Intel's Misleading CPU "Spec" and TDP Ratings

[Music] welcome back to hardware unboxed here we are with the monitor steve setup once again it's been a little while since we've brought this thing out but we have something interesting today to talk about and that is intel's tdp related stuff b560 related stuff and that is all because of your video steve so why don't you just run us through a little bit about here what we want to talk about today well for those of you who missed the video it probably makes sense to go watch that video first but i suppose the summary is i was doing some b560 vrm temperature testing because that's been highly requested started on z590 but with b560 making more sense because now it supports memory overclocking and then i'd say the cpus that make the most sense uh the locked parts or the lower end parts like you know your 11 400 type series 11 700 they're really good value now because very little overclocking headroom all the things you guys are probably aware of so i set out to do the b560 vrm temperature testing and found that it wasn't as straightforward as i thought it might have been i start on the higher end boards they operate as expected so we're something like a core i7 eleven seven hundred we're looking at a four point four gigahertz all core frequency but some of the other boards three point two three two point nine uh so it's a bit all over the place out of the box and then some of them you could remove the power limits and get the 4.4 gigahertz some of them did 4.4 gigahertz with extremely high temperatures that content's still yet to come and then others weren't able to achieve 4.4 and achieve stuff like 4.2 which is fairly close but it's still not there so it was quite important which board you bought and then it leads to the whole mess where some boards you would expect like even mid-range boards like around 150 us mark which you know used to be a lot of money to spend on a motherboard they would run with the power limits enforced so it was a disaster really for those using the the more desirable 65 watt processors because as we found in that video the performance uplift can be as great as 50 between boards so that's quite incredible really so we made the video covering that uh and i would say the feedback was largely positive around 99 positive feedback on that one so that's great and the reason we're making this video isn't because to directly address any negative feedback though a lot of these questions will be brought up more because of either confusion or the negative type stuff but just because there's a lot of misconceptions around this and we'd like to address them and a lot of it is reasonable enough for the users to be confused about it because it is quite confusing and intel has i'd say deliberately made it confusing and that's what we're going to work through so i think one of the first things we can start on is i think the intel spec that's probably what comes up the most i think with most of the people that were a bit confused about our video and who was to blame or what was going on didn't really understand what the intel spec is and again you can be excused for that because intel does not make it clear what their their broad or as i said in the video their loosely defined specification is so most people think of the intel spec the spec as the tdp spec which you know you have pl one pl two uh pl2 enforced uh using a turbo timer and that's the specification that intel defines that's their what would you say uh gamers nexus refers to it as the guideline or their guidance yeah so essentially when intel is you know producing they produce a document that they send to oems who make motherboards who make desktops who make even laptops for laptop processors and it has a list of their sort of default values which as you say gamesnexus sort of refers to as guidelines or something along there and you'll have in that table i'm maybe i'll be able to find it and put it up on the screen but it will have a pl1 and it will usually state that as the tdp of the processor so that would be 65 watts for a locked part then there'll be a pl2 which will be the turbo boost period and then there'll be a towel which is the turbo boost duration so the combination of all these things tells you how high the power can go when we're boosting in the boost state how long that boost state can be and then what power level do we drop down to once we're beyond the boost period and intel provides some default values for that but that's not the end of the story is it no so there's been a lot of videos over the years that explain all the stuff that tim just talked about and regular users they kind of glaze over and go i've heard all these things and it doesn't make a whole lot of sense to me it's quite complicated and it kind of is and it is tim said it's not the only specification so these processes which we used to talk about more have a clock multiplier table it was much more simple when they had like four cores so if there was only there was four different states so i've got some notes here i've got the clock multiplier table for the 11700 i suppose it's important to note that the default clock multipliers are something that intel no longer publishes so it makes it quite difficult to work out at what frequency without power limits the cpus should be targeting and stuff like all core is kind of important information to know because with tdp limits enforce that can move around and it can be different from board to board depending on how efficient the vrm is how well they've done with their voltage tuning to get within that 65 watt limit all that sort of stuff so anyway in the case of the core i7 11 700 when you're using one core it'll run at a 49 times multiplier so with the base clock that gets you 4.9 gigahertz two cores same frequency so one and one or two cores heavily utilized the cpu will sustain 4.9 gigahertz and that is the published max boost frequency so whatever you see like you know 4.9 or 5 gigahertz or 5.3 that's usually one or two calls so if your workload requires three or four calls you're at 4.7 gigahertz so it drops down 200 megahertz there then we've got five cores and six cores so using five or six cores that'll drop to 4.6 so another 100 megahertz gets shaved off seven cores 5.4 gigahertz and then the all core with all eight cores active which is what i saw in my blender workload with the boards without power limits 4.4 gigahertz so if there's no power limits in place those are the frequencies the cpu will run at depending on the workload and that's really how they used to work when we had up to four cores so you've got the max boost of 4.9 and then the all call there are 4.4 however with 65 watt parts the base frequency with in the case of the 11 700 is 2.5 gigahertz and i think well why does it have a base of 2.5 if the all call so when all eight calls are active it's 4.4 gigahertz that's that's a long way away from 2.5 and that comes back to the tdp so and this is where it gets all very messy so when you've got this this power limited spec