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

  • theuglyman0war - Saturday, March 4, 2017 - link

    I'd like to see a lot more older i7 extreme editions covered all the way to westmere so I can sell clients on new builds with such a comparison.
  • mapesdhs - Sunday, March 5, 2017 - link

    Which older i7s interest you specifically?
  • theuglyman0war - Saturday, March 4, 2017 - link

    Checking what I paid last month for i7-7700k at Microcenter...
    Although I did get the motherboard combo price sale they "usually" offer...
    The supposed $60 off for $319 is the cheapest price I found with a quick survey of new egg, amazon etc... And only $20 less then what I paid! Hardly A slashed priced answer shot across the bow by Intel! Not by a long shot!
    I thought I was going to recommend the new cheap price to all my customer's new builds but I am pushing RYZEN and AM4 for a real combined price that makes a difference. ( the cheap price for enthusiast Am4 is enticing but the loss of PCI lanes is of concern for extreme cpu comparison anyway. Not so much compared to i7-7700k though which brings the comparison back to 16 lane parity! )
  • theuglyman0war - Saturday, March 4, 2017 - link

    Could anyone actually point me to the amazing slashed deals that "BEAT" what I couldn't get last month by a long shot?

    ( which was $349 BEFORE rebate. In other words it's not like there were not sales last month as well. And I see nothing now that really amounts to AMAZING compared to last month? )

    Pretty dam insulting from somewhere in the pipe? Not sure if it's Intel. Or it's resellers clinging on to greedy margins not reflecting the savings to save their own ass's and bottom line due to stock considerations? Which iz no excuse considering the writing was on the wall. Someone needs to do a lot better. A heck of a lot better. Particularly considering I was thinking I could jes laff off AMD with an Intel savings and now have egg on my face! :)
  • rpns - Saturday, March 4, 2017 - link

    The 'Test Bed Setup' section could do with some more details. E.g. what BIOS version? Windows 10 build version? Any notable driver versions?

    These details aren't useful just now, but also when looking back at the review a few months down the line.
  • jorkevyn - Saturday, March 4, 2017 - link

    why they don't get 4 channel for DDR4 memory? I think, if you get that you will may be the real I7 6950K Killer
  • sedra - Saturday, March 4, 2017 - link

    have a look at this:
    "Many software programmers consider Intel's compiler the best optimizing compiler on the market, and it is often the preferred compiler for the most critical applications. Likewise, Intel is supplying a lot of highly optimized function libraries for many different technical and scientific applications. In many cases, there are no good alternatives to Intel's function libraries.

    Unfortunately, software compiled with the Intel compiler or the Intel function libraries has inferior performance on AMD and VIA processors. The reason is that the compiler or library can make multiple versions of a piece of code, each optimized for a certain processor and instruction set, for example SSE2, SSE3, etc. The system includes a function that detects which type of CPU it is running on and chooses the optimal code path for that CPU. This is called a CPU dispatcher. However, the Intel CPU dispatcher does not only check which instruction set is supported by the CPU, it also checks the vendor ID string. If the vendor string says "GenuineIntel" then it uses the optimal code path. If the CPU is not from Intel then, in most cases, it will run the slowest possible version of the code, even if the CPU is fully compatible with a better version."

    http://www.agner.org/optimize/blog/read.php?i=49&a...
  • HomeworldFound - Saturday, March 4, 2017 - link

    Everyone here already knew that ten years ago.
  • Notmyusualid - Sunday, March 5, 2017 - link

    Indeed it was.
  • sedra - Sunday, March 5, 2017 - link

    it is worth to bring it up now.

Log in

Don't have an account? Sign up now