Gadgetory

in tells TDP has long been questioned but this particular generation puts the 95 watt TDP under fire as users noticed media outlets measuring power consumption at well over 100 watts on most boards it isn't uncommon to see the 900 K at 150 watts or more in some AVX workloads like blender thus far and away exceeding the 95 watt number aside from TDP being an imperfect specification for power there's also a lot that isn't understood about it including by motherboard manufacturers apparently all manufacturers are exceeding Intel guidance for the turbo boosting duration in some way which is causing the uncharacteristically high power consumption that produces unfairly advantaged performance results the other end of this is that the 900 K looks much better in some tests but it also looks a lot hotter in thermal tests with the lack of turbo duration guidance before that this video is brought to you by Thermaltake and the view 71 enclosure the view 71 is a full tower case that's capable of fitting three video cards and most configurations it's also one of the better cooling cases in our recent case testing bench lineup the view 71 has hinged a tempered glass doors on either side that make it easy to open and show off and it comes with at least one rain fan though you can get the RGB version if you prefer learn more at the link in the description below getting into this then as a reminder we had a power draw chart during our streaming benchmarks in the 9900 Kay review you might remember that in that chart the 99 hundred K sort of boosted up in power consumption and eventually dropped down to 95 watts or thereabouts where it stayed for the duration of the test and that is expected performance based on the specs that Intelli has put out but in most workloads it turns out with other motherboards or even with the same one that we use but with different settings the Intel CPU far and away exceeds that 95 watt number it'll go up to 150 or even 200 watts in some semi stock scenarios with different motherboards this comes down to a few things primarily to motherboard vendors or manufacturers exiting Intel guidance and it's them doing so without really telling anyone so you have to be savvy enough to figure out and then you need to figure out that it's happened down on all the boards so when you see this happening to everywhere you start to question your own sanity like is this is it even wrong if all the CPU is doing this power draw on every motherboard that what is the spec and that's what we're looking at today so the ACS board when we tested it we tested with XMP - which is the default setting not the optimized setting and we chose no - the prompt which disables MCE now this time MC is a bit different it's sneakier instead of just going all Kord the single core max turbo which would be 5.0 gigahertz it is instead going to be correct all core turbo so it looks fine it looks like MCS not even doing anything but the sneaky part is that it's blasting the turbo duration with msie disabled on the Asus board and this is the only one we've found so far that does it properly disabled it sets the correct turbo boost limitation for time and that means that the power after a fixed period of time no more than 100 seconds will throttle down to the 95 watt power limit and that's something we need to define here as well all the other boards as you'll see in our testing today violate this power limit and that's where we get this huge discrepancy and that's where the Intel 95 watt TDP that you all have pointed out is questionable at best so first you need to know about power limits one and two parallel Emmitt two is how much power the CPU can use in bursts Intel's official document describes this as quote a threshold that if exceeded the power limit to rapid power limiting algorithms will attempt to limit the spike above PL to power limit 1 is the maximum frequency under a sustained load given in watts Intel's document describes this one as quote a threshold for average power that will not exceed recommend to set to equal TDP power PL 1 should not be set higher than thermal solution cooling units that's an exact quote from their document in other words Intel recommends a 95 watt setting but does not mandate it once ewm a power or exponential weighted moving average power is roughly equal to power limit there is no more power budget for turbo boosting and frequency down clocks as a result of this we have a chart for that one as well that's when it steps down to power limit one levels exhibiting the lower frequency will see in testing later this is the Intel specification if it were following at what we're seeing on some motherboards you'd actually instead see the power and frequency blasted at those same peak frequencies and PL two levels that you're seeing in the chart in the same image tau is the turbo time parameter which is dependent on ewm a and follows the same curve although there's a typo in there from Intel Intel allows for up to 100 seconds of P l2 or infinite PL one power is allowed for PL 4 to be 10 milliseconds max by Intel spec the hardware defaults for PL 1 is 95 on an 8 core CPU or 119 Watts for PL 2 for the limited duration allowed and that's also for an 8 core Siegen bursts up to 119 watts with PL 2 or even higher with PL 4 but with PL 4 it's never more than 10 milliseconds with PL 2 it's not supposed to be more than 100 seconds and then finally note that TDP is kind of nebulous usefulness in general because as a number it doesn't really mean a whole lot doesn't mean literal power consumption it's again it's it's a bit it's a bit nebulous it's a Intel's measurement of TDP is more about cooler ability to dissipate heat than anything it's not a literal power metric it can be significantly different from the actual power metric which makes it difficult to use so we were talking to Shannon Rob another reviewer about all this from urine 3d and he saw similar stuff that we did so let's get into the charts and look at what we found with the different motherboards for the turbo duration issue power consumption is the most important for this one scene as it is a TDP rating discussion we're measuring power in two places we have EPS 12 volt measurements have the power supply cables via a current clamp then we also have hardware info software measurements the to ensure that we have reliability and can trust the data between each measurement we're also going to be showing the power within the first 30 seconds of load which will better illustrate at what point any