a specialized tablet OS.
If convergence is the key why did Microsoft split it's resources three ways between phone, RT, and Windows 8? Why bother with RT at all when they already have a good ARM OS that could easily have been adapted to larger devices? Tablets are a niche form factor for content consumption. They have their uses, but the future of the 'PC' is phones, not tablets.
I use my phone constantly throughout the day. If I could come home, plug in a single cable and have it connect to a my 1080p screen and USB mouse/keyboard and use it like a PC for email/youtube/web browsing it would be fantastic. It's already possible with newer Android devices. Microsoft should be working on that rather than diddling around with large touch screens, touch mice, touch laptops... Microsoft needs to start focusing on their phone OS rather than ticking off desktop users with touch-centric gimmicks they don't need.
All of the ARM licensees have the ability to develop completely new architectures based on the common core. This is what you should be looking for from ARM-based platforms over the next couple of years. You're also talking about teams that have been working with the architecture for only a fraction of Intel's experience with x86 and they've experienced significantly larger growth and a faster ramp-up than any of Intel's x86 work (Intel's own internal work with ARM even shows a healthier growth curve by a significant margin). Apple's only been designing ARM units in house for about 4 years now, and they're arguably better at it already than TI, who's been doing it for several times longer. There's also a couple of R&D shops out of Israel to keep an eye on that seem to be going even faster than Apple. In the short term, there's going to be some limited fragmentation of ARM as rapid developers build their own instruction sets for it, but in the long term, ARM will subsume these and roll them back into the mainline architecture. These teams "reinventing the wheel" right now are helping to build the next iterations of ARM, even if that's not the goal they set out to accomplish.
It's true that they catch up quickly and then overtake their competitors in the same spaces, but again, they let their competitors do the work of choosing where they invest in. Their new IGPs are a direct response to AMD's APUs (which AMD led the way with and invested in first) and their current low-power strategy is a direct response to ARM's prior investment in the space. Intel's only driver to invest has always been "someone else is doing it better than us. Let's stop them." They're not agile enough right now that they'd be able to face a process that iterates upon itself rather than letting the winds of the existing market determine direction. I'm not saying that ARM is this way now, but the distributed development architecture gives them the potential to be.
Cool, thanks for the share. Intel would do well to add some of their core developers to the Linux Kernel SMP team, since most of those devs have started focusing on ARM/MIPS over the last couple years.
Intel isn't even close to approaching ARM in the cost-per-unit realm yet. Performance per watt and performance per dollar are not the same. You can get Allwinner A10s for less than $7 even at low quantity, and Apple is having TSMC fab their next-gen CPUs for single-digit dollars apiece. This lower per-unit cost also gives the actual device manufacturers the ability to rapidly iterate and pivot with a lower base investment. It's great that their products are approaching competitiveness on paper, but if buyers aren't afforded the same flexibility, "parity" is still a ways out.
Either way, Microsoft doesn't seem to have paid much attention to either side of our discussion, to get the thread back on the rails. If they thought Intel would overtake ARM that quickly, why would they have even bothered with Windows RT instead of leapfrogging it completely? If they think ARM is the rising star, why release such a crippled, lackluster product based on it? It seems to me that their entire ARM tablet strategy is a flub, regardless of which way the market actually shifts.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
I think the future of the PC is all of the above. Yes, smart phones are going to be the most numerous of devices but most of them will be of the cheaper variety and not used for more complex tasks though in time I'm sure true PC phones that can dock and even run desktop software with x86 will come out.
As for larger touch screens, again I think it's part of the mix. I certainly would rather use my large screen tablet for reading and note taking. And there's even uses for touch on larger screens, not so much for vertically mounted desktop screens but they do have there uses for art, white boarding, business uses, etc.
Nope. These are proprietary efforts, they aren't rolled back into ARM.
Like most companies Intel does some mix of going it's own way, and some mix of watching competitors.
Intel started doing IGPs, even before AMD bought ATI. It is only natural that Intel would keep improving IGP, and now that they utterly dominate on desktop CPU, to direct more resources to IGP.
