Quantcast
[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology


View post   

[ Toggle deleted replies ]
File: 187 KB, 656x286, 1534874067690.png [View same] [iqdb] [saucenao] [google] [report]
68163135 No.68163135 [Reply] [Original] [archived.moe] [rbt]

>> No.68163151

>>68163135
maybe at 30fps 720p

>> No.68163161

>>68163135
Fanless is a thing of the past back when your CPU clock was several Mhz

>> No.68163175

>>68163135
You always could have fanless gayming, it always is just a matter of using big enough heatsink.

>> No.68163190

>>68163175
he probably wants a gaybo microatx build

>> No.68163201

>>68163135
only if you live somewhere below 40 degrees with the windows open

>> No.68163258

>>68163135
5nm is not cooler or better than 14nm. its just smaller

>> No.68163264

>>68163190
I'm sure it could be achieved within a typical microatx case dimension. The only real problem would be it'd have to be pretty much custom made, as any off the shelf solutions would cut it in those sizes.

>> No.68163277

>>68163258
50% reduction in heat compared to 7nm

>> No.68163285

>>68163161
>What is a smart phone
>What is a tablet
>What is ARM

>> No.68163287

>>68163135
Gameboy is fanless

>> No.68163289
File: 533 KB, 1600x1266, thermalright-hsc-601-case-panel_b.jpg [View same] [iqdb] [saucenao] [google] [report]
68163289

>>68163135
Fanless gayming has always been possible.

>> No.68163292

>>68163258
That's incorrect, not entirely but incorrect. If we're talking same architecture with just a shrink, at the same clock - it would be cooler as it'd consumer less power. In reality though a process shrink usually comes with extra transistors to improve something and/or higher clock, so the power consumption, and therefore heat produced, remains pretty much the same.

>> No.68163299

>>68163135
Depends on the game, some old games will run on 5w cpu, some will not.

>> No.68163305

>>68163135
just stick some heatfins on a intel atom cpu, i'm pretty sure you can at least run cs 1.6 on that

>> No.68163309
File: 12 KB, 326x294, 1537827242251.jpg [View same] [iqdb] [saucenao] [google] [report]
68163309

>>68163258
THE ABSOLUTE STATE OF /g/

>> No.68163328
File: 33 KB, 218x210, oy.png [View same] [iqdb] [saucenao] [google] [report]
68163328

>>68163289
Shh don't let the goyim know, we still have all those fans to sell!

>> No.68163348

>>68163289
>prototype
have fun never opening your case ever again

>> No.68163356

>>68163285
>What is ARM
Yes, go ahead Anon. Please explain ARM

>> No.68163371

>>68163285
>What is ARM
something that will never match x86 Sandy Bridge IPC

>> No.68163437

>>68163371
>Implying only Intel can play gaymes.
>Implying there are zero games for mobile
There were legendary games built for 8bit systems zoomer. OP said gayming not your autistic definition.
>>68163356
>What is Google
>What is startpage

>> No.68163446

>>68163371
20 years down the line when we're emulating sandy bridge on arm on risc-v desktops at native speed you'll think back to this moment

>> No.68163452

>>68163437
>What is Google
>What is startpage
Thank you for proving you're a thoughtless NPC

>> No.68163482
File: 16 KB, 198x254, index.jpg [View same] [iqdb] [saucenao] [google] [report]
68163482

>>68163452
>Implying that searching for another anon makes me less of an NPC

>> No.68163499

>>68163482
Oh, so you dropped your "what is ARM" without being able to give the simplest rundown about it.
Well done, Anon

>> No.68163509

>>68163499
(you)

>> No.68163544

>>68163446
>20 years down the line when we're emulating sandy bridge at 1FPS

ftfy

>> No.68163560

>>68163446
Given that we're reaching the limits of silicon technology, 20 years is overly optimistic. There will be a period of stagnation before they have something truly ready to replace it.
It's more likely that ARM will die off and be replaced by something else by the time it could be powerful enough to emulate a Sandy Bridge level x86 at realtime performance.

>> No.68163566

>>68163201
Name one place that is consistently over 40 degrees.

>> No.68163594

>>68163135
Why do you want fanless?

>> No.68163606

>>68163566
Your case with an i9-9900k running inside

>> No.68163723

>>68163606
You mean like a hour after you turn it off?

