I suppose it depends what the 8 cores are. If they are Asymmetrical then sure, 2 gpu, 2 low power and a quad full bore then they can be powered up or down as you like. But outside of benchmarks do domestic os or apps use so much parallel threading?
Qualcomm exec on eight-core mobile chips: They're 'dumb'
If you have been wondering when the chip designers at Qualcomm were going to come up with an eight-core processor, as have competitors Samsung and MediaTek, you can stop wondering. They're not. Why? Because octo-core chips are "dumb." "We don't do dumb things," Qualcomm SVP and marketing headman Anand Chandrasekher said in …
-
-
-
Friday 2nd August 2013 23:36 GMT Eddy Ito
According to the video on their site they are 8 equal cores and they have the ability to switch on or off as many as needed for the task at hand. That would give them 8 levels of granularity in the CPU core count plus any clock adjustments. I assume all the cores run on the same clock but I could see them clocking 2 banks of 4 cores differently to increase the potential power tweaking but 8 clocks seems a bit extreme.
-
Saturday 3rd August 2013 04:38 GMT Zola
According to the video on their site they are 8 equal cores and they have the ability to switch on or off as many as needed for the task at hand.
Assuming you are talking about the MediaTek True Octa-Core, the video, nor any of the documentation, fails to make it clear if it's using equal cores or not - MediaTek don't even mention the architecture (Cortex-A7, A12, A15 etc.) of the ARM cores in any of their documentation.
However the "Optimized ARM big.LITTLE" tab at the top of the MediaTek True Octa-Core page gives the game away - the MediaTek True Octa-Core is using big.LITTLE MP, which would only be the case if all 8 cores were NOT equal. Most likely there are 4xA7 and 4xA15 cores, but with big.LITTLE MP all of the 8 cores can be in use at any one time.
It's also funny how they absolutely cane their competitor (Samsung) by pointing out their ability to support only Cluster Migration and not CPU Migration (as we now know, thanks to the broken CCI -400 in the Exynos 5410). Perhaps MediaTek will need to revise their marketing material once the 5420 comes to market, assuming the CCI-400 is fixed of course...
And with a fully functional CCI-400, the Exynos 5420 should also be able to support big.LITTLE MP, just like the MediaTek True Octa-Core, as this is mostly a kernel/driver software thing with Linux support arriving only in the last few months. No doubt Android will gain big.LITTLE MP support in due course, although it's not going to be of much use to those poor international Galaxy S4/Exynos 5410 users.
-
-
Sunday 4th August 2013 11:14 GMT Anonymous Coward
8 CPU core mobiles might be 'dumb', but as long as the leading mobile OS is a bloated mess based on an emulated Java runtime requiring constant background garbage collection, and a bloated and legacy monolithic kernel based underlying OS then powerful hardware is going to be a requirement....
Google are taking the old Microsoft approach - build it to be as inefficient and clunky as possible to help sell new and faster hardware. Microsoft have now left that model in the past, but Google it seems have adopted it with some enthusiasm...
Just compare a cheap Windows Phone with a 'landfill' Android - they might have similar specs, but performance is like night and day. Window Phone currently leaves Android in the dust for performance and efficiency...
-
-
-
Friday 2nd August 2013 23:19 GMT Anonymous Coward
I agree. 8 core phones are pretty dumb.
And I say that as a mobile app developer, working at the sharp end with heavy realtime graphics.
Put more CPU cores on the chip, and yes, they'll get used.. up to a point. The OS will use a 2nd core effectively. 4 cores, and you're seeing diminishing returns from the OS side. A mobile OS is designed for low power, meaning it does as little as possible. It might check email while you're browsing the web. It doesn't need 8 cores for that.
Apps: well, some games might use more cores. Not many. Not much else would use it. The few that might probably won't, because most phones still have 1 or 2 cores, and the devs want a big market. What else? Video encoding? No, there's a hardware encoder on the chip. 3D rendering? On a phone? No. There's a GPU for regular 3D.
So what are the 8 cores for? That one time when you want to revisit the 90s and draw a mandelbrot?
-
Saturday 3rd August 2013 04:46 GMT Zola
Re: I agree. 8 core phones are pretty dumb.
So what are the 8 cores for? That one time when you want to revisit the 90s and draw a mandelbrot?
No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS. With all 8 cores going gangbusters on whatever tasks I'm performing on my full desktop OS.
Try not to think about the limited smartphones of today, which are running operating systems that massively lag the prodigious hardware they have available to them.
Some firms (eg. Ubuntu) have embraced the "hybrid" approach and it's quite compelling, while potentially disruptive for an already ailing PC hardware industry. It remains to be seen if the other mobile OS vendors with fingers in the PC/desktop OS pie, ie. Apple and Microsoft, will ever follow the lead of Ubuntu - due to conflicts of interest I find that incredibly unlikely, so it's just as well I have no interest in what they do.
-
Saturday 3rd August 2013 06:39 GMT Steve Davies 3
Re: I agree. 8 core phones are pretty dumb.
You will need all 8 cores to handle all the data that your phone needs to send to the NSA/GCHQ/FSB etc.
With all this trcking and snooping that is either going on now or will happen very soon, keeping my desktop and phone as far apart (comms wise) as possible makes perfect sense to me.
-
Saturday 3rd August 2013 12:55 GMT Anonymous Coward
How separate are comms from phone and from desktop?
"keeping my desktop and phone as far apart (comms wise) as possible makes perfect sense to me."
You may have posted this before the latest revelations that the UK backbone international carriers are providing GCHQ (and thus GCHQ's paymasters in the USA) free unrestricted unsupervised access to their core networks.
It'll make little difference in that context whether your desktop and phone are separate or not.
"BT, Vodafone Cable, and the American firm Verizon Business – together with four other smaller providers – have given GCHQ secret unlimited access to their network of undersea cables. The cables carry much of the world's phone calls and internet traffic."
http://www.theguardian.com/business/2013/aug/02/telecoms-bt-vodafone-cables-gchq
The article also names Global Crossing and Level 3, amongst others who are co-operating with the program - names which won't be well known to typical Grauniad readers but might be better known among some of El Reg's audience.
-
Sunday 4th August 2013 08:35 GMT amanfromMars 1
Question? How separate are comms from phone and from desktop? ..... Answer*
* In Palace Barracks, Hollywood-type Charades on Parade, do they meld together for the command and control of power and minds for master key manipulation of primitive native codes, which be the smarter virtual drivers for all dumb ignorant animals and arrogant beings?! ........ or they could and should, if not already doing so stealthily for those who server intelligence selflessly for outrageous reward.
