ATX and the future of motherboards

How much longer are *TX, specifically ATX likely to stay the standard for? we know btx failed and *TX is still able to add shit like m2 ports rather easily.

Other urls found in this thread:

asrock.com/mb/intel/H110M-STX/
asrock.com/ipc/overview.asp?Model=H110-STX MXM
accellcables.com/products/ultraav-usb-2-0-to-hdmi-adapter?variant=729676593
displaylink.com/integrated-chipsets/dl-1x5
asrock.com/mb/Intel/Z270M Pro4/index.asp
twitter.com/NSFWRedditGif

Unless some major standards change like ram dimm type or gpus no longer needing pcie slots or moving to something different shit will remain the same as it is.

The case ultimately decides how big your rig will be.

There are some atx that are smaller than matx cases.

of course, but its a 22 year old standard at this point.

atx regardless, its the mounting points aand height width ratio im more concerned about

>of course, but its a 22 year old standard at this point.
It just werks (tm) though.

>of course, but its a 22 year old standard at this point.
so? water has been a standard since the dawn of organic life. doesn't mean it needs to be replaced. atx is a standard that has worked incredibly well and has shown to be very versatile to new technology. such as easily adopting for quad channel configurations, m2 slots, and extra.

there's mini stx: asrock.com/mb/intel/H110M-STX/

and its mxm-baring kin: asrock.com/ipc/overview.asp?Model=H110-STX MXM

just wish they went with a reverse-side pcie slot or riser or something. mxm cards are rare and stupidly overpriced, though it at least keeps it slim with the right case.

>but its a 22 year old standard at this point
And it'll continue to age as we continue to use it.

Old Standards are not bad standards because of their age.

Why create yet another standard when nano, pico, and mobile ITX already exist?

There seems no reason to change it any more than there is a reason to change DIN A4.

Sensible size for humans, sensible size for content.

I think VIA might have royalties for Nano, Pico and Mobile ITX, so being jews, Intel created STX

Is there any reason to get ATX or bigger if you only ever intend to use a single GPU?

Are the CPU, RAM, GPU, and storage soldered on to that nano-ITX board? That kind of kills the "build your own PC" idea, right?

It's pretty common in smaller boards. It's just the choice in socket, it's not like a BGA CPU socket is part of the standard.

Why fix what isn't broken?

>this 1/4-20 bolt is an 80 year old standard, shouldn't we change it?

...

Doesn't look broken to me

So how do you propose we deliver power to the motherboard? Are we going to build a tesla coil off the front of the PSU and somehow keep it from frying the board?

>tfw used to buy nothing but FULL ATX
>tfw smartphone faggotry really is noticable in the PC parts market now
>tfw parts variety gets more and more reduced

>tfw full ATX is now more or less "PRO" area and therefore unreasonably priced

I fucking hate buying tiny ass motherboards

Are mITX builds really as loud as people say they are?I was thinking it would be fun getting a 1060 or 1080 mini and building my rig around it but will it be loud and hit as shit?

But what's the advantage of full ATX over micro ATX? More PCIe slots?

the fucking FUCK is your problem, faggot?

that's one of the easiest shits to connect.
if you brought up the fucking CPU fasteners that almost make you break the motherboard or bend the board so much the CPU gets almost lose, yes, that fucking sucks but not that connector. it is fine.

Exactly this.
I have tons of hardware I stuff into my builds, therefore I prefer full ATX.
On top of that everything smaller as full ATX feels cheap to me. I'm autistic that way

Meant CPU cooler fasteners, not CPU fasteners.

ATX remains the standard.

The smaller form factors suffer from expensive boards, and late introduction.

It seems there are still too many people who have a big tower case, even though they will only get 3 harddrives and run 1 GPU with CPU in the end.

>he uses intel's stock cooler
At least you won't have to buy a heater

What do you need with 5-7 PCIe slots? I mean, SLI/CF no longer scales really at all beyond 2x, NVMe SSDs are pretty much as fast as PCIe SSDs for most use cases, a mATX motherboard should come with at least one NIC, 6-8 USB, as well as a bunch of SATA and an on board RAID controller. Unless you have an ungodly amount of USB devices or some shit there's no reason for all that PCIe.

Regardless of the cooler you use your CPU will put off the same amount of heat.

I have a R9 290 in a mini ITX case. It's very quiet when idle or under a light load, but when the GPU fans spin up it gets pretty loud.

I mostly like to keep all my graphics cards and put the 4 most powerful of the bunch in the box and use them for CUDA computing and it works really well

Why wouldn't I?
As far as I can tell it has been waaay inside the tolerances so far and not too loud.
But I guess that's mostly because I don't overclock and I undervolt that shitpile as much as possible (without any problems) and therefore it stays way cooler than stock and the fan doesn't even spin to full speed after an hour of full load.

That's not how it works user your cpu is not the sun. If you keep it cool it'll stay cool. If you put a block of metal on top of your cpu the thermals won't be the same as an actual cpu cooler.

Oh, you're using this as a workstation. In that case why are you complaining about ATX boards being more pro oriented?

>I have no understanding of physics or thermodynamics: the post

That is exactly how it works.

