>>47669938 This isn't aimed at servers, this is aimed at the Macbook and future similar low-end laptops.
>fanless >fewer watts = more battery life >equivalent performance to Core M >ARMv8/arm64 offers a standardized platform with none of the "durr you need a BSP" bullshit of 32-bit ARM, so you can just use a generic kernel >all on less silicon (and therefore cheaper to build) than x86
Basically these things will make god-tier netbook/tablet/ultraportable laptop CPUs and there's already a GNU/Linux ecosystem in place with a full port of Debian 8.
>>47670091 >30-40 of these cores will still be fucking slow, nowhere near worth $1k USD. either you fucks are shilling or fucking clueless. You have no idea how fast the recent a57/a72 cores are. They're still behind haswell per thread but the size and cost differences are going to hurt intel badly.
Even if you need two entire a72 cores to compete with intel cores, intel charges $4300 USD for an 18 core xeon. 36 a72 cores should be about as expensive to produce as a highend gpu. Intel will suffer badly when it comes to price/performance. Also that is being pessimistic about ARM performance, rumors persist that Intels cpu interconnects are not great and people are seeing non X86 CPUs scale better in large groups beyond 20cores.
There's a very real chance intel will be suffering in both power use and price as ARM server mature.
>>47670215 >>the wattage is an accurate measure of power. >http://en.wikipedia.org/wiki/Power_%28physics%29 >In the SI system, the unit of power is the joule per second (J/s), known as the watt in honor of James Watt, the eighteenth-century developer of the steam engine. in terms of computing power, i would absolutely not consider it powerful compared to contemporary offerings from intel.
>>47669938 I'm a fan of ARM's as microcontrollers and microprocessor enthusiast. Despite that fact I would really want to see some real concurrency for Intel. Bigger concurrency on the market means better prices and better quality.
Performance per watt is whats highlighted here more than anything. ARM is saying that a Cortex A72 based SoC can compete with a Broadwell Core M while drawing a fraction of the power. It doesn't matter if intel's core has higher IPC if its not clocking high enough at low enough power. ARM is showing they have a fucking monster. Even more impressive is that they're expecting this level of performance from TSMC's 16nm FinFET node which is a far cry from intel's 14nm Trigate.
A72 based chips fabbed on Samsung's 14nm FinFET LPP and TSMC's 16nm FinFET+ will make for astoundingly energy efficient devices that deliver real performance.
>>47670304 >How long until Apple ports Os X to arm? It's already done, they can end X86 mac sales tomorrow if they wanted to. They're just waiting until it's convenient to dump intel. Apple won't even pay $1 or $2 extra per unit to keep important suppliers out of bankrupcy, what do you think they will do to intel who charge them $300-$400 for CPUs that probably should cost $50?
>>47670443 >>all these people missing the point >ARM is showing they have a fucking monster what the fuck man. YOU are missing the point. a <1W cpu isn't a fucking monster. if it could outperform ordinary desktop cpu's, sure then i might call it a monster, but as it stands it's so far away from "directly taking shots at intel", " incoming butthurt next year for intel user". it's not even a real benchmark, it's simulated for fuck's sake. it's going to throttle in any real device, just like how current ARM chips throttle in real devices.
>>47670467 > it's going to throttle in any real device No. The A72 was designed specifically to end throttling issues. A chip doesn't need to draw a lot of power to be a monster, you tech illiterate retard.
>>47670544 http://www.fool.com/investing/general/2014/01/29/dont-buy-amds-arm-server-hype.aspx >The upcoming ARM-based Opteron A1100 scores 80 in SPECint_rate >According to Intel, its own Atom C2750, with a 20W TDP, scores 106 on this same benchmark kek
But they don't bother to manufacture chips with as big die as the Intels, because they focus on low power and efficiency, not processing power.
If ARM were to match Intel in power consumption and die, they'd be obviously more powerful.
That's expected of course, since x86 is all about '70's compatibility, while ARM is a clean, more optimized new architecture that aims for performance.
And mind you, ARM chips aren't even on the same level of fabrication process as Intels, because of the headstart the latter had. If they match them there too, we'll have an architecture that is better than x86 in all respects (power, energy, etc.) except for compatibility with 16bit chips.
>>47671867 Are you really going to tell me that while porting OS X to ARM for iOS they didn't create an in house ARM version of desktop OS X? With all the work Apple's putting into their CPUs I'd be surprised if they didn't have OS X ported for ARM.
>>47671967 >if ARM is to become successful Anon, ARM is successful by income and by far the most successful CPU designer out there by volume. Really, *by far*.
> since it makes their platform more viable Nothing you do will really matter much.
Android already beat Windows in terms of devices and time share people spend on them. And that one is predominantly ARM.