You may have posted this before the latest revelations that the UK backbone international carriers are providing GCHQ (and thus GCHQ's paymasters in the USA) free unrestricted unsupervised access to their core networks. …. AC Posted Saturday 3rd August 2013 12:55 GMT
A possible commonly held misperception/misconception, and therefore most probably thought quite real enough in the field to be believed as true, AC, but/and it is not inconvenient for GCHQ, and therefore would they have no need to tout a correction, to have its customer clients/intelligence deficient base market thought to be paymaster rather than starving wretch in dire straits need of fulfilling main course meal. It is a major mistake which disadvantages one considerably, and even catastrophically, to forget and not remember or not simply realise …. Nothing is ever as it seems in markets and special applications [in] programs which deal intelligence product in the Great Game Genre
And further to "Deep inside ARM's new Intel killer" … http://www.theregister.co.uk/2011/10/20/details_on_big_little_processing/ …. and the more complex ARM Cortex-A15 MPCore with 15-to-25 stage[d] pipeline …. schematic diagram here, http://regmedia.co.uk/2011/10/20/arm_a15_pipeline_large.jpg ….. and big.LITTLE partnering processor cores for virtualisation of reality which renders to Systems Admin and SMARTR Admin Systems on/in/with AI Loded Silicon, Remote Future Command and Control with Futures and Derivatives Markets Command and Control, please consider and have a wonder on the nature of things if and/or when
Fetch = Phish for Information
Decode = Phorm with Intelligence
Rename = Float to Market
Dispatch = Support in Supply
It may very well be, that then is the product undeniably a firm favourite for strategic, mission critical leaderships with Global Operating Devices, whether in nimble mobile units or gargantuan static installs/applications/operations/huge server farms? ........ for those into Mastery of the Universe and all who fly in her and follow in her wake to the future ....... which you might like to deny and argue is not a return journey for some in the present with no interest in preserving the past which is a dead space place.
And yes, that is an unambiguous direct titanic challenge to 5 in Holywood to actually do something positively engaging and autonomous, and HyperRadioProActive with IT too, rather than just remain terrified and petrified as do all awaiting new orders which will never ever come from old establishment guards in charge of protection and perpetuation of the status quo.
Wake up, smell the coffee. Times have a'changed and new spaces rule all the old places.
-
-
-
Saturday 3rd August 2013 12:45 GMT Anonymous Coward
Re: I agree. 8 core phones are pretty dumb.
"No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS. With all 8 cores going gangbusters on whatever tasks I'm performing on my full desktop OS."
To an extent. I fully agree that a dockable/docked smartphone running a desktop-capable OS is A Good Thing (for everyone except MS and Intel,obviously). But...
How many people actually make much use of multicore capability in today's desktop workloads? How much will that change in a dockable smartphone?
I'd guess it's a near-negligible proportion, and is likely to mostly stay that way.
For most purposes (exceptions apply) it's better to have a smaller number of faster cores than a larger number of slower cores. But the clock speed limits have been hit in semiconductor manufacturing a while ago, be it x86 or ARM or whatever. So for benchmarketing and "who's shiniest" purposes we've all been told that more cores are what the world needs. And, for a tiny number of workloads, they are.
Sensible number of cores, at a lower price point, might be an interesting concept. For customers or brave manufacturers anyway.
-
Saturday 3rd August 2013 16:39 GMT Anonymous Coward
Re: I agree. 8 core phones are pretty dumb.
"No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS."
Ok, so you already have a keyboard at your power dock, or are you really going to use 2 fingers to run a desktop? What about the monitor, are you going to rock out a 4 to 10 inch one for a desktop, with your fingers obstructing half of it? What if you took a smartphone and removed the monitor and touch capabilities, then you could pack a whole lot more cpu punch into the same size. Well, those are called thin clients, or even Raspberry Pi's...
I agree with the sentiment of using a mobile for all computing needs for normal people, but there is just so much tech. missing to make that a reality anytime soon. For example, holographic keyboards based on iR tech or whatever. Holographic displays that shoot up just like in, if not exactly like in, R2D2 from Star Wars. So 8 core CPU for mobile, sounds great for the year 2050 (or realistically given companies current R&D, year 4125).
-
Saturday 3rd August 2013 18:20 GMT Anonymous Coward
Re: I agree. 8 core phones are pretty dumb.
"I agree with the sentiment of using a mobile for all computing needs for normal people, but there is just so much tech. missing to make that a reality anytime soon"
Does not compute.
My smartphone (or tablet or whatever) will already talk HDMI, as will any decent DVI-capable monitor (or, if at home, TV).
My smartphone (or tablet or whatever) will already talk Bluetooth for the real keyboard/mouse (or maybe USB instead).
My smartphone (or tablet or whatever) will already take power from USB.
The docking station could well be a convenience rather than an essential. The office desk (and/or the home) has the required facilities, and where the required facilities aren't available, improvisation (or the phone/tablet alone) will do.
"if you took a smartphone and removed the monitor and touch capabilities, then you could pack a whole lot more cpu punch into the same size. Well, those are called thin clients"
Surely a thin client is a display, a keyboard/mouse, and a LAN connection, with minimal local storage and minimal local compute power? Something elsewhere (something currently fashionably called VDI, apparently) does the storage and serious computing, in the server room rather than on the desktop.
"or even Raspberry Pi's."
A Raspberry Pi is a learning tool (amongst other things), it is surely not a road warrior's machine. It's not really even a convenient thin client. It's cheap and interesting though, which is usually a plus. But probably not here.
-
-
Saturday 3rd August 2013 18:45 GMT Mikey
@Zola
"No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS. With all 8 cores going gangbusters on whatever tasks I'm performing on my full desktop OS"
Well, sure. This does seem to be the way things are headed and all. But one thing to consider here is heat. The more cores you have burning power, the hotter things will get. The phone/tablet casing can only sink away so much before things start to get out of tolerance, so the device itself will never quite reach proper desktop levels of grunt. Not unless you're happy to have a phone/tablet with a mounting point for a large external heat sink, which would require thermal paste and a large smooth surface to conduct the heat through.
I'd say we're due for a halfway house scenario, myself. Your device has your apps and files on it, and the base station has all the extra grunt required to go the full desktop monty, as well as all the peripherals and the like. To which end, we'll still need desktops of a sort, but at least you can hot-desk easily.
I'm looking forward to all the reports of trouser-fires from a full blown 8-core unit forced into 100% CPU utilisation via a nice piece of rogue software... hehehe.
-
Saturday 3rd August 2013 20:58 GMT Anonymous Coward
Re: @Zola
"[thermals] never quite reach proper desktop levels of grunt. "
True, but the most demanding thing done on a great many desktops and laptops is run the antivirus, or maybe decode a video. Video decode on phone/tablet will be done in GPU. Printing can be interesting if done locally rather than on a printserver, but local printing sounds unlikely in a corporate/"hotdesking" environment. Interactive stuff (email, most web, etc) will still be spending most of the time idle.
"reports of trouser-fires from a full blown 8-core unit forced into 100% CPU utilisation via a nice piece of rogue software."
Don't modern x86 laptops come close to that already when working hard, rogue or otherwise? Software control of fan speed, now that *would* be fun.
-
-
Sunday 4th August 2013 09:59 GMT Robert Sneddon
Re: I agree. 8 core phones are pretty dumb.
How much RAM is in your smartphone? How much cache do the CPU cores have access to? If you want to rock a desktop OS like gangbusters then lots of RAM and a good helping of cache will really speed things up, and those are noticeably lacking in all the smartphone and most tablet designs I've seen. Lessee, the iPhone 5 has 1GB of RAM, not sure how much CPU cache. Androids like the Galaxy S4 have 2GB RAM so its a bit better but still quite anaemic by modern desktop OS standards. About the only affordable tablet that looks like its ready for real work on a desktop is the MS Surface Pro with an i5 CPU (3MB of cache) and 4GB of RAM, but then again it's not a phone.
Frankly it would be better for you to keep your phone in your pocket and rock a real desktop machine like gangbusters instead. Besides what would you do if you got to the office and found you'd left your phone at home?
-
Sunday 4th August 2013 11:23 GMT Anonymous Coward
Re: I agree. 8 core phones are pretty dumb.
