Gadgetory


All Cool Mind-blowing Gadgets You Love in One Place

AMD R7 1700 SMT Off + Overclock Benchmarks

2017-03-14
our first to rise in CPU reviews highlighted performance of the chips with SMT disabled the test that we came up with pre-release based on historical testing with Intel hyper-threading as stated in the original content and in some old Intel benchmarks of ours hyper-threading hasn't always been a net positive overhead sometimes detracted from average FPS by a couple percent at times and so we added SMT tog wins who our rise in testing but there's been non-stop demand to apply SMT zero and an overclock to the 1700 to show and in the best possible light we're following up today on the audience requests and we'll be walking through the results here before getting to that this is brought to you by thermal takes powered 900 enclosure built like a display case for PC components the tower 900 can fit full custom water loops and Orient's the system to put all components on display like a showroom learn more of the link in the description below this test is built around user requests so keep that in mind going into the content we are using the r7 1700 the CPU that we most pray is out of the current lineup of r7 CPUs as it does more or less outperform for the price the 1800 X especially when overclocked we're using that and we're showing it sort of in the best white by combining all of our auxilliary tests in these 1700 reviews that includes SMT off and overclocking so the two of them together now in a benchmark rather than just doing one or the other because there was a time limitation so we will be looking at that today we're overclocking we're turning off SMT and as a reminder from the 1700 review we are increasing the memory clock with the overclocked 1700 that means that this is not a perfect a be test a perfect a be test would be where we compared really just one or two things off and on really one thing technically but the thing here is that the 1700 our model was not capable of exceeding 2666 megahertz without overclocking the CPU itself or changing the reference clock or something like that so that means our testing looks at the performance stock with a lower memory speed 2666 versus the 29 33 megahertz of the overclocked 1700 because that's what we were able to achieve and as stated in the 7100 review the point was really to push it with the overclock and not do an a/b test and see what is possible under better conditions we are continuing that today especially because that's what was largely requested by folks reading and viewing the content so keep all that in mind we have three games tested with four states of the r7 1700 with another few games tested with three states of the r7 1700 the main games to look at our battles of one-offs r2 and total war Warhammer the last of which previously showed tremendous performance gains by disabling SMT on all rise and CPUs we've tested thus far that's all three that exists for full test methods check T 1700 review the article and you can also check the new article link in the description below between these two you'll find most of the information that you need if not all of it Patrick Lathan ran these tests for GN and wrote the article so be sure to check out his work as well in that link battlefield one had a minor increase from 135 FPS average overclocked and SMT enabled to 130 5.7 the average FPS with SMT disabled and the overclocked applied with the stock r7 1700 we're looking at 129 FPS average as the base and disabling SMT increased the average by about 2.6 percent and averages again we're not looking too close yet overclock can increase it by about 4.7 percent and doing both increased it by about 5.2 percent now I'll point out that percentages are not always perfect for measurements because they can make things sound a lot bigger or smaller than they are in reality at this point comparing with them is sort of silly because the fact is disabling an SMP on the overclocked 1700 can does less than 1 FPS average which is within range of errors 0.7 FPS average gain for this particular benchmark because one we're running it manually it is not an automated benchmark and to Battlefield 1 has some unique behavior with data streaming as we've explained in previous content pieces that said the improvement in frame x is noteworthy as we've stated several times in these rise in product reviews or studies the difference demonstrated between the overclock that's on t1 and OCSP 0 is not visible in so far as user experience but it's certainly measurable and repeatable in terms of benchmarking and logging disabling SMT continues to prove beneficial for frame times in this title and we'll see that to be a trend in a couple of the other games tested the really interesting results came from watchdog 2 and as usual total war Warhammer Bob talks to prove to be one of the most CPU intensive games in our graphic settings study for the game where we looked at optimizing the settings within the game for different CPUs the game also allowed the 1700 with SMT 0 and overclocking to attain a relatively competitive 91.3 FPS average placing it above and these own 1800 X and below Intel's aging at stock i7 4790k this performance is mostly notable when looking at these seventeen hundred's own results watchdogs to didn't initially perform any better with SMP disabled but when overclocked disabling the SMP performs a bit better and jointly we're looking at about a fourteen point two percent better performance than stock that's with OSI and SMT disabled ocbn at four gigahertz total war Warhammer total war hammer is one of the games that specifically cited by Andy as been in the process of optimization and we're eagerly awaiting any patches given these results we're seen an 11% increase over SMT zero with no OC thirteen point eight percent over eight plain overclock and a massive twenty nine point two percent increase in average FPS over the out of box stock seventeen hundred with 2666 makers memory in terms of raw numbers that's a jump from 120 to 155 FPS average went under the best condition but again keep in mind that there's also a memory speed change as requested by viewers and readers in the previous content though it's really not that responsible in this particular game for the gains of seen 150 4.7 FTS average places the 1700 just below the last gen I 560 600 K in our tests but if these fluctuations are any indication there's