Zen: New Core Features

Since August, AMD has been slowly releasing microarchitecture details about Zen. Initially it started with a formal disclosure during Intel’s annual developer event, the followed a paper at HotChips, some more details at the ‘New Horizon’ event in December, and recently a talk at ISSCC. The Zen Tech Day just before launch gave a chance to get some of those questions answered.

First up, let’s dive right in to the high-level block diagram:

In this diagram, the core is split into the ‘front-end’ in blue and the rest of the core is the ‘back-end’. The front-end is where instructions come into the core, branch predictors are activated and instructions are decoded into micro-ops (micro-operations) before being placed into a micro-op queue. In red is the part of the back-end that deals with integer (INT) based instructions, such as integer math, loops, loads and stores. In orange is the floating-point (FP) part of the back-end, typically focused on different forms of math compute. Both the INT and FP segments have their own separate execution port schedulers

If it looks somewhat similar to other high-performance CPU cores, you’d be correct: there seems to be a high-level way of ‘doing things’ when it comes to x86, with three levels of cache, multi-level TLBs, instruction coalescing, a set of decoders that dispatch a combined 4-5+ micro-ops per cycle, a very large micro-op queue (150+), shared retire resources, AVX support, and simultaneous hyper-threading.

What’s New to AMD

First up, and the most important, was the inclusion of the micro-op cache. This allows for instructions that were recently used to be called up to the micro-op queue rather than being decoded again, and saves a trip through the core and caches. Typically micro-op caches are still relatively small: Intel’s version can support 1536 uOps with 8-way associativity. We learned (after much asking) at AMD’s Tech Day that the micro-op cache for Zen can support ‘2K’ (aka 2048) micro-ops with up to 8-ops per cache line. This is good for AMD, although I conversed with Mike Clark on this: if AMD had said ‘512’, on one hand I’d be asking why it is so small, and on the other wondering if they would have done something different to account for the performance adjustments. But ‘2K’ fits in with what we would expect.

Secondly is the cache structure. We were given details for the L1, L2 and L3 cache sizes, along with associativity, to compare it to former microarchitectures as well as Intel’s offering.

In this case, AMD has given Zen a 64KB L1 Instruction cache per core with 4-way associativity, with a lop-sided 32KB L1 Data cache per core with 8-way associativity. The size and accessibility determines how frequently a cache line is missed, and it is typically a trade-off for die area and power (larger caches require more die area, more associativity usually costs power). The instruction cache, per cycle, can afford a 32byte fetch while the data cache allows for 2x 16-byte loads and one 16-byte store per cycle. AMD stated that allowing two D-cache loads per cycle is more representative of the most workloads that end up with more loads than stores.

The L2 is a large 512 KB, 8-way cache per core. This is double the size of Intel’s 256 KB 4-way cache in Skylake or 256 KB 8-way cache in Broadwell. Typically doubling the cache size affords a 1.414 (square root of 2) better chance of a cache hit, reducing the need to go further out to find data, but comes at the expense of die area. This will have a big impact on a lot of performance metrics, and AMD is promoting faster cache-to-cache transfers than previous generations. Both the L1 and L2 caches are write-back caches, improving over the L1 write-through cache in Bulldozer.

The L3 cache is an 8MB 16-way cache, although at the time last week it was not specified over how many cores this was. From the data release today, we can confirm rumors that this 8 MB cache is split over a four-core module, affording 2 MB of L3 cache per core or 16 MB of L3 cache for the whole 8-core Zen CPU. These two 8 MB caches are separate, so act as a last-level cache per 4-core module with the appropriate hooks into the other L3 to determine if data is needed. As part of the talk today we also learned that the L3 is a pure victim cache for L1/L2 victims, rather than a cache for prefetch/demand data, which tempers the expectations a little but the large L2 will make up for this. We’ll discuss it as part of today’s announcement.

AMD is also playing with SMT, or simultaneous multi-threading. We’ve covered this with Intel extensively, under the heading ‘HyperThreading’. At a high level both these terms are essentially saying the same thing, although their implementations may differ. Adding SMT to a core design has the potential to increase throughput by allowing a second thread (or third, or fourth, or like IBM up to eight) on the same core to have the same access to execution ports, queues and caches. However SMT requires hardware level support – not all structures can be dynamically shared between threads and can either be algorithmically partitioned (prefetch), statically partitioned (micro-op queue) or used in alternate cycles (retire queue).