>> No.68163761

>>68163264
Why are there no cases that double as integrated heat sinks for internals? Would get dusty quickfast, but dead quiet

>> No.68163843

>>68163277
Won't increases in transistor density swamp any savings in heat/power output as far as dissipation cooling is concerned?

>> No.68164157
File: 299 KB, 1500x1000, streacom.jpg [View same] [iqdb] [saucenao] [google] [report]
68164157

There already is. The 2700, 2600x, 8700, 9600k, all fit within the CPU wattage budget of a Streacom DB4 with the extra CPU cooling kit. A 1600 can stay at its boost clocks permanently in a 20-25c room using the cooling kit, so a 2700 and 2600x (with their massively improved efficiency) definitely can.
The RX 570, VEGA 56 (-50% power limit and undervolted it can beat a 580's perf by a 25%+ margin at 100w), 1060, and 1070 (with a power limit), can all do same thing with the GPU cooling kit.
I'd say that's a pretty decent gaming PC. It's also got room for SFX-L PSUs, 2.5", 3.5" drives, and mITX boards. If you want to contest part choices, feel free, but consider:
* Ryzen's perf/watt peaks around 65-105w. Intel's CPUs only outperform their tier equivalents at around 150-250w, so they're not the best choice for a passive cooler with a cap of 105 watts.
* Turing may be a better choice for GPUs, I'm not sure how its perf/watt is in the 105w range. VEGA is amazingly efficient in the 100w range. An anon in /pcbg/ tested his 56 with a -50% power limit and nothing else, and it beats the 580's Time Spy Extreme score by 30-40% (IIRC it was around 6000). Additionally, the VEGA M matches a desktop RX 570 or a laptop 1060 while consuming less than 75w. Polaris and Pascal are also good in that range, but VEGA is a better choice for maximum performance in all likelihood.

>> No.68164253

>fanless gayming

no one heard about water cooling or what?

>> No.68164297

>>68164253
I thought I could go offline in peace, but then I had to see this retarded post.

>> No.68164337

>>68164253
Still have the pump, which is technically a fan. It's just suspended in water. You typically also use fans on the radiator, although with a 1080mm radiator (3x360, 9x120) with very low fin density you could probably run most components on just the pump. The alternative is running everything passively using convection, which has a few options. Streacom, CR-100A, Morpheus 2 (for low-wattage GPUs), could all passively cool high-efficiency components.

>> No.68164374

>>68163135
>muh games
cringed

>> No.68164390

What are you faggots doing with your PCs? I have multiple case fans to go with my CPU cooler, all Noctua. The only thing that you can hear more than a foot away is the GPU, and only when gayming. I don’t understand why everyone has so much trouble with fan noise.

>> No.68164853
File: 97 KB, 1112x1227, bingo.jpg [View same] [iqdb] [saucenao] [google] [report]
68164853

>>68163509

>> No.68164929
File: 32 KB, 816x275, ipc-1.png [View same] [iqdb] [saucenao] [google] [report]
68164929

>>68163371
>Says Anon, blissfully unaware that Apple is already exceeding Intel in IPC.

>> No.68164947

>>68164929
>Apple is already exceeding Intel in IPC.
Is that supposed to be impressive? I mean, even AyyMD of all companies will exceed Intel in IPC with Zen 2 in a few months

>> No.68164964

>>68164947
Well, the Anon I replied to said ARM will never approach SNB IPC, so. Also, they're leading by like 150-200%, and Zen 2 will not do that.

>> No.68165000
File: 24 KB, 601x500, 4.jpg [View same] [iqdb] [saucenao] [google] [report]
68165000

>>68164929
>Cherrypicked synthetic benchmarks

How NPC of you. You're the only retard on /g/ that shills this bullshit, everyone else knows better Crapplefag.

>> No.68165034

>>68165000
We're talking about IPC, so any imaginable benchmark would be synthetic by definition.
Also I don't shill Apple, I'm well aware that the system around the CPU sucks, in everything from software from thermals. But that doesn't prevent their CPU core to be architecturally good. Even if the A12 numbers were half of what they are presented as, it would still be pretty impressive and very close to Skylake-X.

>> No.68165060

>>68165000
>Cherrypicked
Yeah, just the entire SPECint suite.

>> No.68165061

>>68165034
Well come back and post when an OS with the usability and overall power/influence Windows has that can take advantage of said IPC.

>> No.68165073

>>68165061
Who cares about that? We were discussing IPC here, brainlet.

