Clock for Clock: Ryzen, Broadwell-E, Kaby Lake
Though we have a lot of performance evaluation to contend with, one thing we want to start with is a look at how that new Zen architecture compares to the primary Intel competitors in raw IPC. The Instructions Per Clock capability is one of two key metrics used to define the performance a processor (the other being frequency). IPC is an area where AMD had very much fallen behind Intel over the last 10 years and it has been one of the areas that AMD has been very outgoing about its messaging. It went as far as claiming IPC matching based on testing using Blender back in August of 2016, a claim that raised hopes and eyebrows.
So, before we dive into the meat of our testing, I setup the three competing platforms with all processors running all cores at 3.5 GHz. Our tested products are the new AMD Ryzen 7 1800X, the Intel Core i7-7700K and the Core i7-6900K. We should note that while the 1800X and 6900K are 8C/16T parts, the 7700K is only 4C/8T which will give it a dramatic deficit in the multi-threaded tests.
Audacity MP3 encoding is one of my new single threaded workloads and we see interesting results right off the bat. The Ryzen 7 1800X completes the task in 241 seconds, while the 7700K does it in 182 seconds, giving Intel’s mainstream flagship an advantage of 24%. The Core i7-6900K based on the Broadwell microarchitecture, a full step behind Kaby Lake, is 20% faster than Ryzen. In our Cinebench R15 testing the 7700K is about 10% faster while the 6900K is 8% faster.
Two things stand out immediately. First, the delta of the Kaby Lake and Ryzen gap is quite different between Audacity and Cinebench R15. I would err towards Cinebench being the more accurate representation of the global IPC difference, but Audacity is a true application and workload that many people use (even video encoding has audio rendering). Still, the 8% gap between the 6900K and the Ryzen 7 1800X at 3.5 GHz tells me that AMD’s claims of equal IPC appear to have been overstated.
The multi-threaded results are interesting. First, let’s focus on Cinebench R15. While the 7700K easily gets bested by the Ryzen 7 1800X due to a core count doubling, the 6900K result is more intriguing. Remember that the 6900K had an 8% advantage in the single threaded result; this score shows the Ryzen CPU with the 1.5% advantage. Clearly something has made up for the IPC delta between the cores and we are assuming we are seeing improved implementation of SMT and better inter-core communications with cache. (Also interestingly, the 4C 7700K performs at exactly 50% the level of the 8C 1800X, indicating that the same changes and effects are applying.)
The same holds true in Handbrake where the Ryzen processor is faster than the 6900K, even if by a small amount. The Hitman gaming result is less flattering, showing a significant drop from both Intel processors to what AMD can provide. Though 94 FPS is just fine for most users, the 7700K is 10% faster! And while we are showing 1080p results (not 4K like AMD has done) we are running at the Ultra presets, not “low” to overly emphasis the importance of the CPU.
So, did it ever occur to the
So, did it ever occur to the reviewer that the a bit slower performance in some software (games included) is actually due to poor optimizations?
The industry used the last decade or so to specifically optimize for Intel.
Ryzen is fairly new by comparison, but it demonstrated that it got up to 30% increase in performance through simple patches in games.
Audacity and many other software like it are not optimized for Ryzen architecture.
They are taking advantage of every possible trick in Intel’s hand, and yet barely anything or none of it actually benefits Ryzen performance-wise.
Plus, the Infinity fabric in Ryzen is sensitive to RAM speeds.
2400 MhZ speed on RAM is simply inadequate for Ryzen… 3000 MhZ would be better as that would raise it’s performance by about 10%.
Other than RAM speeds, software optimizations are required to take advantage of Ryzen’s capabilities.
Otherwise, you might as well be comparing apples and oranges right now.
It actually shows that Ryzen via ‘brute force’ is highly competitive for all Intel’s products… just imagine what might happen if we get developers to actively support for Ryzen – of course, this will probably require time as devs usually optimize for hardware they are paid to optimize for – and as we know, both Intel and Nvidia have deep pockets to sway devs to support their own hardware specifically and make AMD look bad (when in reality, its anything but).