ARM
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
ARM GENERAL THREAD
To be honest i belive that arm is the shit, i have rpi and some orange pi, i had server in scaleway. And i looking futher to make small arm server farm and to learn loadbalacing and multiple servers computing.
Another thing is that a lot of u are paranoid about Intel mangament, botnet within motherboard and rootkit shit, so arm is the anwser.
To be honest i am looking after to some desktop arm, maybe i will make some shity 7 inch rpi laptop or wrist comupter, i wouldn't be to hard, but still it use a lot of power and i don't have idea for box and design xD
the last was saged, i can't make threads xD
>>51389397
Could arm really replace x86 and all the software on it?
>>51389611
i believe it and shill for it, practly all the packets i was looking for had arm version or just make install without any problem, ofc gaming and win shit won't work, but it's cheap and opensource sort of xD
>>51389746
>xD
I'm thinking of buying an arm board, My goal is messing with arm asm and maybe even write an operating system for it.
What is a recommended board for this purpose? (as in good documentation and datasheets available)
What does ARM even offer compared to x86 beside a lower price?
>>51389871
RISC
>>51389905
>Implying it matters
http://www.eetimes.com/author.asp?section_id=36&doc_id=1327016
>>51389871
>What does ARM even offer compared to x86 beside a lower price?
Nothing, price is literally the only real reason they are used ever.
>>51389871
lower power consumption.
>>51389988
We don't know the limit of possible microarchitecture optimizations on ARM though, x86 is already near it's limit, while ARM was never optimized for targeting desktop, so we don't know how much faster they could be if power consumption was not an issue
>>51390091
That line is getting very thin m80. intel already released the Z8700 Atom x86_64 processor with a SDP of 2 watts and an approximate max tdp of 6 watts. It only cost $35 too.
ARM better step its game up, who knows how much more efficient the successor to 14nm cherry trail will be.
>>51390110
>We don't know the limit of possible microarchitecture optimizations on ARM though, x86 is already near it's limit, while ARM was never optimized for targeting desktop, so we don't know how much faster they could be if power consumption was not an issue
Are you actually implying devs would optimize software when given a powerful i3-i5 level performance ARM processor? Look what happened when desktop GPUs got more powerful, devs have started to not give a shit about optimizations since they can target R9-270X or greater GPU performance.
>>51390186
>It only cost $35 too.
but u most include cot of shitty intel hardware and guest what, intel have monopol on chips too, so that's another one 30-50 usd
>>51390186
>posting this same shit again
The entire Samsung Galaxy S6 draws less than 5w when running a GPU bench at max clocks in its highest possible voltage state.
A Atom chip that draws 6w just by itself is not even remotely close to competing against ARM.
>>51390269
>but u most include cot of shitty intel hardware and guest what, intel have monopol on chips too, so that's another one 30-50 usd
I don't understand. Intel only wants $35 for a single Z8700 Atom. Pic related is proof.
>>51390342
And guess what?
The Exynos 7420 used in the S6 is estimated at just $29.50, and it was the first production 14nm LPE chip they ever produced. Price was only that high because yields were low.
ARM chips are not expensive, not even high performance ones like the Exyno 7420 or A9.
Continuously, in every single thread, shitposting about intel's laughably misleading SDP and prices are not doing you any good.
>>51390342
sorry for not speking cleary BUT U NEED ITEL MOTHERBOARD FOR IT
>>51390341
>>posting this same shit again
>The entire Samsung Galaxy S6 draws less than 5w when running a GPU bench at max clocks in its highest possible voltage state.
That's the GPU stress test which is irrelevant in this CPU thread. Though I am impressed the 7420 GPU can draw so little power, not sure how much power the 16 execution units on the Z8700 draw under load.
>A Atom chip that draws 6w just by itself is not even remotely close to competing against ARM.
Do you know what the max TDP of the exynos 7420 is?
Anyway yeah ARM is still more power efficient, I never denied that. But now it's a 6 watt atom vs 3-4 watt ARM. Before it used to be a 20 watt x86 cpu vs a 5 watt ARM.
What I'm trying to say is x86 processors are getting more power efficient and can one day match ARM power efficiency if not surpass it.
>>51390376
Must not be that expensive since the Predator 8 tablet (using the Z8700) is only $300.
My autism is triggered.
>>51390441
Shit eating retard, learn how to read.
That is entire system power draw. Entire system.
It was done specifically to show how much power would be pulled in a totally unrealistic scenario that users would never experience. It is pulling more power than it ever could with all CPU cores fully loaded. The 7420 is not a 3-4w chip, it is a 2w chip as is the A9. That is the reason why intel claims a 2w "SDP" trying to make their chip sound comparable when in reality it isn't even remotely. Having three times higher power draw is not comparable.
Atom chips are absolutely nowhere near ARM in terms of energy efficiency. They never will be either.
>>51390372
>ARM chips are not expensive, not even high performance ones like the Exyno 7420 or A9.
I never denied that boss >>51389871
> Continuously, in every single thread, shitposting about intel's laughably misleading SDP and prices are not doing you any good.
The SDP is misleading, I accept that but I use it since it's the only power consumption spec intel gives for the Z8700.
