PCI-SIG Announces PCIe 6.0 Specification

[AleksandarK]

PCI-SIG today announced that PCI Express (PCIe ) 6.0 technology will double the data rate to 64 GT/s while maintaining backwards compatibility with previous generations and delivering power efficiency and cost-effective performance. The PCIe 6.0 specification is actively targeted for release in 2021.

PCIe 6.0 Specification Features

• Delivers 64 GT/s raw bit rate and up to 256 GB/s via x16 configuration
• Utilizes PAM-4 (Pulse Amplitude Modulation with 4 levels) encoding and leverages existing 56G PAM-4 in the industry
• Includes low-latency Forward Error Correction (FEC) with additional mechanisms to improve bandwidth efficiency
• Maintains backwards compatibility with all previous generations of PCIe technology

"PCI Express technology has established itself as a pervasive I/O technology by sustaining bandwidth improvements for five generations over two decades," Dennis Martin, an analyst at Principled Technologies, said. "With the PCIe 6.0 specification, PCI-SIG aims to answer the demands of such hot markets as Artificial Intelligence, Machine Learning, networking, communication systems, storage, High-Performance Computing, and more."

"Continuing the trend we set with the PCIe 5.0 specification, the PCIe 6.0 specification is on a fast timeline," Al Yanes, PCI-SIG Chairman and President, said. "Due to the continued commitment of our member companies, we are on pace to double the bandwidth yet again in a time frame that will meet industry demand for throughput."

To learn more about PCI-SIG, visit www.pcisig.com.

About PCI-SIG

PCI-SIG is the consortium that owns and manages PCI specifications as open industry standards. The organization defines industry standard I/O (input/output) specifications consistent with the needs of its members. Currently, PCI-SIG is comprised of over 800 industry-leading member companies. To join PCI-SIG, and for a list of the Board of Directors, visit www.pcisig.com.

PCI-SIG, PCI Express, and PCIe are trademarks or registered trademarks of PCI-SIG. All other trademarks are the property of their respective owners.

View at TechPowerUp Main Site

 

[cucker tarlson]

x670 boards at $300 for the budget one coming up

 

[zlobby]

That is just an announce. Even 5.0 is not yet completely ratified.

 

[Zareek]

I seems like they went on vacation for 5 years after PCIe 3 and came back with their pants on fire!

 

[Tomgang]

Wow there boy settle down. PCIe gen 4 has just been released and PCIe gen 5 is stil not even released and now we al ready talk about PCIe gen 6.

This makes my brain hurt. Teknologi advance to fast now. New tech are ready even faster than the older one can be released on the marked. THIS MAKES MY BRAIN TREMBLE.

 

[dKenGuru]

Yea, let's rise number in every year.

 

[bug]

Gee, they dodged a bullet there. A few days later and PCs with PCIe 4.0 would have been already released.

 

[cucker tarlson]

Wow there boy settle down. PCIe gen 4 has just been released and PCIe gen 5 is stil not even released and now we al ready talk about PCIe gen 6.

This makes my brain hurt. Teknologi advance to fast now. New tech are ready even faster than the older one can be released on the marked. THIS MAKES MY BRAIN TREMBLE.

not really.tbh it's stagnant as hell.the only thing that keeps advancing at blistering pace is prices.

I mean we've got pcie 4.0 rx5700 that's only just beating 2070 and pci-e 4.0 drives that advertise some stellar sequential transfers but I bet they couldn't beat a quality pci-e 3.0 drive in random speeds anyway.

I'm still on a board that i bought in 2014,cpu from 2015 and sata ssds and I don't really see any of those really slowing my pc down at all.

 

[Tomgang]

not really.tbh it's stagnant as hell.the only thing that keeps advancing at blistering pace is prices.

About the prices. I can only agreed to that. Prices on cpu' s as well as GPU's has sky rocket to where its litterly stupid.