which as tim said isn't the only spec the all core frequency will be brought down to fit within that 65 watt power limit so it's sort of a package power if you use something like hardware info and we saw some boards if when they were the voltage was tuned very well they would run at a 3.2 gigahertz or core so that's well down from the 4.4 but somewhere as low as 2.9 so the voltage tuning wasn't very good there or the vrm wasn't very efficient a lot of wasted power it's probably more voltage than anything really because of the way it works but yeah you're still looking at like a 10 variance for tdp limited motherboards and of course as i said even 3.2 gigahertz as well or it's a big jump up to four point four so we're seeing up to 50 performance boosts in some instances so that's an explanation of the difference between the tdp and the clock multiplier table and where people get confused is they think that the tdp tdp specification is the specification but it's not intel have a loosely defined specification so as long as you're running at either those tdp power limits or up to the clock multiplier table anything at either extreme or anywhere in between is all within the intel specifications so very confusing it leaves it leaves a lot of room to move in terms of performance because as i gave the example which i feel like a lot of people didn't watch that video right through to the end where i showed from i think it was the seventh generation the eighth generation the ninth generation then we skipped to the 11th where because intel's been stuck on this 14 nanometer plus plus process for so long normally you would move to a new node that's more efficient increases density and you get your performance and efficiency improvements that way but because they've been stuck in the one spot trying to add more cores and make them a bit faster increase the the max turbo having to drop that base right down and it's creating this huge discrepancy between the base clock and the boost clock especially for parts that they're lower tier parts like the 65 watt ones so i think the example was something like it was around 16 15 16 between the base and boost with 7th gen and it's crept up ever since then i think it's like almost a hundred percent now with eleventh gem with with parts like the eleven seven hundred so as i said there's just a huge room to move there in terms of performance so you can be within the spec anywhere and get radically different performance and that just makes for a so a pretty awful user experience really i think a lot of the confusion comes as you say down to people believing that the spec is one thing as opposed to a range of things so if you think about intel specification as a range then it becomes much easier in my opinion to understand what intel allows the processor to do so as you were talking about the minimum specification that intel allows is basically like a guarantee of performance they say you're going to get at least the base clock and you're going to be using it's provided that you have enough power to run at the tdp which for the 11700 that we've been using as an example is going to be 65 watts so that defines sort of the minimum guarantee so when let's say you're a big business customer you want to make sure you're getting a certain level of performance you want to do a lot of validation based on a certain level of performance then you need to know what is the minimum that you could possibly get so that's where that specification comes into play and then for more enthusiasts like us like people who would buy z590 motherboards and run big coolers intel has a range that you can run all the way up to and that's where we start to see the clock multiplier table and anything in between there like for example you might have a let's say you have a motherboard that is running power unlimited but your cooler isn't very good so it's going to run at you know it could be running at the temperature limit for example which would then cause power limits to come into play it might run in the middle of that range or you might get a board where you know it it can't sustain 250 watts but it can maybe do 150 watts or then it's going to run again somewhere in the middle of that range and all of these things are allowed by intel and as you say they've kind of trying to have the best of both worlds here by advertising a really low minimum that allows people to get away with using low-end vrms using really crappy entry-level coolers and say that that's allowed while also giving enthusiasts the ability to run them at ridiculously high frequencies while using tons of power and by not advertising either of these things particularly well and just allowing people to go oh well this is a 65 watt processor that can run at 4.4 gigahertz which is not necessarily true with both of those things enabled at the same time then yeah they're kind of getting it makes it look like their process is much more efficient or alternatively much faster than it really is under one of those conditions being limited to yeah exactly and as reviewers it's been confusing for us so over the years as as this has evolved so it's no no uh surprise i suppose that the audience is confused as well and perhaps we haven't done the best job of explaining it over the years and a lot of people are focused on it's easy to focus on what the td tdp spec is because intel tells us what that is and then it's been assumed that it's the motherboard uh manufacturers that are just making up their own spec or running without power limits and they're doing that at their own free will which certainly isn't the case but then if you go and ask intel you know is this board running at an all core of 4.4 gigahertz you know following the multiplier table is that it within spec they will tell yeah that's within spec but they don't it's not clear to the viewers to people buying it that that is within spec so it creates all of this confusion so yeah we're trying to address that as best we can but it's also difficult to explain all the different operating parameters that you'll run into with these boards and it makes it more confusing when you have amd in the market as well who uses totally different tdp calculations so their tdp is not comparable to intel's tdp and they don't have the same sort of boost behavior with precision boost as intel does on their platform so whatever you see on amd cpu and you go okay well i'm running my cpu it's running you know 105 watts it's boosting up to the precision boost limits and then you go over to an intel platform it's like oh well now i have to figure out whether i'm going to be running it at 65 watts and getting a certain clock speed or running at 200 watts and getting a different clock speed the confusion there is definitely real especially if you're transitioning from one platform to the other and we've seen similar things in the past almost in the opposite way with things like amd's maximum turbo frequencies when they talk about you know all this cpu can do 4.9 gigahertz but they don't necessarily mean