Multiple companies can realize where the market is going (integrated graphics) and work their own multi-year roadmaps to win those future markets. I think it has been obvious for a long time that discrete graphics are going to keep moving toward niche. This prompted Intel to work on improving it's in house IGP and AMD to buy ATI (to leapfrog Intel).
Forgot this one:
http://h18004.www1.hp.com/products/servers/options/phi/index.html
You can buy Intel 50+ (x86) core cards for use in supercomputing, they are 1+ Teraflops, and x86, so easier to program for many tasks than GPUs.
But these kinds of core counts aren't really applicable to consumer computing.
Hence my comment relegating them to the very low end of the market.
Citation needed.
My initial thoughts was RT only made sense as a prod against Intel. A threat to get them moving faster on better mobile chips. Maybe it actually did work somewhat in this respect. But it serves no purpose now.
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
I personally think ARM is going to replace AMD.
But it could go either way, competition between ARM fabricators helps speed up ARM progress, second any progress would be patented and eventually others will license it if it is effective and ARM can cross license with Samsung etc.... It's kinda like arguing that nothing AMD does gets into Intel CPUs, it is not accurate.
Right now almost all the mobile apps etc... Are optimized for ARM CPUs and regardless of which has the most power that has an effect, in order for intel to overcome that effect they need to either sell cheaper or have superior performance and thus far they have done neither. On top of that several major players such as Samsung have a large vested interest in not using intel chips. Nothing says they will not use their size and power in the mobile market to keep producing phones with ARM chips even if they are inferior, apple has done it successfully for decades.
But it could go either way, competition between ARM fabricators helps speed up ARM progress, second any progress would be patented and eventually others will license it if it is effective and ARM can cross license with Samsung etc.... It's kinda like arguing that nothing AMD does gets into Intel CPUs, it is not accurate.
Right now almost all the mobile apps etc... Are optimized for ARM CPUs and regardless of which has the most power that has an effect, in order for intel to overcome that effect they need to either sell cheaper or have superior performance and thus far they have done neither. On top of that several major players such as Samsung have a large vested interest in not using intel chips. Nothing says they will not use their size and power in the mobile market to keep producing phones with ARM chips even if they are inferior, apple has done it successfully for decades.
wonderfield
Supreme [H]ardness
- Joined
- Dec 11, 2011
- Messages
- 7,396
To add to that: http://anandtech.com/show/6993/intel-iris-pro-5200-graphics-review-core-i74950hq-tested
This article does a good job explaining why Intel does iGPUs. It's not because they're trying desperately to compete with AMD. Primarily, it's due to internal economic factors and customer demand, primarily from Apple. Apple wanted better iGPUs from Intel, so Intel starting directing more of their resources to appease their own customers.
I would think so, which makes the new Samsung Galaxy Android Tablet kind of puzzling:
Inside Samsung's Galaxy Tab 3, an Intel chip
There were several Android-Intel wins at Computex, and this is still the 5 year old Atom core.
Bay Trail is the big jump, supposed to offer up to 3X performance of Clover, or up to 1/5th power (not at same time). Bay Trail product is supposed to arrive by year end. This should be way out in front of A15.
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
Right now Samsung is growing so well I believe they cannot possibly supply all their own parts. Even their phones and even their premium phones are not flush with their own chips. But that comment was more directed at a point in time where intel is able to surpass arm on price and performance leaving a surplus of ARM capacity. Someone like Samsung would likely take advantage of that from 2 angles.
The software uptake is what drives upstream adoption of the individual proprietary instruction sets. Intel was forced to make x86-64 part of the comprehensive x86 set even though AMD developed it because software builders started to incorporate it. ARM will do the same thing as the instruction sets start to diverge, though Apple is likely to be an exception to this. Samsung, Qualcomm and TI rely enough on mainstream software to have this sort of influence. Once a proprietary instruction set with tangible benefits makes its way into something like the Linux kernel, you can expect ARM to integrate it into the next major revision, but also for other vendors to start incorporating it before ARM upstream even does.
