Future
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.
You are currently reading a thread in /g/ - Technology
You are currently reading a thread in /g/ - Technology
Will ARM be the next step on cpu level freedom?
>>54124314
Fuck no. ARM are champions of binary blobs.
>>54124426
basically we will be tied to at maximum core duo 2 quad and the best amd processor of 2013
>>54124812
Pretty sure there are ARM SOCs out there that don't require proprietary blobs to boot and be used. Not sure about being able to take advantage of everything on the SOC without proprietary blobs though.
lol no
intel will soon release goldmont atoms and better core-m chips. ARM is as good as dead.
>>54124426
this
RISC-V and OpenPOWER/SPARC are the best bets for anything free and actually useful, as opposed to some gimped proprietary out the ass phone chip lazily thrown onto an unexpandable SBC
>>54125117
literally what the fuck does the TOP500 have to do with any of this
trying to shill an architecture for the desktop/mobile markets using supercomputing is like trying to sell a retard a geo metro because chevy wins a lot in NASCAR
Nah too weak and consumes too much power. Supercomputers and servers need both of those things and ARM can't deliver. Not only this but x86 also dominates the desktop and laptop market especially because of x86 software compatibility. To add insult to injury x86 is starting to take over on the tablet market. Now all ARM has is the phone and embedded market.
Things are looking real gloomy for ARM right now especially with intel releasing apollo lake (tablet/phone x86 SoCs).
pic related: the biggest "fuck you" to ARM intel has ever released.
>>54125124
Sadly SPARC should be avoided too now since it's Oracle's property.
>>54125141
It's a sign of what the best CPU architecture is. Super computers require massive raw power but good energy efficiency at the same time. This is also what tablet/laptop manufacturers want. You figure out the rest.
>>54125289
Found the intel shill. Too bad they didn't fire you this time.
>>54124314
Honestly, you should use SPARC or POWER if you want to use a free chip on your desktop/laptop.
>>54124314
>mfw this thread
>>54125336
x86 is by far one of the worst architectures we have.
Intel has poured billions of dollars to fix all the mistakes and make x86 almost as good as the RISC machines. In the end they gave up and hid a RISC machine that they stole from DEC to make the Pentium and newer "x86" processors.
>>54125354
Whatever you say armcuck
http://www.anandtech.com/show/10256/intel-unveils-apollo-lake-14nm-goldmont
>>54125418
Your butthurt about it doesn't make x86 any less magnificent. It's not like ARM has been free of sin either. In the end people want x86 software compatability and performance, ARM can't deliver either.
>>54125421
Atom is getting so good it's making Core M pointless.
>>54125539
Only Apple's ARM solution is comparable to Intel's low end tablet chips. Everyone else is cheating Geekbench scores via SHA1 hardware acceleration and the droidcucks lap that shit up.
>>54125289
Xeon-D laptops when?
Ehh I wish but it will never happen maybe some day they can be on par with current processors. Speaking of which atom by intel is getting pretty good
>>54125321
>what is OpenSPARC
>>54125336
>It's a sign of what the best CPU architecture is
no, not really, many of the fastest supercomputers throughout the years until very recently have used a lot of very unorthodox architectures like vector processors and bit slicing or more mildly weird shit like Itanium chips that are absolute garbage for desktop use (let alone mobile/tablet use, good look scaling a connection machine into a phone)
x86 didn't win out in that market because of what little merit it had buried under all of that legacy bloat, CISC/RISC identity crisis kludge and shitty RAS/enterprise features, it won out because it was cheap as fuck compared to traditional HPC staples like POWER, SPARC, MIPS and Alpha and finally "good enough" for the role, which anymore emphasizes on GPU compute which is much faster for floating-point heavy and parallel-optimized supercomputer workloads than any CPU will be anyway
that's not even mentioning that the highly optimized workloads supercomputers carry out is about as far as you can get from the god awful legacy kludge you're running on the gaming shitbox you made that shitpost from
>>54125418
>and hid a RISC machine that they stole from DEC
what the fuck are you talking about? P6 doesn't seem very similar to the Alpha at all
>>54126261
SPARC is dead
>>54124314
OpenARM Foundation when?
No.
Arm is a low power architecture.