sustained 4.9 gigahertz like intel might mean so the fact that these two companies are sort of advertising the same sort of things both of them have a tdp both of them have maximum clock frequencies but then that can mean totally different behavior for the user ends up being quite confusing yes exactly at least in the case of amd while we have made videos that look at their boost behavior and and help correct some of the issues there with the the variants we saw across a wide range of motherboards they were able to fix that with updated agiza code but at least you were getting pretty much the same level of performance very very fine margins there it wasn't even 10 i don't think there might have been some bad example boards but generally it was it was very low uh percentage differences whereas here you can buy one motherboard for 150 us and then a different motherboard for 150 u.s and then one board if you slot 11 700 on it you load xmp which most you know tech savvy users will know how to do most of them will do you would hope you load xmp on both boards one for sustained workloads can be up to 50 percent faster which is well we called it a mess because i think it is and while if you you know are a bit more tech savvy you can use intel's official software like xtu or you can go into the bios and you can manually remove those power limits and in many instances that will restore full performance but again it depends on the quality of the board and even some of the boards where it does restore full performance you can be running at very high vrm temperatures so a bit of a minefield there is in which boards will actually do it and which ones are running into vr and thermal problems so when i've been discussing this with some of the people that have had criticisms of the video or further questions i've pointed them in the direction of a really good uh interview done by ian cuttress over at a nantech where they interviewed uh i forget the the name of the intel employee now but they interviewed him and he basically explained that what is inspect behavior and what is out of spec behavior and basically it's very simple anything that runs above the base clock up to the default multiplier table is within spec as we've been saying so perhaps we can link that interview in the description of this video just so it's a good interview yeah there's a lot of stuff covered in there and yeah intel's officially saying that that stuff is within spec or is in spec uh but they're not putting that on the product page or making it clear to to buyers so that's i guess what lands us in this mess yeah and i guess that brings us to what is out of spec behavior because if there's something that's inspect then there's got to be something that's out of spec and intel's been pretty clear that overclocking is obviously when it becomes out of spec behavior so you cannot overclock on a locked processor whereas of course you can run up to the clock multiplier tables and remove power limits on a lock processor so as soon as you start overclocking and you've got a k series chip then an overclocking flag gets enabled and that means that you're running out of spec so if you want to run your you know your 11 900k at five gigahertz all core then you're overclocking it and that means that you're running out of spec conversely you can also have out of spec experiences on the other end so below the tdp limit and that's generally because you've either got well insufficient cooling or a motherboard that has insufficient vrm capabilities and then you get throttling so throttling also can be out of spec behavior so that means something else has gone wrong but everything in between those two things tends to be the zone of in spec which as we've talked about is anywhere from yeah the base clock of the tdp right up to the clock multiplier table and yeah intel tends to be pretty clear with this in terms of that particular interview and also you know obviously the fact that you cannot overclock on the locked processes yeah it's exactly right so stuff that i imagine a lot of people will be familiar with stuff like the asus mc so multi-core enhancement feature that runs all cores at the maximum single turbo frequency so in the case of this process it would go from 4.4 gigahertz to 4.9 if you know this cpu supported overclocking which it doesn't so and that would be anything to clarify that would be that would be overclocking yes because when you have eight calls active it's meant to be you know 4.4 gigahertz and then you're raising it to 4.9 so you're adjusting you're increasing the clock multiplier table uh to that to the highest value that is there and then of course if you go above that which with a k uh skew part to you know five point whatever then yeah it's out of spec so that that's basically so there really isn't in the case of the 11th gen there isn't too much room to move out of spec the cpus are pretty close to what they can really do and again that brings us back to the same thing of intel trying to have the best of both worlds as we've seen amd with precision boost getting very very close to the limits of their processor with you know trying to run it basically as fast as it can go all the time intel has been doing very similar things we used to have more overclocking headroom with k series processors but over time as they try and close the gap with amd they want to increase the clock multiplier table which in turn because they've been stuck on 14 nanometer increases the power consumption but then they don't want to advertise a processor that is a 200 watt processor because that's right how does a 200 watt processor look up against a 105 watt or 95 watt processor from amd it looks like they're using a lot more power which they is true but they don't want to advertise that so we're currently in this situation as we've been talking about where yes they don't really want to admit the truth about their processes in some way by making this these parts 65 watts i mean you'd say for example 65 watts is a little bit ridiculous for a part like a 11700 it just is not really a 65 watt part well i agree it's also a little bit clever they've the situation they've found themselves and this is kind of a really good solution for them because if you look at a lot of benchmarks a lot of the applications that reviewers and even people benchmarking their own system tend to run like cinebench for example i'm talking about you know cpu benchmarks that look specifically at cpu performance they don't run for you know five minutes 10 minutes half an hour unless you put them on a loop so if you want a cinebench score especially these higher end parts they finish it pretty quickly so they spend a lot of time in the maximum power state there and it really helps with their score but then if you're looking at thermals or power consumption generally people will loop those tests or run a test that's longer so it does in a sense help them get best of both worlds and take gaming for example which a lot of reviewers and a lot of people watching reviews are really