We also have dual schedulers, one for INT and another for FP, which is different to Intel’s joint scheduler/buffer implementation. 

CPUs, Speeds, Pricing: AMD Ryzen 7 Launch Details The Ryzen Die
Comments Locked

574 Comments

View All Comments

  • mikeZZZ - Friday, March 3, 2017 - link

    Anadtech, can we please run closer to real life scenarios such as a gaming benchmark with a file compression benchmark running at the same time. Even gaming enthusiasts run more than one program at a time. For example, file decompression in the background while playing a game, or baseball game streaming in a small window while playing a game. You already have many individual benchmarks, so why not go the extra but significant benchmark of running two? We know this favors the higher core CPUs (maybe even Ryzen 7 1700 over all other lower core ones CPUs) but that is closer to real life and should be very meaningful to someone wanting to make an informed purchase.
  • ValiumMm - Saturday, March 4, 2017 - link

    Would also like to see this
  • UrQuan3 - Friday, March 3, 2017 - link

    Just want to put out a quick comment about benchmarking with Handbrake. In dealing with Broadwell-E, and especially ThunderX, I've found that Handbrake often doesn't scale well past about 10 cores, and really doesn't scale well past 16 or so. What seems to happen is that the single-threaded parts of Handbrake tend to dominate the encode time. In extreme cases, ultra-fast and placebo will take almost the same amount of time as x264 is consuming input faster than the rest of Handbrake can generate it. On ThunderX, I've found I can complete four 1080p placebo encodes in the same amount of time that I can complete one. I would expect a similar result on a 48 core Intel, though I do not have access to one beyond 24 cores. Turbo boost would hide this effect a bit.

    I am not knocking using Handbrake for benchmarking. The Handbrake and ray-trace results are the two that I care about most. I just thought I'd give a heads up about this limitation. You can check CPU usage statistics to get an indication of when you are running up against this limit.

    Oh, and I am very excited to see multiple ray-tracers in your runs. Please continue.
  • Meteor2 - Saturday, March 4, 2017 - link

    Presumably though you can have several x264 jobs running simultaneously on that hardware? So while your time to encode a certain piece doesn't decrease, you have more total-throughput (e.g. encoding several different bitrates for adaptive streaming). Should give good efficiency too on a larger Broadwell-E or a ThunderX.
  • UrQuan3 - Tuesday, March 7, 2017 - link

    Exactly. It's the first time I've thought about installing a queue manager for a single computer.
  • jade5419 - Saturday, March 4, 2017 - link

    I agree with this. In my experience Handbrake has a core / thread limit.

    I have a Z600 system with dual Xeon 5570 @ 2.93GHz, 6 core / 12 threads (total 24 threads), 48GB of RAM and a Z620 system with dual Xeon E5-2690 @ 2.9GHz 8 core / 16 threads (total 32 threads), 64GB RAM.

    The two systems transcode video at the same speed using Handbrake 1.0.3. Monitoring CPU usage shows all threads of the Z600 at 100% utilization whereas the CPU utilization on the Z620 is approximately 80%.
  • Notmyusualid - Sunday, March 5, 2017 - link

    Ever tried running GTA5 on 28 cores?

    It doesn't work. You have to adjust the game 'launchers' core affinity to < 26 cores or it won't even load.

    Given this discovery, I expect there are many more applications out there, that may crap-out as we see more and more cores come into the mainstream.

    Just a thought.
  • mapesdhs - Sunday, March 5, 2017 - link

    I'd love to know why this happens. I'm guessing something dumb within Windows.
  • Outlander_04 - Friday, March 3, 2017 - link

    There is more than enough good news to make me want to buy a 6 core Ryzen when they become available .
    Likely that will be the sweet spot for gamers
  • 0ldman79 - Saturday, March 4, 2017 - link

    I'm looking forward to seeing Ryzen updated in the bench.

    There aren't any apps or benchmarks that cross over between the FX series and the Ryzen series, so we can't do any side by side comparison.

    Great review guys. Looking forward to the six core Ryzen. I think just like the FX series the six core will be the sweet spot.

Log in

Don't have an account? Sign up now