Because to be honest I just have all those graphics cards running because I'm a poorfag and this way at least I get to a halfway decent speed with all of them combined.

Of course those prices wouldn't worry me if I actually became pro with a nice paycheck but not there yet or maybe never will. Graphics stuff is still just a hobby of mine.

Is there any reason to get ATX or bigger if you only ever intend to use a single GPU?

The real problem with the ATX standard is the way GPUs are supported via expansion slots. Those slots were never intended for high heat dissipation components, and stuffing a 250W GPU with its entire cooling apparatus onto a 40mm by 300mm PCIe card just doesn't work anywhere near as well as it could.

We have tons of room for these massive coolers with 140mm fans for a 150W CPU, but only a fraction of that for the GPU. We really need a new standard that takes into account the thermal requirements of modern systems.

This is a good point. They should really put the main GPU PCIe slot at the bottom of the board so you can have more space for the cooler.

4x lane coolers when?

I currently have a mATX board and I won't be going bigger anymore. There have been laptops for decades now and I don't like the idea that PC needs to be fridge-sized no matter what. It feels dated. If a computer can be packed into as small as a laptop/phone/tablet then we definitely are able to shave a little from a full tower without it being useless.

a single slot on the back of the mobo, making the card parallel could be an option, allowing for the gpu heatsink to rise out like the cpu's, but then you just get one card. There could be maybe one more slot if the mobo was big enough, perhaps with a new slot as well to better use the new technique

incoming 100 years of paint drawing

you'd have an oddly sized case though if you kept the current pci ports, but shave those off with a itx/stx varient and you just have the back port for vid card

I know the feel. I joined the m-itx group because I thought I only needed 1 gpu and with the right case I could carry around the house a portable monster desktop...

Then 2 months after it was powered on for the first time, a fellow VM buddy of mine told me about GPU passthrough and now I'm aching for an actual ATX or EATX motherboard with 3 PCI-E slots (one for current GPU on modern OSes, another for XP and another for 98SE).

gotta wait untill the next time I upgrade my entire rig. Will migrate the 4790k onto an ATX by then

whats the point of gpu pass through?

Using a GPU in a VM

I think itx with m.2 on the back is all most builders need. I have no preference for a smaller computer so go with ATX but itx is sufficient and nice. If they can get a 1080 and 7700k with a 2.5 drive and two m.2 devices into the base of a laptop why use a massive case for the same hardware? Something like the Nzxt manta can cool sufficiently to enable maximum Pascal oc, which is always ten percent, and a ryzen at 4ghz or Intel at 4.8

how is it usually done? i know it cant be cpu based surely? then is the point only for gaming on linux or something?

>have VM
>dedicate GPU to VM
>use GPU in VM

>then is the point only for gaming on linux or something?
It's PCIe passthrough, it's for when you want to assign real PCIe hardware to a VM

i get it but im asking what other applications it has. how is the cpu performance with vms anyway? say you have a 8 core cpu could you allocate 4 of those cores entirely to the vm?

>Works beautifully
>Universally accepted, so compatibility is top-notch
>Constantly evolving/changing to fit our needs throughout it's existence

>"B-b-b-but it's so ooooolllllldddddd."

Fuck off niggers. TX will be the standard until the very foundation of PCs changes drastically and they are built completely differently. Don't need to worry about changing something that works exactly as it should and exactly as we need it, even if it's been around for a while.

Pretty sure it's native (if not, close to native) performance

can't you use the integrated GPU for linux and the PCI-E one for the VM?

Agreed. mATX is great for 99% of use cases. It's too bad the case designers haven't really caught up.

>i get it but im asking what other applications it has.
Most who need GPU in a VM use it for gaming, but it could also be used for compute if you wanted to do it on a different or sandboxed OS.
>say you have a 8 core cpu could you allocate 4 of those cores entirely to the vm?
Yes, this is how it works. The management of the VM might utilize a few % of the CPU, I'm not sure.

I think this is how it is usually done (if the CPU/mobo has iGPU).

well darn tootin i think im going to try it, im igpuless however, so ill have to get a cheap secondary card. is it possible to power only one gpu when the vm is not running, then power the other for it when it is AND switch which gpu is being used for the main system?

You can allocate as many CPU cores as you want. PCIe passthrough is a feature on some Intel CPUs (because they're Jewish) and almost all AMD CPUs that allows an OS in a VM have direct access to the PCIe lanes you give it, so rather than going through a middle layer, the virtual machine has direct access to the GPU, or whatever. x86 has a number of virtualization extensions, including PCIe passthrough, that makes virtualization extremely efficient. You can get near native speeds with the right configuration.

>is it possible to power only one gpu when the vm is not running, then power the other for it when it is AND switch which gpu is being used for the main system?
Hypothetically yes, but in actuality no.

The applicable standards all support what you want to do, however the implementation, mostly from GPU vendors and somewhat from OS developers, prevents it. Due to the way hardware needs to be initialized at system start up, it's difficult to make GPUs that can be passed around by OSes, or at least AMD/Nvidia haven't gotten to it yet. The only way to switch which OS they're assigned to at the moment is through a reboot.

so does the vm have to start on boot to initialise the secondary gpu then?