Loongson is more likely to take over high performance computing.
>>54126378
>ARM is a low power architecture
Literally a meme.
>>54126378
Except that it's not, see core-m and cherry trail atoms. ARM is now actually more power hungry than x86 in terms of performance-per-watt.
>>54126346
not any more dead than POWER
the M7 seems like a pretty damn nice chip, though I'm hesitant to take Oracle's official numbers at face value
>>54124314
nah man, the future is that new arch ibm came up with recently
>>54126534
Just Atom, Core M consumed as much as a regular Ultrabook CPU when turboboosted. The 4.5W SDP is when the CPU is clocked at 1.1GHz, it will gobble up 21W when fully turboboosted.
>>54126552
>it's so revolutionary that I can't even remember it's name
>>54126567
>Just Atom, Core M consumed as much as a regular Ultrabook CPU when turboboosted. The 4.5W SDP is when the CPU is clocked at 1.1GHz, it will gobble up 21W when fully turboboosted.
4.5w is actually the TDP of most core-M chips not the SDP. In addition while core-M chips can consume more than 4.5W of electricity they can only do it in bursts as the battery heating up and increase in thermal stress put the user and Core-M chip at risk.
What's funny is that even though core M chips are throttled in sustained cpu intensive tasks they still outperform a 2.24GHz A9 chip (ie iPad pro) in geekbench multi-core floating point tests and use less electricity too.
We would have already seen core-M chips with a TDP-down of 3.5W since android can run on x86 already (see zenfone 2). But alas jewtel wants $300-400 for these puppies and few people would bother with an x86 phone if it costs $1,000+ and had the same performance as your eight-core 4ghz snapdragon housefire on ~$500 flagship android phones.
>>54126753
*We would have already seen core-M chip phones with a TDP-down of 3.5W
>>54125289
>Nah too weak and consumes too much power.
Stupid cunt.
x86 is for compatibility with ancient chips and ancient software, not power.
(and the ancient software part is moot now, that new OS releases with incompatible APIs break compatibility with abandonware anyways, no matter that the chip could theoretically run it).
ARM, especially the new 64-bit instruction set, is a new, powerful and optimized architecture, more akin to SPARC and POWER than to commodity x86.
ARM is a clean slate, compared to x86.
Nothing divine or extraordinary, just a fresh start, trying to avoid the mistakes of the past and incorporate what we learned the past 40 years, into a new architectural design.
Imo, it's an embarrassment for x86 (performance-wise, not compatibility-wise), that ARM in less than a decade, and using previous gen fabrication processes, almost caught up in processing performance with x86 chips, that had 40 years of continuous development.
All the while beating them in power draw, efficiency, thermal design, etc. .
The only reason it is 2 years behind Intel (at the moment), is that Intel had 35years headstart.
ARM positively outperforms x86 in every aspect, aside from compatibility with 8-bit Motorola chips from the '70s.
That's x86's place in the future;
Running abandoware in niche scenarios.
For everything else, there's ARM.
>>54126873
see >>54126460
Please tell me what ARM chip has the processing power of a Xeon-D 1540 and energy efficiency. Go ahead, I'll be waiting.
Also, both 32-bit ARM and x86 CPUs were introduced in 1985 you massive retard.
>>54125117
Must be why Intel laid off 12000 employees today.
>>54124314
It's next level of vendor-locking and not effective of energy usage.
>>54126945
>>54125117
The Itanium bump is pretty funny. I had to check the Wiki article and apparently Intel "may" be developing a new one. Literally why. Maybe with the recent news of layoffs the Itanium can finally die for good.
Is a Intel atom generally better than a mediatek chip.
Say a nexus player vs fire tv?
