PCI Express 4.0 Brings 16 GT/s And At Least 300 Watts At The Slot

[InvalidError]

I'm also concerned with backwards compatibility. I doubt a GPU like the 750 Ti for example would like a 300W slot.

Unless the PCIe slot pinout has changed, a 750Ti will be perfectly fine in a 300W slot - the card only draws what it needs. Where compatibility issues will arise is trying to put a 300W card in a 75W slot since the motherboard most likely won't be able to handle that.

If sending 300W to the PCIe slots require having a bunch of extra 12V and GND cables going to the motherboard, might as well stick to 12V connectors on the cards themselves.

 

[Memnarchon]

"If you look closely, there is a single 20-pin power connector like any other modern motherboard. There are also four 8-pin and two 6-pin power connectors."
Am I the only one that no matter how many times I look closely, I can only find a 20pin for the motherboard and 2 x 8pins, one for each CPU???

Unless I am missing something, the 2x8pins + 2x6pins connectors are not placed on this motherboard.
Well at least they have the space to place 2x8pins + 2x6pins connectors...

 

[Guest]

That OCulink looks cool, but it better not have two cables coming out of it like the illustration. Because that's stupid, SATA Express stupid. If you really need two cables, then have them surrounded by another rubber tube or braid, so that it's only one cable. I bet some autistic engineer thought "two cables are cool, double power". Well, no, two cables isn't cool, it's stupid.

 

[daglesj]

Good to see market churn at play.

 

[turkey3_scratch]

They're going to need to redesign the ATX cable in order for 300W.

 

[amateramasu]

That OCulink looks cool, but it better not have two cables coming out of it like the illustration. Because that's stupid, SATA Express stupid. If you really need two cables, then have them surrounded by another rubber tube or braid, so that it's only one cable. I bet some autistic engineer thought "two cables are cool, double power". Well, no, two cables isn't cool, it's stupid.

It looks like the illustration is just showing an internal cable to the faceplate, just like USB works right now, not 2 cables from the pc -> peripheral.

 

[josejones]

From the article above:

"NVMe SSDs using PCIe 3.0 x4 recently reached the usable limit of the interface, such as the Samsung SM961 that reads data at 3,100 MB/s."

 

[bloodroses]

For anyone concerned about the high wattage output from the slot with legacy 3.0 cards; as long as the voltage is the same, it will work fine. The way that electronics work is that they only draw what they need for amperage, so having an excess will not harm anything at all (Watts = Volts * Amps). Different voltages on the other hand will cause it to fry. It is the reason why a 1200 watt power supply works fine on a computer that only needs 300 watts (although not exactly the most efficient).

 

[wifiburger]

yes more bandwidth, more standards to compensate for poor coding,
for some reason I don't think compilers will take full advantage of this insane bandwidth,

of course sony will take a low end gpu have register modified and restructure things around to make it perform as good as a high end video card without 300$ motherboard and 800$ video card,
i'll pass

 

[Aspiring techie]

You misspelled Solomon in the paragraph right under the server photo.

 

[elbert]

But will it allow 3 and 4 way 1080 sli?

 

[bit_user]

But will it allow 3 and 4 way 1080 sli?

It won't impact the GTX 1080, not least because that's a PCIe 3.0 card.

SLI uses a separate interconnect, making PCIe largely irrelevant, although it could go over PCIe. But, as GPUs get faster, they're likely to consume much of the additional PCIe bandwidth in single- and dual-card configs.

Anyway, I wouldn't count on more than 2-way configurations ever making much sense, for the majority of games.

 

[josejones]

Is there anything that's been close to saturating the bandwidth of PCIe 3.0? GPUs don't even saturate 2.0, and I don't think SSDs are that fast yet. I have to assume the biggest reason for the next generation is just the power delivery. Probably just doubled the bandwidth now because they'll need to eventually and we have the ability now.

From the article above:

"NVMe SSDs using PCIe 3.0 x4 recently reached the usable limit of the interface, such as the Samsung SM961 that reads data at 3,100 MB/s."