No

However do you imagine you would mount a motherboard with giant coolers sticking out from both sides?

>SLI/CF no longer scales really at all beyond 2x
Most M-ATX boards don't even come with two full-length PCIe slots, though.

After building a computer for the first time, this pic is like porn to me

It's already dead imo for the general consumer.

Nobody buys ATXes except for gaymurs and nerds, a small niche.

Everyone else will be using thin clients.

If you were asking about people who need "enthusiast-level" hardware, then I think modular PCs will be a thing.

Like a main unit which contains the CPU/RAM/SSD/mobo with I/O stuff and an external GPU / audiocard / w.e specific hardware you need that doesn't have to be sit on the mobo to do its job.

>most
What's your point? If you wanted to do SLI you would buy the micro ATX board that had two of the appropriate slots.

>could you allocate 4 of those cores entirely to the vm?
Why would you pre-allocate cores to VMs? Just use the ordinary host OS scheduler.

Only if you don't care at all about GPU performance in the host system.

i don't think a lot of people do GPU intensive stuff on linux here

Half the point of having a full desktop PC is expandability, though.

but the heat

Even so, given that most video cards that matter use two slots, that leaves you with no expansion at all beyond the video cards.

When I build a computer, I like to at least try and keep my options open. Perhaps one day I'll have reason to play around with GPU passthrough, some FPGA dev board, some newly invented storage medium or who knows what. M-ATX just hardly leaves room for anything at all.

panel in the middle, cut outs for the back slots, pretty much the same as how it works in those dan cases or any current case that uses a riser to move the card to the back of the mobo. I'm just saying put slots back there instead of needing a riser and reserve the top pci slot for just a half-height card or no slot at all (more m2-slots or something)

>top
meant front side, backside for gfx, front for misc, if doing something like stx/itx. going atx could allow two slots on the back, but then you'd have 4+ normal slots on the front probably empty (maybe an additional flat slot like the back to make a slim atx case? eh)

>I want an extreme gaymen machine with shitty SLI technology plus I want to be free to install other cards, even though USB 3.1 provides plenty of bandwidth for almost any other device.

>gaymen machine with shitty SLI
Not really, I just use a second old GF 210 card to connect to the TV.

>GF 210 needs more than an x1 slot and more than 1 slot width.

It does need more than a 1-width slot since the connector is wider than that, yes. Obviously, an open slot would be fine, but those are about as common as 16x slots.

You could plug a TV into a fucking USB 2.0 Displaylink adapter.

accellcables.com/products/ultraav-usb-2-0-to-hdmi-adapter?variant=729676593

slot width I was referring to is on the back of the case. if your x1 slot has a closed end to it, cut it off.

The point being that 3 slots isn't very much -- especially not when the primary video card eats two of them -- if you want to keep any options open for the future.

depends on the case & cooling

my 6700k in an ncase m1 is quiet using noctua u9s but it gets pretty warm under sustained load (eg: building chromium which takes an hour or some shit)

the thinner cases like node 202 probably have more noise/thermal issues to deal with but the ncase m1 is just a tiny tower

>3 slots isn't very much
MATX is normally 4 slots.

>options for the future
For what in the future? 99% of peripherals worth having are available in a USB 3 form.

How does that work? 1920×1200×3bpp×30fps is 207 MB/s, far more than USB 2 can transfer.

>MATX is normally 4 slots.
I just checked out some random cards, and like 1 out of 10 had 4 slots.

>99% of peripherals worth having
Depends on what you consider worth having, I guess. Expandability is the principal feature of a desktop PC, why limit it?

Displaylink compression.
displaylink.com/integrated-chipsets/dl-1x5

Works rather well, it's in all the USB docking stations.

I don't see much of a difference between EATX and ATX other than the socket and extra ram banks. What am I missing here?

First one i pulled up had four. Maybe you're looking at some kind of gaymen shit.

asrock.com/mb/Intel/Z270M Pro4/index.asp

>what am i missing
Nothing? Different EATX boards have different feature sets of course, but that sounds right.

>expandability is the principal feature

Why not buy a van instead of a car? Why limit yourself?

Why did BTX die off so quickly? Seemed like an ideal solution for air cooling and gaymen.

I'm not disputing that they exist. I'm just agreeing with that it's sad that full ATX boards are being pushed as some kind of "pro" thing and are going up in price.

A van isn't really more expandable. It's not like you can fit more features to it or anything. Also it's a trade-off, whereas there's virtually nothing you can do with mATX that you can't do with ATX.

Oh, so it's some kind of lossy compression? Seems doubtfully nice for actual desktop stuff. Does it really work well with that?

I had a real giggle at this. You should probably learn the very basics about the reality we live in.

In my experience it's seamless 99% of the time, and I haven't even tried the USB 3.0 version. Somehow it even supports using the 3D acceleration capabilities of your GPU on the connected monitor.

Check out nvidias nvlink.

Theres propably a new "ctx" standard in the horizon with a new gpu connect standard similar to nvlink.

Protip: Its essentially a gpu lga slot.