>> No.68165092

>>68165073
Well apparently Apple cares. That's why they use Incel CPUs for their computers

>> No.68165105

>>68165092
I never denied that. Still not pertinent to an IPC discussion.

>> No.68165111

>>68165105
Let's put it that way.
We'll compare IPCs the day ARM will use instructions as complex as x86

>> No.68165117

>>68163258
New nodes will have power reductions at the same frequency, though most of the gains are for density.

Also have more density means that the processor can be built wider and run at a lower frequency (a GPU with 2x cores but running at half the frequency will consume much less power and generally have same performance, due to the non-linear relationship between power and frequency)

>> No.68165132

>>68165111
Since the ratio of number of instructions required to do any particular piece of work to that required on x86 very typically exceed 1:1 on account of it being RISC, the actual IPC ratio is if anything even greater. The table rather shows "work" done per cycle.

>> No.68165139

>>68165111
>>68165132
Also, don't overestimate the complexity of x86 instructions. Compilers never ever use the more complex instructions. The most complex thing a compiler will utilize is a memory operand, which results in a single extra load instruction on ARM.

>> No.68165185

>>68165132
>>68165139
Alright then.
You leave me no choice: backwards compatibility

>> No.68165189

>>68165185
Again not pertinent in IPC discussions.

>> No.68165236

>>68163135
that's not how it works

while smaller newton meter nodes bring lesser power consumption, the bigger available space will be used for more transistors, bringing the heat back up for increased performance

*sigh* why don't you brainlets just stay on the /v/ sub

>> No.68165252

Sure, you could go fanless right now, but it just so happens that a fan will always win versus having a gigantic passive heatsink. Cheaper less space demanding, can dissipate more heat, allows you to clock higher, etc.
This will be the case with 5nm too.

>> No.68165255

>>68163161
Gtx 1050ti is fanless and runs every game in the world and most on max at min 40fps.

>> No.68165368

>>68165111
>instructions as complex as x86
You say that as if it were a good thing.

>> No.68165371 [DELETED] 

>>68163135

>> No.68165462

>>68163161
my laptop is fanless and has a quad core at 1.6ghz

>> No.68165539

>>68163135
If you clock it to 1GHz max then yes. Maybe you could get away with 2GHz and a HUGE heatsink.

>> No.68165927

>>68163289
Is that the PC case? Imagine having to remove and re-apply thermal paste every time you want to open your case to change something.

>> No.68166371

>>68163566
Australia during the summer, cunt.

>> No.68166671
File: 62 KB, 756x617, 1529424475969.png [View same] [iqdb] [saucenao] [google] [report]
68166671

>>68163135
we need graphene for that.

>> No.68166694
File: 52 KB, 1100x693, n64.jpg [View same] [iqdb] [saucenao] [google] [report]
68166694

>>68163135
fanless gaming has been a thing for quite a while now

>> No.68167187

>>68163566
He’s talking about 40 degrees in freedom units you dirty metricnigger

>> No.68167409

>>68163843
No the shits smaller nigga its cool just be happy

>> No.68167436

>>68165368
Not really a problem unless you're a compiler writer.

>> No.68167466

Meanwhile over at Intel.
>>68167445

>> No.68167477

>>68167466
>gamer toddlers
lmao

>> No.68167567

>>68163289
neat car radiator

>> No.68167605

>>68163285
>what is ARM
you mean "what are embedded systems"
fanless operation is not inherent to an instruction set you dumb pieces of shit

>> No.68167813

>>68163135
You can do that now with Fluorinert. No cooling hardware required. Just seal your computer in a tank of that shit and everything will be forced to stay at 52c or below. Just make sure you don't breathe it.

>> No.68167884
File: 1.19 MB, 2500x1619, Silentium-X99-14.jpg [View same] [iqdb] [saucenao] [google] [report]
68167884

It's possible now.
It's ~1000€ more than a fan cooled PC with the same components, but these things can be configured up to a GTX1080

>> No.68168450

>>68165000
Delusion at its finest.

>> No.68168481

>>68163135
do you play games while sleeping? I can’t even hear my nh-d15 and 1080ti when i game.

>> No.68168526

>>68163135
You can do that now with a 2700 and a heavily undervolted Vega.
You get a fanless cooler for the 2700, and for the GPU. Undervolt Vega to 900mv and set a -50% power limit.
That's without some expensive ass case where the whole thing is a radiator.