 

[bit_user]

The performance shifts from 3.0 to 4.0 while x16 devices like video cards get a massive headroom increase that will aid in providing the bandwidth needed for next generation virtual and augmented reality.

Sure enough, Hololens uses it to connect the HPU and SoC.

http://www.tomshardware.com/news/mcirosoft-hololens-components-hpu-28nm,32546.html

 

[firefoxx04]

You people are always crying about not needing faster. At some point we will need pcie 4.0. Fact. I'd rather have the supporting technology in place when the gpus and other peripherals are ready for use that will saturate it.

Yall are the same people that cried about dual cores being enough. Probably arguing the same thing about quad cores right now. More is better. Faster is better. Options are better.

 

[ern88]

So to get 300 watts per slot. You will have to add probably what a 8 pin power cable I'm guessing?

 

[turkey3_scratch]

The PCB itself will have to be made reinforced and stronger. To quote a member on Jonnyguru:

The_Mask
Probably a marketing dude who didn't get it. Maybe they're gonna change the max to 400W or 500W. But no way that's possible without external power. You need extreme PCB layers for that, way to expensive and totally useless.

Which makes sense, because if you think of it the wires don't just stop at the motherboard, the PCB continues it on.

 

[alextheblue]

The most interesting part is the standardized new-gen OCuLink cabling. I've never been a big fan of M.2 myself but U.2 isn't practical for most desktop users. A new cost-efficienct cable for the 2.5" form factor was exactly what I was hoping for. Of course, I'd like to see it used for external graphics cards as well.

The only issue I see is the dual copper/optical standards. Seems like that would split the market. That will be OK if it's split consumer/professional. Time will tell.

 

[bigpinkdragon286]

The additional wattage at the slot is a terrible idea. As it's been mentioned before, we're already supplying 6 or 8-pin power to add-in boards. This doesn't eliminate extra power cables, just unnecessarily changes where we plug those cables in. As soon as you move the power connections from the add-in board to the motherboard, you break backward compatibility with older motherboards that can't supply high wattage to the slots.

 

[WFang]

This makes more sense now, after the update.. Routing 300+W on PCB traces to the type of connector used for PCIe never seemed like a very practical idea. There is a reason for the need to run that sort of power directly in high-gauge wires to beefy pin-connectors on the cards.

 

[manleysteele]

This is my understanding of PCIe 4.0 power subsystems. First, as a side note, 2 8 pin connectors has never been an approved format for PCIe 3.0 devices. PCIe tops out at 1 8 pin connector and 1 6 pin connector. Anything above that is not to standard. You get 75w from the bus, 75w from the 6 pin connector and 150w from the 8 pin connector for a total of 300w. For a long time, the PCI-Sig has had a 1 6pin connector and 2 8 pin connectors for a total of 450w for PCIe 4.0. There has been no change to the 75w ceiling for the bus, so far as I know. Adding a 2nd 6 pin connector to the card would bring the total to 525w. I don't know what usage case this configuration would support, but lack of imagination by me is no bar to a perceived need from someone else. In any case, my understanding of the slot power configuration remains at 75w. x32 has been ready for a long time with no takers that I know of, but that may change soon. 8 port NVME controllers come to mind. The enterprise space has usage cases for things we won't see in the consumer space for quite a while, if we ever see them. For all I know, enterprise systems may be lousy with x32 controllers.

 

[turkey3_scratch]

The article was just updated yesterday. There was a misunderstanding, there is no "300W" from the slot.

 

[oddbondboris]

iphone 6 and 6s at least use pcie for storage interfacing

 

[bit_user]

x32 has been ready for a long time with no takers that I know of, but that may change soon. 8 port NVME controllers come to mind. The enterprise space has usage cases for things we won't see in the consumer space for quite a while, if we ever see them. For all I know, enterprise systems may be lousy with x32 controllers.

I'm no server expert, but I've never seen nor heard of a x32 card or a system with a x32 slot. Dell's dual-processor servers didn't even bother to include x16 slots, until GPU computing started heating up.