Intel Announces Compute Express Link

cageymaru

cageymaru

Fully [H]
Joined
Apr 10, 2003
Messages
22,076
Intel has announced Compute Express Link (CXL); a new industry open standard that allows for high-speed communications between CPU-to-Device and CPU-to-Memory interconnect in next-generation data centers. CXL creates a high-speed, low latency interconnect between the CPU and workload accelerators, such as GPUs, FPGAs and networking. CXL maintains memory coherency between the devices, allowing resource sharing for higher performance, reduced software stack complexity and lower overall system cost. The technology is built upon the PCI Express (PCIe) infrastruture and leverages the PCIe 5.0 physical and electrical interface to provide advanced protocol in three key areas: I/O Protocol, Memory Protocol, initially allowing a host to share memory with an accelerator, and Coherency Interface.

Some think that Intel is ready to embrace Gen-Z without embracing it outright. In the CXL announcement there are quotes linking CXL to Gen-Z. Robert Hormuth, Vice President & Fellow, Chief Technology Officer, Server & Infrastructure Systems, Dell EMC said, "Dell EMC is delighted to be part of the CXL Consortium and its all-star cast of promoter companies. We are encouraged to see the true openness of CXL, and look forward to more industry players joining this effort. The synergy between CXL and Gen-Z is clear, and both will be important components in supporting Dell EMC's kinetic infrastructure and this data era." Kurtis Bowman, President, Gen-Z Consortium said, "As a Consortium founded to encourage an open ecosystem for the next-generation memory and compute architectures, Gen-Z welcomes Compute Express Link (CXL) to the industry and we look forward to opportunities for future collaboration between our organizations."

Monday's announcement could give Intel an advantage for its upcoming graphics cards, since the chip-maker will be building CXL compatibility into its popular server processors. It is reasonable to argue that the CXL is new and that Intel has a lot of market pull. At the same time, the CXL partner list is nothing less than underwhelming. Intel got big names of its customers, but it did not get other accelerators.
 
Skull_Angel said:
Is it wrong that the first thing that comes to mind is "Look, more direct attack vectors!", just because this is coming from Intel?
Click to expand...
Intel Security.jpg
 
"Intel an advantage for its upcoming graphics cards,"

You guys think those will sell better or worse than those SSDs Intel started making a few years ago?
 
Well, I doubt Mellanox really appreciated them making OmniPath which directly competes with their products.

Another interesting thing to note is how ComputeExpress will be sharing the PCIe PHY much like Infinity Fabric does. I mean, ultimately it's really just a fast SERDES at that point and making them able to run different protocols is an obvious step, no it's not exactly surprising.
 
An attempt to put AMD on the defensive by making it an open standard after sinking in enough development to get ahead - never mind all the back room dealing to get equipment and code out that performs well with and poorly without.

Still, if it is real and not just a paper thing that never really works it is a nice idea.
 
Skull_Angel said:
Is it wrong that the first thing that comes to mind is "Look, more direct attack vectors!", just because this is coming from Intel?
Click to expand...
As it's likely data center focused with accelerators, security may be less of a concern. Disabling security mechanisms on a closed system in favor of performance isn't a horrible choice.

Elf_Boy said:
An attempt to put AMD on the defensive by making it an open standard after sinking in enough development to get ahead - never mind all the back room dealing to get equipment and code out that performs well with and poorly without.

Still, if it is real and not just a paper thing that never really works it is a nice idea.
Click to expand...
OpenCAPI should be AMDs open version. Infinity just allowed them to ship it faster. As mentioned above, it's just SERDES over PCIe from an electrical standpoint. The protocols may be roughly compatible with some tweaks to the controllers.
 
Anarchist4000 said:
As it's likely data center focused with accelerators, security may be less of a concern. Disabling security mechanisms on a closed system in favor of performance isn't a horrible choice.
Click to expand...

I understand that the intended use is for data centers, and I really have no issue with soft/firmware options for disabling any security in favor of performance, but what I'd be concerned about would be hardware vulnerabilities left open to achieve a little performance. In a closed system, you're right, it shouldn't be a problem, but I'd be worried about the knucklehead newbie designing a networked data center around it if that was the case.
 
Charlie (semiaccurate) basically says he think CXL is intel's version of https://www.ccixconsortium.com/, of which AMD and Mellanox are already a part of. For intel's sake I hope their's is substantially better in some regard, otherwise I don't really understand the point.
 
Still waiting for the day Intel announces a reasonably priced CPU with a decent amount of cores and reasonable power usage and heat output. Less in-silicon security flaws would also be greatly appreciated. Th day may be far off.

With that out of the way, I wonder who will implement CXL other than Intel. Any thoughts?
 
Just what seems to be an update on this subject at wccftech. An interesting read that I recommend.

"I couldn’t help but notice that in the CXL presentation Intel gave today lies a hint about the future of its Xe GPU ambitions. While it was not explicitly stated anywhere – it seems to have a pretty obvious implication to me; time to coin a term: Coherent Multi-GPU. CXL is Intel’s new interconnect layer that is designed to solve a lot of issues with the PCIe protocol and one of the major reasons why Multi-GPU never took off properly is due to the lack of – you guessed it – coherency. I think it is very likely that we will be seeing Xe GPUs running in “CXL Mode” in the future."

https://wccftech.com/intel-xe-coherent-multi-gpu-cxl/
 
  • Like
Reactions: Aix.
like this
lostin3d said:
Just what seems to be an update on this subject at wccftech. An interesting read that I recommend.

"I couldn’t help but notice that in the CXL presentation Intel gave today lies a hint about the future of its Xe GPU ambitions. While it was not explicitly stated anywhere – it seems to have a pretty obvious implication to me; time to coin a term: Coherent Multi-GPU. CXL is Intel’s new interconnect layer that is designed to solve a lot of issues with the PCIe protocol and one of the major reasons why Multi-GPU never took off properly is due to the lack of – you guessed it – coherency. I think it is very likely that we will be seeing Xe GPUs running in “CXL Mode” in the future."

https://wccftech.com/intel-xe-coherent-multi-gpu-cxl/
Click to expand...

Intel-CXL-Coherence-Bias-Intel-Xe-GPU.jpg


Essentially, Intel is trying to create a heteregenous computing infrastructure that allows CPUs, GPUs, accelerators, FPGAs and pretty much anything that can go on a PCIe port to be glued together as a cohesive whole
Click to expand...

Very interesting. Wonder if this 'Coherence Bias' architecture, once embraced and extended, will be the Meltdown/Sceptre of 2025 - Now with GPU acceleration!
 
You must log in or register to reply here.
Back
Top