The New Core-to-Core Communication Paradigm

Within the last week, Intel released a surprising announcement regarding the capability of its upcoming Skylake-SP processors ahead of any official announcement and release, and the relevance of it extends through to the Skylake-X chips as well. With a multi-core design, how cores communicate can be vital, when sharing data between them or by accessing each other’s cache via snooping or having to go out to main memory.

In previous generations, Intel has championed its ring bus strategy. Pretty much every commercial multi-core chip Intel has released uses some form of ring bus, and it boils down to a single communication ring on which all the cores are connected. If data needs to a certain core, it travels on the ring until it reaches the core it needs. The idea is relatively simple, and has worked quite well. It does mean that there is a range of core-to-core latencies however, depending on how far away the core is.

Imagine connecting to a network of computers, but you could only connect to the two computers either side of you. If you needed to speak to a computer that was 5 nodes away, the data would have to travel through each of the nodes to get there. Welcome to the ring bus. The speed of this ring bus was typically denoted by the ‘uncore’ frequency.

With Intel’s three-die program on its enterprise chips, each one used different variants of the ring bus topology. On Broadwell-E, the 10-core LCC design used a single ring bus for all 8 cores and the DRAM controller/PCIe root complex, whereas the HCC and XCC designs both had dual ring buses that used router communication methods to talk between each other. Now obviously talking to a core on the other ring bus added latency, and was sometimes considered a ‘non-uniform memory access’ (NUMA) environment on chip, with extra latency if you needed data from memory on the other ring bus. (Typically NUMA is applied to dual socket platforms, where the DRAM on the other chip has a large delay.) Intel emphasized that the Broadwell-E design was memory location aware, and tried to keep relevant data local to the core it was needed on.

With the larger XCC CPUs for Broadwell-E, the ring buses become quite cumbersome: you end up with one ring bus servicing 12 cores, and another ring bus servicing 12 cores (due to the layout with other elements of the chip).

As more and more cores get added, what is the solution? More ring buses? How about a 2D array of ring buses? Or even a mesh or a network?

The announcement that Intel made in the past week was that Skylake-SP processors will use a mesh topology, whereby each CPU in the design will be connected to the ones above, below, and to the side. Earlier in the week I mentioned on Twitter that a mesh in this context, or as drawn above, is just an array of ring buses in a 2D arrangement, although new information has come to light.

There are several things to talk about the mesh here, so in no particular order:

Mesh Me Up

Intel already sells chips with mesh connectivity: the latest Xeon Phi Knights Landing many-core processors. These massive (~646mm2) chips have 72 cores arranged as 36 pairs, six memory channels and 16GB of MCDRAM, and using a ring bus (or a succession of ring buses) is probably not feasible.

For KNL, Intel calls the mesh the ‘untile’. Taken from the KNL optimization guide:

‘At each mesh stop, there is a connection to the tile and a tag directory that identifies which L2 cache (if any) holds a particular cache line. [In Knights Landing] there is no shared L3 cache within a physical package, [so] memory accesses that miss in the tile must go over the mesh to the tag directory to identify any cached copies in another tile. Cache coherence uses the MESIF protocol. If the cache line is not cached in another tile, then a request goes to memory.’

Knights Landing can run in a variety of mesh modes, due to the lack of an L3 but the presence of an intermediary MCDRAM layer, certain modes to do with the memory hierarchy help with different codes. It is not known if Skylake-SP will have the same feature set.

When distributing cores along a mesh, another way to think about it is as a big routing environment. There are two ways – each core could talk to a central processing router which then sends information as it is needed, or each core acts as a networking point itself, forwarding packets in the direction it is needed. The first method, commonly known as a crossbar method, is easier to implement but can be bottlenecked very easily, not to mention power hungry if the whole crossbar needs to be enabled all the time. The second method, where every core is a router node, looks like what we have already in the image Intel provided.

The announcement Intel made was done officially by Dr Akhilesh Kumar, one of Skylake-SP’s CPU architects. He has been at Intel for a number of years, and digging back through releases with his name on popped up a very interesting paper regarding how to implement a router-like mesh in a multi-core environment, published in March 2014 which is about the right time when Skylake-SPs mesh was being developed. We haven’t had confirmation from Intel that this is how Skylake-SP’s mesh is actually like, though it is worth covering what is in the paper.

For anyone interested, the paper is called MoDe-X: Microarchitecture of a Layout-Aware Modular Decoupled Crossbar for On-Chip Interconnects, IEEE Transactions on Computers, Vol. 63, No. 3, March 2014 P622.

MoDe-X: Is This Skylake-SP’s Mesh?

As mentioned above, the easiest way for a mesh/router is through a crossbar. When the design instead uses each core as a node in the mesh, the result is essentially a ‘mini’ decoupled crossbar in each core acting as the director for data coming in and out of the core. With each core in control of its own actions, the design becomes modular. Put these elements together, and you have a MOdular DEcoupled Crossbar, or MoDe-X.

Each core has an integrated network interface and a router, with each router connected to the four routers around it (or 2 routers for corners and 3 routers for edges). A data packet can come in one of five directions – north, south, east, west, or from the core the router is attached to. The data packet then needs to be sent to one of five directions again: north, south, east, west, or to the core the router is attached to. If each core is essentially a node on a 2D map with an x-ordinate and a y-ordinate, if a packet needs to go to (0,0), the router can send it on its way as it knows which nodes are nearest to it. There are algorithms designed to reduce congestion, such as only moving in the x-direction until you are correct then start moving in the y-direction, but a basic system will have buffers and queues and will know how busy the local network congestion is.

