Gadgetory

intel's new i 790 700 k is available for about four hundred to four hundred $30 which lands it between the 9900 k price at around 550 on a good day and the 8700 k is $370 price point we got ours on a bit of a high side of the curve but you can get them as low as 400 on new egg for example these days and we were looking to test the new eight core eight thread cpu versus the not that old 8700 K and the hyper threaded 9900 K of a similar spec until made a big move away from four core eight thread CPUs and the incumbent pricing structure with the 9700 K acting as the first case qi7 in a while anyway to lack hyper threading before that this video is brought to you by us and the Jian store one of the best ways to support us and get something in return is to grab an item from the Jian store like one of our blue beer glasses with a gold halo featuring the popular tear down logo or one of our critically acclaimed mod mats with wiring diagrams and a GPU tear down grid we also have a brand new black and blue mousepad that many of you requested now available on the store we keep selling out of things thanks to you all thought we've been working hard to get it back in stock go to store cameras next to sunette to grab something today the 9700 K is a weird one because if you look at it first instinct is kind of wait a minute they just turn off hyper-threading so what's the deal because that's what happened it's it's okay to have product segmentation we've talked about that in the past these types of architectures can cost billions of dollars to make if you look at something like and videos newest architectures for example and in order to get any return on that ever you do have to segment it so we can understand and respect that to a certain degree now what's harder for people to really understand and accept is do you just flip the toggle to turn off hyper threading or what's going on because you've got a ninety nine hundred K for 150 bucks more and it's more or less the same product as the 9700 K just there's some numbers changes nothing major it's still the same silicon ultimately and then really the big difference is there's no hyper threading so you end up without SMT and so that gives you an eight core a thread part whereas basically every previous i7 was four cores eight threads so in terms of pure thread count you end up about the same core count obviously goes up but the doesn't really necessarily mean anything depending on what kind of workload you have and price goes up which makes it feel kind of funny to buy a 9700 k if you used to bind the previous i sevens where it's more than the $340 range and i mean it just it was the best other than h EDT so the stack has completely shifted now now you have functionally a low-end h EDT cpu in the desktop market the 9900 k and you have the 9700 k dropping a core feature of the i sevens but instead adding more physical cores so i guess the question then is is it a lateral move or worse is it a downgrade from the 8700 k where you had six cores 12 threads so you have more threads and that is one that a test and blender would be great to illustrate something we have these results they because in blender you can really start looking at thread impact to performance we're having more cores doesn't necessarily mean everything sometimes just having threads even if it's just sort of pseudo threads throughout some tea that helps a lot so that's something we're looking at today with you 9700 k versus the 8700 k then of course we also have overclock numbers in there as well we only pushed ours to five point one gigahertz we did have trouble with this particular cpu so our 9900 k did pretty well once we had it deleted liquid metal and under a big water loop it was doing something like five point three gigahertz i actually we hit five point four briefly so that one did pretty well for us the 9700 k we were hoping it would supersede the performance of the 9900 k because having fewer things in AO having fewer threads means you're dealing with less heat in a concentrated area so it should theoretically overclock better problem is that you're still relying on silicon bending at the end of the day so you don't want the silicon lottery well something you can do about it so other than go crazier we can deal with this one and put it under chilled water and that it would do pretty okay but let's go through the benchmarks for games through bent blender and then we also have some thermal testing on here as well comparing versus the u 99er k both stock and d lidded to give you an idea of what kind of cooler you might need to deal the CPU f 1 2018 produces extremely high frame rates something not commonly found in other titles this game becomes CPU bound rapidly as our RT X xx atti allows all of these CPUs to fully stretch their legs as best illustrated by these 79-80 x ii at 4.6 gigahertz hitting 307 FPS average it's a bit high the point is that the bottleneck in here is elevated enough to really look at performance differences with CPUs to a point