heavily interested and focused on the gaming performance and games won't like in the case of something like the 11700 it won't peg all cores at 100 load for the duration of the test the cpu might only be at 20 30 percent which is off that's really often the case in a whole lot of games when you've got an 8-core cpu so this cpu is just putting those calls up to the absolute maximum frequency and if you remove the power limits there's very little to be gained there in extra performance so some people argue does it even matter but when you get down the track and the games start using more and more of the cpu the gap between whatever the competing part is that doesn't have the same problem will widen and you can take an older cpu as an example like if you bought something like a core i5 8400 that would that performed really well when it first you know hit the market but if you run it 65 watt limit now i imagine that gap has widened to something like a ryzen 7 3700x or whatever you may be comparing it to yeah and there is some justification for doing it the way that intel's doing it apart from it obviously being a clever marketing tactic in that they do understand that a lot of the general apps that someone uses are going to be your sort of short burst apps like if you're just web browsing you know at times you're just going to load something up it's going to need to do a little bit of a burst which will run at that maximum pl2 state it'll consume 200 watts for like a second it will run at the maximum turbo frequencies then it will drop down and you won't need to be using the processor anymore so it makes sense that if they're going to say a processor is 65 watts that for burst applications you want to average the power out over a long period of time so that on average using 65 watts but that might be using 200 watts briefly and then zero watts or idle state like 30 watts or whatever it is for a longer period of time you get the average under 65 watts so that's generally how intel has i guess justified these sorts of power limits and it makes a lot more sense in mobile platforms where yeah you know you're you're running a laptop with a 45 watt processor over time it's going to heat up it's going to run very close to its thermal limits you can't just run it at 100 watts indefinitely but you might be able to run it for 100 watts for a little bit before dropping down and that's where we see these limits really come into where it makes sense but on a desktop platform they're kind of i guess trying to make their cpus more efficient for users while still giving them the performance they need in boost states and i just don't necessarily think that it applies as well to a desktop platform and it creates issues in the future as well because for example there's nothing stopping intel from making like a 25 watt desktop processor and calling this 25 watts but it runs at 200 watts some of the time so there's really no there's no lower limit that they could keep going down and down and down in terms of tdp rating and keep making their clock speeds lower and lower and lower i mean we've already seen that you know comparing 7th gen to now there's nothing stopping them from making a 1 gigahertz processor if they really wanted to all of that said if the tdp spec was the spec so and again if if it was the spec motherboard manufacturers would have to adhere to that specification it's not the wild west where asus can get just overclock intel's processes intel you know would cut their allocation they have to follow because intel doesn't want their partners coming up with their own poorly validated specifications that create stability issues and god knows what for their customers they want it to just work and they want to be stable and they want the customers to have a good experience so next time they buy a cpu they're more inclined to buy with the company that gave them the good experience previously so there's just i i i know for a fact like they're not making up their own specifications so they might add features like mce there may have been brief periods of time where that sort of stuff was you know they got a bit cheeky and enabled it by default but that was always corrected you know intel won't let them do that because if you let motherboard manufacturers do whatever they want they're going to try and deliver the most performance they can out of the box to make their boards look the absolute best and they'll do it at the cost of stability they just will and we've seen that not just with mce we've seen that before with things like motherboard manufacturers enabling enabling bclk overclocking on like locked parts or enabling memory overclocking on b series motherboards before that was an allowed feature which of course a motherboard manufacturer can just let that happen but usually you'll see that bios update come out and then very quickly be retracted to remove that functionality because intel does not want out of spec behavior and those two things are out of spec behavior unlocked processes they don't want that happening so they make the corrections so they're not just allowed to do whatever they want yeah so yeah now that we're clear on that if the tdp spec was the spec it would mean that the intel cpus perform as you see in the reviews and they have quite a substantial amount of overclocking headrooms that'd be something you'd really have to explore because it's something they offer that their competition does and yeah that comes at the cost of you know huge increases in power consumption and difficult to cool processes but it is a thing you you can get as good or better performance than their competition perhaps through overclocking so it changes the story there whereas at the moment as tim said a few times they're trying to have the best of both worlds so is it an efficient process that's quite clever with its power management or is it brute forced and just runs to a clock multiplier table and uses as much power as it needs to do that and the answer is it's both pick one some boards adhere to the tdp some follow the clock multiplier table both are within spec that is the point that is the problem we blame we we put the blame squarely on intel yes 100 and i think if anything intel is giving oems something that they can get a benefit from by having this very wide range of official inspect behavior because it allows a motherboard manufacturer to make an extremely cheap and crappy b560 motherboard because they don't have to design a board that supports 200 watts of processing they can support a board that does 65 watts or 125 watts 125 watts being the maximum that you can get on these sorts of platforms so all they have to do is meet that bare minimum run at 125 watts we will support the 11 900k that's it whereas if the spec was say higher and they were running their process at 200 watts at the clock multiplier table suddenly oems need to spend a lot more money on their motherboards and a b560 motherboard that costs you know 110 dollars right now is just not feasible because the vrm has to be so much better