Jump to content

CXL 1.0 Specification released

LukeSavenije

Sources: @AluminiumTechAnandtech

 

With the move from single core to multi-core performance, a new war has started with how data is moved between compute resources. The Interconnect Wars is here, and it just got even more complicated. We’ve seen NVLink, CCIX, and GenZ as a host-to-device and device-to-device high-speed interconnect, from different companies with many different features. Now CXL, or Compute Express Link, is Intel's take on it.

 

Quote

image.png.8876e822c2dd2be593f6cdbeef82de7a.png

image.png.fef29acc1a2ca562c8081122319a0e88.png

image.png.9d35d4f763beb52e0bace177f4bbbe7b.png

 

 

Quote

At its heart, Compute Express Link (CXL) will initially begin as a cache-coherent host-to-device interconnect, focusing on GPUs and FPGAs. It will use current PCIe 5.0 standards for physical connectivity and electrical standards, providing protocols for I/O and memory with coherency interfaces.  The focus of CXL is to help accelerate AI, machine learning, media services, HPC, and cloud applications. With Intel being at the heart of this technology, we might expect to see future Intel GPUs and FPGAs connecting in a PCIe slot in ‘CXL’ mode. It will be interesting to see if this will be an additional element of the product segmentation strategy.

 

it surely is a interesting move, but i'm not sure if it's needed. with many other interconnects already availible, the choice for intel to make a own is a weird idea in my opinion.

Edited by LukeSavenije
added sources
Link to comment
Share on other sites

Link to post
Share on other sites

Just now, LukeSavenije said:

A interconnect standard made by intel

 

as mentioned in the article above...

*PCIe 4.0 but it's rebranded

Ryzen 7 3700X / 16GB RAM / Optane SSD / GTX 1650 / Solus Linux

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, NunoLava1998 said:

*PCIe 4.0 but it's rebranded

actually 5.0, but you're sort of right

 

except that it's like nvlink: you connect devices with each other

Link to comment
Share on other sites

Link to post
Share on other sites

19 minutes ago, LukeSavenije said:

actually 5.0, but you're sort of right

 

except that it's like nvlink: you connect devices with each other

Afik pcie 5.0 hasn't been defined fully yet so it can't be 5.0. Probably a typo and should be 4.0.

Link to comment
Share on other sites

Link to post
Share on other sites

5 minutes ago, CarlBar said:

Afik pcie 5.0 hasn't been defined fully yet so it can't be 5.0. Probably a typo and should be 4.0.

Most likely no. This is for their PCIe 5.0 generation parts. Though what's the difference between this and CCIX or GenZ? Beyond just Intel wanting their own name brand on something.

Link to comment
Share on other sites

Link to post
Share on other sites

5 minutes ago, CarlBar said:

Afik pcie 5.0 hasn't been defined fully yet so it can't be 5.0. Probably a typo and should be 4.0.

according to anandtech writer Ian Cutress it's 5.0 both in the whole article and the attached pictures... so I'm not sure if it's a typo

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, Taf the Ghost said:

Most likely no. This is for their PCIe 5.0 generation parts. Though what's the difference between this and CCIX or GenZ? Beyond just Intel wanting their own name brand on something.

i haven't found anything tbh...

Link to comment
Share on other sites

Link to post
Share on other sites

So is this just Intel’s response to AMD’s Infinity Fabric?

Current LTT F@H Rank: 90    Score: 2,503,680,659    Stats

Yes, I have 9 monitors.

My main PC (Hybrid Windows 10/Arch Linux):

OS: Arch Linux w/ XFCE DE (VFIO-Patched Kernel) as host OS, windows 10 as guest

CPU: Ryzen 9 3900X w/PBO on (6c 12t for host, 6c 12t for guest)

Cooler: Noctua NH-D15

Mobo: Asus X470-F Gaming

RAM: 32GB G-Skill Ripjaws V @ 3200MHz (12GB for host, 20GB for guest)

GPU: Guest: EVGA RTX 3070 FTW3 ULTRA Host: 2x Radeon HD 8470

PSU: EVGA G2 650W

SSDs: Guest: Samsung 850 evo 120 GB, Samsung 860 evo 1TB Host: Samsung 970 evo 500GB NVME

HDD: Guest: WD Caviar Blue 1 TB

Case: Fractal Design Define R5 Black w/ Tempered Glass Side Panel Upgrade

Other: White LED strip to illuminate the interior. Extra fractal intake fan for positive pressure.