"No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS"
Ubuntu are late to that party - Microsoft already have that capability in Widows Phone 8 / Windows 8 / Windows RT. The exact same kernel underpins all of them....
-
Sunday 4th August 2013 11:37 GMT Matt Bryant
Re: Zola Re: I agree. 8 core phones are pretty dumb.
"....No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS....." No you won't. For a start, what are you going to be actually doing that requires even quad ARM cores? Most desktop use today and in the future is well within dual-core capabilities - mainly browsing and a bit of word processing or spreadsheet work, maybe the odd PDF - so much so that I am amazed when people insist to me that they need an octo-core CPU in their current desktops. They do not. If you are doing something really heavy then you will be using a heavyweight desktop, not a smartphone and dock. I am currently running a virtualized environment with four VMs (three servers and a client) quite happily off a dual-core desktop with a four-year old CPU, the biggest constraint being RAM, not the number of CPU cores. I definitely do not want to lug around a smartphone with even that level of capability as I do not need it, it's already there on my desktop (which has a BIG screen and can display lots of information, not a pocket-sized one).
-
Sunday 4th August 2013 13:04 GMT Anonymous Coward
Re: Zola I agree. 8 core phones are pretty dumb.
"I definitely do not want to lug around a smartphone with even that level of capability as I do not need it, it's already there on my desktop"
Good for you. You have a properly equipped desktop which your employer is willing to dedicate for your use. Not everyone computes like that. Quite a few wandering souls would benefit greatly from having one client device which, with appropriate support, would do for both on the road and office based use, or even just avoid needing dedicated desks in an office/call centre/etc. That's part of the reason why, for the last decade or three, laptops have had docking stations. The fact that you apparently can't see this doesn't mean the "use case" doesn't exist.
Agreed about dual core being mostly enough for most people though.
-
Sunday 4th August 2013 19:30 GMT Matt Bryant
Re: AC Re: Zola I agree. 8 core phones are pretty dumb.
"....You have a properly equipped desktop which your employer is willing to dedicate for your use....." Actually that's my home desktop, nothing to do with my employer. My work "system" is a laptop with a lot less grunt, but it is capable of connecting over VPN to a VDI setup and on to a whole host of servers, so it is already mobile and much more capable than any smartphone is going to be for years, thanks. I have had the ability for over a decade to do documents and work on a Blackberry, but tried not to as it is simply a stupid thing to do on a small screen. Until smartphones can do some form of projected screen to give you the ability to actually read and work on docs they will not replace desktops or laptops or even tablets.
-
-
-
Sunday 4th August 2013 18:06 GMT Henry Wertz 1
Re: I agree. 8 core phones are pretty dumb.
"No, it's for when I drop my Smartphone into a mains powered dock and start running a full desktop OS. With all 8 cores going gangbusters on whatever tasks I'm performing on my full desktop OS."
Still unnecessary. A while ago, I installed a Linux chroot environment, and ran openoffice and the gimp off a 1.2ghz single core ARM in a Droid 2 Global (using X11 to display the apps remotely -- I was not trying to run them on a 3" or whatever screen.) It took a few seconds for OO to load, but CPU use was near zero, the delay was due to the microSD card. The apps ran downright snappy, and kept CPU use near zero even when I started whipping through some menus and stuff to try to use some CPU time. You're not using WIndows here, you don't need 8 cores for a desktop environment.
Honestly I agree with the Qualcomm guys, an 8-core mobile chip is dumb.
-
-
-
-
Friday 2nd August 2013 23:36 GMT Anonymous Coward
No. A low-performance CPU will use much less power than a high-performance CPU running at a lower clock speed.
The high-performance CPU has a ton of extra logic that simply isn't necessary for low-performance scenarios. Out-of-order everything, instruction scoreboarding, register aliasing, bigger caches, branch predictors, blah blah blah.
The idea for big.little (however it's capitalized) is pretty good.
-
-
Friday 2nd August 2013 23:56 GMT Anonymous Coward
"It is, but can anyone think of a case where you need quad core little?"
No, not at all. I can only assume that it makes something about the design easier. Maybe when it transitions from one set of cores to the other, it can "divert" the instruction stream on a 1:1 basis in a way that's transparent to the OS...?
I don't know if this is practical but it SEEMS like a better idea to just have one set of asymmetric cores... 2 little cores, 4 big cores, and allow the OS to control which cores are powered up and which cores are running which processes...
-
-
Saturday 3rd August 2013 05:16 GMT Zola
No. A low-performance CPU will use much less power than a high-performance CPU running at a lower clock speed.
And equally, a high-performance core may use less power than a lower performance core, simply because it can complete the task more quickly.
It's the race to idle, the more quickly you can finish a task (and switch off the core) the less power you use overall. So a task that runs for 2 minutes (potentially maxed out) on a low performance core but only 30 seconds on a higher performance core may actually use the same or even less power on that higher performance core.
-
Saturday 3rd August 2013 06:29 GMT Anonymous Coward
"It's the race to idle, the more quickly you can finish a task (and switch off the core) the less power you use overall."
Yes, good point. But a lot of times cell phones are in a state where they only need to wake up a few times per second, do a small amount of calculation, and go back to sleep. That's where a low power core is preferable. It takes much less power to spin one up/down vs. whatever power is needed for the small amount of computation.
-
Saturday 3rd August 2013 09:24 GMT Piro
I'm pretty sure race to idle is no race at all
The more power you throw at the problem, in general, the more smoothly you are doing the same task.
Instead of rendering at 30 fps, you may be hitting 60, using a lot more power.
If you're playing a game, or any intensive task, there is never a race to idle. Only constant power usage.
-
-
-
Friday 2nd August 2013 23:46 GMT Sorry that handle is already taken.
Cortex-A7 is significantly more efficient in terms of both power and silicon than Cortex-A15. Hence "big.LITTLE"; when you don't need the peak performance of the A15 you can drop back to something that uses less electrical power per unit of processing power.
That is, of course, assuming you manage to implement it properly. Samsung had to re-spin the Exynos 5410 as the 5420 due to the cache coherency issue that affects battery life, but not before using the 5410 in the Galaxy S4.
-
-
Friday 2nd August 2013 23:47 GMT Anonymous Coward
Yes, definitely dumb
Why should my phone have twice as many CPU cores as my Core i7 desktop computer... ?
Much more valuable to the consumer would be to focus on increasing per-core performance. According to some benchmarks, current state of the art ARM cores are still only about half as fast as Intel "Core" cores on a per-clock basis. There's a lot of room for improvement.
-
Saturday 3rd August 2013 11:24 GMT Anonymous Coward
Re: Yes, definitely dumb
I don't really know why you've been downvoted here. People should look at their task manager from time to time. I'm a pretty heavy PC user, my quad core PC is generally idle for most tasks other than gaming with the occassional spike on a single core.
When given the shocking battery life of modern phones, personally I think phones would be better sticking at 2 -4 cores, 2xlow power, 1 dedicated to the UI, 1 for smooth multi-tasking, i.e. downloads/app installs/email sync so it doesn't affect the UI and the 3rd (and 4th) core being a silly powerful battery sucking monster for gaming and such.
I'm still on the Nexus S, granted I don't game, but the UI does get a bit jittery when downloading/installing apps/send e-mails with attachments. A 2nd core would rectify this, but I definitely don't need 4+.
