PCIe 3.0 is Here; PCIe 4.0 Already in the Works
Tons more bandwidth for future applications.
Last Monday at a morning press conference in Santa Clara, PCI-SIG, the Special Interest Group overseeing the PCI Express standard, officially announced the availability of PCI Express 3.0 products, and stated that work on the PCI Express 4.0 standard has already begun.
PCI-SIG chairman Al Yanes stated that compliance workshops held in April and June included 23 add-in cards and 19 system boards which meet PCI Express 3.0 specifications. Yanes further stated that major products are underway from many PCI-SIG members, including Agilent Technologies, National Semiconductor, PLX Technology, and Tektronix.
The current PCI Express 2.1 standard enables a speed of 500 MB/s per lane, in each direction. Thus, the 16 lane connection commonly used for PCI Express graphics cards currently has a transfer speed of 8 GB/s. The PCI Express 3.0 standard will double the above transfer rates, enabling a speed of 1 GB/s per lane, or 16 GB/s for a 16 lane connection.
Despite the announcement of PCI Express 3.0 products, the biggest news came in the form of two additional announcements: A “Cable Workgroup” has been established to create a new cable standard, possibly with the intention of competing with Intel’s proprietary Thunderbolt technology. Also, early work on the PCI Express 4.0 standard has already begun.
Al Yanes stated that the cabling technology will have a goal of 20 inch cables for use with servers, 14 inch cables for use with desktops, and 10 inch cables for use with mobile devices. The cable is initially expected to push the limits of copper wire technology, but the door is being left open to include optical technology. Yanes stated “We have crossed every bridge we have come to… but we don’t want to rest on our laurels”.
Intel has been non-committal as to whether it will license Thunderbolt to PCI Express add-on card developers. The technology is certainly feasible: a four lane PCI Express 3.0 connection slot, operating at a speed of 4 GB/s (or 32 Gigabits per second), could easily host a 10 Megabit per second Thunderbolt PCI Express add-on card. However, Intel spokesman Dave Salvator told us recently, “There are no plans to do a Thunderbolt expansion card at this time.” This could mean that Intel plans to license Thunderbolt only as a motherboard feature, not as an add-on card feature. It’s certainly feasible that Intel’s plans are what has lead PCI-SIG to decide to develop their own new cabling technology.
As for PCI Express 4.0, PCI-SIG is confident that they’ll be able to once again double transfer speeds when the technology is introduced. Historically, the development of a new PCI Express standard occurs about once every four years, and PCI-SIG stated that the “4 year sequence” would likely continue to be the expected development period.
PCI Express 3.0 technology is expected to be used to power InfiniBand, Ethernet 40G/100G, ultra high speed Solid State Drives, and, of course, high end graphics cards.
- Quake Turns 15; Carmack & Hollenshead Speak
- DDR3 Modules to Near 90% Market Share in 2011
- Duke Nukem Forever Coming to Mac
- Deals June 27: Logitech MK550 Combo $39.99 FS
- Blizzard Investigating World of Warcraft for iPad
- This USB Drive is Powered By SandForce
- L.A. Noire Announced for PC
- Google to Face Federal Antitrust Probe
- Winner of the DSLR Father's Day Camera Bundle
- Acer's AC700 Chromebook Shipping This Week
- Patriot, Super Talent Launch SSDs with SF-2200
- Blizzard Launches WoW ''Starter Edition''
- Sony Vaio Z Has External Radeon Light Peak GPU
- Nvidia Boasts New Laptop GPU With DX11 Crysis 2
- Latest Windows 8 Rumor Says RTM by April 2012
- Deals June 29: 21.5'' Dell IPS Multi-touch $214
- Interplay: Bethesda Knew Fallout MMO Was More Than A Name
- Perry: Gaikai Uses Modern, Not 5-yr-old Hardware
Sweet can't wait to get me a mobo/graphics card that supports PCIe 3.0
good job... and a standard that makes sense...
now just to get a video card that uses all of the PCIe 3.0...
now just to get a video card that uses all of the PCIe 3.0...
... and a CPU to handle all the bandwidth.
I'll wait for Ivy Bridge and mobo/GPU with PCIe 3.0 for my next upgrade.
Is that supposed to be 10 GIGAbits per second for the Thunderbolt?
So is every one happy now that the three 3's are all officially out?
Usb 3.0, Sata III (6gb/s), and (finally) PCI-e 3.0.
Has PCI-e 2.0 been maxed out yet, in everyday desktop application?
So is every one happy now that the three 3's are all officially out? Usb 3.0, Sata III (6gb/s), and (finally) PCI-e 3.0.
Yes, Rainman is happy now
If board makers would flat-out drop 2.0 and fully adopt 2.1, video cards that require 2 x 6-pin (or 1 x 6-pin + 1 x 6+2-pin) would be a thing of the past because the motherboard could actually supply 150W through the PCIe slot itself.
PCIe technology is evolving faster than manufacturers seem willing to adopt it.
Has PCI-e 2.0 been maxed out yet, in everyday desktop application?
nope? loll
Has PCI-e 2.0 been maxed out yet, in everyday desktop application?
Nope. Not even gpu's are maxed out with pci-e 2.0
Very nice, tho current lanes aren't saturated yet. Even so technology marches on! Anyone know if the new spec will also include higher power delivery specs considering some GPU's are breaking the current one?