Intel's approach to IGP was neglected garbage ("Intel Extreme Graphics") until there was a real viable alternative. Then they pushed hard and came further in 2 years than the previous 5 combined, but again, only because someone else did it first. Intel's core strategy is buit on being the "best mover" rather than the "first mover," and it's going to come back to bite them.
You don't need to be at the low end of the market to iterate quickly. You don't need to be at the low end of the market to recognize cost savings through smart (rather than raw volume) purchasing. You don't need to be at the low end of the market to spend money wisely and invest in better performance-per-dollar. It's not like Allwinner's products are at the bottom of their market segment, either, and Apple's A7 is likely to be near the top of the deck like the A6 was when it was released.
Why wouldn't they have just given the money to Intel and said "build us this?" That's basically what Apple did over 5 years ago when they decided to switch to Intel. They approached Intel with a large sum of money and said they would invest in Intel scaling up to deliver what they needed in the desired quantities. To begin with, Apple was getting exclusive access to Intel parts several months in advance of their channel release. Apple even had special models of Core 2 Duo processors made that Intel never sold to anyone else, like a P7350 with VT-x enabled. Microsoft has more than enough weight to throw around that they wouldn't need to shoot themselves in the foot and create a failed product just to get another vendor to move in a particular direction. They could have just told Intel what they needed and invested in the infrastructure required to make it. I suppose this angle might hold weight if they were hedging their bets with AMD too, but like you said earlier, they're barely viable at this point, even in the mobile space.
That is pretty thin reason. Surely they can get some third party ARM chip for less than an Intel chip for what is mainly an ARM ecosystem product.
A lot of Android software is managed software, but performance apps like games often use the Native Development kit, often only outputting Native ARM code. These either won't run on Intel, or need to be emulated.
The Intel chip must provide some quite compelling advantages to merit changing to a non-native CPU architecture.
If it is scoring wins today with the Old chip, Bay Trail is going to destroy...
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
I don't agree with this a company can very easily be a faster more powerful mover and be more effective since they do not waste resources on all the failed products. Look how they beat down AMD after the P4 fiasco. In fact just to make a point pretty much every company that exists today as a major player is not a first mover, apple, MS, google, intel, facebook they all copied and killed their competition not a single one of them ever invented or was first to any of their most popular core products. Once they get big they know they can pull it off even better through shear inertia.
Nice article. The chart illustrates exactly the rapid ramp-up I'm talking about. I think Anand is giving Apple a little more credit in the process than they deserve, but I agree for the most part. I wasn't trying to say that Intel invested in GPUs to compete with AMD (for they had little need to compete with them on any front during most of the timescale occupied by the chart) only that they had no desire to be the first ones on the scene with a good integrated GPU. They didn't get there until some other vendor had already shown the viability of the market idea, and then they just moved in and ate the market share. As the article mentions, Intel hired in a bunch of GPU developers from other firms that already had experience doing this. These engineers knew what they wanted (more space on the die for the GPU) but Intel didn't give it to them until there was already a market segment for them to occupy that had been pioneered by someone else, even if it was other firms (SGI?, VIA? and ATI) where the engineers had been hired away from. It's not about competition, it's about waiting for someone else to dip their feet in the water first. The only real recent counter-example I can think of is ThunderBolt, but it's not the best since it's another one of those Apple-ordered items and has low market penetration.
None of these changes were in instruction set, they are all code compatible. They are just internal tweaks doing things slightly differently.
That is nothing but a one sided view. They have both been going in the same direction for years. AMD bought ATI to leapfrog Intel, and now Intel has all but caught up again.
I don't know where you think it is going to bite them. It looks that by next year they will probably have the best Desktop CPUs, Best SoCs and quite possibly Best IGP as well.
wonderfield
Supreme [H]ardness
- Joined
- Dec 11, 2011
- Messages
- 7,396
Maybe, but it's pretty clear that Apple was in a position to benefit enormously from Intel's stronger iGPUs. The power savings there versus having discrete GPUs is pretty astounding.
Granted, everyone's in a position to benefit from that, but no one cares more deeply about both notebook battery life and overall GPU performance than Apple. PC OEMs care about one or the other, but rarely do they care about both. Given that, I'd like to think that Apple was pretty integral to the equation.