-
Saturday 3rd August 2013 15:44 GMT FutureShock999
Re: Yes, definitely dumb
You seem to have missed a lot of history here. Historically, ARM cores have fewer pieces, parts, and transistors than Intel, and have therefore run slower, but with much less power. ARM isn't about speed - it is about work per watt. Intel comes from a desktop background where it was always about having the faster processor, period.
To address their relatively low power (in both ways) cores, ARM has gone multi-core in a big way. Whereas Intel has gone down the power-saving route in a big way - reducing clock, idling parts of the chip, etc. And they have developed some very cool tech for that.
I personally think that the ARM approach is simpler, and will therefore yield a more cost-efficient solution. For a phone, or tablet even, there are a lot of tasks that are limited by the speed of the communications interface, checking for email in the background, even downloading a web page, etc. That will be slower on mobile than a laptop or desktop for quite some time, even with LTE. ARM's approach is optimised for that kind of slow, steady work.
Improving the per core performance would help on games, etc., but even with power saving, it still isn't as efficient as a slower, but steady engine. Think of driving your car - you CAN take an F1 engine, put it in your car, accelerate madly to 150 mph, and then coast back down to below the speed limit, repeat, repeat, repeat, and get an average speed of 80 mph. Or, you can put your bog standard production-line auto engine, and steadily maintain 80 mph - which do you think is more efficient really?
-
Saturday 3rd August 2013 17:33 GMT Anonymous Coward
Re: Yes, definitely dumb
"You seem to have missed a lot of history here. Historically, ARM cores have fewer pieces, parts, and transistors than Intel, and have therefore run slower, but with much less power. ARM isn't about speed - it is about work per watt. ... To address their relatively low power (in both ways) cores, ARM has gone multi-core in a big way."
You seem to have missed a lot of *recent* history on the subject. Everybody is tripping over themselves these days to offer ARM cores with more performance per core, including ARM. Just in the last few years we've gone from pre-Cortex ARM cores to Cortex-A8, A9, and now A15, while in the meantime firms like Qualcomm and Apple are now making their own cores with higher IPC (Krait, A6, etc.).
Which is frankly necessary because many (most?) cell phone tasks are limited by single-core performance. Notice how all the benchmarks to do with rendering web pages and running JavaScript aren't affected one whit by the number of cores the CPU has? That's because it's all single-threaded and you need to make per-core performance faster. Running this software with 348257348905723945 cores will not speed it up at ALL.
-
Monday 5th August 2013 06:27 GMT P. Lee
Re: Yes, definitely dumb
Once mufti-processing was introduced to android you have multiple dalvik processes running. With cheap simple cores, its easier to replicate them and spread your apps between cores than to juggle thing around, especially as a process might be maxing out a core. Is your email client busy processing email? No problem if your browser is rendering its current page on a different core, your ebook reader is busy rendering another xml page on a different core and you left a game running too. It is unsurprising that the ex-Intel man is pushing single thread performance. Intel did very well in the desktop market with single-thread performance. However, in that market, the CPU's were overpowered for most business use, with AV creating a very important race-to-idle situation when paired with a high-power CPU. In any case, 4 low-power/4 high power (normal use / high-power usage) is not really octocore.
Also note that phones are a bit of a test-bed for arm development. Without being able to compete with intel in terms of fabrication and single-thread performance, ARM producers might like to use multi-core (much as AMD does) to provide overall performance. For the vertically integrated corporations, the marginal cost of ARM cores is very low, compared to buying off intel who lose a CPU sale as they add more cores to one package. If ARM producers are heading into the server market, they might look at high-core-counts over single-thread performance, even as Intel does with xeon and i7.
-
Monday 5th August 2013 12:14 GMT Anonymous Coward
Re: Yes, definitely dumb
"your browser is rendering its current page on a different core, your ebook reader is busy rendering another xml page on a different core and you left a game running too."
I presume you're a hardware person? Those applications require something called software too.
To make that multicore setup work well, you need a decent OS with a decent scheduler. And the scheduler has to understand the potentially complex implications of things like process->processor affinity (e.g. because if a process moves from one core to another it may lose some or all of its cache).
Life is easier for the hardware designer and the software designer if there is one processor and multiple processes being scheduled on it. Easier is almost always better.
Obviously this falls down where the sole processor isn't man enough for the workload (or where the scheduler isn't behaving appropriately for the workload/environment combination).
There was a chap called Cutler once. Iirc he had an interesting approach for a scheduler which knew how to keep interactive users happy whilst also allowing compute bound stuff a reasonable crack of the whip. Whatever did happen to him...
-
-
-
-
Friday 2nd August 2013 23:58 GMT Don Jefe
Al Dente
'that is the equivalent of throwing spaghetti against the wall and seeing what sticks," he said..
The only way to tell if spaghetti is done is to throw some at the wall. They won't stick until they're al dente. Obviously this man has no knowledge of pasta and should keep his opinions on such matters to himself.
-
Saturday 3rd August 2013 10:56 GMT Harvey Trowell
I get where you're coming from, but the only way?
I know there's a fine tradition of chucking spaghetti about the gaff, but I'm a fan of fishing a piece out and eating it myself. Another way is to slice through a strand with your thumbnail and check both the resistance on the way through and the appearance of the core, but that way you don't get to eat it. After all, al dente does mean "to the tooth" rather than "to the thumbnail", or "to the splashback tiling" for that matter. Paris, because she's no stranger to splashbacks.
-
-
Saturday 3rd August 2013 00:04 GMT Zola
"Dumb" until Qualcomm are able to produce an 8 core chip
This is a classic case of a company not being in a position to create a particular product, so until they are in such a position one option is to slag off the competition that have already made a similar product.
Nokia did the same when they were stuck with single-core chips, vehemently claiming there was no need for multiple cores. Until, that is, Microsoft got their act together and added support for dual-core chips and since then Nokia have never produced another single-core Smartphoone. See how that worked?
Chandrasekher should be ashamed of himself for using such a low brow argument to justify his own companies inadequacy.
The icon is for Chandrasekher - must try hard when discussing product execution delays.
-
Saturday 3rd August 2013 00:53 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
It isn't rocket surgery to create an 8 core chip. It is literally almost as easy and cut and pasting additional cores to go from dual to quad or quad to oct. 99% of the difficulty was already passed in both the hardware and the software going from single core to dual core.
Quad core CPUs for phones is ridiculous overkill and completely unnecessary. Going to 8 core is ludicrous. Use the extra silicon area where it can make a different, for the GPU.
That GPU upgrade is much more necessary, since saddling phones with the unnecessary upgrade from 720p to 1080p more than doubles the number of pixels they have to push around, so they need more than double the GPU just to maintain performance parity! The way things are going in the insane resolution race, I wonder if we'll see a '4k' phone in a few years...
I guess fools like you who make purchase decisions based on bigger numbers in the specs are why Mediatek is busy making a "true" octo core. I thought it was Apple buyers who were supposed to be the brainless idiots swayed so easily by marketing?
-
Saturday 3rd August 2013 04:04 GMT Zola
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
It isn't rocket surgery to create an 8 core chip.
The difficulty Qualcomm likely have is that they are using their own architecture for their multi-core SoCs, and not using standard ARM IP.
So stuff like big.LITTLE has to be re-invented by Qualcomm, stuff like the CCI has to be re-invented and updated to support more and even heterogeneous cores.