You could "help" ARM by making a bunch of extra servers with it that aren't prices like existing NAS and stuff, but really, it's not necessary. Everything is already pretty much happening and entities with far more manufacturing power than you are behind it.
>>47672088 It really was shitty though, any processor that requires tons of extra work is by definition shitty. It's not supposed to make it more difficult to make the lives of developers more difficult.
Sure a few studios like Naughty Dog did a lot with it, but they didn't have a choice and they had direct access to a lot of resources most developers didn't. They work directly with the team that designed the PS3's APIs.
>>47672069 Well, you can run checklists and stuff off an android. You do communication. And spreadsheets and office stuff also is already happening. Most people and their employers will like doing that on stupid simple Android.
So that leaves maybe 2% of the user base who is not a plain consumer or "light" computer worker. They'll be the ones that will certainly continue to use Linux, or if Microsoft changes things around, perhaps also Windows. But really, I suspect Linux.
I also don't doubt that maybe 3-5% of the devices will continue to be servers and stuff. But these will almost certainly mostly run Linux anyways, and they'll really not care if it's ARM or x86 as long as compute power is sufficiently good and power draw low.
Not saying the transition to ARM is absolutely certain, but if current pricing policies and all that continues, it is gonna happen. Never mind that China clearly also wants off US dependence in terms of computers, and ARM is clearly the way to make that happen.
>>47672160 And more importantly, they're fast now. I usually tend to call the processing power most people need (for office, web browsing, porn, social networks, HD movies, and all that usual stuff) about "Pentium 4" - tier.
The ~$100 models for these boxes are now clearly a good notch above that already.
Basically, they really can work as "desktop computer" type of device for most people.
But to be fair, probably half of the most interesting ones on the market now are suddenly x86 Intel-based. They obviously cut prices and supplied the Chinese manufacturers. And Microsoft is also trying to get along for that ride specifically, they get legal OEM Win 8 and 10 very cheaply.
So it's not all entirely decided already or anything.
All you gtartds are assuming that Intel is resting on it's laurels and won't adapt and ARM will be the end of them.
Except it won't.
Most ARM code and libraries are slanted towards mobile devices, a market that Intel has only recently been putting some serious thought into.
For ARM to really threaten Intel, they has to muscle their way into high performance computing, an environment that ARM has no experience with.
Further, ARM would need to convince developers that an ARM desktop CPU has a chance and should have programs written for it. Debian has a full port and thats fine an dandy but thats not going to convince developers that the ARM chip in a desktop environment will work. If you can't get OS developers on board, then application developers have no interest.
And the final nail, ARM would have to convince corporations that are running programs written exclusively for x86 to either get a new ARM compatible version or rewrite it for an ARM based system. Corporations are penny pinchers inherently and will stick with what they know will work. That's why so many companies ran Windows XP well past 2009.
>>47672469 No fucking shit, that was my point >>47672088. But now you're basically saying effort = bad, which is not true at all. Also I don't know about you, but I'm talking about the Cell for high power computing, not for a little gaymen box
All this shitposting about processing power. >I still haven't sent any arm processors that can process 10 bit h.264 videos without turning into a fireball. >I have yet to seen any attempt from Intel to improve mobile CPU s.
>>47670511 >How will it affect gaming performance for Steam and other games? Negatively, since nothing works on arm and never will because x86 is too big to fail. Nobody is going to write AAA games for arm, all consoles and windows are x86 and that wont change for a long time.
Mobile ARM SoCs are low power chips because they're designed for it. The transistor library implemented by the design software, specific types of insulators used, and the metal stack are just a few big factors in separating a mobile chip from a high power desktop chip. ARM IP absolutely can scale upwards in clocks. A many cores vendor a couple years ago clocked Cortex A9 cores up to 3ghz which is a massive uplift over the speeds that were implemented in mobile parts.
>>47673945 The CPU in the Wii U actually is like a cross between PowerPC and POWER 7 characteristics, it's not just a triple core GameCube CPU with quadruple clockrate, and the rest of the hardware is far more than GameCube as well. It's current gen. Also PowerPC isn't dead.
>>47673992 If you look at the chain of replies, it was dealing specifically with steam and gaming (especially in pc). In that space, powerpc is dead with only nintendo and its 1st party recycled ips keeping it barely alive.
x86 is the platform of choice for anything gaming and that isnt changing anytime soon. 2/3 consoles adapted to it, that alone should tell you that x86 is the future of pc and pc-like gaming and that arm will not be relevant in anything other than smartphones and tablets for years if at all.
>>47674031 you underestimate the legacy factor, nobody will switch to arm overnight. it will take years and years after arm makes something truly worthy switching to
good luck telling your boss you need to throw out all those x86 xeons and switch to arm because they use less electricity but its going to cost millions of dollars and will only pay off after 20+ years
>>47671916 Application incompatibility most likely forced them to core m over any arm superchip. Although the rosetta powerpc emulation when the first core duo intel macs came out was actually really good considering what it was.
i still remember the mooing fans though on my black macbook, what a piece of shit it was.