How are you all so new?

>> No.68168587

>>68163289
that's a big heatsink.

>> No.68169473

>>68163135
higher density->higher temps

>> No.68169662

>>68163289
imagine slapping that on a ryzen 3 2200G on a mini-itx motherboard

>> No.68169673

>>68165927
*blocks ur path*

https://www.innovationcooling.com/products/ic-graphite-thermal-pad/

>> No.68169802

>>68169673
>35 W/m-K
Alright, around 3 to 4 times the conductivity of the average paste, but I bet it's also way, way thicker compared to applied paste.

>> No.68170024

>>68165255
Minimum 40 FPS? Maybe at 720p.

>> No.68170090

>>68170024
the only Intel certified resolution, everything above is anti-semitic

>> No.68170414

>>68170024
No I'm talking about 1080p.
Newest horizon 4 runs at max with average 55fps.

>> No.68170434

aw shiet. I wanted to buy a 7 nm Ryzen laptop. should I wait for 5 nm and microLED screens? how far away is that dream combo?

>> No.68170525

>>68167436
Or a processor maker, which is why it affects us all.

>> No.68171636

>>68170434
>just wait

>> No.68171972

>>68164947
poomd won't exceed anything, street shitter

>> No.68172233

>>68164853
(You) isn't on there you fucking braindead droolie

>> No.68172475

>>68163135
lol, no.
the smaller a technology gets, the higher the energy density gets too.
those fucking things are going to be HOT!

>> No.68172556

>>68168587
for you.

>> No.68172585

>>68172475
>the smaller a technology gets, the higher the energy density gets too.
smaller transistors need less power to switch
if energy density scaled up with shrinking why are old CPUs on old nodes equally as hot and power demanding?
this is low quality bait, try harder

>> No.68172633
File: 59 KB, 711x575, 1375200183048[1].jpg [View same] [iqdb] [saucenao] [google] [report]
68172633

>>68163437
Great back up to your claim.

>> No.68172714

>>68172585
He isn't wrong. Google dark silicon, current nodes have much higher density than older nodes did.

>> No.68172729

>>68163761
There's a couple:
NSG-S0 (not released, kickstarter)
Streacom DB4
Compulab Airtop 2 (G)

And I believe Streacom makes others. And the Intel 'SoCs' for their microPCs can be passively cooled. Laptop 'SoCs' could probably be passively cooled too.

>> No.68172865

>>68172585
"i don't understand transistor scaling" the post.

>> No.68172956

>>68172865
I see nothing wrong with his post. We're always working around heat limitations and small transistors allow for less heat and thus more density.

>> No.68173078

>>68172956
http://www.cs.mun.ca/~paul/transistors/node4.html
see:
power per unit area (for the device!) stays constant
BUT, current density, contact resistance, and interconnect resistance ALSO scale, and they scale faster! (heat comes from I2R)
the fact that YOU see nothing wrong with the post doesn't make it less retarded, it's still wrong.
overall, smaller technologies put off more heat per unit area, because they are more than just the fets.

>> No.68173166

>>68164929
A12 is still not anywhere close to i5 2500k IPC

>> No.68173244

>>68173166
>posts figures and numbers
>"no it's not"
So what's your counter-evidence, dear faget?

>> No.68173428

>>68167813
>Fluorinert
enjoy cancer

>> No.68173460

>>68169673
liquid metal is better at 70+

>> No.68173516

>>68172729
hdplex

>> No.68173831

I looked into building/buying a low power fanless system for shitposting, like an intel NUC in one of those cases that are basically a heatsink or a mini ITX system. But stuff like your RAM etc is probably better off getting a little bit of airflow. If you just use an oversized cooler with a huge fan on it + a silent case, you can get a nearly silent system that still gets airflow so you're not just frying everything all the time and get great performance. Fanless is more expensive and really only worth it for special applications (like an asbestos factory or some shit).

>> No.68173991

There are 90TDP fanless CPU coolers, 1050Ti can be bought fanless OOTB, I think there was some 1060 fanless conversion kit, you could also DYI. I bought 1050Ti though because 1060 was too expensive. Playable frames @1080.

>> No.68174104

>newton meter

>> No.68174151

>>68173831
>RAM needs moving air
Not true.