some room for easy optimisation by Creative Assembly we'll be keeping a close eye on total Warhammer in the near future and hopefully it'll serve as an indication of just how much software changes can fact and these hardware performance when it comes to risin and SMT optimization we're now entering territory where we don't have SMT zero plus stock tests but can still compare against the overclock plus SMT test time is limited here and we don't have it the huge team required to run infinite tests as some folks seem to think we do but the least impressive result on our bench was in ashes of the singularity with DirectX 12 which showed a miniscule within range of error decrease in performance over a plane overclocked ideally we'd see a significant performance decrease by disabling SMT but no effect is a good start you can look at our 1800 x sm t 1 and 0 tests for some further information on how a Rison behaves in this title note that this particular workload is intensive on the CPUs so we'll need to defer to titles which achieve higher framerate to better see the fluctuation Metro last light and GTA both showed minor performance increases of 1 to 2 fps in average nothing for a real user to consider disabling SMT / dumping SMT continues to improve our frame times as shown in previous tests so those changes aren't always actually visible to the user so again that's totally worth disabling SMT / you're jumping between BIOS probably for nothing for the most part charge support these are in the article below if you want more detail on Metro last light and GTA 5 specifically we only performed one synthetic or render test here as this was primarily an issue with gaming performance not much point in looking at render performance Cinebench isn't a real workload so to speak but it is an Andy favorite and somewhat a simulation of rendering and it takes the full advantage of Andy's new 8 cord 16 thread our 7 CPUs as such multi-threaded scores are the relevance numbers here not single threaded as we covered previously our 1700 at 4 gigahertz with paths empty enabled scored at 1764 CV marks this is significantly higher than Intel stock I 760 900 K which is three times the price by the way as that CPU is responsible for scoring 1470 point 5 marks mileage may vary as the 4 gigahertz overclock wasn't stable in longer render tests but three point nine was and the 6900 k2v there did score 1823 when it was similarly overclocked to about where its limit was without some C disabled however these 1700 scored 1250 2.5 that's an average number still impressive but as expected it's much lower than when allowed to render on all possible threads disabling SMT is limited in gains compared to a straight overclock but is a more guaranteed gain the 1800 X is blown out of the water in price to performance by the 1700 as stated in the original 1700 review and that is especially true when overclocking as for disabling SMT and overclocking those gains are limited across the board except for total war Warhammer which seems a bit of a standout title watchdogs sue has some interesting results but nothing worthy of longer discussion which total war certainly is disabling SMT and overclocking has limited gains in average FPS for the most part though does show some more gains in a few games in the frame times especially at the low end so this isn't the 16 to 30 percent increase that some people were speculating in the comments we're really not seen anything close to that for the most part except for total war Warhammer which is again somewhat of a standout title here overall yeah there's some performance improvement with overclocking and SMT disabled but for the most part you'll see the greatest gains by just doing one of them and then ideally leaving SMP on because just kind of refresh everyone the argument we've been making for the most part we stand alongside these 1,700 as a mixed workload CPU it is fantastic in terms of doing things like rendering blender premier stuff like that we've already gone through all that in that review the 1800 X at the price not so great for gaming in terms of value it turns out the 1700's a lot more arguable in that way and if you look at it from a mixed workload user standpoint where maybe you're doing some content creation streaming alongside gaming the 1700 is a pretty good choice the thing here though is that if you are getting to the point of disabling SMT and overclocking to get the best gaming performance it sort of eliminates the entire purpose of conclusion because again the conclusion is built on the idea that the cpu is a good mixed workload performer if you turn SMT off you lose all of that suddenly it's worse than comparable Intel chips because you've killed half of the threads and things like blender they don't care about whatever these games seem to care about when it comes to SMT on or off or performance all they care about is more threads is better so you see blender rendering one tile for each thread you have 16 tiles rendering obviously that's a lot faster than eight so the argument basically becomes well why you disable SMT and the answer is because when you're benchmarking you want to look and see where is the potential Headroom where is the potential choke point and what can we learn from that data it does not mean in this case that you should disable SMT on a 1,700 that you purchased and use it for gaming because if you're doing that again value is not great and you kill the entire production argument we're doing production take the hit to gaming performance and be happy with the production performance because it's a 1,700 again you overclock it it's a good performer that does the 1,800 X in terms of price to performance by a long shot it can equal the 1,800 axis performance if overclocked assuming all overclocks can achieve the 3.8 to 4.0 gigahertz range on the 70 standard we'll see how that works out as they continue to bend chips in the future but hopefully that answers the question so keep an eye out for more content the 1080 TI hybrid stuff is either online already depends the video goes live or will be shortly subscribe for that leave a comment below if you have further questions links in the description below patreon.com slash gamers Nexus to help us out directly thank you for watching I'll see you all next time you
We are a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for us to earn fees by linking to Amazon.com and affiliated sites.