where it may become sort of almost useless at this type of framerate but we're still gonna look at it the i7 9700 cane stock ends up at 272 FPS average putting it functionally tied with the 9600 K at 5.2 gigahertz but ahead of the stock 9600 K by about 6.2 percent the stock 9900 K operates at less meaningfully ahead of the stock 9700 K with limited gains of about 3.3 percent overclocking the night 700 K pushes it to roughly stock 99 hundred K performance levels compared to the 8700 K the stock i7 9700 K ends off at about 10% ahead an average frame rate although overclocking gets the 8700 KS and nearing equivalence when at 5.0 gigahertz for anyone moving off of something like an i7 2670 9 TK and will highlight those CPUs for you you're looking at noteworthy performance improvements even when considering at 2600 K overclocked heavily Rison remains confined to the lower half or so of the charge as a result of the lower frequency and this game does seem primarily bound by Amdahl's law here's a frame time chart for f1 2018 at 1080p showing the 9700 K 5.1 gigahertz CPU 8700 K 5.0 gigahertz stock CPU and r7 2700 at 4.2 gigahertz functionally at 2700 X or close enough to it as a reminder of frame time plots show individual frames over time providing the truest representation of how long it took to render each frame we're looking at frame to frame intervals on the y axis labeled in milliseconds time to render with frame number on the x axis lower is better but more consistent is better than strictly lower the 9700 K manages to be both lowest and most consistent may the most fluid courier of frames in this test average frame to frame interval is about 4 milliseconds for the 9700 K with the 8700 K at about 4 point 2 milliseconds the 2700 at 4.2 gigahertz competes reasonably and average is about five point three to five point seven millisecond frame times all three tests encounter spikes upwards of 16 to 20 milliseconds but none of them are jarring to a point of ruining the experience they are noticeable though and most humans tend to notice about an 8 to 12 millisecond delta frame to frame as for 1440p results the top end results are truncated by the GPU limitations imposed by a higher resolution maxing out the ceiling at 240 FPS average the 9700 K ends up ranked alongside the 9900 K and Center 90 DXE that's because we're now hitting a bottleneck with the RT X 20 atti resultant of resolution increased to 1440p if you're playing at high resolutions and commonly GPU bound this chart illustrates the performance fall-off from becoming limited by another component lower end devices end up performing about the same as their slower than the GPU and so not limited assassin's creed origins at 1080 p medium stands is our next benchmark assassin's creed uniquely makes use of both additional threads and additional frequency whereas most games lean entirely on frequency this is made obvious by our 79 80 XC at 4.6 gigahertz our performance cpu is like the 9700 K at 5.1 gigahertz a significant frequency Delta but also a significant core Delta and you can kind of do the math there the 9700 case stock CPU ends up at 127 FPS average marginally better than the 87 heart K at 5 gigahertz there's about a 12% to uplift over the stock 8700 K a noteworthy gain and this comes down to stock frequency increases in the 9700 K for reference the Intel i7 2670 gigahertz does about 76 FPS average and the am the r7 2700 at 4.2 gigahertz does about 104 FPS average right around where the 7700 K is so that marks most tears of the chart for you far cry 5 positions the i7 9700 K at 149 FPS average right behind the 9600 K at 5.2 gigahertz note that far cry has also had routine issues with frame times and I five CPUs something we are actively researching now that we have our 9700 K in hand that said what we can learn is that the game benefits from higher frequencies in a direct fashion it's just that it also benefits from having eight threat the 9700 k ends up running at 159 FPS average when clocked to 5.1 gigahertz putting it between the two ninety-nine hundred K SKUs on our chart the 8700 KS stock CBN's have a 141 FPS average which plants it below the 9700 K but not by a meaningful amount no one can reasonably detect the average frame time difference of 0.4 milliseconds between these two devices and consistency is about the same for each so that's good the 2700 at four points u gigahertz ends up at around 111 FPS average which is within error margins of the 2600 at the same frequency these two AMD results illustrate a frequency and IBC limitation not a thread limitation plotted instead as frame times we can look at the 9700 K 87 hard K and 9900 K stock performance the 9700 K ends up fairly consistent with a few spikes 20 milliseconds and 16 milliseconds throughout testing it's not bad overall and we'd have to run a lot more passes here to determine how much those spikes are within the usual error margins the 87 arcade stock CPUs end up similar to the 9700 K in average frame time but ultimately fall behind in the