Also: On a totally unrelated note, Intel can suck my dick for not including eDRAM on high-end socketed Haswell parts.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
This may have been true a few years ago but certainly not today. Intel's drive towards better battery life and GPU performance is being driven by phones and tablets where Intel is kind of in competition with Apple at this point.
wonderfield
Supreme [H]ardness
- Joined
- Dec 11, 2011
- Messages
- 7,396
Consider this: The only Haswell parts with high-end Iris Pro silicon and eDRAM are those in a BGA package destined for OEM desktops and MacBook Pro-class notebook products.
Who makes both desktops using BGA-packaged CPUs and MacBook pro-class notebook products?
Who makes both desktops using BGA-packaged CPUs and MacBook pro-class notebook products?
InternationalHat
[H]ard|Gawd
- Joined
- Aug 13, 2004
- Messages
- 1,481
8 is poor on a tablet compared to iOS and android anyway. It's halfway between a desktop and tablet interface and does neither well.
I'm not surprised to see Surface RT floundering and I don't imagine "surface pro" or any of its clones will really sell outside of the nerd circles that convince themselves they need it. 8 is not even a blip on the radar when talking about overall tablet sales.
Outside of business users in the field already in a totally MS shop, there's not much upside to microsoft's offerings.
I'm not surprised to see Surface RT floundering and I don't imagine "surface pro" or any of its clones will really sell outside of the nerd circles that convince themselves they need it. 8 is not even a blip on the radar when talking about overall tablet sales.
Outside of business users in the field already in a totally MS shop, there's not much upside to microsoft's offerings.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
Actually in one survey for last quarter Windows 8/RT showed up as 7.5% of tablet shipments worldwide, that's more than a blip considering that Windows 7 registered at nearly zero by the same firm in the same quarter the prior year.
wonderfield
Supreme [H]ardness
- Joined
- Dec 11, 2011
- Messages
- 7,396
A list that doesn't include such manufacturers as HTC and Nokia. Phone handset manufacturers, in other words.Sony, Apple, HP, Dell, Lenovo, and any other manufacturers who make all in one desktop systems?
That's the point. These high-end graphics parts aren't going into either tablets nor into phones. Intel isn't putting eDRAM in any SoC destined for tablet or phone use. In fact, all current Intel SoCs don't even use Intel graphics: they use out-of-date PowerVR GPUs, the highest-end of these chips getting only the dual-core 544.
The theory that Intel's pursuit of graphics performance is about phones and tablets is a theory backed by absolutely zero compelling evidence.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
Some of these high end parts will end up in high end Windows 8 tablets, hybrids and convertibles. My point was that what's really driving Intel and power efficiency and GPU performance are tablets and mobile. They own desktops and laptops and that market isn't growing, even Mac sales have flat lined the last six months. And the next Atom will use a derivate of the HD 4000 as its GPU. Not going to see a Mac with an Atom in it anytime soon.
MS Windows had a 7.5% share of the tablet market... about 2.5 years ago: http://www.idc.com/getdoc.jsp?containerId=prUS24093213 That's mostly a leftover share of a much smaller market, which was changing as Apple ramped iPad 1 shipments. And then, finally with a real "tablet OS", MS pretty much disappears into the noise (under 2%) this last quarter:
(IDC seems to have lumped all manufacturers' Windows Tablet PCs into the "Microsoft" in that chart)
(IDC seems to have lumped all manufacturers' Windows Tablet PCs into the "Microsoft" in that chart)
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
So what your graph shows is that MS share was decreasing and now it is increasing.
wonderfield
Supreme [H]ardness
- Joined
- Dec 11, 2011
- Messages
- 7,396
The only chips with Iris Pro graphics have 37W and 47W TDPs. That puts them well out of the acceptable range for tablet inclusion.
Razer may decide to be a little silly and integrate one into a next-generation Edge, but you can probably count the numbers of Edges they're selling each quarter on two hands. I find it hard to believe that Intel was so aggressive with GT3, including four times the amount of eDRAM Intel engineers decided was ample, because Razer might make a tablet with one of those chips and might sell seven of them.