>> No.68174720

>>68170525
I don't really get your meaning. Incompetent designers will fuck up anything regardless of the inconsequential RISC/CISC label on the specification.

>> No.68174919

>>68174720
Yes, but competent designers, on the other hand, will produce a better result when starting from a reasonable design.

>> No.68175020

>>68174720
>the inconsequential RISC/CISC label on the specification
It is certainly true that the "RISC vs. CISC" issue makes no sense, but mostly because CISC designs are so varied as to make the term as such meaningless. There are much better CISC ISAs than x86, like S/360. But x86 objectively does make for a poorer starting point for processor implementation.

>> No.68175099

why would you want to go fanless instead of trying to squeeze every ounce of performance out of your parts?

>> No.68175145

>>68163161
I play Dota 2 on my fanless 22 nm celeron N2840

>> No.68175201
File: 384 KB, 1600x1200, fortnite-review-hero.jpg [View same] [iqdb] [saucenao] [google] [report]
68175201

>>68163135
Here's my fanless gaming PC, OP.

>> No.68175301

>>68175201
>"Personal Computer"
You mean "personal media-consumption device", Anon.

>> No.68175328

>>68175301
>multi core CPU, GPU, several GB of RAM, and 32GB to 512GB of permanent storage built in
>not a computer
Face it, anon. They're pocket computers.

>> No.68175381

>>68175328
Yes, carefully set up in such a way that you can't do anything with them but disgusting media consumption
>b-but you can totally install your own bootloader and run a loonix server on them
On some models, yes, but noone does. Including you.

>> No.68175441

>>68175381
>Yes, carefully set up in such a way that you can't do anything with them but disgusting media consumption
You can do damn near anything with a phone, if you want to program or something nothing is stopping you

>> No.68175551
File: 70 KB, 478x700, 123344-n0.jpg [View same] [iqdb] [saucenao] [google] [report]
68175551

>>68167884
Anyone got pictures of the insides of these Deltatronic Silentium systems?
Curious to see what it's like since they seem to be specifically using old discontinued super cheap all-steel Chieftec cases for them.

>> No.68175590

>>68163843
The heat generated will be more localised, but the overall amount of heat would be reduced, unless you pack more transistor into the chip.

>> No.68175624
File: 916 KB, 245x285, okay-1.gif [View same] [iqdb] [saucenao] [google] [report]
68175624

>>68175441

>> No.68175882

>>68175624
>let me just post le epic reaction gif that'll teach him

>> No.68175916
File: 2.94 MB, 480x270, happiness-2.webm [View same] [iqdb] [saucenao] [google] [report]
68175916

>>68175882
I'm glad that you admit yourself beaten. Now go back to your media consumption, NPC.

>> No.68176036
File: 21 KB, 355x355, oliet bangs.jpg [View same] [iqdb] [saucenao] [google] [report]
68176036

>>68163161
My Core2 Quad Q8200 was passively cooled with a scythe ninja (1st gen I suppose)
Never hit more than 74C

I with they still made good coolers, not the ones now that have the fins stacked WAY too close to eachother.

>> No.68176058
File: 47 KB, 500x673, 1538348926149.jpg [View same] [iqdb] [saucenao] [google] [report]
68176058

>>68175301
>>68175381

>> No.68176106

>>68163161
I love the fact that this post got so many reactions from retards bragging about their powerful CPU/GPU sporting a passive cooler 20 times the size of the chip when the post was about CPUs from the past which had no such things.

>> No.68176128
File: 133 KB, 255x243, intelinside.png [View same] [iqdb] [saucenao] [google] [report]
68176128

>>68163135
Just because you can, doesn't mean you should.

>> No.68176278

>>68176106
They instead had a ceramic package large enough to act as heatspreader thanks to the pin spacing.

>> No.68177808

>>68174919
What even defines a "reasonable design" in your eyes? My basic assumption would be something more straightforward, simple and "logical" with little added complexities, and mature architectures like x86 are certainly full of countless extensions, compromising design decisions and compatibility measures that might not make sense to someone designing a new architecture completely from scratch, but you could also say the same of any established architecture that has had to adjust to changing times and demands.

I don't really think you're wrong on the fundamental idea that a seasoned or otherwise "competent" engineer would have an easier time working with a more simple and straightforward design (or perhaps whatever "reasonable" means to you,) but I'd put forward that an engineer that shies away from any sort of complexity and can't effectively function with a less "reasonable" but still well-documented design with nearly 40 years of examples to follow and history to learn from does not qualify as "competent" either.