PCIe 3.0 is Here; PCIe 4.0 Already in the Works; PCIe 2.0 Still Not Taken Advantage of.
Has PCI-e 2.0 been maxed out yet, in everyday desktop application?
Yes, if you are looking at enterprise solutions which outweigh the high end GPU market. HBA and NIC tech is pushing the limits and creating a demand for more I/O and bandwidth.
It's not that I don't agree that pushing the standard is a good thing, but the fact that we can't even saturate the current interface makes this a lot less exciting.
It seems like the closest things we'll get will be the extreme SSDs that are essentially drives on the expansion slot (doesn't OCZ have one now?)
Below is an exert by "Another DimWit" From the toms hardware forums. Thought it might be interesting to share.
"PCIe 2.0 delivers 5 GT/s, but employs an 8b/10b encoding scheme which results in a 20 percent ((10-8)/10) overhead on the raw bit rate. PCIe 3.0 removes the requirement for 8b/10b encoding and instead uses a technique called "scrambling" in which "a known binary polynomial is applied to a data stream in a feedback topology. Because the scrambling polynomial is known, the data can be recovered by running it through a feedback topology using the inverse polynomial and also uses a 128b/130b ((130-128)/130)encoding scheme, reducing the overhead to approximately 1.5%, as opposed to the 20% overhead of 8b/10b encoding used by PCIe 2.0. PCIe 3.0's 8 GT/s bit rate effectively delivers double PCIe 2.0 bandwidth"
am I the only one that hasn't really seen any thunderbolt products?
PCIe 3.0 is Here; PCIe 4.0 Already in the Works; PCIe 2.0 Still Not Taken Advantage of.
Id rather them be way ahead of the game then behind and have issues.
I'm wondering... Will manufactures find the "need" to develop and promote the use of SLI/CrossFire? At this speeds on this article, I just can't see why.
I'm wondering... Will manufactures find the "need" to develop and promote the use of SLI/CrossFire? At this speeds on this article, I just can't see why.
I think you're confused. The reason you don't see faster video cards isn't because they max out the x16 slot. In fact, most cards won't even max the x8 slots.
am I the only one that hasn't really seen any thunderbolt products?
Check out the new Apple PCs, they're the only products I've seen that use the ports. Also does anyone actually understand the picture in the article- it doesn't seem to actually correspond to the article. Also the bit-interfaces on graphics cards- how do they relate to this new
standard- would they have to be upgraded before the new interface could be utilized completely? Also since onboard GPUs are ON the mobo how does their speed with the CPU relate vs a discrete card.
I think you're confused. The reason you don't see faster video cards isn't because they max out the x16 slot. In fact, most cards won't even max the x8 slots.
there was an article here a while ago, and it did 4x 8x and 16x by used of tapeing off cards. and if i remember right the difference between 4x and 16x was between 8 and 12% instead of the 75%+ you would expect if the bandwidth was actually used.
I'm glad that they're updating so early. Its WAY better to update the standards before bandwidth limits are reached rather than wait for them to be reached and to slow things down. It took too long for USB 3.0 to come out, so its good that PCIe 3.0 before its needed mainstream.
Everybody said the same thing about SATA when SATA II came out.
So a modern computer today has DDR3, USB 3.0, SATA 3 and now PCIe 3.0
Cool.
Is that supposed to be 10 GIGAbits per second for the Thunderbolt?
Actually they can go up to 50gb.
Check out the new Apple PCs, they're the only products I've seen that use the ports. Also does anyone actually understand the picture in the article- it doesn't seem to actually correspond to the article. Also the bit-interfaces on graphics cards- how do they relate to this new standard- would they have to be upgraded before the new interface could be utilized completely? Also since onboard GPUs are ON the mobo how does their speed with the CPU relate vs a discrete card.
Please don't put "Apple" comments on a monumental news story like this, it just ruins the mood.
How many years will it be until you think PCIe 4.0 will be useful? What kind of GPU power would we have?
The fact of the matter is, PCI-e 2.1 is faster than any current hardware could efficiently utilize. Rather than working on these new standards that push bandwidth limits even higher, why don't we put more focus on being able to actually utilize said bandwidth?
Id rather them be way ahead of the game then behind and have issues.
Very true... you could think of it that way, too.
Just imagine what a game could look like, if it was made with absolute PC optimization in mind, made for nothing other than taking full advantage of current and future hardware.
I wonder if anyone is thinking the same thing about that as I am right now.
Crysis.
You know, the game that was used to test hardware up until a year ago? It was so beautiful, even being made back that long ago. Now compare that year's hardware to what we have now. Just imagine how beautiful a game we could make if we kept that same proportion...
The fact of the matter is, PCI-e 2.1 is faster than any current hardware could efficiently utilize. Rather than working on these new standards that push bandwidth limits even higher, why don't we put more focus on being able to actually utilize said bandwidth?
Except we are pushing the limits of current PCI-e 2.0 / 2.1, just not in the desktop department. Enterprise FC SAN's currently use 8GFC as the standard per-port speed with 16GFC already approved and on the near term horizon. Heavy virtualization has lead to host systems needing 2~4 HBA's with 2x FC ports each along with 2~4 4x 1Gbe or 10Gbe ports depending. With this heavy requirement for large I/O the current generation of PCIe isn't going to cut it. Not to mention stuff like fusion-io drives already tax PCI-e bandwith and those have "Enterprise use only" written all over them.