The default way to design the router is to take the date from every input, decided where to send it, then send it. The MoDe-X design does it a little differently, and essentially splits the directions up into three groups: one group is north/south, one group is east/west, and the local core is the third group. The following diagram is a bit complex, but we’ll go through it:

On the left are the five inputs: E, W, L (local), N, S.
On the right are the five outputs, W, E, L, S, N

The first column of ‘shared’ boxes determines where the data wants to go: does it need to traverse a row, a column, or stay local. From each input, there are only two arrows going to either the top pair or the bottom pair of ‘shared’ boxes. Depending on contention and allocation, the data is then sent to the second column whereby if the data needed to change columns, it could either go to Sout or Nout and if it needed to change rows into Wout or Eout. In both circumstances the data could go to the local node.

In the research paper, there is a six-input design which modifies the local input in order to make the wiring in the second half of the design easier (and it lowers power), but the image above is a fair representation of the idea. There are other additions in the paper as well, with layout aware arrangements (how exactly you put it together in silicon) as well as feeder wires such that if data needs to continue in the same direction it can do so with less effort.

The paper is very dense (it took me a while to get to grips with it), but worth a read if you are interested. I found it easier to consider it more like a router, than an on-chip mesh. As mentioned, this is a 2014 paper, and shows simulation data on an 8x8 grid of cores, where this design is both lower power and smaller area than a generic mesh design or competing designs. There were some areas where the generic design had marginally better latency, but the power conservation numbers (due to the wiring reduction and an aggressive power gating scheme to disable certain parts for up to 99% less power use) makes it a more attractive option (especially in multi-core CPUs). Intel was doing simulation work in early 2014 (which means 2013), and the paper mentions that they were in the process of putting these ideas into silicon. It wouldn’t be too far-fetched to consider that this is what we are seeing today.

I Keep My Cache Private: Moving to a 1MB L2 and a Non-Inclusive L3 Analyzing The Silicon: Die Size Estimates and Arrangements
Comments Locked

264 Comments

View All Comments

  • halcyon - Monday, June 19, 2017 - link

    Thank you for the review. The AVX512 situation was a bit unclear. Which models have which AVX512 support of the now released chips (and the future, yet to be released chips). It would be interesting to see AVX512 specific article once the chips are out and we have (hopefully) some useful AVX512-optimized software (like encoding).

    For myself, decision is easy now to postpone, until ThreadRipper is out. The thermals are just out of whack and for my workloads, the price is not justified. Here's hoping Threadripper can deliver more for same price or less.
  • Bulat Ziganshin - Monday, June 19, 2017 - link

    all skl-x has avx-512 support. i9 cpus has double fma512 engines, but as i guess - that's only difference, i.e. remaining 512-bit commands (such as integer operations) will have the same throughput on i7 and i9
    but i may be wrong and it will be really very ibteresting to check throughput of all other operatins. probably Anand can't do that and we will need to wait until Agner Fog will reach these spus
  • halcyon - Monday, June 19, 2017 - link

    Thanks. Many questions remain: AVX512F vs AVX512BW, which are supported now, which in the future? What is the difference? How does it compare to Knights Landing? Are number of AVX units tied to number of cores? What is the speed differential in AVX512 loads? What is the oc limitation from AVX512 support? etc.
  • nevcairiel - Monday, June 19, 2017 - link

    Those Knights Landing/Knights Mill specific AVX512 instructions are actually very specific to the work-loads you would see on such a specialized CPU. The instructions chosen for Skylake-X and future Cannon Lake more closely match what we already know from AVX/AVX2.

    For the "types" of AVX512, basically its split into several sub-instruction sets, all containing different instructions. AVX512 will always include F, because thats the basis for everything (instruction encoding, 512-bit registers, etc). BW/DQ include the basic instructions we know from AVX/AVX2, just for 512-bit registers. SKL-X supports all of F, CD, BW, DQ, VL.

    Wikipedia on AVX-512 has some more info on the different feature sets of AVX-512:
    https://en.wikipedia.org/wiki/AVX-512

    Its generally safe to just ignore the Knights Landing specific instructions. They are very specific to the workloads on those systems. The AVX-512 subset used for Xeon "Purley" and SKL-X is more inline with the AVX/AVX2 instructions we had before - just bigger.

    For software, x264 for example already got some AVX512 optimizations over the recent weeks, it might be interesting to test how much that helps once all the launch dust settles.
  • halcyon - Tuesday, June 20, 2017 - link

    Thank you very much!
  • satai - Monday, June 19, 2017 - link

    Any idea, why is Intel so much better at Chrome compilation?
  • IanHagen - Monday, June 19, 2017 - link

    I'd like to know that as well. Ryzen does particularly fine compiling the Linux Kernel, for example, as seen in: http://www.phoronix.com/scan.php?page=article&...
  • tamalero - Monday, June 19, 2017 - link

    Optimizations?
    I still remember when Intel actively paid some software developers to block multicore and threading on AMD chips to boast about "more performance" during the athlon X2 days.
  • johnp_ - Tuesday, June 20, 2017 - link

    It's just the enterprise parts. Kaby Lake S is behind Ryzen:
    http://www.anandtech.com/show/11244/the-amd-ryzen-...
  • Gothmoth - Monday, June 19, 2017 - link

    if powerdraw and heat would not be so crazy i would buy the 8 core today.

    but more than the price these heat issues are a concern for me.. i like my systems to be quiet....

Log in

Don't have an account? Sign up now