 

unRAID server (Plex, Windows 10 VM, NAS, Duplicati, game servers):

OS: unRAID 6.11.2

CPU: Ryzen R7 2700x @ Stock

Cooler: Noctua NH-U9S

Mobo: Asus Prime X470-Pro

RAM: 16GB G-Skill Ripjaws V + 16GB Hyperx Fury Black @ stock

GPU: EVGA GTX 1080 FTW2

PSU: EVGA G3 850W

SSD: Samsung 970 evo NVME 250GB, Samsung 860 evo SATA 1TB 

HDDs: 4x HGST Dekstar NAS 4TB @ 7200RPM (3 data, 1 parity)

Case: Sillverstone GD08B

Other: Added 3x Noctua NF-F12 intake, 2x Noctua NF-A8 exhaust, Inatek 5 port USB 3.0 expansion card with usb 3.0 front panel header

Details: 12GB ram, GTX 1080, USB card passed through to windows 10 VM. VM's OS drive is the SATA SSD. Rest of resources are for Plex, Duplicati, Spaghettidetective, Nextcloud, and game servers.

Link to comment
Share on other sites

Link to post
Share on other sites

3 minutes ago, sazrocks said:

So is this just Intel’s response to AMD’s Infinity Fabric?

pretty much

Link to comment
Share on other sites

Link to post
Share on other sites

7 minutes ago, Mira Yurizaki said:

Obligatory:

standards.png

 

ASUS X470-PRO • R7 1700 4GHz • Corsair H110i GT P/P • 2x MSI RX 480 8G • Corsair DP 2x8 @3466 • EVGA 750 G2 • Corsair 730T • Crucial MX500 250GB • WD 4TB

Link to comment
Share on other sites

Link to post
Share on other sites

CALLED IT. 1 moment..

 

EDIT:

https://discordapp.com/channels/375436620578684930/375436621450969089/551066019180052480
image.png.aad47ecf77191f4f6a8ae29f21a2c32c.png

 

CXL specific connector when?

PLEASE QUOTE ME IF YOU ARE REPLYING TO ME

Desktop Build: Ryzen 7 2700X @ 4.0GHz, AsRock Fatal1ty X370 Professional Gaming, 48GB Corsair DDR4 @ 3000MHz, RX5700 XT 8GB Sapphire Nitro+, Benq XL2730 1440p 144Hz FS

Retro Build: Intel Pentium III @ 500 MHz, Dell Optiplex G1 Full AT Tower, 768MB SDRAM @ 133MHz, Integrated Graphics, Generic 1024x768 60Hz Monitor


 

Link to comment
Share on other sites

Link to post
Share on other sites

5 hours ago, Taf the Ghost said:

Most likely no. This is for their PCIe 5.0 generation parts. Though what's the difference between this and CCIX or GenZ? Beyond just Intel wanting their own name brand on something.

According to charlie boy, pretty much nothing between this and CCIX, yet...

 

https://semiaccurate.com/2019/03/11/intel-releases-compute-express-link-spec/

MOAR COARS: 5GHz "Confirmed" Black Edition™ The Build
AMD 5950X 4.7/4.6GHz All Core Dynamic OC + 1900MHz FCLK | 5GHz+ PBO | ASUS X570 Dark Hero | 32 GB 3800MHz 14-15-15-30-48-1T GDM 8GBx4 |  PowerColor AMD Radeon 6900 XT Liquid Devil @ 2700MHz Core + 2130MHz Mem | 2x 480mm Rad | 8x Blacknoise Noiseblocker NB-eLoop B12-PS Black Edition 120mm PWM | Thermaltake Core P5 TG Ti + Additional 3D Printed Rad Mount

 

Link to comment
Share on other sites

Link to post
Share on other sites

5 minutes ago, S w a t s o n said:

According to charlie boy, pretty much nothing between this and CCIX, yet...

 

https://semiaccurate.com/2019/03/11/intel-releases-compute-express-link-spec/

AMD and some of the ARM Server makers will be shipping servers with CCIX before Intel has working silicon for testing of CXL. Intel can force certain adoptions, but if the ecosystem is already operating, it's Itanium again.

Link to comment
Share on other sites

Link to post
Share on other sites

8 minutes ago, rcmaehl said:

CALLED IT. 1 moment..

 

EDIT:

https://discordapp.com/channels/375436620578684930/375436621450969089/551066019180052480
image.png.aad47ecf77191f4f6a8ae29f21a2c32c.png

 

CXL specific connector when?

You can't go much smaller with the Slots we have right now. To go smaller, we need a different slot system, wich is till based on the Original stuff from the late 70s...

