ITT: we redesign the PC
>>51213417
cables on 3/4 sides tard
We base it on spring scales instead of Turing machines. Tension on the spring equates to various numerical values instead of just using some tape to be a 0 or 1. This means various charge levels, options, switches etc. And more complex logic.
>>51213417
That's pretty much what we're using rotated, only of course you made it harder because mobo connections have to go around the gpu.
I 'ad a go at it
>>51213417
>>51214504
Jesus christ apple tries too hard. I mean, it's compact, but cooling and upgradability are worse. Plus the power button is under the cables.
>>51214504
Thats pretty gay
>>51214520
This looks like complete shit, just about guaranteed to malfunction.
0/10.
why didn't BTX catch on?
>>51214593
because it was always just a dell and hp meme to begin with
>>51214593
Layout doesn't work for CPUs with an integrated memory controller.
>>51214558
i made it in my own image
>>51214736
kek
>>51214273
But the entire point of the turing machine was that it can be implemented in a much simpler fashion and emulate any machine such as the one you mentioned. Sure, you'll have analog measurements, but a good turing machine could simulate it to a greater degree of precision than you'd need anyway.
>>51213417
what about crossfire/sli?
Ok /g/, where do I register it?
Am I ready to work on NASA?
>>51214445
Where's all of the space for the chipset?
>>51214504
Nice try Ahmed, but you just put a computer in a trash can
What the fuck am I even reading kid
>>51215422
in between the CPUs and pci cards, its just a rough diagram though. What do you think of it?
>>51214273
You mean quantum computing? It's been done anon.
Work in progress.
>>51215534
Forgot to mention:
Left is front, right is back. Dual chamber tower design.
>>51215511
>no DIMM
>>51215666
do you not see the 6 slots per cpu
>>51215666
devil dubs checked
>>51215121
reminds me of older PC's where the only port on the motherboard was for the keyboard.
>>51215449
Top fucking kek.
>>51214593
ATX works better for expansion card cooling
>>51214623
Don't forget gateway
>>51215534 I like this
>>51214504
Revolutionary
It's magic
>>51214520
We will be able to simulate that shit using a Turing machine or equivalent within a decade
The human brain is much too slow to adequately run a Turing machine
>>51215121
>putting CPU and RAM on PCI Slots
expect it to have higher delay and always lose in comparison
go to hell with that shitty design
>>51213417
Proposal for GPUs:
> empty I/O shield for higher airflow
> output pixels shunted through PCIe to mobo outputs
> optional sister PCIe card with just extra outputs
I mostly like where this is going:
>>51215534
but VRAM generally needs to directly flank the GPU to save DDR interface power and simplify PCB routing.
also, this seems like it would be substantially bigger than standard atx mini-towers.
>>51216061
the concept was for an enthusiast build in something with a size closer to a Corsair Air 540. the GPU board being optional and still fitting standard GPU's in the PCIe slot if needed. the VRAM is probably unnecessary anyway, thanks to HBM...
>>51214745
Well a turing machine/computer could theoretically simulate anything imaginable. I'm just saying that we have a physical version of something that's beyond the 0 and 1 switch idea..
>>51215532
Fuck me
>>51216286
maybe we could make PCI gpu's with replacable VRAM and actual gpu chips instead
>>51216607
as much as I like that idea, we'd still be limited to the clearance of the PCI slots. GPU's draw double the power of CPU's but only have half the cooling...
>>51216290
>theoretically simulate antything imaginable
not using electricity like we do. Try making electric turing machines that emulate a living cell or a virus.
You'll need components smaller than Lorentz radius. Which means electrons would be unpredictable at that scale. Also, what would you use for the physical structure?
Beowulf cluster of 16 hand soldered homebrew 68k SBCs running Gentoo
>>51216960
clusterfuck.tiff
>This thread
>>51215534
>board interconnect as far away as possible from the CPU
Will never work. There is a reason that the PCI-E slots are literally an inch away from the CPU.
Nice try Jews, I'm not going to show you my design, just for you to rip it off.