Yes, it's probably (relatively) easy to go from a dual-core to a quad-core and then an octa-core design when all the cores are 100% identical, but in the current 8-core SoCs not all the cores are identical (which would be very bad for power consumption) and it's this point which is most likely making life that bit more challenging for Qualcomm. They'll need to be considering how they solve the power problem as they ramp up the number of cores, and if they decide on using big and LITTLE cores they need to work out how they achieve it - design another new architecture for either the big or LITTLE cores? Not quick and not cheap.
This is all most certainly not rocket science, but because of the route Qualcomm has taken to market - ploughing their own furrow which has served them well up until this point - they now have a lot of work to do to catch up with the likes of Samsung and Media-Tek who are simply using off the shelf designs from ARM.
Quad core CPUs for phones is ridiculous overkill and completely unnecessary. Going to 8 core is ludicrous.
Generally, I'd agree - for todays smartphones.
But these 8-core SoCs aren't only going to be used in smartphones, they'll be absolutely fine in tablets with much bigger batteries, and I want an 8-core SoC in my "hybrid" smartphone of tomorrow which will run a full desktop OS when docked.
The big.LITTLE concept makes a lot of sense when thinking about "hybrid" products that sip small amounts of power while offering moderate processing capabilities when mobile, but are able to provide much more processing capability when docked and on mains power.
Octa-core processors may not be the most sensible option today with relatively dumb smartphone operating systems but in the future, with more capable operating systems and form factors, their benefits should be more obvious.
Qualcomm claiming 8-core SoCs are "dumb" is just deflecting attention away from their current ability to compete.
-
Saturday 3rd August 2013 06:38 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"I want an 8-core SoC in my "hybrid" smartphone of tomorrow which will run a full desktop OS when docked."
Uh, why?
My Core i7 desktop computer "only" has 4 cores and my laptop "only" has 2 cores, and both work great.
I have no idea why people seem to think more cores = more performance. Of course it does for certain very specific applications like Handbrake or raytracing or compiling huge projects but most software does not benefit from multiple cores at all.
Dual core CPUs are nice because a lot of software CAN max out a single core and then it's nice to have an "extra" core to make e.g. processing user input snappier. Four cores is a case of quickly diminishing returns... for the vast majority of the time, I bet users wouldn't be able to tell if they were using a computer with 2 or 4 cores. 8 cores is just ridiculous unless you're doing some extremely niche stuff and routinely find yourself wishing for more than 4 cores.
-
Saturday 3rd August 2013 09:01 GMT Tim Parker
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"I have no idea why people seem to think more cores = more performance. Of course it does for certain very specific applications like Handbrake or raytracing or compiling huge projects but most software does not benefit from multiple cores at all."
You either have an application that can use lots of cores or, surprisingly, lots of applications that use one (or a couple) of cores.
"I bet users wouldn't be able to tell if they were using a computer with 2 or 4 cores. 8 cores is just ridiculous unless you're doing some extremely niche stuff and routinely find yourself wishing for more than 4 cores."
One of the applications might be a 'kernel', the others can engage in something called 'multi-tasking'. That scenario is not awfully niche. It's somewhat less relevant in the phone, at least for heavy-weight processes, beyond a couple of cores but can still be very useful as far as responsiveness is concerned - especially if you get the match between process requirements and processor capabilities reasonably suited.. hence things like big.LITTLE (although i'm not entirely convinced by 4 x LITTLE either yet)
-
Saturday 3rd August 2013 09:22 GMT Chris Miller
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
Your description is accurate for servers, but most desktop applications only need significant CPU when you're interacting with them, and most people can only interact with one at a time. I'd rather have two fast cores than 4 slower ones, thanks.
The problem with a phone is that, if you somehow managed to load up 8 cores simultaneously, you'd just flatten your battery that much sooner. And not many people are (or would ever want to be) running compilers or ray-tracing programs on their phone.
-
Saturday 3rd August 2013 12:56 GMT Tim Parker
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"Your description is accurate for servers, but most desktop applications only need significant CPU when you're interacting with them, and most people can only interact with one at a time. I'd rather have two fast cores than 4 slower ones, thanks."
Yep - I agree. I was commenting on the OP who said they couldn't understand people equating more cores with more performance - that's a fact of life for some work-loads and the fact a desktop can mostly get away with a single core and a bunch of RAM doesn't stop it being true.
"The problem with a phone is that, if you somehow managed to load up 8 cores simultaneously, you'd just flatten your battery that much sooner. And not many people are (or would ever want to be) running compilers or ray-tracing programs on their phone."
Completely agree - on phones they are mostly a waste of time.. you really only need a decent processor for the task switching/processing and something handling the rendering loads (which can be mostly off-loaded).
-
-
Saturday 3rd August 2013 10:30 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"One of the applications might be a 'kernel', the others can engage in something called 'multi-tasking'. That scenario is not awfully niche. It's somewhat less relevant in the phone, at least for heavy-weight processes, beyond a couple of cores but can still be very useful as far as responsiveness is concerned"
Multitasking can be accomplished very well with a single core--desktop computers did this for decades. On my old single-core iPhone 4 I would often run a music player app and a turn-by-turn satnav app in the background while doing something else (text messages, browsing the web, etc.) in the foreground and everything ran smoothly.
The knee-jerk idea of "I'm running two apps so this would obviously be much better if each app had its own core" is irrelevant without a discussion of how much CPU time each app requires.
Right now my desktop computer is running 117 processes and 853 threads. I'm multitasking my balls off but the CPU load is less than 3% and I would certainly not benefit from having more cores.
-
Saturday 3rd August 2013 12:37 GMT Tim Parker
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"Multitasking can be accomplished very well with a single core--desktop computers did this for decades."
That's time-slicing and it was - and is - piss-awful unless the processes are doing bugger all or have tiny stacks. I can work fine - hence the comment I made about throwing CPUs at phones being not relevant - but if you have multiple processes _actually doing processing_ then lots of cores is invaluable. Multiple cores for development such as the type I do, is pretty much essential to get anything done in finite time - however I realise that's not everyones user case , that said time-slicing won't work here.
-
Saturday 3rd August 2013 12:59 GMT Tim Parker
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
Meant to reply to this in the previous one..
"The knee-jerk idea of "I'm running two apps so this would obviously be much better if each app had its own core" is irrelevant without a discussion of how much CPU time each app requires."
yes - agree - but I thought that was a given. I was commenting about the suggestion that more cores don't mean more performance, when clearly there are many use cases, not just on heavily loaded servers, where that is not true. A phone is one example of something that doesn't need shed-loads of cores in reality however.
"Right now my desktop computer is running 117 processes and 853 threads. I'm multitasking my balls off but the CPU load is less than 3% and I would certainly not benefit from having more cores."
Indeed - a lot of the time my workstation is like that, then again a lot of the time it spends with a load way over 7 (8-core AMD 6100). That being my point, you don't need lots cores all the time but when you do there is no substitute (modulo sufficient RAM and IO for your task) - time-slicing a beefy CPU can be horrendously slow compared with separate compute units, cache thrashing alone can drive performance down the tubes.
-
Saturday 3rd August 2013 13:35 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"cache thrashing alone can drive performance down the tubes."
Really? It obviously looks plausible at first glance, so let's look in a little more detail.
Start with the single-CPU picture.
App1 is running cosily with its own code and data filling the cache.
Eventually, OS scheduler clock ticks, OS scheduler runs, OS scheduler decides to give App2 a go (or App1 yields and App2 gets to run).
How long does it take for App2 to fill the cache with its own code and data ? If it takes any significant fraction of the time between scheduler runs, the system design is wrong, surely? (Cache fills are too slow or the scheduler is running too frequently on compute-bound tasks or both).