>>54126873
magazine columnists and laymen suffering from a severe case of dunning-kruger syndrome have been regurgitating this dumb shit for nearly thirty years since pure RISC architectures like ARM, MIPS and SPARC began making headway in the late '80s
the reality is that running "abandonware" (do you even know what this word means?) isn't a "niche scenario" and the majority of computer users aren't trendy hipsters who pay out the ass for the latest shiny pieces of software, and don't give a shit about unnoticeable performance/e-dick gains from liquidating their software base every 10 years
>(and the ancient software part is moot now, that new OS releases with incompatible APIs break compatibility with abandonware anyways, no matter that the chip could theoretically run it).
legacy software is not solely composed of shitty 16-bit kludge games
well-behaved 32-bit NT software will run just fine in modern Windows versions
the same even goes for 16-bit software in 32-bit versions of Windows (see pic related)
>Imo, it's an embarrassment for x86 (performance-wise, not compatibility-wise), that ARM in less than a decade, and using previous gen fabrication processes, almost caught up in processing performance with x86 chips, that had 40 years of continuous development.
aside from the part where the ARM architecture is 31 years old, SPARC and MIPS were also pretty much home-grown by small companies that weren't nearly the juggernaut Intel was, yet their early iterations were workstation-class and outperformed x86 chips heartily.
it's not difficult to make a new architecture that will out-perform desktop shitbox chips, it's more a question of whether it's worth the effort to bother with it
>>54126998
HP pays them to, it has a slightly better RAS featureset for mission-critical systems and it's also the only architecture that can run current versions of VMS or HP-UX
>>54127236
What's funny is that Windows 10 with all this legacy support is now actually lighter than Android 5.0+, iOS7+, and OS X, weighing in at 3.9GB for a full install. Most current phones lose about 5-7GB of the storage to the OS, OSX El Cap uses 15GB.
>>54128731
>Windows 10 only requires 3.9 GB of hard drive space
>minimum system requirements call for 20 GB of hard drive space for the 64 bit version
So you really expect me to believe that they require you to have 5x as much hard drive space as you actually need at a minimum?
>>54126355
>OpenARM Foundation when?
You can already make open ARM chips, all the ARM MCUs are well documented with open software. It's just the phone SoC companies that behave like idiots. Freescale is pretty good with their CPUs but their A72 chip has no GPU and is expensive because it's loaded with 10gbs networking equipment.
If enough people asked you could probably get one of the companies to make an 8 core A72 chip with a decent gpu, quad channel ddr4 and many PCIe lanes. Something like that would only cost about $100 if mass produced.
But at the moment they don't think there's a market for it.
>>54126261
>what the fuck are you talking about?
http://www.nytimes.com/1997/10/28/business/intel-and-digital-settle-lawsuit-and-make-deal.html
>>54125730
Check out the Thinkpad P70. It's got a Xeon and supports up to 64GB of ECC RAM.
>>54127236
ARM was built to be as cheap as possible while still delivering decent performance, not to compete with x86. That's why they're taking over the MCU market with their Cortex series.
>>54130951
interesting, but if anything that implies the Itanium used jacked DEC IP, the Pentium Pro was already old news outside of servers by the time that happened
>>54131136
maybe now but I don't believe it was back then, though I can't really remember why Acorn decided to jump on the RISC bandwagon
>>54126552
IBM's z196 is the fastest CISC processor available now https://en.wikipedia.org/wiki/IBM_z196 and before DEC went 64bit with Alpha, VAX was the fastest CISC processor on the market. x86 has always been worse.
>>54126998
>Itanic
>>54132612
Probably because a RISC machine is easier to implement and requires less silicon to produce.
>>54132925
Don't forget the 68k, either.
I wonder if the PC would have taken off like it did if they had gotten the go ahead to base it around a 68000.
>>54133349
apparently when they were trying to ditch the 6502 they just couldn't find anything suitable for a graphical system that met their goals, so they just went and built their own
ARM was definitely cheap to implement though
>>54126873
>>54135544
IBM would have been using something better than MS-DOS.
It would have gotten virtual memory sooner instead of that horrible 286 segmented mode.
There would be no 640K limit or A20 gate.
The BIOS botch wouldn't have happened.
Microsoft would not have become a big OS company.
IBM was also considering using the IBM 801, a predecessor to the RT PC and PowerPC. That might have ran Unix.
>>54135544
Are there any modern 68k machines?
>>54132612
When Itanium was first released, DEC was already on its deathbed.
On the other hand, the first Alpha was released in 1991, and the Pentium Pro architecture was released in 1995. So Intel had 4 years to steal DEC's IP.
>>54138847
If DEC was getting licensing fees from Intel for each x86 CPU they sold, they could have survived longer.
