Gadgetory

we just published deleting and thermal testing results for Intel's i9 7900 X now being followed up with the same task for the i9 79 80 X II and I 970 960 X we'll be looking at thermal performance of the 16 core and 18 core Intel CPUs power performance and some production workloads in our initial review of the i9 79 60 X and 79 80 X II this includes comparative data for Andes at thread Ripper 1950 X 1920 X and other Intel comparative chips like these 7900 X that came out recently before getting to that this video is brought to you by synergy the software that lets you share a keyboard and mouse between multiple systems if you have limited desk space and multiple computers to command synergy removes the need for separate peripherals or a KVM and works as over the network software use our link below to get 50% off the home or provision with SSL so we're still travelling from the Linus thing but the goal of today's review is going to be focusing on the deleting and thermal testing that we did prior to leaving we did the most of that so we have the most data there that included some power testing and then with those tests came validation and blender Cinebench premier firestrike a couple of other applications some productions some synthetic so we really digging through all those numbers today in the review of these CPUs and the prices of the seven-line ATX e and 70 960 X are a bit higher than the Saudi 900 x7 900 X is $1,000 part the 79 and 80 X e is about 2000 and the 79 60 X is about 1700 then you have things like thread upper down at around a thousand as well for the 1950 X and a bit lower for the 1920 X so if you're not caught up to speed on the deal it in liquid metal benchmarks and thermal interface benchmarks that we did recently check that video and content as well that will inform you on all the background for this and then if you want testing information how we did the deal id's what liquid metal we used all that type of thing then check the link in the description below for the article where we've got all the testing laid out and the process for the previous content piece which was the 7900 acts so let's dive straight into it we're looking at these 16 core and 18 core parts today starting with D lighting and thermal tests we can look at prime95 28.5 power torturing at 3.6 gigahertz and 1.15 VI D fixed the voltage ID is a bit higher than what Otto would output but it will ensure that all CPUs tested will remain stable at that specified voltage and means that voltage remains fixed Otto voltage causes fluctuations that would invalidate results but of course they tend to be more optimized for out-of-the-box performance just this is what we don't want right now is fluctuations and optimization we want fixed performance for the Tim and for the liquid metal to better determine how the impact from the thermal interface plays out as we've already covered the 1700 X results in a standalone content piece we'll focus on the 79 60 X and Saturday 980 x e today the 79 60 X and its stock configuration with Tim and additional silicon adhesive and at 93 degrees Celsius with liquid temperature an area and a set X maximums pack of 60 C this is the most concerning and from the standpoint of getting Intel to act on its poor thermal solution is the most powerful OMS and sis like Dell Alienware high power and cyber power rely on ASA TAC and other companies for affordable CLC solutions for workstations and HDD t platforms the facts that we're nearing the 60 C maximum operating liquid temperature on an open-air bench means that under heavier working conditions and with 240 to 280 milliliters CLC's we'd be nearly at or breaching the 60 Celsius threshold when in a case this may result in some behind-the-scenes politicking between Intel and OEM partners in the future and is the most likely path to getting some kind of change in the thermal interface replacing the Tim with liquid metal drives us down to 78 degrees Celsius with the kraken X 62 liquid temperatures and now at 45.5 Celsius that's a reduction of about 13 to 14 C for this particularly power intensive test starting on ATX e plots at 90 Celsius with its stock configuration with the liquid metal variants brought down to 72 to 73 this reduction is particularly massive measuring in at about 17 degrees Celsius drop in average CPU core thermals by switching to a better thermal solution on the package here's the secret with this test though the 79 to 80 XE did not finish the test concluded abruptly when two cores hit the t.j.maxx value of 104 to 105 degrees Celsius which we can illustrate with a versus time plot for the temperature and so we had a shutdown or a crash event that prohibited the test from completing to the same duration as all the others using blender Next and overclocking the CPUs wheel and the 79 60 X at 4.6 gigahertz with a 1.2 to voltage ID and are forced to use a 360 millimeter there I'll take flow radiator with 3 Corsair magnetic levitation fans in order to prevent throttling with the 360 millimeter radiator and a 60 DBA worth of maglev fan Arsenal the 79 60 X with Tim operates at around 85 degrees Celsius liquid metal and the 280 millimeter radius as much worse fans mind you even manages to outperform the Tim and 360 millimetre radiator on the stock CPU this alone is enough reason for Intel to consider better alternatives a quieter smaller cooler is achieving better thermals when using superior interface material even though it has worse fans using the 360 millimetre Thermaltake flow radiator and Corsair fans with a liquid metal application we're now at 76 to 77 degrees Celsius a reduction of about 9 degrees from the Tim and 360 millimetre solution 79 80 X Eve meanwhile operates at around 87 to 88 Celsius when using Tim