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 ESTExecution, Load/Store, INT and FP Scheduling
The execution of micro-ops get filters into the Integer (INT) and Floating Point (FP) parts of the core, which each have different pipes and execution ports. First up is the Integer pipe which affords a 168-entry register file which forwards into four arithmetic logic units and two address generation units. This allows the core to schedule six micro-ops/cycle, and each execution port has its own 14-entry schedule queue.
The INT unit can work on two branches per cycle, but it should be noted that not all the ALUs are equal. Only two ALUs are capable of branches, one of the ALUs can perform IMUL operations (signed multiply), and only one can do CRC operations. There are other limitations as well, but broadly we are told that the ALUs are symmetric except for a few focused operations. Exactly what operations will be disclosed closer to the launch date.
The INT pipe will keep track of branching instructions with differential checkpoints, to cut down on storing redundant data between branches (saves queue entries and power), but can also perform Move Elimination. This is where a simple mov command between two registers occurs – instead of inflicting a high energy loop around the core to physically move the single instruction, the core adjusts the pointers to the registers instead and essentially applies a new mapping table, which is a lower power operation.
Both INT and FP units have direct access to the retire queue, which is 192-entry and can retire 8 instructions per cycle. In some previous x86 CPU designs, the retire unit was a limiting factor for extracting peak performance, and so having it retire quicker than dispatch should keep the queue relatively empty and not near the limit.
The Load/Store Units are accessible from both AGUs simultaneously, and will support 72 out-of-order loads. Overall, as mentioned before, the core can perform two 16B loads (2x128-bit) and one 16B store per cycle, with the latter relying on a 44-entry Store queue. The TLB buffer for the L2 cache for already decoded addresses is two level here, with the L1 TLB supporting 64-entry at all page sizes and the L2 TLB going for 1.5K-entry with no 1G pages. The TLB and data pipes are split in this design, which relies on tags to determine if the data is in the cache or to start the data prefetch earlier in the pipeline.
The data cache here also has direct access to the main L2 cache at 32 Bytes/cycle, with the 512 KB 8-way L2 cache being private to the core and inclusive. When data resides back in L1 it can be processed back to either the INT or the FP pipes as required.
Moving onto the floating point part of the core, and the first thing to notice is that there are two scheduling queues here. These are listed as ‘schedulable’ and ‘non-schedulable’ queues with lower power operation when certain micro-ops are in play, but also allows the backup queue to sort out parts of the dispatch in advance via the LDCVT. The register file is 160 entry, with direct FP to INT transfers as required, as well as supporting accelerated recovery on flushes (when data is written to a cache further back in the hierarchy to make room).
The FP Unit uses four pipes rather than three on Excavator, and we are told that the latency in Zen is reduced as well for operations (though more information on this will come at a later date). We have two MUL and two ADD in the FP unit, capable of joining to form two 128-bit FMACs, but not one 256-bit AVX. In order to do AVX, the unit will split the operations accordingly. On the counter side each core will have 2 AES units for cryptography as well as decode support for SSE, AVX1/2, SHA and legacy mmx/x87 compliant code.
Facebook
Twitter
RSS
574 Comments
View All Comments
BurntMyBacon - Friday, March 3, 2017 - link
@ShieTar: "Well, the point of low-resolution testing is, that at normal resolutions you will always be GPU-restricted."If this statement is accepted as true, then by deduction, for people playing at normal (or high) resolutions, gaming is not a differentiator and therefore unimportant to the CPU selection process. If gaming is your only criteria for CPU selection, then that means you can get the cheapest CPU possible until you are not GPU restricted.
@ShieTar: "The most interesting question will be how Ryzen performs on those few modern games which manage to be CPU-restricted even in relevant resolutions, e.g. Battlefield 1 Multiplayer."
I agree here fully. Show CPU heavy titles to tease out the difference between CPUs. Artificially low resolutions are academic at best. That said, according to Steam Surveys, just over half of their respondents are playing at resolutions less than 1080P. Over a third are playing at 1366x768 or less. Though, I suspect the overlap between people playing at these resolutions and people using high end processors is pretty small.
Average frame rate is fairly uninteresting in most games for high end CPUs, due to being GPU bound or using unrealistic settings. Some, more interesting, metrics are min frame rate, frame time distribution (or simply graph it), frame time consistency, and similar. These metrics do more to show how different CPUs will change the experience for the player in a configuration the player is more likely to use.
Lord-Bryan - Thursday, March 2, 2017 - link
Who buys a 500 dollar cpu to play games at 720p res. All that talk is just BS.JMB1897 - Friday, March 3, 2017 - link
That test is not done for real world testing reasons. At that low resolution, you're not GPU bound, you're CPU bound. That's why the test exists.Now advance a few years into the future when you still have your $500 Ryzen 7 CPU and a brand new GPU - you may suddenly become CPU bound even at QHD or 4k, whereas a 7700k might not quite be CPU bound just yet.
MAC001010 - Saturday, March 4, 2017 - link
Or a few years in the future (when you get your new GPU) you find that games have become more demanding but better multi-threaded, in which case your Ryzen 7 CPU works fine and the 7700k has become a bottleneck despite its high single-threaded performance.This illustrates the inherent difficulty of comparing high freq. CPUs to high core count CPUs in regards to future potential performance.
cmdrdredd - Saturday, March 4, 2017 - link
"Or a few years in the future (when you get your new GPU) you find that games have become more demanding but better multi-threaded, in which case your Ryzen 7 CPU works fine and the 7700k has become a bottleneck despite its high single-threaded performance."Maybe, the overclocking scenario is also important. Most gamers will overclock to get a bit of a boost. I have yet to replace my 4.5Ghz 3570k even though new CPUs offer more raw performance, the need hasn't been there yet.
One other interesting thing is how Microsoft's PlayReady 3.0 will be supported for 4k HDR video content protection. So far I know Kaby Lake supports it, but haven't heard about any of AMD's offerings unless I missed it somewhere.
Cooe - Sunday, February 28, 2021 - link
Lol, except here in reality the EXACT OPPOSITE thing happened. A 6-core/12-thread Ryzen 5 1600 still holds up GREAT in modern titles/game engines thanks to the massive advantage in extra CPU threads. A 4c/4t i5-7600K otoh? Nowadays it performs absolutely freaking TERRIBLY!!!basha - Thursday, March 2, 2017 - link
all the reviews i read are using NVidia 1080 gfx card. my understanding is AMD graphics has better implementation of DX12 with ability to use multiple cores. I would like to see benchmarks with something like RX480 crosfire with 1700x. this would be in the similar budget as i7 7700 + GTX 1080.Notmyusualid - Friday, March 3, 2017 - link
http://www.gamersnexus.net/hwreviews/2822-amd-ryze...cmdrdredd - Saturday, March 4, 2017 - link
Overclocking will be interesting. I don't use my PC for much besides gaming and lately it hasn't been a lot of that either due to lack of compelling titles. However, I would still be interested in seeing what it can offer here too for whenever I finally break down and decide I need to replace my 3570k @ 4.5Ghz.Midwayman - Thursday, March 2, 2017 - link
Here's hoping the 1600x hits the same gaming benches as the 1800x when OC'd. $500 for the 1800x is fine, Its just not the best value for gaming. Just like the i5's having been better value gaming systems in the past.