total frame throughput the 8700 K runs slower on average if comparable with the 9900 K significantly more consistent almost to a points of creating sort of a heart rate interval it's very consistent at the beginning there we've also applauded the 2700 at four points u gigahertz for reference giving a test comparable to the 2700 X the 2700 at 4.8 gigahertz falls behind in both frame time and total frame through hood it's still over 60 FPS which would be around the 16 point six seven millisecond line on the y axis and it is overall consistent in frame delivery much like the 9900 K and 8700 K it's just a little bit slower but still consistent unsurprisingly we see mostly the same ranking order at 1440p we're truncated at the top end by the GPU which puts a cap on maximum performance in this instance the 100 K 9700 K and 9900 K are all rough equals the 2700 still ends up at around 111 FPS average when overclocked which is what you'd expect because we're capped by the CPU for the 2700 not by the GPU which is what we see at the top end of the chart civilization 6 provides an important look at turn time performance rather than framerate performance instead testing Grand Campaign a I turn completion times as an example with 5 AI players that each take 10 seconds to process a turn if you've never played this type of game before you would be waiting 50 seconds before you are allowed to make your next play which can be frustrating or get old as the turn time is dragged on for our chart the 9700 K at five point one gigahertz ends up within margin of error of the 9900 KF five point two gigahertz we were unable to produce meaningfully different results between these two and test variance puts them as functionally equivalent that's good for the 9700 K although Civilization 6 has never been particularly thread intensive to begin with this is best illustrated by looking at these stock r5 2600 versus the stock r7 2700 where you see the 2600 outperforming the 2700 by slim margin that's because of the higher stock frequencies and shows that the extra threads there really don't help in Civilization 6 illustrating what type of game Civilization six is the stock 9700 K ends up at eleven point seven seconds return with a stock 9900 K again tied with the 9700 K the 8700 K at five gigahertz ends up at eleven point eight seconds also functionally the same although the stock 80-some 100k ends up outside of error at twelve point five seconds per turn blender gives us a look at rendering performance for a tile based production 3d rendering workload this is very different from how something like premiere would handle encoding so we'd advise definitely not trying to extrapolate this performance to video production applications for blender the 9700 K at 5.1 gigahertz ends up completing the GN monkeyhead render in 25 minutes or the GM logo render in 28 minutes this place is it around where the stock 2700 is but was significantly higher power consumption at sixteen point five amps or about 198 watts sometimes 195 depending on when you check the test overclock in the 2700 to 4.2 gigahertz similar to wear a 2700 X would land puts it at 23 minutes for the monkey heads and 27 minutes for the logo this is where Rison starts to pull ahead and is also where the 9900 case shows its performance advantage over the 9700 K because of that threat difference the same is true for the 8700 K which leverages its extra threads on the six core 12 thread approach to lead the 9700 K in each test with stock and overclock this illustrates that it's not all just about physical cores eight cores is beaten by 6 cores 12 threads on the same processor family it's still an entire family so architecture is not really a big factor here thermals on the intel i7 9700 k are on the screen now for auto settings keep in mind that we heavily rely upon the motherboard used as the temperature is most directly impacted by Auto voltages we're using the ACS at Maximus 11 hero for this one with MC e disabled for the Maximus 11 hero there's Auto settings put the 9700 K at 27 degrees Celsius over ambient when loaded to 100% with a blender workload and as a reminder that's over ambient if you're new here that's not the direct temperature reading keep in mind that this is with MC e disabled so it's following the boost duration specification set by Intel and sticking to a power consumption of 90 watts liquid temperature is at around 6 degrees over ambient during this test pass and for power consumption it starts out at about nine point five amps and then drops to seven point five amps which would put you at 90 watts go into about five gigahertz and 1.35 volts 5 point one gigahertz on the 9700 K we measure an average all core load of 54 degrees over ambient with power consumption at about 188 watts and liquid thermals at about 10 degrees over ambience this place is the 9700 K is noticeably cooler than our 9900 K sample once soldered which measured at 64 degrees over ambient and that's with the solder the thermal Delta isn't for any reason other