I think in the end what I'm saying is that, if you've truly got the drive and passion to become an electrical engineer, compiler writer or go into some other similar systems-level profession, whether you're working with x86, MIPS, ARM, SuperH or whatever else isn't really going to make a significant difference in the end. I can agree with the basic idea that the increasing complexities of many architectures can increase the chance for vulnerabilities, but such things are unavoidable no matter what you do (as recent Spectre/Meltdown hysteria shows) and living in constant fear of them to the point of holding yourself back into the stone age just isn't a good strategy for the majority of applications.
>>68175020
Nothing I can really disagree with.

>> No.68177873

>>68163135
>fanless gayming PC
we already achieved pc-less gayming pc with Video Game as a Service (VGaaS). See google cloud gaming service or nvidia geforce now.

>> No.68177949

>>68175328
It depends on what your definition of "computer" is. Pretty much anything with a microprocessor or other complex ASIC can be considered a "computer" by literal definition, including your alarm clock, microwave timer or bluetooth-enabled sex toy. Some of those kinds of appliances have some pretty powerful chipsets in them, but do you talk about your wireless speakers or smart TV in a thread like this and never shut the fuck up about how "amazing" they are? Not usually, because while they may be computers by definition, they don't act like them. They're appliances. Usually when we discuss "computers" on /g/ we're talking about personal computers.

Smartphones are definitely a little more than dedicated appliances as you can run whatever application you can develop on them, but they're still incredibly dumbed down, with very poor methods of data manipulation and input, minimal screen real estate, restrictive operating systems and software bases mostly composed of cloud-enabled shovelware and crappy games. They may be as "general purpose" as a true personal computer on paper, but in practice they are basically a consumption appliance. It's what almost every part of them is designed to do, it's what their developers encourage with things like walled gardens, it's what they are and always will be, and trying to act like they're anything else just because they've got shiny big numbers on the spec sheet is just being dishonest.

>> No.68178082

>>68177808
>What even defines a "reasonable design" in your eyes?
There are probably many schools that can arrive at a reasonable design, and I wouldn't really want to single out any one of them as the One Way (though RISC-V is pretty good).

However, the problem with x86 is that it contains several elements that are pretty unreasonable. The worst, obviously, being the horrible instruction encoding. The only reason both Intel and AMD have gone with µop caches is because fast x86 decoding is so horribly power-hungry that a whole extra cache structure is worth having just for the privilege of getting to shut down the decoders during loops. But there are other issues that also makes x86 sink compared to other alternatives, particularly:
>Severely over-restrictive memory model in SMP systems
>No explicit separation between D- and I-caches.
>For all that is said about how the saving grace of x86 instruction encoding is smaller code size, it actually does a pretty bad job at that and most compressed RISC variants tend to be significantly smaller

So in short, it's not just about making the engineers' or compiler writers' job harder, but about making choices that make the architecture intrinsically impossible to implement as well as architectures that don't make those choises.

>> No.68178441

>>68164929

Oh. Is that why we don't see Intel chips on Macs anymore?

>> No.68178477

>>68170434
Buy the 7nm

>> No.68178489

>>68164929

Well, that's retarded. That's like bragging about having the best muscles for swimming but not being able to swim.

>> No.68178582

>>68178082
I can respect the criticisms, I've heard arguments go either way on the encoding but I'm not really knowledgeable enough myself to attempt to take either side.

As far as the last paragraph goes, I think I can see where you're going, but still I can't say I feel it bothers me much. x86 since the early 2000s has outplayed dozens of competing architectures that many here would call (and probably justifiably) superior, and pushed them into extreme niche markets or outright extinction. Non-technical and financial reasons of course played a big part, but in the end it seems demonstrable that the many engineers who have worked with x86 over the years haven't had a problem implementing it at least well enough to remain useful and competitive, especially now where things are simply so fast that architectural deficiencies just aren't as prominent anymore outside of a few niche applications, especially when it comes to personal computers and high-end servers which are usually what we're thinking of when we discuss these kinds of things.

I don't think I'd hesitate to disagree that x86 is far from what you'd want to base a new architecture or processor upon from a purely technical standpoint, I'm just not really into the outright demonizing of it or some of its fundamental ideas that seems to be especially ramping up since spectre/meltdown with general criticisms that often just sound like they're regurgitating marketing bullshit off of a keynote slide or tech news article.