>Misleading prices
Can you please elaborate
>>51390522
FOR FUCK SAKE INCLUDE MOTHERBOARD, x64 x86 use MOTHERBOARDS and they eat too, and they cost too
EVEN ITX
>>51390063
what about power comsuption?
>is a legit question, this inb4 is cause this is /g
>>51390522
>That is entire system power draw. Entire system.
That was only the GPU being stressed though. Can you show me a link where all 8 cpu cores are stressed and show how much power they draw? Or does this mean this SoC is only capable of pulling 6 watts from the socket and will throttle frequencies to stay at 6 watts of power consumption?
>It was done specifically to show how much power would be pulled in a totally unrealistic scenario that users would never experience. It is pulling more power than it ever could with all CPU cores fully loaded. The 7420 is not a 3-4w chip, it is a 2w chip as is the A9. That is the reason why intel claims a 2w "SDP" trying to make their chip sound comparable when in reality it isn't even remotely. Having three times higher power draw is not comparable.
That is pretty jewish of them, I must admit.
>Atom chips are absolutely nowhere near ARM in terms of energy efficiency. They never will be either.
For now. They could be someday, intel is still interested in entering the mobile market. Why are you so sure of yourself anyway?
>>51390726
Armshills don't want you to know but xeons have better power efficiency than any shitty arm chip on the planet.
If arm chips were so fucking efficient as the armshills claim then servers would use arm chips exclusively. Think about that.
>>51390554
The Z8700 is a mobile chip you dumbass. Manufacturers are responsible for the mobile motherboard, not intel.
>>51390342
ARM on suicide watch!
ARM is for poorfags
>>51391056
Basically, I'm amazed they even collected enough money for a shitty raspberry pi
>>51389397
Don't they already have x86 processors for phones now?
>>51391192
They do. The zenfone 2 is like 200 buckaroos but has the performance of a note 3. Unfortunately it has a lower batterry life due to the screen or the atom processor. I'm not really sure though.
>>51389397
Strange thread anon but I'll send some arms
>>51390798
Try actually reading the page that was from, you enormous dipshit.
Those are synthetic projections made by the author, not real measured values.
Ryan Smith created this nonsensical metric when evaluating the prior Exyno 5433. ARM cores do not behave this way in the real world.
>>51391460
Do you have the link? Also do you know the max tdp of the SoC if all 8 cores are stressed heavily?
>>51391495
Its a big.LITTLE chip. Both of the core clusters are not used simultaneously.
Jesus you're fucking stupid.
>>51391522
Then how is it able to score 5,000+ on geekbench? Are the four A57 cores that fucking powerful?
>>51391557
pic related proof
>>51391600
>single big core scores 1480~
>hurrrrrrrffffff how does 4 cores reach 5000~ durrr what is multicore scaling guys
>>51391600
>7122 multi-core integer performance
Jesus fucking Christ that's based.
>>51391626
So then this >>51391522 nigger is lying.
Both quad core clusters CAN be used at the same time, right?
>>51391657
A single A57 core is scoring 1480~. Perfect multicore scaling across all 4 of the A57 cores would yield a score of 5900~
It only shows a score of 5100~ because multicore scaling is not perfect.
You are so fucking retarded you should honestly be killed for it.
>>51390498
CAN'T UNSEE
>>51391680
Jesus christ so the answer to >>51391557 is YES?
Why haven't we made an 8 a57 core ARM chip then?
>>51389871
-higher flops per watt in low power devices
-embedded industry standard
-ISA/design is licensable
>>51391727
>Why haven't we made an 8 a57 core ARM chip then?
They probably have, but it not be very useful for phones, it would be a microsever chip.
>>51391772
The Opteron A1100 is exactly that.
>>51391802
MOAR COARS!
>>51390726
Not any higher. The iPads have been worse on Battery Life / Battery size ratio than the Macbook Airs for a long time. The iPad Pro fares worse than the Macbook Pro.
>>51391738
>-higher flops per watt in low power devices
Half-precision only. And you have to be full retard to use half-precision for anything where you remotely care about accuracy - can't represent even a 4 digit number properly with it.
>>51391913
armshills btfo
>>51391977
>>51391913
You're both clinically retarded.
Nvidia is the only company who made a point to advertise their FP16 performance in claiming that the Tegra X1 was a 1TFLOP SoC.
The X1 isn't even remotely relevant, and ARM's Mali GPUs as well as Imagination Tech's latest line all primarily function with FP32 as per usual.
If you want to use half precision you have to code specifically for it, and its done everywhere. Nvidia is pushing for mixed precision (FP32, FP16, even FP8) to save power in their future desktop GPUs.
>>51392186
>Nvidia is pushing for mixed precision (FP32, FP16, even FP8) to save power in their future desktop GPUs.
Is AMD doing this faggotry as well?
>>51392235
Volcanic Islands aka 3rd generation GCN and newer arch has better support for half precision float point and integer ops.
If they don't support it Nvidia will push reviewers to use some absurd benches that favor their cards only.
This is fine for handheld devices, even laptops, but it has no place in discrete GPUs. The only reason why they're doing this is to manipulate bench results.
>>51392339
>>51392186
fucking njewdia again
