Gadgetory

today we venture into the untold story of liquid metal we normally talk about how liquid metal and d-lighting improved thermals and overclocking Headroom by allowing for higher vcore at lower temperatures but today we're going to show you how liquid metal allows for a reduction in noise levels by a perceived 2x while still allowing equivalent thermals to stock Tim but with higher clocks overall that'll be our concluding chart for today thrash to leave power and thermals for the Intel i7 8086 k before that this video is brought to you by NZXT is new h 500 case which we recently found to have an impressively effective cooling setup that is entirely negative pressure when stock the H 500 is the successor of the S 340 and s 340 lead offering high build quality that's all steel and glass and Cale management features that are also a top class for the $70 compact mid tower case H 500 is a part of NZXT Zanu H series lineup which also features options from mini ITX micro ATX and full ATX builds learn more at the link of the description below originally when asked if for when we would review the 8086 k we basically said no we're not going to it's kind of pointless and that still stands because ultimately it's an i7 8700 K except it's pretty selected so these are the highest Bening 8700 KS that's it that's really the whole difference of the CPU at the 8086 K so everything else gaming performance all that stuff stands for the 8700 KS the same exact processor except for overclocking results and we decided to do an overclocking livestream which will be already done by the time this video goes up where we basically bend the processor and see what it can achieve so we decided might as well do some D Lydian thermals and noise test and while we're at it and use this processor as a vessel to prove an important point that is overlooked with liquid metal which is that liquid metal isn't just good at improving your overclocking Headroom reducing thermals allowing higher v core it's also good at just keeping the same thermals with lower fan speeds you can do the trade-off and end up with significantly lower noise levels and that is kind of an untold side Madeline something I want to talk about today so the 8086 case in background Intel made a limited amount of these it's not that limited we know how many it is and it's a lot so not super limited but they made they basically pre-selected a bunch of 8700 KS and that's the 8086 is for a good bit more money the bigger takeaway again here is fan speeds for liquid metal we're using thermal grizzly conductor not as we normally do and then we're also doing a really important thing here we're just scraping off the silicone adhesive in order to reduce the gap between the die and the IHS that's where a lot of the thermal improvement comes from not just from the liquid metal so pay attention to that as well we'll start with power consumption here go through thermals and then get to our noise discussion at the end for power consumption testing as always the methodology will be detailed in the article link in the description below but for the basics we measured power consumption at the EPS 12-volt cables going into the motherboard this measures the cv power consumption and includes the vrm power consumption as well is much more isolated than wall meters and way more accurate than wall meters but still includes the vrm power consumption in there additionally our test cases include Cinebench single and multi threaded indicated by n t and one t in the charts times by fire strike physics fire strike combined as a gaming benchmarks standard it's the same kind of workload split between CP and GBS gaming blender which is av ax prime 95 29.2 which is a VX and we used 8 k in place FFT is for prime95 29.2 which is one of the most thermally and power abusive test you can do so that will stand as our basically hitting the thermal limits of the CPU test with blender being a bit more realistic but still a B X so prime95 is the most abusive let's start there for the thermals and the power tests running 8 k FFTs in place we started with a complete auto test including default bios settings in the Maximus 10 motherboard and only enabled XMP on our corsair 3200 kit the result was 130 watts of power consumption at the EPS 12-volt rails overclocking trivially it's a 5 gigahertz and 1.3 volts which was sufficient to sustain this clock resulted in a 194 watt power at the rails 5.3 gigahertz at one point four or five volts measured 250 watts getting awfully high at this point and starting to require some extra help on cooling the vrm fortunately the Maximus board has built up enough vrm that over temperature is not yet a concern our 5.3 gigahertz and 1.41 volt test did not pass prime95 8k and neither did the test the 101 BC LK they both failed this particular application for blender using GM's in-house built