I'm not saying any of that towards you, though, thanks for not being a dick about it. I'm always interested in reading legitimate comparisons and criticisms like these.

>> No.68178713

>>68178441
Whoever pretended IPC was the sole determiner of performance and value in a system?
>>68178489
Noone argued that it made the A12 the best chip. But it does seem to make it the best architecture.

>> No.68178838

>>68178582
Neither am I arguing that x86 is completely useless or anything. Rather the opposite, if anything; x86 has grown a lot since the 8086 and become much more orthogonal and useful with the many extensions since then. However, that still doesn't change the fact that a more reasonable architecture has potential to be better, and that the only reason they haven't is, indeed, that the x86 vendors have had, by far, the greatest research budgets to pour into their implementations.

And the reason why Intel despite that still never managed to challenge ARM in the low-power market is probably to no small extent architectural.

>> No.68178943

>>68178838

ARM cannot and will not replace x86_64, ever. For the simple reason that desktop users don't have to worry about keeping power consumption as low as possible as a priority (where ARM shines) and that cisc is just better a some stuff over risc and that shit won't change no matter what.

>> No.68179004

>>68178943
>that cisc is just better a some stuff over risc
Name one single thing that is not backward compatibility.

>> No.68179083

>>68163151
Ok

>> No.68179187

>>68178838
>and that the only reason they haven't is, indeed, that the x86 vendors have had, by far, the greatest research budgets to pour into their implementations.
I don't know if I can really fully agree with that. Plenty of serious technical competition to x86 at one time or another came from smaller companies, and some from larger ones, and some even from within Intel itself. A big budget definitely helps, but I don't like the idea that it's just a matter of throwing money at it until something good comes out, it buys tools and talent, but it doesn't necessarily give them insight or ideas.
>And the reason why Intel despite that still never managed to challenge ARM in the low-power market is probably to no small extent architectural.
I think market factors played a pretty big role, too. What really made a low-power Atom stand out against the scores of SoCs from established companies that had already been entrenched in the market for years? It feels really similar to the difficulties ARM has in servers and desktops for me, for all the retooling, recompiling, retraining and renegotiating among other things, what were you really going to get out of it?

>> No.68179318

>>68179187
>but I don't like the idea that it's just a matter of throwing money at it until something good comes out
Neither do I, but there's no arguing that it helps and helps a lot, if only for getting the top talent to the project. There's no doubt there were (and probably still are) a lot of very talented people at Intel, and I don't want to accuse them of just going for the money, but where there's money, there tends to be genuine glory too.
>some from larger ones
Not one of them, not even IBM, could even remotely compete with Intel in terms of the budget they could pour into processor development alone, though.
Point being, I guess, that I certainly agree there is enough variability in implementation that a good implementation can trump, and trump by far, a good architecture. But the best implementation of a worse architecture will intrinsically be worse than the best implementation of a better architecture.

>I think market factors played a pretty big role, too.
Of course there were many other factors as well, but to this day, Intel has not managed to produce anything that can compete with the truly low-power ARM designs on a technical level. Atom chips are an absolute joke in terms of power consumption, perhaps not compared to high-performance ARM cores like the Cortexes A76, but definitely compared to the actual low-power ones like the Cortex A7 or, at which point the comparison just becomes outright silly, the Cortex M series
Also, speaking of market factors, I also don't really think there can be any doubt that there has been market pressures for low-power x86 tablets and the like. The smartphone vendors already entrenched in ARM wouldn't go use x86, but the current ultra-light x86 laptop vendors would certainly want to produce even lighter and less power-hungry designs, and Intel simply has nothing for them.

>> No.68179352

>>68179004

>Name one single thing that is not one of the most basic and important thing is to consider.

K.

Cisc Is not limited to operate on CPU registers.

Cisc is not as dependant on good code, considering the current flood of Pajeets in tech (and Africans in the near future) you can't trust the whole thing on the compiler doing a good job.

Defuckingbugging

>> No.68179374

>>68179352
>Cisc Is not limited to operate on CPU registers.
But in practice they do anyway. There is literally no difference between an instruction with an implicit memory operand and an explicit load instruction preceding the instruction in question, execution-wise.
>Cisc is not as dependant on good code
What is this even supposed to mean? There's no difference between CISC and RISC in this regard.