you know you can't have motherboards with no heatsinks anymore you're going to have to put heatsinks on it that adds cost you have to add more phases you have to add better components and that allows current the current system allows those oems to sell boards in that very low power range which is fine if the customer that buys that board is also buying a locked processor running it at the power limits it's a lower end part so they're not expecting high performance but it's not great if that customer then wants to upgrade down the line to an 11 900k and are expecting the performance seen in the reviews so by doing this again one of the main factors that has been a competition between intel and amd's motherboard pricing where generally speaking you know b550 motherboard may be a bit cheaper than a b560 motherboard it would have been an even larger gap in terms of motherboard pricing if the intel spec was more defined towards the top of the range that they currently allow which would make intel again less competitive and this is where the 14 nanometer issues the power consumption issues start to affect them in more ways than just oh look it uses a lot of power it's things like motherboard cost and cooling costs and all that sort of stuff yeah exactly and sort of circling back to my original video it also creates this weird situation where with an entry-level motherboard running the 1100 700 in a lot of workloads it's slower than a better b560 board running the 11400 so it's this weird crossover where you can upgrade your cpu and not really get any extra performance yep so depending on power limits and that sort of stuff because obviously a 50 performance uplift is well over the additional calls you get there so it just creates this huge mess for motherboard buyers well intel buyers in general i would say yeah absolutely and reviewers as well because it's a very difficult situation for us to make a review that says this is how this processor performs but then you could go and buy a motherboard that we didn't use for testing particularly if we're testing z590 with these cpus and then you're going and buying a b560 motherboard and then you're not getting the same performance which completely changes everything we've said in the review it changes how competitive a processor is versus its competition it changes things like power consumption data it that does favor intel in that situation but that's important as well it changes our recommendations for potentially things like coolers and that's not a good situation to be in where in this enthusiast market where a lot of people building their own pc do research and watch reviews to know the parts they want to buy are potentially getting screwed basically in terms of the performance that they're may ultimately getting when they're buying their system and again intel is effectively allowing this to happen and really it would be great if they didn't do that so then everyone's on a level playing field when it comes to performance and then you don't need to go and make a giant video where you buy like 30 b560 motherboards and test them all because you'd have no idea which one is going to run the power limits and which one's going to run the clock multiplier table like that's just ridiculous yeah well it it creates the layers of problems that we've got because as you say there's there's boards that run at the clock multiplier table there's balls run at the power limits and it's not really price dependent there can be really expensive b560 boards that adhere to the power limits but they are capable of running without the power limits so there's that but then there's also some boards really aren't capable of running with the pallets removed they'll either vrm throttle thermal throttle or they will do it but the vr will be running well over 100 degrees so if you're trying to build a cheap editing rig or something that wants to heavily utilize your cpu some sort of productivity task not great if you're running your motherboard for long periods of time at those very high temperatures uh yeah you might be able to get another board for the same price that wouldn't have had that problem or maybe if you were aware of that up front you would have spent 10 or 20 dollars more to get a board that can can sustain that performance without you know risking very high thermals on the vrm and you get you get a very different experience with amd for example because because their processes are so much more efficient you start to have these things where as you've talked about you can basically buy any b550 motherboard obviously there are some crappier examples and some better examples but generally speaking the performance range that you see there is they're all about the same because there's no need for one part to run it at a low tip low power limits and another part to run it really high because the power consumption is manageable with an amd processor even something like a 5950x is consuming you know not 200 watts well below that when we're looking at peak power consumption so yeah the efficiency differences at least over time i think we've come to understand more and more the implications of the difference in power consumption and the difference in efficiency on a desktop platform previously people have just i guess dismissed it a little bit in terms of oh this doesn't really matter because ultimately you can just use a bigger cooler and you know it's a desktop desktops don't need to be power limited it's all fine but then what we're seeing with this generation is really the head of it where actually it's not super fine and there's a lot of complications that it comes into and amd being the more efficient platform has allowed them all sorts of advantages that maybe we didn't really expect to play out in this way yep yeah i don't want to make this too much of an amd versus intel thing because i know where that leads us but i know what you're saying this does lead me to a comment that i saw quite often whereas people were saying you know we're making a mountain out of a molehill because you can just remove the power limits and most of those boards will run at the full performance which i've mentioned a few times now as i okay that is true for a lot of boards that will be the case but that doesn't make it any less of a disaster in my opinion it doesn't make it you know it's not clear which boards will do it and it's not clear you have to do that like a lot of people were shocked there were plenty of people that watched that video that watch a lot of our content and they were shocked that removing the power limits could result in such a massive performance uplift for core heavy workloads and you know saying that that should be common knowledge or you're stupid for not knowing that well i don't really like that attitude towards it because i'm sure there's a lot of things that you may not know about in life that you'd rather made it you'd rather that was made clear than hidden from you and you found out the hard way so and like when you go in when you go into a motherboard