And once the cache is (re)filled with App2's code and data, the single CPU case can in principle be faster than a multiCPU at the same clock speed, because the unavoidable hardware and software overhead needed to support multicore (or symmetric multiprocessing, aka SMP, as it used to be called back in the day) can be completely avoided in a uniprocessor system.
Now consider the 2CPU "SMP" case. How much performance-reducing overhead (hardware and software) is introduced by having to keep 2 sets of hardware and software coherent? Is this overhead always present, even when there is only one runnable process? (Hint: Yes it is)
Not saying that multicore/SMP is bad for the applicable workloads.
Am saying that careful consideration is needed before believing much (never mind all) of the multicore hype.
Further discussion most welcome if I missed anything.
-
Sunday 4th August 2013 16:19 GMT Tim Parker
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"How long does it take for App2 to fill the cache with its own code and data ? If it takes any significant fraction of the time between scheduler runs, the system design is wrong, surely? (Cache fills are too slow or the scheduler is running too frequently on compute-bound tasks or both)."
I think this hits the nail on the head - although a more in depth discussion with regards to the exact type of work-loads could take us a long time... For a SMP machine there is an overhead to keep the cache status in step, you're right, although it's overhead is not comparable to cache refill times in the cases i've seen and some theoretical banter (some of it a bit hand-wavy admittedly), that being the issue. Cache refresh is slow compared to cache retrieval in general, usually a couple of orders of magnitude, and has been for a while - that's just due to the comparitive speeds of the memory sub-systems (i'm assuming something similar to the on-die/module RAM L1/L2 and external DRAM model). Cache refill for a bunch of applications is dependant on lots of stuff, mainly amount, but for simplicity lets say they all have roughly similar requirements - which is essentially a fixed cost per switch. If you have to clear the cache when you context switch you pay that penalty, otherwise you're OK - and L2/L3 caches are going up in size in the consumer space (mainly Intel beefing up their L2) so that's good for a single processor. If you're getting a lot of cache usage per processs, quite common in a quite a few workloads, then the number of processes it takes to blow it out goes down and you start paying the penalty sooner. As the number of processes goes up the work to cache refill penalty ratio goes down in a uni-processor machine. For multi-processor situations, if the processes are altering a lot of global state then a lot of the CPU caches are being dirtied and you're taking some hit on reloading the global data - which may or may not be significant, but will slow you down. If the processes are not hitting global data then the individual caches aren't getting refilled (so much) and the over-head of switch is heading towards zero as you're not switching out (as much)... assuming a decent scheduler with half a clue about processor affinity that is, something that seems in short supply still...
..anyway - the point I was trying to make is that slicing on a single CPUs becomes more and more inefficient as the number of active processes goes up, the same is somewhat true about multi-processor setups, especially with processes altering data common to other processes, but degrades at a much slower rate - and the hardware book-keeping usually not a big overhead, especially when compared to the hit sustained when being forced out to external memory (DRAM).
The big killer for performance in multi-processor (not just SMP) environments we find now is disk IO - for which basic workstations with single SATA drives have the most obvious problems.. some workloads, like compilation, you just hit a wall in scaling really, really quickly (few CPUs, say 3-4 sometimes). A few years back, a Sun T2 with 64 virtual cores and half decent, but single, UW SCSI got up to a dozen times speed up before stopping... solid states drives now make all the difference in getting any decent speed-ups in build/test for the stuff we're running.
All in all, i'm not saying that single, chunky CPUs can't give useful multi-process performance but for a lot of work-loads - in particular when you have relatively 'independent' processes - a half decent multi-processor setup with a reasonably good scheduler will scale much, much better.
-
-
Saturday 3rd August 2013 15:52 GMT Chemist
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"Indeed - a lot of the time my workstation is like that, then again a lot of the time it spends with a load way over 7"
Agreed, I'm rendering 1080p/50 video in H.264 and both cores are running melt @ ~70% cpu whilst letting me browse this forum smoothly. If I had more cores available I'd be transcoding already rendered video to 720p/25 for playback on a laptop and if I had more I'd be setting up another editing session by generating proxy clips.
-
Saturday 3rd August 2013 17:40 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"yes - agree - but I thought that was a given. I was commenting about the suggestion that more cores don't mean more performance, when clearly there are many use cases, not just on heavily loaded servers, where that is not true. A phone is one example of something that doesn't need shed-loads of cores in reality however."
Uh, yes, and that's been my point this entire time, namely that it IS dumb to have an 8 core mobile processor because of the workloads on mobile devices.
-
-
Saturday 3rd August 2013 13:03 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"That's time-slicing and it was - and is - piss-awful unless the processes are doing bugger all or have tiny stacks."
Better not tell that to the HYPErvisor people, as they and their followers have built their world around reinventing the timeslicing concept.
"Multiple cores for development such as the type I do, is pretty much essential to get anything done in finite time"
It's almost always simpler and more efficient to deliver a particular workload on a single core with a decent OS and scheduler than on a multicore. Of course if the single core isn't fast enough, your 2nd choice option is multicore, but that immediately introduces hardware and software overheads which the singlecore approach doesn't have.
-
Sunday 4th August 2013 11:24 GMT Destroy All Monsters
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
> Better not tell that to the HYPErvisor people, as they and their followers have built their world around reinventing the timeslicing concept.
LOLNO. You seem to be new to this "Computer" thing.
Virtualization is not about timeslicing. It is about configuration management.
-
Sunday 4th August 2013 12:52 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"Virtualization is not about timeslicing. It is about configuration management."
It's the PFYs and PHBs that are new to "this computer thing", and also to the "cost of ownership" thing.
In the early days of virtualization, my recollection is that it was all about "use one physical server to timeslice the resources between multiple logically-separate OS instances, and save money by saving space and energy and...". Obviously they didn't word it quite like that but that was the clearly recognisable approach for those that remembered. If that's not what you mean by "configuration management", perhaps you could expand?
-
-
-
Monday 5th August 2013 12:54 GMT Mike Dimmick
Time-slicing
Time slicing is no problem at all if the majority of the threads on your system are blocked, waiting for something to happen (e.g. user input, a network request to complete). The battery killers are the apps that poll to find out if something's happened, rather than subscribing to an event that tells them something has happened. It's down to the OS to provide such a notification system, and for developers to use it rather than polling (the OS typically has to provide a way for the app to find out information when it starts up, or to make decisions in response to another notification - it can't *just* have events).
Also, apps should not waste CPU time (hence power) calculating things that the user cannot currently see. On iOS, Windows Phone, and Windows Runtime ('Metro') apps on Windows 8, if an app is not in the foreground, all its threads are suspended. You have to specifically register distinct code to be able to run in the background. The OS only gives these background tasks a limited amount of time to run before killing them, to prevent runaway code killing the battery. Audio players, turn-by-turn navigation or location-tracking apps need to register so they aren't suspended, and Apple and Microsoft check that these permissions/capabilities aren't requested by apps that shouldn't have them when verifying them for store inclusion.
Android allows apps to create as many threads as they like, and doesn't suspend background apps. You might 'need' more cores on Android simply because background apps are unnecessarily wasting CPU time (and battery power).
The rule of thumb is that you need more cores if you constantly see more than 90% usage across all the cores. It's very unlikely that a single active app plus the OS rendering, and a few background tasks (that are throttled anyway) can actually saturate that many cores. Windows Phone and iOS devices top out at dual-core.
