AMD Ryzen TDP Deep-Dive & What Cooler Manufacturers Think of "TDP"
HTML-код
- Опубликовано: 17 сен 2024
- This video talks about AMD's definition of TDP in great detail, deep-diving on Ryzen thermal design power, power consumption, heat load, and best CPU cooler choices.
Sponsor: Buy Thermal Grizzly Conductonaut (Amazon - geni.us/LCcsv)
Article: pending publication after the video!
Most people have likely seen "TDP" on a cooler box at some point, and it unfortunately has no useful correlation to "TDP" on a CPU box -- Intel or AMD -- and mostly correlates to formulas used to work backwards into a number that's nice for marketing. There are some useful applications for TDP numbers, but for consumers, it neither measures power nor heat (particularly accurately), so different metrics are needed to determine the best CPU cooler for a product. The most common mistake is for people to use TDP to compare processor power consumption between parts, especially cross-brand (e.g. AMD vs. Intel TDP), and this is a fair mistake to make when considering the unit of "Watts" in the TDP figures. Unfortunately, electrical power doesn't ever once show up in the AMD TDP formula, so this isn't really a useful metric for comparing power consumption.
In this video, we're defining TDP, particularly as used for AMD Ryzen parts (e.g. AMD Ryzen 3000 CPUs on Zen 2, but it's the same for the others). We won't be talking about Intel today. The focus is on TDP, what it is, where it's useful (or not), and what cooler manufacturers think of the Thermal Design Power metric that they're provided.
We have a new GN store: store.gamersne...
Like our content? Please consider becoming our Patron to support us: / gamersnexus
** Please like, comment, and subscribe for more! **
Follow us in these locations for more gaming and hardware updates:
t: / gamersnexus
f: / gamersnexus
w: www.gamersnexus...
Host, Additional Research: Steve Burke
Editorial, Research: Patrick Lathan
Video: Josh Svoboda, Andrew Coleman
Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.
Major research piece for us. We spent a few weeks going back-and-forth with AMD and with cooler manufacturers, then dug through all of the reviewer guides through Ryzen 1st Gen. All your support on content like this is appreciated -- it takes a long time and we really enjoy it! It's fun stuff to dive into. Article will come later. Support us directly by grabbing a modmat, toolkit, shirt, or glass from the GN store: store.gamersnexus.net/
You might also like our newest HW News video, which talks about some upcoming Ryzen parts: ruclips.net/video/VNfrGEBtncQ/видео.html
I run my 1800x on a corsair h100i on full load 44c
Thank you. Much appreciated
@@tsheplin If your 1800x functioned like mine at stock, it stays cool like that because under a full load, avx or non-avx, it stops between 95 and 100w Package power draw. avx will make clocks go lower but it will still stay limited in that fashion. I'd like to see some stock screenshots where that isn't the case.
@Gamers Nexus
TL;DR AMD's fomulas on how to quantify thermal dissipation. Various thermal states of electrical and metallic components and the variability of the thermal medium...also called "AIR" drastically changes the characteristics of thermal dissipation.
In other words, your Heatsink won't dissipate 95w (thermally) if your ambient thermal medium [AIR] is at 140F. This also *all* changes depending on the performance characteristics of the metal and interface of the "Item" you use as a heatsink.
It's not rocket science. "thermal radiation" (heat) wants to equalize with the rest of the cosmos...but it can only do so (?efficiently?) in the right conditions. [Insert somewhere the notion of Entropy]
@@coinexchange1761 You're right. It's not rocket science. It's marketing science.
Stopped studying for a thermodynamics test for this. End up studying anyway
Can't escape Heat Death!...
I stopped doing homework for theoretical electrodynamics xD
Energy can neither be created nor destroyed
my condolences thermodynamics are a lost cause for you
LOL
Must watch video for everyone! Loved the Watt is Watt part :D I guess there are engineering Watt and marketing Watt
Watt? :P
But Watt is not only electrical, might not be used to SI units...
en.wikipedia.org/wiki/Watt
So basically if a 1800w stove that heats food to above 150C, my 1800w vacuum cleaner should melt after 3 minutes, because no plastics can survive 150C+.
And the CPU is this magical device that consumes 100% of the power it gets. It's has infinite resistance, and no electrical data is passed to the other devices.
Got it!
@@Quetzalcoatl0 a vacuum cleaner is not really a good counter example as I think it has a fan disposing of the heat ;)
@@Quetzalcoatl0 1)Vacuum cleaner converts electrical energy mostly to mechanical tbh ....
2) If CPU were to consume 100% power it can get, then it would be equal to max output that PSU can output through motherboard. If it had infinite resistance, then no power would be drawn.
3) Your statements are so bad, that you should go study physics a little.
"There's no rgb cpus... Yet.."
Steve do NOT give them ideas!
My exact reaction XD
Corsairs already on it
only when it comes with RGB thermal paste, lol.
We absolutely need another heat source on the CPU
@@XavierFox42 just follow the lights, lol
I can't wait til we have a better, agreed-upon standard for what TDP is trying to convey
agreed! i wish for an actual industry standard for out-of-box TDP.
@@StaelTek yeah! I'm sure they'll get there some day, and everyone will look back on the current era like it's some antiquated, uncivilised mess
There is really no need, as long as you keep in your head that it's just a rule of thumb and that you allways have to get a cooler that is slightly overkill.
It's the same with PSUs, if your system will max draw 400 watt, than you atleast get a 500 watt PSU, but a 600 watt would be better, cause it gives you headroom.