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
It shows Tablet PCs decreasing, then a small bump in shipments when Windows 8/RT were released. Then the table above it shows pretty bad Surface & Surface RT sales (900,000 shipped in Q1'13), which is apparently a decline compared to the prior quarter, and out of line with its main competition.
IOW, not good news for a market growing at a nearly 150% rate (20 million Q1'12 vs 49 million Q1'13).
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
You are showing your bias, clearly you are trying to compare completely different statistics and draw conclusions from it. In one graph MS is listed along side Samsung, but Samsung sells both windows and android tablets. You cant sit there and quote completely different numbers and stitch them together.
You want to use the graph then you have to admit that 2.5 years ago MS had a 0% share because they didn't make any tablets. But instead you quote the OS and try to make it evidence of a shrinking market share then throw up surface shipments which are not the total sales of windows 8 tablets.
You want to use the graph then you have to admit that 2.5 years ago MS had a 0% share because they didn't make any tablets. But instead you quote the OS and try to make it evidence of a shrinking market share then throw up surface shipments which are not the total sales of windows 8 tablets.
That's just the available data on IDC. You can read the details on the site. I didn't make it up. If I had better data I would have posted it. I was responding to the 7.5% share claim the Wall made.
I think Windows 8 tablets are doing fine, considering the disadvantages they currently have (high prices, poor Metro ecosystem to drive tablet sales, etc). My posts in this thread have been about very poor Windows RT sales, which unfortunately are not split out in the public IDC data. The table of data shows the problem, regardless.
I think Windows 8 tablets are doing fine, considering the disadvantages they currently have (high prices, poor Metro ecosystem to drive tablet sales, etc). My posts in this thread have been about very poor Windows RT sales, which unfortunately are not split out in the public IDC data. The table of data shows the problem, regardless.
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
The table of data that does not split out RT sales shows the problem with RT sales?
Yes. Based on prior MS public statements (1 million Surface RT units shipped in Q4'12) and 400,000 Surface Pros sold in Q1'13. Halving sales in one quarter for a launching product (Surface RT) is a horrible sign.
Damn, some people are living in a fantasy land of "if I close my eyes and cover my ears, bad news doesn't exist."
A few days ago MS also tossed on another freebie for the twice discounted in under 2 months Surface RT: free touch type cover. $100+ accessory. Free. lol, a sign that things are not going badly.
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
Surface RT is still $499 at MS store I don't see even one price drop let alone 2. You should work for apple the way you craft words.
Surface RT is being discounted in retail. MS itself announced the first $50 price cut back in April. In late May, retailers have dropped the price by another $50 (Staples, Best Buy, etc), which is unlikely to just be a coincidence.
But since you don't like reality, bye. This is a waste of time. PLONK!
But since you don't like reality, bye. This is a waste of time. PLONK!
rudy
[H]F Junkie
- Joined
- Apr 4, 2004
- Messages
- 8,704
So link me to the $399 surface RT with free touch cover.
BladeVenom
Supreme [H]ardness
- Joined
- Jun 29, 2005
- Messages
- 7,707
Top 5 tablet makers as per IDC in Q1:
Apple . . . . . . 39.6%
Samsung . . . 17.2%
Asus . . . . . . . 5.5%
Amazon . . . . . 3.7%
Microsoft . . . . 1.8%
Total tablets sold in Q1: 49.2 million units
Android powered 56.5% of all tablets, iOS powered 39.6% and Windows 1.8%
Apple . . . . . . 39.6%
Samsung . . . 17.2%
Asus . . . . . . . 5.5%
Amazon . . . . . 3.7%
Microsoft . . . . 1.8%
Total tablets sold in Q1: 49.2 million units
Android powered 56.5% of all tablets, iOS powered 39.6% and Windows 1.8%
heatlesssun
Extremely [H]
- Joined
- Nov 5, 2005
- Messages
- 44,154
Microsoft doesn't make all Windows tablets.