than the 9900 K drawing significantly more power in this test the two results are not directly comparable other than two salient Lee 9700 k's lack of hyper threading does make it easier to keep cool especially under heavier overclocks by drawing less power this isn't too like for like comparison given that power draw is different for each test but that's not really what we're comparing the 8086 K is also not directly comparable and that it's dye is a completely different size and so it's dissipative properties are different than the 9700 K and 9900 K that said it's still accurate just with regard to comparing how your temperatures might vary from product to product or for purposes of selecting a new cooler or upgrading an existing one thermally the 9700 K isn't that hard to cool under Otto conditions and it can be kept reasonably cool with our X 62 to ADC LC even under overclocked conditions although this is a somewhat conservative overclock it does become a challenge to overclock beyond 5.1 gigahertz and 1.35 to one point four volts with this cooler as we're entering the 90s for some of the individual core temperatures not all of them but as long as a couple are in the 90s in approaching t.j.maxx those will hold back our maximum overclock which comes back to our 99 80 XE content where we show the core to core Delta so for those instances you would need to upgrade the cooler but some of this is also a silicon binning instance as well because you might be able to drive a lower voltage than we needed for ours that's all the data then not really a huge move from the 87 or K in some areas its lateral in some areas like blender it's a bit down actually because of the reduction threads despite other improvements and overall just feels like kind of a wash it's not a huge change and like most CPU releases if you had the most current previous generation from the same company find out a whole lot of point an upgrade and that remains true today the real question is is there a point at what point is there a reason to upgrade so we have data for a 2600 K in there probably a good time to upgrade from that one it's still doing really well it's holding on 4.7 gigahertz overclock you're doing alright but it's definitely getting him haul of the tooth so if you're on something like that then your considerations are something like in the same price category roughly really it's an 8700 K or a 9700 K with a 8700 K being potentially cheaper in some non-us regions especially and with the and let us know what your regional pricing is but depending where you look we've seen the 87 okay on our a bit cheaper and that's true in the u.s. too it's just that for 30 bucks cheaper buying something older you feel kind of bad like should I just spend the extra thirty dollars and get the new one but if the performance is roughly the same it's not like a GPU where driver supports really a concern over time it's just I mean the performance is what it is it's not really gonna change all that much things like Spector and meltdown notwithstanding so should you buy the 9700 K well the 1900 K is kind of out of this conversation because it's just so much more expensive the 8700 K is a bit advantaged in blender but that's about it so for an extra couple bucks if you really care about there's that reduction of frame times then yes the 97 hard K is better if you want something like just blender per once you're not on a recent in tolerating the CPU and you're building a new computer for the first time in a few years and you're looking for 3d rendering and application performance and something like blender then consider strongly rising as a much cheaper alternative for something like a 2700 and then overclock it in 15 minutes or less and you functionally have a 2700 X that outperforms most but not all of the new Intel CPUs for a price reduction in that application and something like blender and rendering not necessarily true for encoding in premiere where Intel and quick sync do have a bit of an advantage there that AMD does not presently have and that's discussion for another time but overall the 97 hard K is hard to come to a firm conclusion on because it's just such a weird part it doesn't really feel great to buy something where you know they just turned off hyper threading and previously it was on but also there are more physical cores so kind of about a net zero for thread count increase which is boring and the frequency goes up a couple hundred megahertz which is boring so hard to say but not exciting enough to run out and buy if you're building a computer anyway don't feel bad about it it's just that yeah the price creeps up a bit so you've got alternatives if you don't like that that's it for this one subscribe for more as always go to patreon.com/scishow helps out directly or go to store iraqi aaron's nexus net if you'd like to pick up a shirt like this one or we just got the dark mouse pads in so it's a black blue nation if you've been waiting for one of those thanks for watching I'll see you all next time