are you sure you want to buy that motherboard you were looking

>>59272914
It's going to be at least a couple of years before the majority of consumer hardware (including CPUs and GPUs) is PCIe 4.0 compliant, so yeah.

I know people working on PCIe 4.0 collaborations right now
Sandisk will be late to PCIe 4.0
AMD (Radeon) does not have any plans for PCIe 4.0 at this exact moment
It'll likely never reach a consumer retail in 2017, that's just unrealistic when so many companies are still looking for partners and don't even have the software/hardware capability to use PCIe 4.0 yet

>>59273052
And there's lots of speculation about the future Intel GPU, (which is actually a real, materialized plan but may not make it like many of intels projects)

>>59273052
Yeah. Until consumer-grade CPUs support it, there's no reason for hardware vendors to make consumer-grade hardware to support it either.

My personal estimate is that there won't be any consumer hardware supporting PCIe 4.0 until 2nd or 3rd quarter next year, at the earliest.

Stupid question: Do we actually have anything powerful enough to saturate PCIe 3.0?

>>59273083
>future Intel GPU
Tell me more

>>59272914
Considering nothing I'm putting in the machine is going to come close to saturating PCIe 3.0, I'll happily buy a motherboard and not wait for 4.0

>>59273224
I work with PCIe clusters at work and run high-performance applications that do.

But at home? No.

>>59273224
Doesn't matter if it exists or not. The backbone communication for the rest of the devices needs to be higher than they can saturate for a while. Gives the technology room to grow

>>59273224
I believe there are capture cards that come close or do, that's actually the only use I can see for it

WOW

will I get a 1 point extra benchmark on something?

SO HYPE TO SPEND 5 GRAND NOW

>>59272914
PCIe 4.0
so, 2% better than 3.0 that is 2% than 2.0?

>>59273309
Isn't every iteration double the bandwidth of the last?

File: Screen Shot 2017-03-06 at 22.53.02.png (221KB, 3240x1434px) Image search: [Google]

221KB, 3240x1434px

>>59273325
>>59273309

>>59273224
Not GPUs that's for sure
Not a lot takes advantage of that insane I/O, GPUs are more or less self-contained so they don't really need as much bandwidth as you would think

>>59273309

It's literally double the bandwidth.
About 60% faster than 2.0 in real life if you want to get autistic about it.

Just kys.

File: 1400719270649.png (613KB, 900x900px) Image search: [Google]

613KB, 900x900px

>>59273083
Oh boy I can't wait for Larrabee v2!!!

File: image.jpg (33KB, 500x342px) Image search: [Google]

33KB, 500x342px

>>59273225
All I overheard was my dad on the phone...
But it is a real plan being discussed with real suppliers, collaborators and Intel company customers right now

>>59273360
>>59273360
>[3.0 is] about 60% faster than 2.0 in real life if you want to get autistic about it.

>>59272914
>graphics cards still dont saturate pci-e 2.0 16x
A single one doesn't at least.

Remember when people thought PCIe 4.0 would support 300W slot power? Remember how motherboard vendors threw a fit because they'd actually have to put in effort for power delivery to the PCIe slots? Remember how everyone was disappointed it only ended up with 75W again?

PCIe4.0 is useless unless you're doing molecular dynamics or protein meshing.

>>59273398
Might showing me a PCIe slot that can fit several GPUs at once for that to become a problem?

And before >hurrdurr don't throw shit in my mouth the lanes are limit
Do remember that enterprise motherboards come with chipsets and PLX switches supporting over 100 lanes so you're never NOT gonna use the 3.0 x16

>>59273401
>PCIe4.0 is useless unless you're doing molecular dynamics or protein meshing.

640K is more memory than anyone will ever need

>>59273401
Why would anybody have thought that?
I mean there's clear obvious limitations as to why we can't have 14 gauge copper wiring spanning from to ATX power slot to each PCIE slot, did people not expect that?
I actually don't believe that people thought it would have 300w slot power, that's just retarded

>>59273401
Fuck motherboard creeps, the I/O , third party chips and likewise traces just keep going down generation by generation and they still charge the same shit for it.

File: Chipset Diagram.png (33KB, 837x789px) Image search: [Google]

33KB, 837x789px

>>59273450
Multiple GPUs behind a PCIe switch would easily saturate the upstream link.

>>59273450
There are cards with two chips on board, rx295 I believe is one

>>59273473
http://www.tomshardware.com/news/pcie-4.0-power-speed-express,32525.html

Few months later:
http://techreport.com/news/30574/pcie-4-0-wont-actually-deliver-300-watts-from-the-slot

Yes they did.

>>59273388
mite b cool

if nothing less the shitposting about Intel's entry into the discrete gpu market would be fun to witness

>>59273473
Riser cards? Though I have never seen it being used for GPUs, mostly for stuff like SATA/SAS

>>59273528
Intel doesn't have the patents or 3D driver legacy to compete with AMD and Nvidia in the discrete market, Intel only started giving a shit about their 3D drivers post-SB

>>59273450
When you use Crossfire there is communication between the GPUs through PCI which eats up more bandwidth so even if both slots are running at 16x, and often they don't, it becomes a bottleneck.

>>59273496
>in other news, 100% of power supplies are incompatible with new motherboard standard

Speaking of power, when are we gonna get rid of the cumbersome 24pin ATX and P4 connector and replace them with something smaller? For fuck sakes we already got USB-PD that can deliver 100W

>>59273642
>get rid of 24pin ATX

Can't be too fucking soon.

>>59273642
>m-muh legacy support

>make the 24pin modular
>m-muh cost

You can't win against jews.

>>59273642
Bigger wire gauging means colder, more durable cables
How often are you unplugging your 12v main?

>>59273642
Until we reach some state that makes ATX a stone on the road, no.

sort of related, how much performance am I losing by having a pcie 3.0 video card plugged into pice 2.0?

>>59273712
It's a pain for SFF builds that aren't using 12V DC

>>59273764
About 1-2% on a bad day.

The spec isn't finalized, and won't be for a long time yet.

if this will bring NVMe SSDs with nearing 1million IOPS then it can't come soon enough.

File: 1405085945808.png (12KB, 560x407px) Image search: [Google]

12KB, 560x407px

>>59272914
>never used PCIe 3.0
>don't feel like I'm missing out on anything
>even the super expensive GPUs barely drop performance on PCIe 2.0
>now there is PCIe 4.0
>mfw

I'm still using PCIe v1.0

>>59273897
What good are NVMe SSDs with 1 million IOPS when they have limited number of writes?

>>59273817
So? We had N routers before they finished that spec.

Remember how 5-6 years ago system memory speeds never made a difference for gaming and now higher speed kits can increase 20% FPS or higher in no small amount of games?

Maybe the same will happen with PCIe4.0 and we might get an unexpected boost.

>>59274054
So does spinning rust, thankfully you won't see that happening in consumer workoads even with TLC and probably even QLC

>>59274054
What good are the cells in your body giving they can only divide so many times?

>>59273388
>My uncle works at Nintendo.

So we are at the point now where we just keep getting new standards that far outpace what we even use at home. Eh whatever it doesn't hurt anything.

>>59274195

Yeah you might as well just kill yourself now.