throttle kicks in this chart shows a lot of range first of all none of these results involve manual overclocking we only ever changed one or two settings one of which would have been XMP out of all these results there's only one that follows Intel specification that's the ASUS Maximus 11 hero with no selected to the enhancement prompt after enabling XMP for this one we measured about 98 watts down the EPS 12-volt cables which is within reasonable margin of error for the 95 watt TDP specification though TDP again isn't really a perfect number this is within spec so far for the first 30 to 60 seconds we observe it dwindling at power consumption from 140 watts to 120 watts before eventually falling off to 98 watts this is due to turbo boost duration policies something we'll show in another chart coming up and this is ultimately what all of the boards should look like for power metrics speaking of that other chart for reference here's what the frequency functionality looks like when operating under settings with MCE disabled via XMP 2 and selecting the no prompt this plot takes place over a 23 minute blender render or there abouts so this is a real-world use case of the 99 hundred K notice that there's a sharp fall-off after about 30 seconds where average all core frequency plummets from 4.7 gigahertz to 4.2 gigahertz notice that the power consumption remains at almost exactly 95 watts for the entire test pegged to 94.9 watts and relatively flat this corresponds to the right axis with frequency on the left it sort of feels like an RT X flashback where we're completely power limited the difference is that this is under the specification despite the CPU clearly being capable of more you'll see that the frequency picks up again when the workload ends leaving us with unchallenged idle frequencies let's get that power chart back on the screen here's the weird thing if you leave the Asus board to just auto without any form of XMP without any MCE modifications whatsoever then the multi-core enhancement feature is left to auto as well this seems reasonable but in this version of multi-core enhancement instead of boosting to the maximum single core frequency which is very obvious when done and violates spec of course asus is extending the turbo-boost eration that's a new form of MCE it's tricky and it's well hidden rather than boost visibly different numbers like applying the single core boost modifier to all cores the board is allowing turbo to last ad infinitum the fact that the board operates outside of spec when auto but inside of spec when XMP is on and with no selected is a really odd choice to say the least you'd expect a more consistent policy with Auto probably following the spec more closely if anything it seems as if a snus hasn't decided if it wants to be within spec and lose to other boards or exit spec and exceed TDP let's get the EVGA outlier out of the way this is what power consumption looks like when you completely disable Intel's turbo boost functionality it's obviously awful as you're not really supposed to do this it runs the CPU way slower this establishes though that turbo is within TDC specifications of course but only when turbo duration is properly controlled by the motherboard manufacturer EVGA did not appear to have an MCE option of any kind and only had the normal turbo toggle so we decided to use it as a vessel for testing that power consumption baseline with no turbo boosting whatsoever next the other end of the chart shows the MSI godlike with enhanced turbo explicitly enabled as pulling 192 washdown the EPS 12-volt cables for reference enhanced turbo is natively set to auto which has an infinite turbo duration and violates TDP under all motherboard settings enabling enhanced turbo explicitly will boost all core frequency to 5.0 gigahertz and is basically a new version of MCE it is however sort of off by default in that auto and off are the same thing and on boost a single core frequency across all cores so really it's often auto violate the turbo boost duration spec and then on is a user activated function that pre overclock see that's the real meaning if we highlight the complete auto MSI configuration no changes at all that has us at 150 watts draw not anywhere close to the power figures cited by Intel technically this makes MSI the ones at fault but we don't know if there any behind-the-scenes politics that could influence things MSI also exits back with enhanced turbo set to auto which keeps the 4.7 gigahertz ratio but also removes the turbo duration limits to say when enhanced turbo explicitly leaves us in the same spot turbo durations still have no power limits and the power is still at about 150 to 155 watts there's there's no limit within the duration that one would expect by Intel spec instead it's basically infinite here's the frequency versus time plot when Emma size godlike is involved for this setup the asus board is still set to follow spec and has MC e disabled the MSI board is under it's Auto enhanced turbo settings which are the same as it's disabled enhanced turbo settings in this example the MSI board is holding strong to 4735 megahertz which is violating the spec in multiple ways sort of cheating for one BC LK is 100.8 which we already discussed for to the turbo duration limit is removed under these Auto settings and all settings actually and so it holds a higher ratio power draw is also higher peg to 152 watts instead of 95 from Asus and that extra 60 watts goes a long way toward performance back to power one more time with ACS selecting yes to the prompt puts us at around 142 Watts right around where the others are at this point the only difference between EVGA MSI and Asus is the automatic voltage provision EVGA has the tightest speed core and so draws less power as you can see in these power numbers but also is exiting the turbo duration spec for reference our initial review is done with the correct setting we use the asus motherboard with XMP - and no selected at the prompt and so our results follow the Intel spec and guidelines that's because we learned about this last time with a z3 70 if we used a different motherboard though our results would be higher and our thermals would be higher significantly and our power would also be higher so that's where some of the confusion came in a couple weeks ago from the reviews we have two performance tests to look at the first one is a truncated test rendering our GM logo seen at only 20 percent resolution you can see the normal GM logo rendered on the screen