-
Monday 5th August 2013 16:45 GMT Tim Parker
(Mike Dimmick) Re: Time-slicing
"Time slicing is no problem at all if the majority of the threads on your system are blocked, waiting for something to happen (e.g. user input, a network request to complete). "
Indeed - most performance questions disappear if almost everything is doing bugger all :)
-
-
-
-
-
-
Saturday 3rd August 2013 10:34 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"My Samsung S3 has 4 cores. But lags all the time when switching from Firefox to Home screen.
it's either RAM or lack of cpu..."
Not necessarily. Could be your "disk" speed or maybe switching to the home screen triggers some kind of blocking network action. And even if the bottleneck is the CPU, throwing more cores at the problem won't fix it unless whatever's causing the lag is multithreaded in such a way that it would be significantly faster with more cores.
But why speculate. A quick web search reveals that there are literally dozens of Android apps that will monitor and report CPU usage. Get one of these apps that displays CPU usage for all 4 cores and see whether or not they're all pegged while you're experiencing this "lag." If not, the problem isn't that you have too few cores.
-
Sunday 4th August 2013 11:30 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"My Samsung S3 has 4 cores. But lags all the time when switching from Firefox to Home screen.
it's either RAM or lack of cpu..."
It's mostly the inherent architectural limitations of Android. It is not an OS capable of consistent real time performance as currently designed...
-
Sunday 4th August 2013 12:58 GMT Anonymous Coward
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"Android is not an OS capable of consistent real time performance as currently designed..."
Define realtime. A decent recent Linux kernel is more than capable of a decent multimedia experience. That wouldn't be the case if it wasn't half decent at RT.
If you mean decent millisecond-level interrupt/scheduler latency and so on - then (1) why is it relevant in a phone or tablet or even typical desktop? (2) pick a properly configured Linux and you'll get far better latency than Windows can manage.
-
Sunday 4th August 2013 18:55 GMT Henry Wertz 1
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
"Define realtime. A decent recent Linux kernel is more than capable of a decent multimedia experience. That wouldn't be the case if it wasn't half decent at RT.
If you mean decent millisecond-level interrupt/scheduler latency and so on - then (1) why is it relevant in a phone or tablet or even typical desktop? (2) pick a properly configured Linux and you'll get far better latency than Windows can manage."
I won't bother defining realtime. I don't think anyone is faulting the Linux kernel (which is not hard realtime, but good enough for videos as you say). It's Android. It will decide to go start garbage collecting, to kill off some background tasks, start up some other background tasks, go do something else for a while,..... you can do exactly the same task, and one time it's instant, the next it takes seconds to complete. Linux, barring naughty device drivers you can keep CPU use below 100% and pretty well say "my app will run within 10ms." Android? It might take 10ms. It might take 250ms. It might take 2000ms. Who knows? There's a few Android programming tricks I've seen programmers recommend to try to make timing-critical stuff smoother, but they appear to be hacky workarounds at best.
-
-
-
-
Monday 5th August 2013 08:20 GMT Zola
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
Uh, why?
My Core i7 desktop computer "only" has 4 cores and my laptop "only" has 2 cores, and both work great.
Not if it has hyper-threading and presents 8 cores to the OS....
But that's beside the point, your 4 (real) i7 cores are significantly more powerful than the ARM cores in mobile SoCs, so having more cores in the mobile SoCs isn't a bad thing.
Of course we'd all like mobile SoCs to have cores as powerful as i7 cores but that's just impractical given current technology. Therefore the best approach is increased parallelism, and more cores (some big, some little) which gives finer grained power control when such control is required (ie. while mobile and on battery power).
-
Monday 5th August 2013 09:38 GMT Matt Bryant
Re: Zola Re: "Dumb" until Qualcomm are able to produce an 8 core chip
".....But that's beside the point, your 4 (real) i7 cores are significantly more powerful than the ARM cores in mobile SoCs....." But the i7 cores are actually TOO powerful for most desktop tasks, they spend most of their time idling whilst the rest of the system (even SSDs) catch up, especially when the user is doing the majority of low-CPU desktop application tasks. In reality, what users need is a few powerful cores, not lots of slow cores.
For a laugh, some of my team replaced a users newish desktop with an ancient single-core Celeron machine for a particularly annoying luser, then they waited for her to complain. After a week they got bored of waiting and asked her for feedback and she had no complaints. All she did was word processing and browsing, she simply did not need the quad-core desktop she previously had.
-
-
-
-
-
Saturday 3rd August 2013 15:36 GMT zanto
Re: "Dumb" until Qualcomm are able to produce an 8 core chip
last time i checked qualcomms quad core beat the shit out of samsungs octa core. this despite everyone prediciting that the a15 would smoke the krait.
untill samsung/nvidia/mediatek come up with an architecture to challenge krait, Snapdragon will be the chip to beat.
FAIL for being such a dumass and falling for the number of cores marketing tripe.
-
-
Saturday 3rd August 2013 00:04 GMT ruscook
“First they ignore you, then they ridicule you, then they fight you, and then you win.”
― Mahatma Gandhi
Oh dear, didn't Intel also ignore multi core when Sun did the SPARC chips. Then when AMD copied Sun, Intel decided that gigahertz (chewing lots of power) wasn't everything and multicore did work. They then played catch up for a while.
Looks like the mobile space may repeat.
-
Saturday 3rd August 2013 03:28 GMT HBT
Except that...
... the next stage in smartphone evolution involves them taking over some of your general computing needs, by docking like a laptop and presenting a proper desktop on an external monitor. See what Canonical are doing.
That will require more cores.
Staggering lack of vision / abundance of cow poo (delete as appropriate) from the QC dude.
-
Saturday 3rd August 2013 06:42 GMT Anonymous Coward
Re: Except that...
"... the next stage in smartphone evolution involves them taking over some of your general computing needs, by docking like a laptop and presenting a proper desktop on an external monitor. See what Canonical are doing.
That will require more cores."
Yes, just like today's desktop and laptop computers all have at least 8 cores.
Oh, wait.
-
Saturday 3rd August 2013 14:08 GMT Steven Raith
Re: Except that...
Your desktop and laptop have less, cores, but they are far more powerful, suck a lot more juice, and have a much higher TDP.
If you want a desktop experience on a mobile-esque device with limited battery power and cooling capacity, you need low powered cores for the mobile experience, and wapping on more cores to help compensate for the lack of 'horsepower' in those cores makes sense for giving a desktop experience that's comparable to a 'normal' desktop/laptop machine - particularly as they are much easier to passively cool, as pretty much any mobile device will be. You can't work around that without active cooling, which at the moment, isn't possible in the mobile (and barely in the tablet) space.
You can't just stick high TDP desktop equivalent cores in a mobile form factor.
Of course, all this relies on someone actually making the mobile-desktop-dockable device workable in the real world - and consumers wanting it.
I'm sort of on the fence on these devices - I like the idea, but I can't see my usage case using it. I have a laptop for sofa surfing, and a monster, overpowered quad core rig in my room for heavy lifting. I suppose I could use one at work, but I'd need a fixed screen, keyboard and mouse. And I already have the laptop....sort of negates it for me.
Could be handy in open plan, hotdesking offices though. Have your works phone and your actual desktop machine in one device...I know a few places personally where that would be a hit.
Steven R
-
Saturday 3rd August 2013 17:45 GMT Anonymous Coward
Re: Except that...
