Wednesday, August 5th 2009
PCI-Express 3.0 Hits Backwards Compatibility Roadblock, Delayed
PCI-SIG (Special Interest Group), the organisation responsible for development of PCI specifications announced that generation 3 PCI-Express (PCI-E 3.0), is off its target launch time from late-2009 to Q2 2010. Although work on the bus is almost finished, there seems to be problems with implementing backwards-compatibility with older generations of PCI-E. Assuming PCI-E 3.0 is standardised in Q2 2010, one can expect implementing products (motherboards and expansion cards supporting PCI-E 3.0) only by a year later.
PCI-E 3.0 packs features that overcome the bottlenecks of PCI-E 2.0, such as the removal of the 8P/10b encoding scheme that added at least 20% data overhead for the 5 GT/s PCI-E 2.0, reducing it to 4 GT/s effective. At 8 GT/s the new bus will have effectively twice the bandwidth.
Source:
TechConnect Magazine
PCI-E 3.0 packs features that overcome the bottlenecks of PCI-E 2.0, such as the removal of the 8P/10b encoding scheme that added at least 20% data overhead for the 5 GT/s PCI-E 2.0, reducing it to 4 GT/s effective. At 8 GT/s the new bus will have effectively twice the bandwidth.
62 Comments on PCI-Express 3.0 Hits Backwards Compatibility Roadblock, Delayed
Actually, no, I don't care. I have zero USB devices. :laugh:
Oh, the irony. :wtf:
Nevermind me.
Nathan Brookwood - principal analyst at Insight 64.
From what I heard PCIe 3.0 is expected to provide also lower power consumption than the previous PCIe 2.0 protocol.
Put a GTX 295 on PCIE 3.0 and it will still consume an elephants share of power. XD
Sigh, the next mobo to move to will be one with PCIe 3.0, USB 3.0 & SATA-III.:respect:
LOL, you could say it's 3G, because it'll be PCIe 3.0, USB 3.0, SATA-III, DDR3 & for AMD users AM3 on HTT3!:roll:
EDIT: Don't forget there are also limitations on the mobos' bus, perhaps sending too much wattage through would overheat/overload the mobos.
FYI, it is SATA 6 GB/s, not SATA III (SATA II is the group that set the specifications for SATA 3 GB/s). SAS already has some controllers that are 6 GB/s capable so it is pretty close to hitting mainstream, me thinks.
2.x= 150W
3.x= 200W (the cap, most will still cap at 150W.. its optional like Tessellation is with DX10)
6-pin connector = 75W, 8-pin connector = 150W, [unknown number]-pin = 300W (not 200W).
if the slots provided more power and cards used it, then you wouldnt be able to use the cards in an older slot design.
since you can use any 2.0 card in a 1.1 or 1.0 slot, its obvious they have not increase the power to the slot - because no cards use it.
the standards probably include the cables.
6-pin = 3 x 25w = 75w
8-pin = 4 x 25w = 100w
slot = 75w
total = 75w + 75w + 100w = 250w
I do believe these figures are important to power supply manufacturers. If they want to sell a PCIE 2.0-ready PSU, they need to make sure it has at least a 6-pin power connector (150w available to the card, total).
Want something faster than an SSD? they have, they called them SSD's as well... lol.
why not ask for 2TB floppy drives while we're at it.
Floppy, floppy, floppy... :rockout:
150watts/12V = 12.5A... thats a lot to run through the mobo
( on a 6 pin cable the minifit molex connectors are rated for around 9A each but the wire if its a 16ga can handle between 13A(enclosed) and 22A(in free air) each. So if you go with the lower spec (minifit ends) a 6 pin cable for a GPU can supply about 27A (27A X 12V=324 watts. I havent played around with the 8pin plugs too much but I noticed they are the same as the 6 pin ones (3 yellows) they just add 2 extra grounds to them.
Broadcast manufacturers like gates (harris allied) QRK and leagues of others all jumped on that technology in the late 60's early 70's and proceeded to guarantee employment for a lot of folks like me. They would build an entire console and transmitter parts as well based on this and if they needed more current just feed extra pins in paralell. new problem, too damn much heat, pins lose thier tensile strength, card pins start arcing, copper starts burning boards. and what once was the smell of success turns into a foul stench that means money in my pocket....
Next time you get to see the bottom of a mobo take a good look a how thin some of the traces are feeding that pci slot, and then there are feedthru points along the way that create more problems in a high current application.
When i got my first 780 chipset board i noticed in bios that you could assign a max wattage of 150w on the biostar bios for the pci slot and i just shuddered to think what would actually happen if a card tried to go there all on the slot power.
they have no choice, thats all an edge card connector will ever hold. but i bet we will see 8 pin molex with 4 b+ and 4 ground as the norm by 2012. just watch, its the only way unless they start making some real efficiant gpus with cooling that hasnt been thought of yet.
so its not so much because of backwards compatibility, but thats all those pins and copper can hold mathmatically and nothing can change that except for a radical new design in circuitbords and connectors, but not with things in existence as we know them today.
keep in mind 75 watts is considered average useage, and 150 is peak MAX , if your card tried to go to max for anything more than a few seconds hold time, it would not take long before you would see smoke-you can bank on that.
But thats what people are saying PCI-E 1.0 and 1.1 have the same power levels. both are 150W - the extra power only comes on cards that moved to 8-pin, or two plugs.