bias i'm just interested how difficult is it to find the right setting to enable you know full power limits or removing the power limits like not you're not only needing to know that you need to do that but then you actually need to go execute it in other words where is the setting in the bios what there's a lot of confusing stuff in there so you need to be you need to go beyond just the knowledge you need to then know you know for this motherboard what's the setting called do i need to use x to you in windows as opposed to a buyer switch is that difficult to find uh it depends on the board i mean as the saying goes knowing is half the battle so if you do know at least you can go on research or work that out whether it's wise to do that will depend on your board so you if you were worried about blowing up your board which can take your cpu with it then maybe that's something you don't want to do but if you think that your motherboards vrm can handle those 150 200 watt loads then yeah you have to work out how you do that some boards are certainly easy than others and it's not even the same for a brand so asrock for example with their better quality boards they'll have an option to buy something like remove power limits or something like that so you just enable that and it removes all the power limits with one option selected msi give you i can't remember exactly what it's labeled but they basically give you the option of choosing there's one of three options so one's like the box cooler which is the tdp limited specification and then a lot of the boards have a tower cooler or a liquid cooler and generally they're both just completely uncapped performance so if you find that menu you can select that and it will remove all the power limits that way that's not super obvious if i'm honest to label it under like another designation as opposed to power limit so yeah and as a seuss you know they're usually a word thing i mean enabling xmp on asus balls in the past has been quite difficult because they like to use their own jargon and different menu systems but you know their boards overall are pretty good in that regard a lot of people like their biases so but yeah it varies from board to board you can download the xtu software and you can remove the power limits that way but again you it's not super tricky but there are more things that you want to max out the timer you want to max out a few d you know a few different options there i think there's about three or four in total that you need to max and it requires you to load into windows and do it which is yeah download the software install uh yeah um so yeah it's it's just extra steps basically that you don't have to do on competing platforms and again it's the fact that you need to know that you need to do that if you want the full performance of the processor yeah it's just not a great situation to be here it's a whole lot of sort of complex overclocking steps that as it turns out aren't overclocking yeah and you sort of like a lot of people when they watch our videos are going to be more technically inclined than your sort of average buyer that's going out and has heard hey you know intel is good value right now you know if you're building a new system intel is the way to go so you go to a store you just pick a b560 board off the shelf you grab your cpu and your memory off you go you'll have no idea about this no idea you if you've never watched hardware unboxed you've never you don't do a lot of research apart from what you know what cpu fits in the motherboard and what ram fits in the motherboard then this is just an added headache that is very difficult to understand and would just yeah you're just not going to know you're just not going to know if you're that sort of more casual end user type person and it's not even just about the performance that you're receiving right now as i said as games become say you bought like an 11400f and in a year or two gaming became much more demanding and the six core processor was starting to see 89 utilization maybe stuttering was getting introduced or you know the fps just wasn't as good as it was and you're running at the tdp limits so you could then spend a lot more money and buy something like an eight core model because you don't want to upgrade your system or whatever it may be so you slot in at 11 700. you could have just removed the power limits on the 11400f and you would have got almost the same upgrading performance by buying a different cpu so that's not great yeah that's it's not just not a great situation so for all of those reasons that's why we called the whole b560 platform a disaster and we're not saying the chipset is a disaster it's more the intel specification and what it does to 65 what parts is the disaster but it becomes a disaster on b560 opposed to z590 because half of the boards run at the power limits and half of them don't at least that's what i've found so far with all of my testing whereas not so much of an issue with z590 because every single brand except for asrock ignores the power limits and just follows the clock multiplier table so you will not see this issue on any msi gigabyte or asus z590 motherboards which makes it a bit easier you can just buy one of those boards slot in any of the cpus and they will just run the clock multiplier table asrock follows the tdp limits on the z590 motherboards i suspect because of a lot of the issues they had with z490 and but they have made the option to remove power limits in the bar so one click thing so there is that so this is a couple of topics to round out this video and the one thing i'm interested in your thoughts in is should reviewers be testing these cpus on the power limited platforms or should they be testing them at the clock multiplier table power unlimited configuration because there's always a lot of discussion around this as to which way is sort of the correct way to be testing and obviously if you test one you're going to not be showing the performance you may be getting in the other configuration so what's your opinion on i know you already test without the power limits enabled most of the time um but where do you see all that stand i i think i'd start by saying there is no right or wrong answer because they are both in spec so there's that to begin with the reasons why we choose to run at the clock multiplier table without any enforced power limits is because we especially in the past have mostly tested on the z-series motherboards because traditionally b series boards we've sort of recommended avoiding because you get stuck at lower memory speeds and yep you know it just hasn't been as good value basically whereas that's all changed now which has made this more complicated which i suppose is why we're having this conversation but with z590 by the board to take z490 for example almost all of the boards ran without the power limits so we thought in almost all instances when we're testing these boards we will be going and manually changing all the power limits and that's actually a more complicated process