specially tailored CPU benchmark we measured 92 watt power consumption for the auto configuration this is relatively close to intel's advertised TDP at 5.0 gigahertz and 1.3 volts we measured just 135 watts a steep 60 watt drop from the prime95 abuse we saw at the same setting both our AVX workloads 5.1 gigahertz though at one point three five volts Dru 149 watts with 5.3 gigahertz at one point for five volts pretty high drawing 198 watts for blender we didn't actually require one point four or five volts this is just for scaling though our five point three gigahertz tests at one point four one volts passed and did so successfully as time consuming 178 watts each even with 101 B CLK both which failed for prime95 fire strike is next including both the physics stand-alone test and the combined gaming workload the combined test is more accurate cv power consumption when playing a video game that splits load across the GPU and CPU the results aren't too distant with combined often a few percentage points under physics stand-alone the most noteworthy results are 5.1 gigahertz at one point three five volts consuming only 105 watt because these cpus are so heavily Bend our volt frequency table allows much leaner power consumption and here's times by just if you're curious this one is more intensive but is still well under blender and prime our 5.3 gigahertz overclock at 1.4 1 volts past at 155 watts but the 101 BC LK configuration failed this test Cinebench is last this one shows at both multi-threaded or NT and single threaded or 1t results for the 8086 k here we see power consumption scaling up to about 197 watts with our more aggressive voltage provision with 177 watts for the 1.41 volt and 5.3 gigahertz configuration compared to stock our Delta is nearly 100 watts versus Auto but performance is significantly uplifted 5.0 gigahertz at one point 3 volts provides the best mix of processing power and power efficiency at 133 watts single threaded consumption skyrockets to 100 watts with the 5.3 gigahertz and 1.45 volt configuration but otherwise remains under 70 watts for all other tests the 101 BC LK configuration did not pass this test with the voltage we set it to for thermal testing we combined the two solutions prime95 8k FFTs which again it's the most abusive in terms of power consumption which means that it follows that thermals will be higher as a result ultimately that's all that matters it's how much power is being consumed by the CPU and being output as heat for the cooling solution to soak in this case including either liquid metal or the stock thermal paste which we sometimes refer to as Tim though they're both technically Tim so for this one prime95 8 K F of T is for the modes of use of thermals and then blender for realistic but still high thermals both of which are AVX these results will show both dee-lighted and stock thermal results we used the same top layer of thermal paste for each set of tests with a crack and X 62 cooler maxed out for rpms we also tested with thermal grizzly conductor not liquid metal applied under the IHS and atop the die this is something we've done regularly over the past year and it often provides 10 to 20 degree performance improvements a lot of the improvement though comes from carefully scraping off all the excess silicone adhesive around the borders but more still comes from the highly thermally conductive liquid metal thermal paste is a polymer compound comprised largely of silicon whereas liquid metal uses indium and gallium to improve thermal conductivity to nearing 70 watts per meter Kelvin depends a bit on the composition of the particular liquid metal they all use slightly different mixtures either way in the 70s for watts per meter Kelvin thermal conductivity versus at44 in tile space is quite an improvement well link to the thermal grizzly conductor not tube below in case you want to buy it for your own D lid let's start with the harder prime95 8k test first the complete stock Otto - measured an average core temperature of 46 degrees Celsius over ambient meaning that we were in the upper 60s when considering ambient temperatures of about 23 degrees Celsius this isn't bad at all and largely is thanks to the low voltage required for an 80 86 K to hold the stock clocks running with thermal paste Tim that is at 5.0 gigahertz and 1.35 volts temperatures increased by about 20 degrees up to 65 point six degrees over ambient this is from overclocking and over voltage although the voltage required is modest when considering our 8700 K needs one point four volts to hold five gigahertz going to liquid metal completely negated this and brought us down to about 47 degrees over ambient equal to the complete stock 8086 K with Tim when left on auto settings and yet this one was at five gigahertz core decor Delta's weren't as bad as we've seen on other Intel CPUs like these 79 ATX II