AMD Ryzen 7 1700X Review; Testing SMT
Now that the initial excitement surrounding AMDís Ryzen processors has settled down and the screams of joy turned down to a muted buzz, it isnít time to take a breather quite yet. While our original review covered the $500 Ryzen 7 1800X thereís still two more very similar processors to cover; one of which is the less expensive 1700X.
By this point about a week ago I was well into testing that 1800X and for my initial impressions, I felt it wasnít a great fit for gamers but it would be an awesome purchase for massively parallel processing tasks. Creative professionals, folks running virtual clients or those who needed home servers could all benefit in some way from what AMD was offering. But was it really necessary to buy into the ďfull montyĒ chip or would the 1700X provide a perfect gateway drug into the world that Ryzen was endeavoring to create? I wanted to find out.
Even though our launch-day review of the Ryzen series covered the ins and outs of the entire product stack, letís dive back in to get an idea about what the 1700X brings to the table. For those of you interested the initial review also took a deep dive into the Zen architecture and AMDís new AM4 platform so if this is your first introduction to Ryzen, Iíd recommend reading that article first.
For its part the Ryzen 7 1700X retains the 8-core, 16-thread architecture of the 1800X with the only real difference being its achievable frequencies. Both the Base and Precision Boost speeds come in at an even 200MHz lower while TDP remains a very reasonable 95W. As with all X-series Ryzen CPUís it also features an additional 100MHz of headroom through XFR. Essentially performance will likely be between 5-10% lower than the flagship X1800 but its $400 price is 20% less. That could be a pretty compelling story in its own right.
Another thing to take into account is that the Ryzen 7 1700X requires less input voltage and when you couple that with lower clock speeds, it will produce less heat than the X1800 and thus be easier to keep cool. As such it will achieve the temperature threshold needed for XFR much easier. With that in mind it isnít hard to understand why I think this may be the darling of AMDís current lineup.
Right before I published the Ryzen 7 1800X review, I received several emails from AMDís PR team discussing how to improve in-game performance of their current architecture. Reviewers werenít achieving framerate levels which aligned with expectations and our results reflected that. Now the word ďexpectationsĒ here is key since I think they were overblown to begin with. As I said about a week ago, be it Intel or AMD those large eight-core, 16-thread processors are poor companions for a desktop system that focuses solely on gaming and light workloads. They canít achieve the clock speeds necessary to be competitive on a price / performance level in lightly threaded scenarios.
One suggestion from AMD involved turning off SMT within the BIOS in an effort to improve performance in certain instances. My gut reaction to that email isnít fit for these pages but let me explain nonetheless. In Ryzenís case simultaneous multithreading essentially allows a processor design with eight physical cores to produce sixteen threads. This is hugely beneficial in programs that can scale across that many concurrent threads, but it can also lead to issues where the necessary optimizations arenít in place.
Not only would toggling SMT be a completely unrealistic expectation for end users, but itís been long known that simultaneous multithreading isnít the end-all argument for CPU performance. Physical cores and the threads they produce often trump ďvirtualĒ ones. This is why AMDís Piledriver cores were able to achieve impressive (for the time) results against Intel within certain situations; some games and applications simply donít play nice with SMT-based designs, be it Intelís Hyper-Threading or Ryzenís paralleling technology.
At the time, I didnít have time to humor AMDís recommendation since it came with less than 48 hours before the NDA lifted. However, I decided to give the whole ďletís turn off SMTĒ directive a shot in this particular review... but with a bit of a twist to level the playing field. Not only did I turn off SMT for the Ryzen 7 1700X, but I also disabled Hyper-Threading on the i7-6900K to insure that this remains a true apples to apples comparison. Hopefully, this will also highlight application support rather than just claiming that any inherent ďissuesĒ are AMD-specific.
Now before heading directly into the benchmarks, one thing I noticed is that the behavior of Intelís Broadwell-E is slightly more nuanced than Ryzen when thereís no multithreading involved. With all eight cores evenly loaded the Ryzen 7 1700Xís frequency remained at 3.5GHz, which is 100MHz higher than its stated Base Clock. The i7-6900K on the other hand took full advantage of Intelís Turbo Boost Max technology and leveled out at 3.7GHz, which is a good 400MHz higher than its Base Clock.
With that being said, Iím not taking overall non-SMT performance into account within this particular article. Rather, I want to see how the Ryzen 7 1700X natively performs and if any of its SMT benefits / pitfalls are mirrored on Intelís side of the fence. It should be an interesting oneÖ..
|Latest Reviews in Processors|