>>51215983
who said it was PCI
>>51215449
underrated post
>>51213417
just a rough layout design, based on the fact that heat rises and should make it easier to dissipate heat passively without having 6 fans going WRMRRMRMRMRMRMRM just to keep things cooled in there. should make the air flow more efficient as well.
The lines are where you put the memories.
>>51220322
>based on the fact that heat rises and should make
actually this fact doesn't matter in cases.
see this video for more ino
>>51220413
https://www.youtube.com/watch?v=g8EN3K-eaVA
>>51220430
>linus tech tips
trashed
r8 double ended graphix for double output speed
>not soldering in and enclosing all parts into one convenient giant CPU die
It's the way to go guys, trust me.
No one wants to be a number crunching autistic that spends thousands to get 5 more FPS in some shitty game.
Just let the computer cost be directly proportional to computing performance.
>>51220430
That video does not cover heat physics
also includes the worst case for air pressure balance
>>51220524
But where would you put the extra cores anon?
>>51218919
And that reason is?
>all these interconnected GPUs
enjoy your shit latency and framerate
>>51220370
>Crossing Memories
>>51220753
>Not getting the joke
>>51220753
>>Crossing Memories
Anime adaptation next fall
>>51220370
GTFO
T
F
O
Connect to monitor
Connect mouse and keyboard through Bluetooth
Have a fully functioning PC, redesigned for the future
>>51215277
>L5 cache 4GB
Only this, will raise the price of the cpu by something like 10k$
>>51215121
Soooo this
This might become a thing if GPUs dont get smaller.
>>51214273
thats retarded. as soon as the spring starts to give, the calibrated tension will become inaccurate
>>51220599
youre so fucking stupid it hurts. thats why
>redesign the PC
Assuming standard home PCs, the exact same thing, except make the processor fucking RISC.
Fuck the x86 line (includes amd64). Been holding us back for years.
No more optical drives.
Integrated GPU only.
Integrate chipset into CPU.
Stacked memory on CPU.
Stacked 3D XPoint on top of the CPU for persistent storage.
External power supply.
Whole thing should not be larger than a Rubik's cube.
>>51213417
Here is a mockup I did for a custom PC real quick in sketchup.
Basically a lot like a PC now but I designed it around a natural airflow of sucking the air from the bottom and out the top. Has the PSU outside the case, along with the drive bays to not interfere with the airflow just passing over the HD's, motherboard and GPU.
>>51221812
I don't even get what is with you wannabe 90s kids and the "x86 is holding us back because it's not RISC" meme, x86 has been internally RISC for twenty years and desktop RISC stagnated and died in the 2000s because it was uncompetitive overpriced garbage. This is coming from somebody who actually uses non-x86 architectures on the desktop.
>>51222798
I don't understand. Is RISC the same think as ARM? Sorry if I sound like a noon.
>>51214593
Yeah. I wonder.
>>51223147
RISC CPUs removed seemingly useless instructions from the set, allowing more room on the chip to make the useful ones better
most architectures are RISC
the problem with x86 is it has unneeded backwards compatibility with 80's computer theory
Intel's been wanting to move away for over a decade but the market, and government, have shut them down every time
>>51223147
Reduced Instruction Set Computing
CISC (Complex Instruction Set Computing) architectures like DEC's VAX, Motorola's 68k and Intel's x86 have a bewildering variety of instructions which can be very powerful, at the cost of making the chip more complicated and slower.
RISC chips like ARM, SGI's MIPS or DEC's Alpha cut out all of this bloat, reducing the instruction set to a small and manageable amount. Where a big, slow and complex CISC design may use a specific, powerful instruction, a small, fast-clocked and simple RISC design will instead use a series of smaller, less complicated instructions to do the same task.
/g/ loves the RISC meme because RISC systems absolutely kicked ass in the '90s, nothing could outrun the SPARC, MIPS and Alpha juggernaut. Intel would later give in and design the Pentium Pro around a streamlined RISC core with a subsystem that translated x86 instructions into smaller, simpler instructions for the core to execute, and it managed reasonably well, Pentium Pros were great performers on 32-bit code (and 32-bit code only).
