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

  • rascalion - Monday, June 19, 2017 - link

    Are the Ryzen numbers in the charts retests using the last round of bios and software updates?
  • Ian Cutress - Monday, June 19, 2017 - link

    As much as possible, the latest BIOSes are used.
    Our CPU testing suite is locked in for software versions as of March 2017. This is because testing 30/50/100+ CPUs can't be done overnight, we have to have rigid points where versions are locked in. My cycle is usually 12-18 months. (Note I'm only one person doing all this data.)
  • FreckledTrout - Monday, June 19, 2017 - link

    Ian any chance once there are a few BIOS tweaks you could say do a mini updated review on the 7820x vs Ryzen 1800x. With Ryzen having latest BIOS as well plus 3200Mhz memory. I'm just curious really how the 8-core guys line up when some of the dust settles and I think a lot of people will be.
  • Ian Cutress - Monday, June 19, 2017 - link

    Any reason why 3200? I'll have Intel people saying it is pushing the Ryzen out of spec
  • jjj - Monday, June 19, 2017 - link

    You could do a memory subsystem scaling review for all platforms, Skylake X, Threadripper, Ryzen (Summit Ridge) and Coffee Lake. Cache, interconnect, DRAM. See where they are, how they scale, where the bottlenecks are, single rank vs dual rank modules and perf impact in practice.Why not even impact on power and efficiency.

    In any case, you'll need to update Ryzen 5 and 7 results when Ryzen 3 arrives , isn't it?

    For DRAM at 3200 it might be out of spec - overclocking the core is out of spec too but that has never stopped anyone from overclocking the memory. Right now 3200 is what a lot of folks buy , at least for higher end mainstream Ofc some will argue that Ryzen scales better with memory and that's why it is unfair but it's a hell of a lot more reasonable than testing 1080p gaming with a 1080 TI since it's a popular real world scenario.

    At the end of the day the goal should be to inform, not to watch out for Intel's or AMD''s feelings.
  • vanilla_gorilla - Monday, June 19, 2017 - link

    >For DRAM at 3200 it might be out of spec - overclocking the core is out of spec too but that has never stopped anyone from overclocking the memory.

    This. Exactly. We're enthusiasts and we always push the envelope. No one cares what the specs are all we care is about what these processors are capable of in the right hands.

    And Ian I think you guys do an awesome job, there's no other place I look for CPU benchmarks. Keep up what you do, we all appreciate it, as well as your willingness to have a dialog with us about the process. Really cannot say how impressed I am by how open and engaged you are, it's really commendable.
  • Ian Cutress - Tuesday, June 20, 2017 - link

    Thanks for the comments :)

    Though on your comments about pushing things out of spec. We have a good deal of readers who want plain stock for their businesses - AT isn't solely a consumer focused site. Otherwise I'd just jack all the CPUs and just post OC results :D Our base testing will always be at stock, and for comparison testing there has to be an element of consistency - testing an OC'ed part against a stock part in a direct A vs B comparison is only going to end up with a barrage of emails being rammed down my throat. There has to be some planning involved.
  • Ian Cutress - Tuesday, June 20, 2017 - link

    I've been planning a memory scaling article, I just haven't had the time (this article was around 6 weeks of prep with all the events going on that I had to attend).

    Note we don't retest stuff every review. With our new 2017 test suite, I've been going through regression testing. Usually regression testing is done once for the full segment until the benchmarks are changed again. I'll look at my next few months (still stupidly busy) and look at the priorities here.
  • FreckledTrout - Monday, June 19, 2017 - link

    Most people can easily buy a 3200 kit for not a lot of extra money. It doesn't take a lot tweaking(well not anymore on AGESA 1.0.0.6) or silicone lottery like an OC, just a bit more cash. From what I have seen with Ryzen it is the sweet spot on price and performance. I would assume Its likely the most chosen configuration on the R7's. To make it fair use 3200 on the 7820x as well. I only ask because Ryzen did way better than I would have thought and would like to see it with 3200Mhz memory and latest updates to see really how close Intel and AMD are on 8-core systems. Then im going to build :)
  • tipoo - Monday, June 19, 2017 - link

    Launch review! Nice work dude(s).

Log in

Don't have an account? Sign up now