"If you want a desktop experience on a mobile-esque device with limited battery power and cooling capacity, you need low powered cores for the mobile experience, and wapping on more cores to help compensate for the lack of 'horsepower' in those cores makes sense for giving a desktop experience ..."
Completely, completely wrong. Adding more cores does not speed up most experiences. It's only a few niche workloads that benefit from more cores.
-
Sunday 4th August 2013 01:13 GMT HBT
Re: Except that...
completely, completely wrong. Adding more cores does not speed up most experiences. It's only a few niche workloads that benefit from more cores.
I typically have multiple browser windows open at any one time, totalling dozens of tabs. And of course the browser runs background processes for its plugins eg Flash, Java.
And the OS likes to run multiple processes/threads in the background, to do things from time to time.
Add in a couple of Office apps and a media player, and you can easily load 8 cores at times. In fact, I imagine that a larger number of slower but lower power cores is the optimum implementation for such a workload (lots of things going on but none requiring extreme CPU power), in terms of providing a consistently smooth user experience with lowest overall power draw.
-
Sunday 4th August 2013 19:37 GMT Anonymous Coward
Re: Except that...
"I typically have multiple browser windows open at any one time, totalling dozens of tabs. And of course the browser runs background processes for its plugins eg Flash, Java.
And the OS likes to run multiple processes/threads in the background, to do things from time to time.
Add in a couple of Office apps and a media player, and you can easily load 8 cores at times."
Garbage.
What you're saying sounds correct and that's why most people would probably agree with you, but do yourself a favor sometime and run Activity Monitor or Task Manager or whatever your OS comes with and you can see for yourself how little CPU time you're actually using, instead of just thinking "hey I'm doing a bunch of stuff, it must TOTALLY be using all my CPU lolz."
The fact that you think your Office apps are using any CPU time at all shows how ignorant you are; they are just sitting there idle, waiting for you to type something. (Unless you're doing an enormous Excel recalc, or Outlook is indexing... which it only does on idle, though, so that's irrelevant.)
Right now I've got 20 tabs open in Chrome (some running Flash ads), I'm running Photoshop, Libre Office, XCode, iTunes, Mail.app, and a dozen other programs, and my CPU is under 3% load. If somebody disabled 3 of my CPU cores it might take me hours to notice.
-
Sunday 4th August 2013 20:32 GMT Anonymous Coward
Re: Except that...
Good point well made. (?)
There's far too much technology decision making and mouthing off happens based on "I know the answer" without actually doing any research to see if the answer is actually right.
Thirty seconds research (or a bit of realworld experience) would have proved your point in advance.
Though to be fair, Flash does frequently seem to be quite capable of keeping one core fully occupied on my Win7 box without actually doing anything of value, which is why I now run with flashblock enabled. Sorry Flash-based advertisers, but you have only yourselves to blame. Whoever you might be. (Doesn't do that on my XP box, for some reason...)
-
-
-
-
-
Monday 5th August 2013 08:20 GMT Zola
Re: Except that...
Yes, just like today's desktop and laptop computers all have at least 8 cores.
Oh, wait.
My Linux desktop has 8 cores (AMD FX-8350). It wasn't expensive to build either. And yes, I do use all 8 cores. Even just a couple of years ago 8 cores in a mainstream desktop CPU would have been pretty unthinkable - progress, eh?
-
-
-
Saturday 3rd August 2013 08:56 GMT Anonymous Coward
Why is everyone assuming that these 8 core chips are for phones?
We now have several of the large players putting ARM chips on servers where the more cores the better. Plus the fact this guy is ex-intel so thinking of anything other than intel on servers is a no-no. Therefore his statement is not surprising when viewed from that standpoint.
-
Saturday 3rd August 2013 12:42 GMT John Savard
Alternatives
In the case of desktop x86 chips, there are obvious reasons for more cores: Microsoft licensing policy, and the fact that it isn't possible to go from 4 GHz to 8 GHz with present technology. In the case of mobile chips, there may be room to increase the speed per core without increasing power consumption prohibitively.
It is true that faster is better than having more cores, but that assumes you have the choice.
-
Saturday 3rd August 2013 15:57 GMT heyrick
Just a small quibble about the analogy...
"You can't take eight lawnmower engines, put them together and now claim you have an eight-cylinder Ferrari."
You're right, you can't put eight mower engines together to make a Ferrari. But we aren't talking eight individual ARM chips here (that would be very silly), we're talking cores on the same chip with a degree of commonality - same data and address bus, same clock... Well, let's take the core of those eight mower engines, the cylinder, and stick them into one lump of metal, feed them from a shared fuel supply, connected to a shared exhaust system, and synchronised by being connected to the same crankshaft.
Gee, starting to sound like a rough description of a V8 engine, isn't it?
-
Sunday 4th August 2013 18:37 GMT Henry Wertz 1
Re: Just a small quibble about the analogy...
"You're right, you can't put eight mower engines together to make a Ferrari. But we aren't talking eight individual ARM chips here (that would be very silly), we're talking cores on the same chip with a degree of commonality - same data and address bus, same clock... Well, let's take the core of those eight mower engines, the cylinder, and stick them into one lump of metal, feed them from a shared fuel supply, connected to a shared exhaust system, and synchronised by being connected to the same crankshaft.
Gee, starting to sound like a rough description of a V8 engine, isn't it?"
Yes, but your analogy doesn't match at all what is going on inside an 8-core chip, the analogy completely breaks down. The 8-core ARM basically DOES have 8 CPUs on it with just some shared glue. It's not going to run any given app at 8x the speed. Only those specially written apps (video encoding, image processing, certain computational models, ....) would split the job and send 1/8th to each of the 8 cores, finishing in just over 1/8th the normal time. The typical Android programming model doesn't use multithreading in any way whatsoever; even if you register multiple broadcastreceivers and so on in your app, and the app does get multiple signals "at once", Android still only runs one handler at at time. Running a single app, your speedup from extra cores would be approximately 0 (a small increase due to garbage collection running on the second core.)
I'm quite sure having 2 cores is nice, for something timing sensitive like a game or video player, it makes it much more likely that it runs on time while any background stuff uses the other core. Beyond that, I just don't know how many people run multiple CPU-bound things at once on their phone. I sure don't.
-
-
Sunday 4th August 2013 09:15 GMT swissrobin
"I typically have multiple browser windows open at any one time, totalling dozens of tabs. And of course the browser runs background processes for its plugins eg Flash, Java.
And the OS likes to run multiple processes/threads in the background, to do things from time to time.
Add in a couple of Office apps and a media player, and you can easily load 8 cores at times.!"
8 threads do not require 8 cores. Even 800 threads do not require 8 cores. Unless the sum of the MIPS required for each thread exceeds the bandwidth of one core, they require one core. It is a fallacy to believe that because two threads run on two cores they somehow run "smoother". There are so many shared resources (pretty much everything except the L1 cache) that whether the two threads compete for "the core" or "everything else" makes no real difference to the end user experience. Unless of course the software is so broken that the scheduler is locked out for 10's of milliseconds at a time.
Academic challenges (e.g. compute pi) apart, most tasks interact with the network, the disks, the graphics, the user, .... all of which serialize access and often with latencies much higher than the typical task switch which is probably doable in a microsecond on a 1GHz CPU.
That said, you're going to get 5 blades^h^h^h^h^h^h^h^h16 cores whether you like it or not, so really not sure why we're arguing about it here :-)