with a liquid metal version at 75 see a reduction of around 12 to 13 degrees Celsius power blends well with thermals so we'll look at those next unless otherwise noted all of our power tests thus far and into the next section are taken at the rails we clamp the EPS 12-volt cables to get those numbers directly and for our standardized blender test we're measuring power consumption of 218 watts with the I 979 60 x16 core 32 thread CPU at stock settings or 492 watts when overclocked to 4.6 gigahertz with a 1.2 2 voltage ID that's 40 amps down the EPS 12-volt cables at that point we measured about 214 watts for the stock 79-80 x ii with in our usual variants of the 70 960 X and also measured nearly 500 watts with the 4.5 gigahertz overclock now of course how worthwhile this trade-off is depends heavily on the actual performance improvement and we'll get to that next comparatively the AMD thread urbanizing 50 X CPU consumes about 144 watts in the same test we saw the 1920 X plot around the same with a 145 watt throughput overclocked in the 1950 X landed at 274 watts though we only ever achieved a 4.0 gigahertz overclock in our initial testing so it wasn't really worthwhile let's look at Blender performance results our 4k animation scene renders in 15 minutes on the AMD 1950 x threader for 16 core CPU in 22 minutes on our stock 7900 X or 18 when overclocked and in 14 minutes on the 79 60 X let's focus on the 14-minute 79 60 X time and 15-minute 1950 X time for now the 1950 X requires 144 watts of power during this render with the 79 60 X requiring 218 watts that's an increase of about 74 watts or around a 50% increase in power for about a 5.9 percent decrease and render time required which was a good thing you want that decrease although some studios might reasonably argue that the speed benefit outstrips power requirement for those particular studios as animators can get more done we see this as increasingly difficult to argue when requiring 50% more power for gains in these single digits overclocking the 79 60 X gets it to 11 point one minute but we're how at around 490 lots of power consumption at this point so that's a render time reduction of 22% versus the stocks under 960 X which is absolutely impressive and worthy of high praise just looking at how much Headroom exists in these CPUs but it's 23% time reduction at the cost of 125 percent more power so the preys sort of evaporates at this point sadly this is very much a your mileage may vary area lower voltages can better stabilize these high frequencies with lower power consumption but that comes down to chip quality and we just couldn't achieve anything better than what we were getting here anyway at eleven point one minutes the seven nine sixty X is twenty seven and a half percent faster than the 1950 X but draws three hundred and forty eight watts more power to do it the 79 80 X II meanwhile completes the render in thirteen point three minutes stock or ten point one minutes when overclocked and these speeds make the 79 60 X far more desirable than the 79 80 X II in this task given its price point and relatively equivalent blended performance again though we must consider the 1950 x4 folks not GPU accelerating the renders or for anyone who has specific scenes that render better with CPUs the 1950 X here offers a better balance of power consumption and thermals in exchange for it's still fast render times they're not faster and they're not the fastest but they're fast and if you want the fastest and you just don't care about anything else including the cost of the CPU and the power consumption which is factored into cost at some point well it looks like you've got a new fastest CPU to consider for interesting it's just not as balanced a victory by Intel but it is a victory we have to give credit for tremendous overclocking Headroom but also must take some away by pointing out that the architecture scaling and efficiency is showing as the clocks increase and as the voltage increases Cinebench power consumption looks like this during the multi-threaded test the 79 60 X consumes about 249 watt stock and about 486 to 490 Watts overclock the 79 ATX e consumes around 230 watt stock for this particular test or around 444 watts overclocked and this with a lower voltage so it consumes less power if you're wandering throughout Ripert meanwhile consumes 151 watt stock for the 1950 ex core 4 core that places the 79 60 X at 64 percent more power than the 1950 X when both are stock and as for how that translates to performance here's a chart the AMD threader per 1950s stock CPU achieves a score of 30 170 CB marks multi-threaded and 165 single threaded the I 979 60 X stock CPU achieves marginally better multi-threaded scores of 31 90 points but has a significant single-threaded lead at 194 points this isn't a surprise at this point this is what we've been seeing with Intel and AMD over the last well a long time now for all intents and purposes though the 7-9 60 X is functionally equivalent in multi-threaded performance to the 1950 X with a less than 1 percent performance lead in this particular benchmark this sub 1 percent gain is at the cost of an additional 100 watts down the EPS 12-volt rails though to be fair in blender we saw a greater gains than were seen in Cinebench but the power consumption was still very high and there's blender tests so not everything has changed overclocked to 4.6 gigahertz the 79 60 X boosted score over stock by 25% 25% again that's impressive insignificant Headroom for a gain out of an overclock that's achievable by an end user in exchange though its power consumption is increased by about an additional 240 watts so this 25 percent performance lead is equivalent to the gain over the 1950 X as it scores similarly as