Exsamples. Intels i7 975X cost at the time 999$ and today intels extreme cpu cost a bargain 1999$. Best deal of the year right

Back in 2009 the fastest single gpu card whas gtx 285 and i paid what is 360 usd with taxes and shipping. Today in the other hand, the fastest single gpu today meant for gaming cost a round 3300 usd for titan rtx and for rtx 2080 ti is around 1600 usd. All price comparison on gpu's is from my own country from now and back then.

So yeah prices has been stupidity high.

 

[Deleted member 24505]

I think PCI-SIG need to take a chill and relax, PCI-E 4 is not even out yet.

 

[Wavetrex]

Both 4 and 5 are still slow compared to NVLink and Infinity Fabric.

Those two links are the reason they are pushing for quick increase in transfer speeds, as PCI-e has fallen behind A LOT.

 

[ncrs]

I think PCI-SIG need to take a chill and relax, PCI-E 4 is not even out yet.

IBM POWER9 CPUs have been shipping with PCIe 4.0 since August 2017.

 

[zlobby]

Both 4 and 5 are still slow compared to NVLink and Infinity Fabric.

Those two links are the reason they are pushing for quick increase in transfer speeds, as PCI-e has fallen behind A LOT.

Only difference between them is that nvlink and IF are PROPRIETARY, if this rings a bell. Anyone can design a custom, purpose-built solution and it will be way faster than PCIe.

PCIe main goals are compatibility, versatility and cost. Thing that are all missing in nvlink and IF.

 

[seronx]

Request Rejected

=> A Reconfigurable 28/56 Gb/s PAM4/NRZ Dual-mode SerDes with Hardware-reuse

Request Rejected

=> A Reconfigurable 16/32 Gb/s Dual-Mode NRZ/PAM4 SerDes in 65-nm CMOS

Request Rejected

=> A 3D-Integrated 56 Gb/s NRZ/PAM4 Reconfigurable Segmented Mach-Zehnder Modulator-Based Si-Photonics Transmitter

High Speed Reconfigurable NRZ/PAM4 Transceiver Design Techniques

oaktrust.library.tamu.edu

=> High Speed Reconfigurable NRZ/PAM4 Transceiver Design Techniques


NRZ(PCIe 5.0) <-> PAM4(PCIE 6) is definitely in the possibility. With the added benefit of PAM4(PCIe) <-> PAM4(GenZ), with same silicon and different AIB.

 

[Solaris17]

The thing is though, this isnt fast at all. Its the same comment I made previously about CPU architectures. This stuff is on private road maps YEARS in advance. Gen 5 is only now seeing the light at the end of the tunnel. The announcement of 6 is just a general outline. It hasn't manifested into anything more than smoke.

Even then, Gen 5 and Gen 6 will hit enterprise and speciality applications before it hits consumers. PCI-E 4 is new to "General Populace" but gen 4 is not "new".

 

[Aquinus]

Even then, Gen 5 and Gen 6 will hit enterprise and speciality applications before it hits consumers. PCI-E 4 is new to "General Populace" but gen 4 is not "new".

Well, even that might be a stretch because even servers in the x86 market don't even support PCIe 4.0 yet. The spec might have been completed two years ago, but it takes time to develop hardware to use it. The most we've seen are some showcases of how it has been done with NVMe and network adapters, but even then the news around PCIe 4.0 has been rather quiet.

 

[Solaris17]

Well, even that might be a stretch because even servers in the x86 market don't even support PCIe 4.0 yet. The spec might have been completed two years ago, but it takes time to develop hardware to use it. The most we've seen are some showcases of how it has been done with NVMe and network adapters, but even then the news around PCIe 4.0 has been rather quiet.

Servers and enterprise generally see things first, but maybe I wasn’t clear enough with the “speciality applications”. Any consortium generally reaps the benefits of the technology in which they sit in for. Even if it isn’t something as obvious as “pci express” or “nvme” the prototype controllers and tech generally finds itself in something even if it doesn’t share a namesake, like proprietary fiber fabrics, or storage mediums.

Some of that proprietary tech never even sees consumer consumption because it is cut watered down or modified into a different product that does not resemble what is used. Like custom blade interconnects etc etc.