>>47674047 You're a moron. Not everything is about MOAR POWA, and it's not about just replacing equpment because new stuff's out, but when that equipment hits its EOL depending on its use ARM could be a big possibility.
>>47674079 They'll probably use ubis again if they swich architectures.
>>47674047 AMD already seems to be pushing into the high performance ARM market with K12. That name cleary indicates they will go for maximum performance and not some low power shit.
When the ARM server CPUs offer superior options than x86 solutions they will be used in some cases. Nobody will throw out their shiny new 18 core xeons for them any time soon but when you build a new HPC farm the ARM cpus will probably be an option.
>>47674317 It would be safe to assume that it _was_ throttling during a benchmark.
These chips are designed for performance as far as computing, they are designed for low power/high battery life/no fans.
It's also first gen; I wouldnt personally get one but I like where they are going. I have a 1.3GHz quad core baytrail and it easily does normal everyday tasks.
Would I compile a kernel with it? No Would I render HD video with it? No Would I do audio production on it? No Would I gayme on it? No(except home-streaming, Which works fucking fantastically on these)
But would I do basic shit like browse, office work, etc? Absolutely.
>>47671279 >booting up your 128 core ARM desktop PC and quick switching between VMs of any x86 OS - VMs that run the OS almost as fast as they would on their native hardware.
http://en.wikipedia.org/wiki/Hardware-assisted_virtualization >Full virtualization is used to simulate a complete hardware environment, or virtual machine, in which an unmodified guest operating system (using the same instruction set as the host machine) executes
Unless there's been some new shit happening, I'm guessing that cross-arch virtualization's gonna be slow as fuck.
It it too much to ask for a 12" ARM powered laptop with a business grade chassis, a decent screen (no shitty TN), that uses the space freed up by the tiny motherboard for a big ass battery that will last for 24h.
>>47674608 yes, cross architecture virtualization would have to run like a full fledged emulator, meaning there will be the massive performance penalty over HW-assisted virtualization that is typical of emulators . But with that many cores (and that far in the future), maybe the performance penalty could be reduced to a negligible amount by then
>>47674760 >>47674745 And, what exactly are you trying to say... the ipad clearly has the advantage(because more coars)... if it was tri-core ipad vs quad-core macbook then maybe you'd have some sort of point.
>>47674725 I know, this chart is far from enough to attempt an approximate power comparison. And of course 4 threads != 4 cores, I just felt like being a dumb shit. >you an AMDfag? I stick to them when I do purchase x86 hardware but I tend to avoid it.
Also what does AMD have to do with that? Their CPUs aren't hyperthreaded.
If they redesigned it for, say, 96w or 140w I think the first step they'd have to do is moar coars. I'm trying to imagine a 40 core 2.8GHz ARM desktop. Good lord.
my Qualcomm Snapdragon S4 Plus MSM8960 dual core 1500MHz server with 2GB RAM actually compiles code reasonably for what it is. (U.S. SGS3 with USB OTG dock providing power to the unit, with a powered hub attached for 2TB external storage and an Ethernet dongle bridged to wlan)
>>47674047 You overestimate the legacy factor. The x86 legacy only makes sense for Windows. If they're using Linux, then it would be pretty easy to switch to ARM. There was a time when nobody would use Windows or x86 on a server. A lot of people used UNIX on RISCs. A time before that, UNIX was considered a toy OS too and real computers ran OSes like OS/370, VMS, and MULTICS.
>>47671508 >>47671974 >Sony dropped the Cell for a fucking 8-core AMD CPU, that's how shitty the Cell was The cell was a powerful CPU at the time. But it doesn't hold a candle to modern OoOE pipelined CPUs.
>>47675122 for highly specialized tasks, a 40-core ARM cpu might be useful. but you're kidding yourself if the cost of manufacturing only scales linearly with the core count, and for general-purpose computing i will sure as hell prefer a smaller number of high-performance cores than a larger number of low-performance core.
>>47674789 >But with that many cores (and that far in the future), maybe the performance penalty could be reduced to a negligible amount by then That's not how it works. Single thread performance is still extremely important because all problems can't easily be parallelized.
>>47675158 >The x86 legacy only makes sense for Windows. No not really. One might think it's just to update everything on linux and you are good to go. In reality you need specific compiler versions and os versions to run your legacy shit on linux.
Thread replies: 170 Thread images: 10
Thread DB ID: 55771
All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the content originated from them. If you need IP information for a Poster - you need to contact them. This website shows only archived content.
If a post contains personal/copyrighted/illegal content you can contact me at firstname.lastname@example.org with that post and thread number and it will be removed as soon as possible.