than going the other way so you have to make sure that everything is set correctly to run at that intel base specification or or their guidance or whatever you want to call it so we've avoided that and i also again don't like it for the reasons i mentioned earlier whereas it can be misleading especially for productivity benchmarks you're showing a cinebench score from maybe an average of three runs or maybe the best of three runs but that's not how that application works if you took the score which you can do now in cinebench 23 if you took the score after a 30 minute loop it can be quite different to the the best run or an average of three because then you're well and truly running at the power limits for that last run or probably the last 20 runs or whatever so we don't like it in that sense as well because it can be a bit misleading on what you would get or it'd be less than what you would get on a z590 or z490 motherboard not running the power limits and then you know it may you you're sort of handicapping it against the competing parts if like the ryzen processors yeah so we just don't do it for that reason and also i suppose it could be a little bit misleading if the base or the all cl or the all core is say 4.8 gigahertz without power limits but then you enable power limits and it drops down to 4.3 and then you enable something like mce or a basic overclocking feature that runs at the max single and you get quite a big performance uplift there so legit i believe there's quite a lot of overclocking headroom with that part but when you take it out of the box put on the motherboard it turns out there's only a very small amount of overclocking headroom so that can change what you think you need to build you know build into the equation in terms of cooling overhead and that sort of stuff but anyway it's basically we just thought this is how they perform out of the box on most boards so that's what we want to show our audience and we're not really interested in doing an oem spec review because we're not reviewing dell and hp systems we're reviewing yeah it's because you guys are going to buy those parts and build your own computer so i think with all these discussions it's like so long as the reviewer that you're looking at is testing everything in the same way then it's not that big of a deal like the testing is still valid because it is part of the intel specification but obviously it would be bad if for example you were testing an 11th gen part without power limits but you tested the 10th gen part with power limits because that would exaggerate the performance difference between those two generations and lead people to believe that you were getting a much more substantial upgrade than you were getting i think there has been some instances potentially with intel's own um internal benchmarks that they kind of show in our press events where they i believe someone discovered one instance where they had done that exact thing which obviously makes their platform look a lot better um but yeah as long as you're testing yeah the sort of everything in a consistent manner it tends to be fine but i guess it does bring up you know now that we know the b560 motherboards do run with power limits in a lot of instances it does open the opportunity for say follow-up content sort of like we were talking about with the nvidia driver overhead issue where people are saying oh you need to test in this way in your day one reviews this does make for good follow-up content you test with the high-end z590 everything you know not limited configuration and then you can follow up with that b560 testing for people that will end up with those motherboards so then you can sort of see where the performance lies for those people yeah i think for this the the k skew parts the 125 watt parts not too much of a big deal there but certainly for the 65 watt parts you really ideally want to test with both yeah the tdp and the clock multiplier performance because yeah it does vary so much that the the the there's no right or wrong but the ideal answer would be to test both ways yeah i think it's as simple as that so so i guess the final thing that i wanted to get into was this is obviously a problem it's a big problem it'd be great for this not to exist for there not to be confusion so how does intel fix this what is something that they could do with say a future generation processor that would solve this so that you're not having these large discrepancies between which motherboards you buy so the tdp becomes maybe potentially more accurate what are sort of your thoughts on what intel should be doing should they be advertising a higher tdp well i think intel's hoping that the problem will solve itself and they'll be in a situation that amd's in they move to a more efficient process node and the problem kind of takes care of itself because well that would be probably the easiest way out for them uh and short of that like say say we're talking about the 11th gen what would be a better way of handling this having at least two different power states so you can choose and boards could be certified to be like you know 65 what or tdp tdp probably the right word anymore but they could be well we've seen this in the past like where i think it was first or second gen ryzen motherboards there was a 320 board i want to say that was only rated for 65 watt so the higher parts like the ryzen 7 parts weren't officially supported by those boards and perhaps didn't work at all but where i'm going with this is you could have the entry level dirt cheap motherboards you know skimpy vrm with no heatsinks and that could be the 65 watt model but if you spend you know 30 40 more you can get the 125 watt model or the 200 watt model or whatever it needs to be so yeah i mean probably not an ideal situation for intel i guess with how they compete with amd but it would make things a lot clearer for people that you know if you want uncapped the full performance of these parts you know you need a certain board yeah i think that would make a lot of sense something along the lines of potentially i mean a lot of people use arc for example intel's database of the cpus which gives you all this information on the clock speeds and tdp and all that sort of thing i think it'd just be good to potentially just have two different sections have a section at the top that says this is the like the minimum performance and then it gives you the base clock and it gives you the tdp and then you could have like a a maximum performance level section or whatever and that would give you your boost clocks and then a power level that you like a typical power level for an all-core workload at the maximum frequency so at least then buyers again i mean obviously as you say it's not advantageous for intel to do this because they don't want to advertise high power consumption levels but this would be a lot clearer for buyers to know that you could either be running at that lower specification or running at the highest specification and as you say potentially a certification process