the stalks of 960 X thus every 980 X II hits 33 88 points multi-threaded leading these stocks every 960 X by 6.2 percent not hugely worthwhile for the extra money for most users but again for those studios where the only thing that matters is being done first and fastest or whatever and freeing up animator time for other things well maybe it matters but that's not really our core audience so hard to say when overclocked to 4.5 gigahertz the 70 and 80 X mm forty-three 98 points which is again impressive but it's also drawing a lot of power just like we've seen with the other tests just quickly here's a look at power consumption during it the single-threaded workload Intel's lead is in the range of 17 percent in single threaded scoring for Cinebench with power consumption at around 70 watts stock so a bit more than the 1950 X with a good bit more performance single-threaded and that's a game that intel has had for a while now power consumption during 3d marks physics tests look something like this the 1950 X consumes about 100 watts when stock the 79 60 X stock CPU consumes about 173 watts when stock and the 79 80 XC was consuming about 144 watch when stock overclocking pushes our consumption metrics up to 280 watts with the higher v ID of the center 960 X or 238 watts with the slightly lower B ID on the Saudi 980 X E and slightly lower frequency in terms of the resulting scores 3d mark 2 seems to hit a point of diminishing returns with the 79-80 x e if you're a benchmark err it seems that these 79 60 X is the best buy for hitting the higher combined scores and 3dmark we're able to hit 35,000 points and fire strike physics with the 4.6 Hertz overclock or 28,600 before the overclock the 1950 X stock CPU hits 25,000 points at its 100 watt lower power consumption resulting in a 70 960 X stock physics score that is about 14% higher fire strike has a somewhat unique requirement of high core count and high frequency both of which are achieved on these 79 60 X but again this is a synthetic application so the usefulness of this data depends on what you are doing we have a lot more power figures than what we've shown so far but it doesn't make a lot of sense to show some of them so prime95 for example we have power numbers for that that are pretty reliable normally but we were having an interesting issue with the asus rampage extreme 6 board and the 79 60 X or 70 and 80 X II CPUs we're not sure which is the culprit here but basically with prime95 28.5 when overclocked over bolted things like that the power cycling was extreme to the point of tripping osep over current protection so this is with going through the digit plus PRM and enabling a 240 percent current limit and things like that but something's Howard was trippin so with the 79 60 X and suddenly on EDX e we found that when overclocking and over bolting them the application would for awhile show maybe twenty seven to thirty six or maybe even forty three amps at the very very high end of it but always what happened at some point is it would jump to 50 amps you can do the math yourself fifty times twelve point three is what you get for watts so in some instances of testing we were getting into the area of 615 watts down the EPS 12-volt cables to the CPU under liquid not liquid nitrogen and thermal and Tim thermal interface material they're all pasted basically so it would trip over CP shutdown and the power numbers are therefore a lot higher in reality than what they were when averaged and we're not gonna bother showing them for that reason but not sure what the cause of that is it seems like a prime95 specific issue but we were hitting in the 40s for amperage with these CPUs when overclocked and other applications like blender so it's it's not only prime it just kind of depends on how far you overclocked what voltage you use things like that let's run through a final benchmark before concluding with Adobe Premiere we saw the following rendering performance metrics we've started finally hitting performance levels in CPUs where our premiere ABC HD 1080p 60 P project is it picking up performance with CUDA but that's not the focus of our tests and ever has been we've typically only included CUDA results to demonstrate that CPU performance becomes insignificant as production is accelerated with GPUs with our particular test but we're seeing the 1950 X very slightly the 70 90 60 X and 79 80 X e all to a greater breaking the previous 20-minute barrier with CUDA starting with the CPU only performance though because that's what we actually focus on the 79 60 X at 4.6 gigahertz impressively completes the render in 33 point 6 minutes for a time reduction of 22% versus the 79 60 stock CPU which had a time of 43 minutes Sura and ATX II is much less exciting completing its render and effectively the same time at 79 60 X and this is the trouble with premiere the software like firestrike has some unique and weird requirements that don't always make sense it wants frequency but it also wants cores but not too many otherwise they won't be engaged the 796 DX it makes far more sense for our single test project than the 79 to 80 X EDA is looking at premiere but the 1950 X also seems to be hitting the same wall at some point and slightly outperforms the 79 60 X stock so it falls a bit behind quite a bit behind when we start GPU accelerated with CUDA in this test but it's hard to say how valuable that is outside of our test scenario and as we've explained in the past the problem with premiere is it's very project sensitive so these numbers really apply to our project it's a 1080p AVC HD project I think we're rendering it at 22 megabits per second and it's one of our youtube videos it was the original EVGA icx review I think of the 1080 FTW or something like that