now it's been playing for a little bit and this is what the scene the final result looks like it's real-world use case of these higher-end CPUs it's one frame from our intro animation to this very video the second performance test is the normal 50% resolution render which takes about half an hour even on a good CPU for the 20% resolution test the shorter one you'll see that only three items are outside of the error margins and are quickly identifiable at the low end of the chart the EVGA FTW with turbo completely disabled this is obviously bad and not something you should be using but we did that that wasn't an EVGA default spec or setting that's where Intel would be at 65 watts with the 9900 K more critically the asus hero with MCE completely disabled selecting no at the XMP prompt performs noticeably worse than the other motherboards Asus gets punished for partially following Intel spec though their auto setting still exceeds it so we can't give Asus too much credit here they're not really the good guys it's just they were less bad than some of the others at 4.2 minutes to render the logo we're taking 11 percent longer to render the logo than the rough average of 3.8 minutes for the other boards here's the full render we only have a few of the 9900 Carranza's on here as it takes so long to run this test it's just enough to make a point our original review used the correct settings to follow Intel spec and so is the worst of the 900 K results that's at twenty five point five minutes which is just ahead of the r7 2700 at 4.2 gigahertz it's about a minute faster it's about three minutes fast for them the 8700 K at five gigahertz and significantly slower than the 70 900 x10 core CPU stock with the MSI godlike and EVGA FTW extended turbo durations we saw a render time reduced by about 9.3 percent that's noteworthy and diminishes the gap between the 7900 X while widening the gap between the twenty seven hundred and twenty seven our X stand in at 4.2 gigahertz finishing in twenty three minutes instead of twenty five point five might not sound like a lot but it's all about scale if you're an animator you're likely rendering thousands of frames a 10 percent requirement reduction in time per frame will add up very fast and so it's misleading the show performance numbers outside of the spec at the same time though if all of the board's violate the spec maybe it isn't misleading if that's the reality then the inaccuracy is in showing the power figures not in the performance figures either way both cannot be simultaneously accurate will leave thermals for another piece we have those as well but we're gonna push that with some other key insights that we ran some extra tests into another content piece this one that you should have a pretty good idea of where things stand now clearly the one instance that we've seen thus far and we're boards we didn't show here but as ROC Tai Chi follows the same behavior as the others except for the ASIS MCE off one so the ASIS board is the only one when you physically go in there and enable XMP and disable MCE that's the only one that really follows the spec everyone expects when they see 95 watts that's what we used in our review we had experience with this from Z 370 but if you didn't have that board or as a user you don't buy that board you're never gonna be up that number and unless you have that board and other boards you really can't compare them and figure out this is clearly what they meant when they said 95 Watts so asus is following the spec to some degree they're not angels here though they're also violating it with Auto settings which just seems completely backwards but either way at some point you're left asking who's really at fault here because intel has guidance that document we showed earlier is from Intel they've set guidance they've set parameters but they're not necessarily being strict enough with them and so a motherboard manufacturers are doing whatever they want now apparently Intel's going to be more strict this generation which is why you see Asus buckling down on it where previously they didn't and in fact asus was responsible for a lot of the sort of trickery that happened in previous generations so Intel might be starting to crack down on this but for now can you blame the motherboard manufacturers or do you blame Intel because motherboard manufacturers are doing whatever they can to make their boards look better than the other but at the same time it makes Intel look better sort of unfairly so really what should probably happen is Intel should drop the 95 watt number that's used and cited everywhere since it only applies under those certain scenarios described earlier and the boards don't really follow it and instead cite higher performance with a higher power consumption because those two go together as opposed to the basically impossible to achieve power consumption under all these other motherboards parameters except for one in one specific setting and and the higher performance there right now they're doing two things in opposition to each other when they need to be doing two things either together proportionately rather than inversely which is what Intel is trying to do with respect so it's a question of who do you blame Intel for not being strict enough with the guidance that they do have or the motherboard vendors for sidestepping the guidance constantly and running basically out of spec and you're left here with basically just ASIS in one setting doing this spec that like running at spec running where they should be so if it's just one under one setting then is it even a spec is it even guidance because it's just making a serious look Pat at that point instead of what everyone else is doing so that's it for this one hopefully that answers some questions as to what what's the deal with TDP why they say 95 watts and it's not not if i watch the answer is turbo duration and you're allowed to boost outside of TDP it's like a hundred twenty watts or there abouts it's one point two five times pl-1 but it's not supposed to be sustained and that's where these manufacturers are kind of cheating and then PL 3 + PL 4 can be exceeded also so that's it for this one you can help us out direct through this type of content by going to store documents access net and picking up one of our products like this shirt or the mod mats you get a patreon.com slash gamers Nexus helps out there as well or get access to our behind-the-scenes videos pushing to this week and subscribe for more I'll see you all next time