What /g/ forgets is that eventually, RISC chips and the systems they shipped with turned to shit and stagnated on the desktop, and RISC workstations during the 2000s weren't the absolute shit-wreckers they used to be.
>>51223294
And there's a reason the market shuts them down, because that legacy compatibility is not unneeded. The real world doesn't give a shit about staying up to date, they give a shit about what works. Had they not have, we could be living in the EPIC/VLIW utopia Intel had long dreamed of with the Itanic.
Is there anyway to construct a power supply that is thin and flat so it can be cooled by a waterblock?
I'd really love to see some innovation in desktop form factor PCs (as in the ones that lay flat on your desk). Towers are cool but it's even cooler when your fully upgradable PC just kinda vanishes underneath your monitor.
I've heard that heat can be a problem with these systems, but couldn't water cooling take care of that?
I think GPU'S are now important enough to have their own socket like a CPU. I think it would be far more efficient and customisable. You could have Dimms for the GPU just like the CPU and cooling could be the same as the CPU, just use the same form factor mounting. I know it wouldn't be possible right now because none of the GPU manufacturing companies would agree to doing so because it would put them out of business. Imagine the preformance gains you could get from making this switch though. The Only problem is the GPU chip form factor with different numbers of pins on each chip, a modular solution would be needed.
>>51223899
...How would migrating from a universal slot to a proprietary socket and lessening costs spent on manufacturing video memory put anyone out of business? Literally all this would do is screw the customer over even more since he would have to buy a heatsink and video RAM along with a GPU, and he would have to totally board swap if he ever wanted to upgrade an older system.
>>51223899
>he doesn't know the GPU is essentially another computer
>wants to fit it into a socket on the motherboard, when there already is a slot for it, on the motherboard
>>51224240
>GPU is soldered
>VRAM is soldered
>cooling limited to two PCIe slots worth of space
>the only way to expand upon a GPU is by buying more, if it's even supported
>basically a prebuilt inside another computer
>>51224324
How will forcing it into a space-wasting proprietary socket change any of this?
VRAM upgrades are a useless placebo, there's a reason cards haven't shipped with memory slots in over fifteen years.
>>51223899
this could be pretty viable thing, gpus could be cheaper sans the gpu ram slots... we could get gpus in different flavours based on different amounts of hbm onboard memory ie. 1gb 4gb 8gb etc...
>>51225345
It would work well with HBM. You could have 2 gpu sockets next to the cpu, all the VRMs for the CPU and GPU would be in 1 spot. You could then cover the gpu, cpu and VRM with one waterblock or large heatpipe cooler. If only you could get 32gb HBM on a cpu as well, you could probably make a 10cmx10cm motherboard.
>>51220541
underrated post
>>51220541
>filename not 68hoursinMSPaint.png
>>51214504
It does look sexy af though.
>>51221755
you just need good material design anon
im learning all about it in android shool
>>51214504
shit is too round at the top and bottom edges
would like an nice matt black case with an soft rubber feel to it
>>51215701
Those are called trips, as there is not two, but three of them, my friend.
>>51226183
>>51225345
>>51223899
This is fucking retarded. THE ONLY reason HBM is so fucking fast is because it is soldered RIGHT NEXT TO the GPU.
In fact, the only reason regular VRAM is so fast is because it's soldered onto the card itself. Soldering and integration is the way to go for performance gain, because signal travel time and loss is massively reduced by soldering and integration. Why the flying fuck don't you think CPUs long ago absorbed memory controllers and integrated them?
Even worse, putting the GPU itself and VRAM in separate slots just means the GPU has to wait for the fucking system bus to be clear even more than it does now (during DMA transfers).
You'd gimp the damn GPU and the bottlenecks would be everywhere. You two are completely retarded.
>>51220541
>AGP slots
>CPU heatsink connected directly to PSU
holy christ
>>51220322
Nigguh dats tight son