>> No.68179466

>>68179374
>What is this even supposed to mean? There's no difference between CISC and RISC in this regard.

He meant compiler-wise. Cisc can get away with a simple compiler but if a risc one has a badly coded one the whole thing shits itself. That's why it took so many years for risc to gain traction, the software just wasn't there to make it work properly.

>> No.68179629

>>68163761
the grillbook pro does not have this problem.

>> No.68179652

>>68179466
>Cisc can get away with a simple compiler but if a risc one has a badly coded one the whole thing shits itself.
No. That is only true for in-order architectures. For out-of-order implementations, the compiler can generally do just as crappy a job on CISC as on RISC code and still get the same results. The same was true for in-order CISC, too.

>That's why it took so many years for risc to gain traction, the software just wasn't there to make it work properly.
At this point it sounds like you're talking about VLIW, however. Good RISC compilers were never particularly hard to get up and running.

>> No.68180203

>>68176128
delid and delap

>> No.68181693

>>68163258
>reducing the distance electricity has to move wont make it run faster
okay...

>> No.68182231

>>68163135
>fanless cpu
>fanless gpu
>ssd
>no more fans
Oh, please.
P L E A S E

>> No.68182513

>>68163135
it's called a snapdildo 855

>> No.68182602

>>68163161
I have a 144MHz ARM Cortex-M4 microcontroller with networking hardware and hundreds of IO pins sitting on my desk that doesn't need a fan. I also have a multi-core ARM chip that runs at 1-2GHz in my phone. It also doesn't have a fan. Are you okay, anon?

>>68163356
not that anon, but ARM makes the core designs then sells the IP to manufacturers, who then either surround those cores with various peripherals, e.g. Cortex-M microcontrollers, or some companies like Apple make their own variations of those chips to their own custom specifications.

>> No.68182895

Just get longer cables and leave it in the living room.

>> No.68182967

>>68182231
Sure, and then you can hear all the electrical buzzing and coil whine, which is even more annoying.

"Silent PC" is a fruitless pursuit, you'll always find noise somewhere. Just get it down to an acceptable level and be done with it. Save your sanity.

>> No.68183521

>>68173428
Like I said, don't breathe it like some kind of savage.

>> No.68185109
File: 164 KB, 620x460, nofan.jpg [View same] [iqdb] [saucenao] [google] [report]
68185109

>>68182231
>fanless cpu
lol

>> No.68185210

Should I buy some stock?

>> No.68185404

So what about fanless PSUs then? I hear there are some that exist, but even they need case fans or something to move at least a little air through those heatsinks

>> No.68185596

>>68185210
Maybe

>> No.68186010

>>68185109
Nice 3 fans in the case, and presumably another one spinning fast on the graphics card.

>> No.68186134

>>68163135
>>68163151
>>68163161
>>68163175
I used a AMD AM1 Athlon 5350 machine as a HTPC for some years. It's a nice 25W TDP APU that's easily cooled by a passive heatsink. Can't into modern games on a 4 core 2GHz APU but can play 1080p video and play things like NES, SNES, MAME and simple old games just fine. I eventually replaced it because it can't into 4k video playback but apart from that it was a nice totally silent system.

If you limit yourself to a 25-35W CPU and a simpler graphics card like a 1030 or 1050 then you can easily do completely fanless. You can't expect to run Fortnite at 4k on such a system, but if you limit yourself to simpler games and lower resolutions it's quite possible.

>>68163594
It's nice.

>>68167884
Nice.

>>68185404
>So what about fanless PSUs then?
There's plenty of them already, all the EVGA SuperNova and mid-high range Corsair PSUs have a "semi-passive" mode. Buy one that's ridiculously beyond your needs and it will never get so hot the fans kick in. Look into it. For example, I used a Corsair RM 650W for my Athlon 5350 APU HTPC. Fans never kicked in. A 250W PSU would probably been enough in terms of power.

>> No.68186183

>>68163289
Can you DIY that?

>> No.68186190

>>68164929
>still spouting bloomberg fud to manipulate stocks.

>> No.68186539

>>68175099
I want my computer to be quiet

>> No.68186591

>>68186190
>actual SPECint numbers
>bloomberg fud
Whatever helps you sleep at night, Anon.

>>
Name (leave empty)
Comment (leave empty)
Name
E-mail
Subject
Comment
Password [?]Password used for file deletion.
Captcha
Action