could could help a lot there for motherboard buying or just knowing the performance in general so yeah i guess it's just really it really comes down to at the end a big lack of clarity around these things like needing to get a person from intel into an interview with a nand tech to explain the intel specification is not really an ideal way to be explaining all of this to enthusiasts like people who are watching our videos tend to be educated and able to understand these things so by trying to hide it away it's not it just leads to these confusing situations where people have no idea what's going on that's right that's why it makes no sense to defend intel this one say this is an intel's fault or you know however you want to defend this it's bad for you it's bad for people buying into these platforms it's you want them to be more transparent you want them to be more open and honest and they're deliberately not doing that right now like this hasn't happened by accident they used to publish the clock multiplier tables this used to be things that all reviewers knew right if you google 11 900k clock multiplier table or all call frequency you've got to rely on reviewers giving that information and i tried to find the 11700 multiplier table i couldn't find it online that information just isn't there anymore you can if you buy the cpu install in your system and load up the xtu software you can you can read that information from the cpu straight away but short of that unless someone else is providing that information for you which is difficult to find and the reason is again it's not by accident they used to tell you what the cpu's expected all-core frequency was and in the case of the 11700 it is 4.4 gigahertz that is the official all-call eight-core load frequency 4.4 gigahertz except some boards it'll be 3.2 some will be 4.4 some it'll be as low as 2.9 and again this isn't by accident intel is doing this deliberately and it is bad for consumers so as far as i'm concerned you should not be defending intel for this at all simple as that yeah yeah i totally agree with that it's definitely a mess that is not it's not a necessary mess it's i mean i guess from a marketing perspective they're kind of trying to do everything they possibly can to compete but in terms of you know what makes sense for consumers it is totally unnecessary and it's just one of those things where it's so frustrating that we have to come and make a video which i'm now just looking at how long we've been recording for been going for an hour talking it we've been talking about this for an hour and people are probably still going to come out of this video being like what's going on yeah i mean there's not really too much more to say on this one i wouldn't have thought we've kind of gone through about how this is a big big mess it's not a great situation for b560 buyers it's not great for reviewers it's not great for intel they're probably loving life because they can just do whatever and apparently that's all okay but yeah well yeah i hope for those who were genuinely confused this video has been helpful i hope some of our explanations again not scripted it's sort of off the top of my head i've got a couple of notes here but hopefully it was easy to understand and it did help clear it up for some of you um that'd be nice yep well i think we'll wrap it up there tim i think that's pretty much it yeah i like doing these discussion videos always enjoy bringing out monitor steve for a bit of fun to talk about the latest stuff and yeah i found your testing in the original b560 disaster video which you have if you haven't watched please go back watch that i found that a really interesting video and certainly the results are not what i was expecting i guess i was to be fair i was expecting at some point boards to be tdp have to follow the tdp specification but i guess i wasn't expecting the performance difference to be as large as you were showing in some applications and and certainly the the low clock speed that some of those boards were running was it was very surprising so yeah it's been a bit of a bit of a learning experience for me as someone that doesn't necessarily test cpus all that often yeah i can tell it was very frustrating and on that note the b560 vrm temperature testing is coming it's that kind of testing in general like doesn't matter what chipset it is or motherboard that stuff takes a long time it generally takes me especially when i'm doing i think i'm doing four or five processes per board i think four and you're looking at like a day's work for one board there but in the case of b560 and a lot of the boards or at least half of them as i said i've got to test the board twice so i've got to test the out of the box specification with 265 what then i think it's two 125 watt processors then remove the power limits and repeat all over again so yeah it's there's a minimum of eight hours of stress testing just there so you can see how it takes one day to test one b560 board if i start in the morning and and go till uh quite late at night so anyway that's the situation there but i've done about seven boards now i've got about seven that i'd like to still cover once that's done you'll see that content and you'll know which boards to avoid and which ones are pretty good value for those of you that want to buy what i think are good value parts like the 11400 the f variant 11700 and the f variant there and this will also apply to 10th gen parts you want to put on those boards as well yeah so look out for that content when it comes on the channel in the next couple of weeks um of course you can support us if you're in the 20 club and you've watched all the way to the end of this long video then maybe consider supporting us on patreon or float plane if you're not already a member there i imagine lots of people who get to this point probably already are so thank you for your support but if you do sign up and become a new member you'll be able to chat to us in discord you'll be able to join our monthly live streams which honestly are pretty similar to this sort of thing usually you're here in person because we're doing q and a's at the same time but yeah very similar sort of discussions we have on those and yeah anything else to say anything else to add no well it's not the live stream will be coming up in i'd say about less than a week from when this video goes live or about that so yeah if you're interested you know now's the time to jump in and get get access to that but as soon as there's a lot of cool a lot of other cool things there but nothing else to add i think we're gonna wrap this one up my throat's getting a bit dry tim uh yeah good stuff just thank you everyone for watching especially if you made it this far i mean that's just awesome can't thank you enough so i'm just going to say that i'm your host steve have yours tim we'll see you see you next time [Music] you

Channel: Hardware Unboxed

Views: 111,719

Keywords: hardware unboxed

Id: WKzNkWfoQyQ

Length: 55min 36sec (3336 seconds)

Published: Fri May 21 2021