but I mean you start throwing 4k projects at it with different codecs or you start using different visual effects or you apply scales and transforms or warps or things like that the numbers get crazy and it's it's just project sensitive and it's it's very specific to single projects so you take the numbers for what they're worth with our benchmark and keep in mind that if you're not producing similar videos to what was used for the benchmark then the numbers don't necessarily apply now you'd hope that there's some sort of linear scaling across the board but that's not always the case particularly if you start introducing weird variables like CUDA and GPU acceleration CPU acceleration should be more the same though that's why we focus on though is not on CUDA so that's gonna wrap it for now we're still traveling as you can see but I just wanted to come out the gate with this review focusing on thermals power and going through some of the core production benchmark numbers the 79 60 X is a much more interesting CPU to us than the 79 80 X II the suddenly on ATX e is one of those flagship Halo type people who want the best will buy it even though they don't necessarily need it people who actually need it well they're kind of out of our scope so hard to say what kind of studios or workstation environments would use that versus like a Xeon or why would you use that for 79 60 X things like that maybe if overclocking is not permitted on those types of business class systems but the 79 60 X is the better balance of the two better balanced financially achieves great clocks if you can get it there with a better liquid cooler but the 1950 X is a major thorn on the side of these now Intel CPUs here are better gaming performers we did streaming benchmarks again they're not really worth showing because we couldn't really get a good way to drop frames on these high end cpus including the 1950 x but the one thing that's consistent as intel's and does hold PS a bit higher than AMD but we're getting into class of cpus where you shouldn't absolutely not be buying them for gaming like do not go out and buy these CPUs for gaming the closest you get to that is maybe you're someone like build Zoid and you buy a 79 60 X 4 firestrike benchmarking that's like the only gaming adjacent scenario where it makes sense but if you're doing production tasks seriously consider the 1950 X it's cheaper it consumes a hell of a lot less power that's really the big thing here it's it's not cost because cost you can kind of justify the extra money if you're a professional and you look at it like I can complete my tasks X percent faster that time is money an extra $700 now is completely insignificant over the next year of time that I'm either cutting off from my employees work or from my own work but these Intel CPUs the HT DT ones aren't always faster than 1950 acts at least not in a significant way they they tend to be a bit faster but the the extent of which that speed is realized in applications in a meaningful way is limited you might be faster by a couple percentage points in some cases but the bigger thing is just how much power it takes to drive these CPUs to the point where they are I mean thread Ripper is just more power efficient that's all there is to it Intel's falling behind here and it's something architectural II some of this is to do with leakage and thins where the thermal solution is just not up to snuff and it's falling behind and leakage it's falling behind in just the requirement of the cooler you can get cheaper coolers with a better thermal interface on there like we show with liquid metal or he stick with it stock he throw 360 on it with some loud-ass fans he call it a day but that's not really a great solution to us so we'd like to see Intel fix that 1950 X is a serious consideration if you're looking at via CPUs if you really need that single threaded performance you can't get better than Intel right now that's still the case so consider that as well the saturnine 60 X is a tremendous overclocker it's a lot of fun to work with you can push it really far but you have to mod it you have to deal with it get a better material on there get a good cooler on it then you can get it places and you need a board like the rampage 6 extreme as well otherwise you're not gonna have a vrm that can drive 600 watts without problems so power is a big concern here it's the biggest concern it's a major criticism against intel's current x-29 CPUs and with power factors in thermals they go hand-in-hand in every sense so that pretty much wraps it up for now for more information as always a link in the description below this is a scenario where Intel's done some good and some bad but probably a little bit more bad on this one than good so it's it's kind of hard to recommend these CPUs right now but we'll see we'll see if they do something for the next round or as we continue to explore them when we get back to home base perhaps there are more use cases where it makes a lot more sense but for the stuff we're showing today hard to justify their great CPUs in terms of performance for raw throughput but once you start looking at the things attached to that performance power thermals money the arguments not so good so thank you for watching as always you go to patreon.com/scishow can access to all that directly it helps a lot you can go to gamers Nexus squarespace.com or Store gamers Nexus dotnet to pick up a shirt like this one we just restocked all of them and I'll be out of hotels soon so that's oh one more note for those of you in the previous video who are concerned about me being sick or something no I'm just I'm recording in the hotel rooms I have to keep it down so we're all good but yeah I'll see you all next time you