"Hell is full of good meanings, but Heaven is full of good works"

Link to comment
Share on other sites

Link to post
Share on other sites

1 minute ago, Taf the Ghost said:

AMD and some of the ARM Server makers will be shipping servers with CCIX before Intel has working silicon for testing of CXL. Intel can force certain adoptions, but if the ecosystem is already operating, it's Itanium again.

Intel is also losing importance right now with the upcoming 7nm Rome, there probably isn't any reason to get an Intel Chip for any reason. Especially with all that Spectre/Meltdown stuff.


With AMD and ARM working together with the CCIX stuff, I doubt that Intel has any chance with their stuff. Especially when there are a ton of Acceleration Cards with CCIX Support - wich are already announced by Altera or some other FPGA manufacturers.

 

Especially since all that stuff will be released somewhere late 2020, if not 2021 while CCIX will be released later this year. That's more than a year for the free and open standard...

"Hell is full of good meanings, but Heaven is full of good works"

Link to comment
Share on other sites

Link to post
Share on other sites

Just now, Stefan Payne said:

Intel is also losing importance right now with the upcoming 7nm Rome, there probably isn't any reason to get an Intel Chip for any reason. Especially with all that Spectre/Meltdown stuff.


With AMD and ARM working together with the CCIX stuff, I doubt that Intel has any chance with their stuff. Especially when there are a ton of Acceleration Cards with CCIX Support - wich are already announced by Altera or some other FPGA manufacturers.

 

Especially since all that stuff will be released somewhere late 2020, if not 2021 while CCIX will be released later this year. That's more than a year for the free and open standard...

Even if Rome was 5x more powerful per socket than Intel's CPUs at the same cost, it'd be years before AMD put a huge dent in Intel's sales. Companies don't buy from Intel or AMD but from SIs and Channel companies, thus the entire ecosystem will still keep producing Intel systems. And, if demand suddenly reached Intel levels, there isn't enough Fab space. Intel is the world's largest Fab company, and by a fairly wide margin.

 

But what it does mean for a certain type of ecosystem is AMD dominance within it for several years. AMD really should be pressing hard to get that off the ground before Intel can force its own standards on things.

Link to comment
Share on other sites

Link to post
Share on other sites

2 minutes ago, Taf the Ghost said:

And, if demand suddenly reached Intel levels, there isn't enough Fab space. Intel is the world's largest Fab company, and by a fairly wide margin.

 

I think that's Samsung actually but intel may relcaim this year

MOAR COARS: 5GHz "Confirmed" Black Edition™ The Build
AMD 5950X 4.7/4.6GHz All Core Dynamic OC + 1900MHz FCLK | 5GHz+ PBO | ASUS X570 Dark Hero | 32 GB 3800MHz 14-15-15-30-48-1T GDM 8GBx4 |  PowerColor AMD Radeon 6900 XT Liquid Devil @ 2700MHz Core + 2130MHz Mem | 2x 480mm Rad | 8x Blacknoise Noiseblocker NB-eLoop B12-PS Black Edition 120mm PWM | Thermaltake Core P5 TG Ti + Additional 3D Printed Rad Mount

 

Link to comment
Share on other sites

Link to post
Share on other sites

21 minutes ago, Stefan Payne said:

Intel is also losing importance right now with the upcoming 7nm Rome, there probably isn't any reason to get an Intel Chip for any reason. Especially with all that Spectre/Meltdown stuff.


With AMD and ARM working together with the CCIX stuff, I doubt that Intel has any chance with their stuff. Especially when there are a ton of Acceleration Cards with CCIX Support - wich are already announced by Altera or some other FPGA manufacturers.

 

Especially since all that stuff will be released somewhere late 2020, if not 2021 while CCIX will be released later this year. That's more than a year for the free and open standard...

isnt altera intel though?

Link to comment
Share on other sites

Link to post
Share on other sites

13 minutes ago, Taf the Ghost said:

And, if demand suddenly reached Intel levels, there isn't enough Fab space. Intel is the world's largest Fab company, and by a fairly wide margin.

To add it's not like AMD can leverage other fabs outside of TSMC, since GF stopped their 7nm development and Samsung is the only other company that can do it (as far as I know), and they've got other things to worry about.

 

So either AMD has to release Rome at a larger fab processor is just deal with low supply.

Link to comment
Share on other sites

Link to post
Share on other sites

Its entirely possible that they might dual source their Zen2 Dies from multiple foundrys, once Samsung got their 7nm process running.

AMD can not have short supply of their 7nm chips.

 

"Hell is full of good meanings, but Heaven is full of good works"

Link to comment
Share on other sites

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×