we have a chart for this one for the 8086 K we think that the reduction in corticoid deltas is because of the relatively small die size when compared to hed T CPUs resulting in tighter thermal performance across the die for auto testing with Tim we measured a quarter core peak Delta of six degrees which is one of the better measurements we've seen overclocked with Tim we measured a quarter core delta peak of 10 degrees which reduced to about eight degrees with conductor not applied let's move on to blender with blender we measured an average load temperature of 36 degrees over ambient when left to auto and thermal paste which again we've labeled as Tim on this tart despite liquid metal also technically being Tim with Tim at 5.1 gigahertz and 1.35 volts that increased 15 degrees up to 51 degrees over ambient using thermal grizzly conductor not with this same volt frequency profile we measured 39 degrees over ambient a reduction of about 12 degrees from the 5.1 gigahertz and 1.3 5 volt Tim test the vigor news here as before is that this also corresponds with only a three degree increase from auto yet we have a 600 megahertz increase in frequency with Tim at 5.3 gigahertz and 1.25 volts which proved stable and blender but not prime we measured a 60 degree over ambient temperature liquid metal at the same frequency and voltage 5.3 1.35 measured just 49 degrees a reduction of 11 degrees Celsius in this test finally to of a point we ran an overclock at 5.0 gigahertz with 1.30 volts and fan speeds of just 60% with this balance of voltage and frequency and using liquid metal we were able to reduce our noise levels considerably down from 51 DBA to approximately 40 DBA this acoustic reduction is perceived as slightly greater than 2x2 a human due to the logarithmic scale of the decibels and illustrates our biggest point with deleting efforts in this particular CPU so that's the important thing that we wanted to convey today even when dropping 10 DBA off of our CPU fan noise which is substantial 10 decibels is roughly perceived as about a 2x change to a human observer because it is logarithmic and 3 DBA is about where it starts to get pretty noticeable to most people just for kind of a scale of reference so even drop in 10 DB out the CPU fan speeds we're still able to maintain a 5 gigahertz overclock part of that in large part actually is because of the CPU of course but outside of silicon quality you still have cooling quality and that comes from our delayed efforts where we scrape off the silicon adhesive it's just kind of the unsung hero of this and then also apply liquid metal to it which is highly thermally conductive so 5 gigahertz overclocked with the same temperatures that are about equivalent to the stock clocks with Tim so that's really the bigger advantage here is if you've never considered liquid metal as a means to do things other than increase your overclock and over voltage Headroom this is kind of the point we're trying to make 51 DBA to about 40 by deleting and scraping off compound and putting on some better stuff instead so further just to kind of note on the 8086 k we're not doing a full review on it we're greatly looking forward to the live stream which will be in the past when this video goes up should be a lot of fun but the kind of big thing here with 8086 K the reason it's worth an overclocked live stream and not a review is because of its Binda nature so this CPU is doing 5 gigahertz trivially at 1.3 5 volts for reference our 8700 K the first one could barely do 5 gigahertz and if it did it was at like 1 point 4 or 5 volts in sort of lighter load application hour later 8700 could do 5 gigahertz at one point four zero volts as opposed to five with 1.35 for the 8086 k and then the stock operating voltage is also significantly lower on the 86 k is it worth the extra money for an 80 86 K Bend CPU probably not if you have a really specific use for it and you're definitely going to overclock past 5 gigahertz and you think that is inherently fun to you sure but it is not inherently natively better in any meaningful way in terms of gaming performance than an 8700 K hence not bothering to review it because you're really at that point all we're doing is testing the OSI performance what we're doing in a live stream anyway so anyway that's the big news is the noise reduction something we haven't actually measured with liquid metal before but have talked about so hopefully that helps you out with some ideas on what you can do for your build next as always go to patreon.com/scishow and exit topside directly stored gamers nexus net to pick up one of our mod mats for a 4x2 modding service for pc building or you can go to the store and pick up one of our restocked anniversary edition shirts in teal we finally got a bunch of them in thank you for supporting and buying them subscribe for more I'll see you all next time