EDIT:: Of course I also think you are right, I don't thing everything is like that. Not all tech necessarily trickles down. For example, I dont think RAM does.

 

[Aquinus]

Servers and enterprise generally see things first, but maybe I wasn’t clear enough with the “speciality applications”. Any consortium generally reaps the benefits of the technology in which they sit in for. Even if it isn’t something as obvious as “pci express” or “nvme” the prototype controllers and tech generally finds itself in something even if it doesn’t share a namesake, like proprietary fiber fabrics, or storage mediums.

Some of that proprietary tech never even sees consumer consumption because it is cut watered down or modified into a different product that does not resemble what is used. Like custom blade interconnects etc etc.

Sure, my point though is that adoption takes time.

 

[metalfiber]

They also released the cooling solution fan for the chipset...

 

[dicktracy]

Just skip PCIE 4 and 5 and move directly into 6.0 lol

 

[beautyless]

Just skip PCIE 4 and 5 and move directly into 6.0 lol

They can skip but not so fast to be ready for 6.0, I think, around 2024.
Would you like to use pcie 3.0 waiting for 5 year.

 

[R0H1T]

I bet this is Intel pushing their agenda with CXL, much like TB & USB 4.0

They can skip but not so fast to be ready for 6.0, I think, around 2024.
Would you like to use pcie 3.0 waiting for 5 year.

They can't skip PCIe 4.0 as it's already here & has been for the last 2 years, IBM had it on their chips but it didn't become mainstream because ARM & x86 didn't support it. With zen2 AMD have leapfrogged Intel in terms of feature/parity & (peripheral) connectivity, but looks like Intel wants everyone else to skip PCIe 4.0 & go straight to rev.5 especially with their push for CXL.

 

[ratirt]

I thought intel is going with PCIe 5.0 with the new processors? I read somewhere about it. Now there's PCIe 6.0. By the time I get my new 3900X Ryzen there will be 9.0 PCIe. They all need to take a chill pill. I feel it's like a race who is going to support the newest technology (not necessarily released). This only makes me sure that they will jack-up the prices and I really don't want that cause they are high as they are right now.

 

[metalfiber]

I knew this article was around here somewhere. It's a good read about PCIe gen 3.0 scaling....


"By this time next year, we could see the first desktop platforms and GPUs implementing PCI-Express gen 4.0 in the market. If only "Turing" supported PCIe gen 4.0, you would have had the luxury to run it at gen 4.0 x8 without worrying about any performance loss. Exactly this is the promise of PCIe gen 4.0, not more bandwidth per device, but each device working happily with a lower number of lanes, so processor makers aren't required to add more lanes."

NVIDIA GeForce RTX 2080 Ti PCI-Express Scaling

It takes a lot of bandwidth to support the fastest graphics card, especially one that can play anything at 4K 60 Hz, with an eye on 120 Hz. The GeForce RTX 2080 Ti could be the most bandwidth-heavy non-storage PCIe device ever built. PCI-Express gen 3.0 is facing its design limits.

www.techpowerup.com

 

[zlobby]

I knew this article was around here somewhere. It's a good read about PCIe gen 3.0 scaling....


"By this time next year, we could see the first desktop platforms and GPUs implementing PCI-Express gen 4.0 in the market. If only "Turing" supported PCIe gen 4.0, you would have had the luxury to run it at gen 4.0 x8 without worrying about any performance loss. Exactly this is the promise of PCIe gen 4.0, not more bandwidth per device, but each device working happily with a lower number of lanes, so processor makers aren't required to add more lanes."

NVIDIA GeForce RTX 2080 Ti PCI-Express Scaling

It takes a lot of bandwidth to support the fastest graphics card, especially one that can play anything at 4K 60 Hz, with an eye on 120 Hz. The GeForce RTX 2080 Ti could be the most bandwidth-heavy non-storage PCIe device ever built. PCI-Express gen 3.0 is facing its design limits.

www.techpowerup.com

If the quoted article shows anything, it's that the 2080ti is bottlenecked by its performance in 4K, not by PCIe bandwidth.