The AMD Zen and Ryzen 7 Review: A Deep Dive on 1800X, 1700X and 1700
by Ian Cutress on March 2, 2017 9:00 AM ESTZen: 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.
Facebook
Twitter
RSS
574 Comments
View All Comments
Meteor2 - Friday, March 3, 2017 - link
...In which case you'd be better off with a 7700K, looking at the benchmark results. Cheaper too.ddriver - Thursday, March 2, 2017 - link
Ryzen offers the same performance at half the cost. More pci-e lanes is good for io, however quad channel memory is pretty much pointless, aside of pointless synthetic benches. Ryzen might not make it to my personal workstation due to the low pci-e lane count, but it has enough to replace my aging 3770k farm nodes, to which it will be a significant upgrade, provided the chip and platform turn out to be stable and bug free,Intel has gotten lazy and sloppy, bricking products, chipset bugs, they haven't really done anything new architecture wise for years, milking the same old cow.
It is rather silly to assume that gaming dictates CPU prices, this IS NOT a gaming product, if your ass-logic is to be followed, the intel needs to drop the 7700k price to 168, because in games it is barely any faster than the i3-7350K, and has the same pathetic, even lower than ryzen, number of pci-e lanes.
This is a chip for HPC, which gaming is NOT. Go back to the kiddie garden, eight core chips are for grown ups ;)
imaheadcase - Thursday, March 2, 2017 - link
People compare it to gaming, because its the main driving for these type of CPUs, its not even gaming specfic but VR, Graphics modeling, etc. You honestly think people are buying these for offices or industry for complex math problems? lolddriver - Thursday, March 2, 2017 - link
It is not "people" but "fanboys", and they cling to gaming because it is the only workload where intel can offer better performance for the price, albeit by comparing products from different tiers, which is quite frankly moronic.Cars are faster than trucks, so who in the world needs to spend money on trucks? That's the kind of retarded logic you are advocating...
Smart people buy whatever suits their needs. Obviously, if all you do is play games you wouldn't be buying ryzen or a lga2011 system. Just get an unlocked i5 and overclock it, best bang for the buck. You must realize that even if you don't, other people use computers for tasks other than gaming. And for a large portion of them ryzen will be the best deal, because it is versatile - it is good enough for gaming too, while still offering significant performance advantage compare to an intel quad in tasks that are time staking, and are very much competitive with intel's 8 and 10 core chips while delivering more than twice the value, which is important for everyone who doesn't have money to throw away.
Claiming that "gaming is the main driving for these type of CPUs" is foolish to say the least, because games don't benefit from that particular type of CPUs. Most of the games can't even property utilize 4 threads. And this is not likely to change soon, because the overhead of complexity and thread synchronization is not worth it for non-performance demanding tasks such as games.
Lord-Bryan - Thursday, March 2, 2017 - link
That's one really well thought out argumentrarson - Thursday, March 2, 2017 - link
Ryzen's versatility and price are the two biggest factors that make it so good. It might not beat the very best gaming CPU that Intel has, or the very best multi-threaded monster that Intel has in every scenario, but it's competitive with both at half the price of the high-end stuff. Hence, while I do game some and want to build a computer to use for gaming, I also do other stuff like audio production that benefits greatly from Ryzen's multi-threaded performance. To me, it's a no-brainer: Ryzen right now is the best bang-for-the-buck chip for someone who wants all-around high-end performance, by far. Maybe not the 1800X, I kind of think the 1700X is a better value, but still, for most people who want multiple-use performance instead of absolute maximum gaming performance, Ryzen is the clear choice.Ryzen's max clock speeds seem, like Intel's, to be hindered by the total number of cores on chip, so it should be extremely interesting to see how the 4- and 6-core chips overclock once they arrive, and what kind of performance they'll achieve. I actually think that, like Intel, a 4-core Ryzen might be a better gaming chip than the 8-core, and if that's the case, then it might be really darn close to Intel's best Kaby Lake, because like you pointed out, most games aren't threaded well at all.
Additionally, from a gaming perspective, it seems like AMD has done more to push technology forward in that respect than anyone else. They've worked on Mantle, Vulkan, FreeSync, TrueAudio, and others. They've always tried to give performance value by offering more cores, but software has been slow to take advantage of them. Intel is content to stagnate by offering extremely incremental increases because performance is "good enough" so developers have no reason to really try to take advantage of extra cores aside from outside use cases. With Ryzen, AMD is pushing chips towards higher core counts (much like they did with the Athlon X2) but this time, they're trying harder to get developers on board and help them achieve good results. So while it always takes forever for software to better utilize the hardware, once the hardware becomes more common the software will start to follow and you'll see the actual gaming performance improve. Is that a valid reason to buy Ryzen today if your sole focus is gaming? Of course not, but it does bode well for Ryzen owners in the future. The performance can only get better. Can the same be said about Intel? Well, probably not if you're using one of the 4-core chips. It's pretty much a known quantity.
I had high hopes for Bulldozer and Ryzen is the exact opposite of what Bulldozer was. I feel like the CPU market has been stagnant for years and now suddenly there's a reason to be excited. This makes AMD competitive again, which will be good for pricing even if you're an Intel fan. It's been a long wait, but it was worth it, this is a good product.
Notmyusualid - Friday, March 3, 2017 - link
http://www.gamersnexus.net/hwreviews/2822-amd-ryze...Makaveli - Thursday, March 2, 2017 - link
+1 ddriver you destroyed that kid with your logic well done.khanikun - Friday, March 3, 2017 - link
Gaming definitely isn't the main driving force for CPUs, as use case changes. I bought a 7700K for my gaming rig. I'd get a 1700 for a VM host, as I'd like to start building a lab again. It won't be this round though. I'd rather wait for AMD to iron our any kinks and buy the next generation. It's something the 7700K could do, but more cores would definitely make it a much better lab.Meteor2 - Friday, March 3, 2017 - link
Nobody buys mid/high-end consumer chips for HPC. They buy them for gaming. A few for video production. That's it.