Also keep in mind that TDP or even power draw numbers given by amd or Intel most of the time, if not always talk about a situation where the cpu is running at base clock speeds, so no boost.
it's not going to happen because it's used as a marketing ploy. it will only change if it's mandated by governments
@@teaser6089 I can agree that there's no need for it, but that doesn't mean that it wouldn't be a welcome change. A quality of life upgrade, one less thing to think deeply about when a standardized analysis could simplify something that doesn't need to be complicated.
I think the PSU comparison isn't quite apt, as the power draw stated is still a definite number. If you want more headroom you can go above, but you at least know your minimum. Seems like more of a grey area when it comes to TDP
ME: "deep dive"?? I thought almost every GN video was a deep dive
STEVE: Hold my cryonaut
you meant conductonaut , right ?
@@Anon20855 nah, hydronaut is better to deep dive.
@@NatureQuest2024 This is an extreme reaction. I get a bit lost too during this one, despite having no problem with buildzoid's videos. But that doesn't mean I think of unsub.
@@NatureQuest2024 well if you can't stand it , just go to a mainstream , dumbed down version of tech Jeesus like the LTT for example
Just go LTT if it makes your head hurt. I just take a beer while grading papers and listen to him in the background as a soothing quiet piece of rational thought i wish i could understand with any sort of competence.
10/10 for unexpected Limmy's show
I don't get it.
How did you not get noticed in eight months Limmy, I'm disappointed
Is ¤▪︎》•□ heavier than ○■°☆
you will in time, limmy. in time.
@@pravda9646 who is this Limmy guy?
@@ADCFproductions idk if im missing a joke but he's the guy in the video who says steel is heavier than feathers
This was long, This was made longer by how complex this piece was, having to go back and make sure I understood the concepts. This was very well needed in the community. Thank you Steve and Team! There will be a beer waiting for you @ The Corner bar soon haha!
Please use the GN beerglass in blue, with the golden Rim. Thx.
PC users:What is TDP?
Companies: *WeLl iT dEpeNdS*
iT DePends
@@zeroclockgaming said everyone when asked "Intel vs Ryzen"
KEKW
@@blueblade455 tech is like a religion
TDP = Total Dollar Pull on your wallet.
Gosh, this was done a couple of years ago. I have to say that Linus Sebastian did a great job explaining TDP, but You guys went above and beyond in explaining TDP. I commend you and your team for gathering the details that matter. I love that you're not afraid of formulas as well. True engineers at work. I love this channel, and so doesn't 1.7 million other people.
Can you please create an online test I can take after one of these lectures for Gamers Nexus University credit?? All grading of course in "X"s where more "X"s are better. I'd like to graduate fully ray traced with RGB honors and a minor in hair care. The Mod mat diploma suitable for framing will be greatly covenanted.
Grading scale is X to X! The more Xs, the better!
it's not a university course, it's required qualification training for working with Gamers Nexus National Laboratories, lol. (Mostly Kidding)
Speaking as someone with a bachelors in Electronic Engineering if I'd been able to take Gamers Nexus as an elective it would have been more useful to me than some of the mandatory courses I did have to take at uni. Guess how many times I've used Turbo Pascal since graduating back in 2000 for example. Or 6802 assembly.
totally agree Mister The Geoff. I've never touched another Apple II E+ or Apple-Basic since them days in school. I won't say it was totally useless for some 'concepts' that were taught, just that it could have been with something most of the industry uses instead of some niche system that is now a dead dinosaur.
P.S. thankfully for my mother (in regards to college), thankfully the professor worked at DEC, so some of the students would just go to the Professor's office at DEC to learn on a 'Real Computer' back in them days (1988/1989 ish). I was the tag-along kid doodling on napkins down in the DEC cafeteria with some other people there, while my mom and other students learned on a VAX. They had no idea what was brewing down in the cafeteria, lol.
@@Zarcondeegrissom Yeah if the Raspberry Pi had been a thing back when I was learning how to program a 6802 to play a tune I'm sure that we'd have been using them instead of some weird propriety thing that had been knocking around the lab since the 1980s.
Im literally blown away by how in depth your work went on this topic - Thanks.
Oh yes, a deep-dive.
My favorite
OMG...Limmy on Gamers Nexus?...never expected to see that.
It'll be Dee Dee fae Yoker next week! :D
!
Les Porter? Did your name always rhyme wae Yoker or did it used to be Smith or something?
Best day of my life...
Just put your cpu under wa'ah, pure wa'ah. problem solved.
Thank you for that GN, always nice to remember that video. :3
TDP = Thermals? Duh, Performance!
That's oversimplified
i like " thermal dependent performance" being AMD's new slogan!
As a computer engineer I really appreciate the deep dives like this that get into the technical details. I find it super interesting and I'm happy to see someone on the outside of these companies doing the research and reverse engineering to figure out how things like this work.
Amen
As an engineer, i wonder why the fundamental value of thermal impedance of the cooling assembly in °C/W never makes an appearance. That's how we normally figure out heatsinks for circuits, by constructing an impedance chain from die to heatspreader or thermal interface pad, from there through the heatsink to ambient air. If we can print the borderline useless value in W/mK on thermal paste, because distance to be bridged depends on a lot of things not the least of which is the viscosity of the paste itself, why can't we use a much more descriptive value for coolers?
TDP should be listed as how much energy you need to dissipate to keep it at a stable-safe temperature while maintaining maximum performance for prolonged period of time (not 10seconds)
power consumption should be listed separately
When "prolonged time" is brought into the picture, it means "steady state" in a thermal consideration, and with all steady state calculations, heat in = heat out, hence it is basically the same amount as power consumption. That means, once we know the power consumption for a CPU, it would help us better pick the cooler we need.
Do note there's a caveat to this: ALL coolers dissipate more heat the greater the temperature difference between them and the environment, i.e. a CPU running at 90C with an air cooler stuck on to it blowing ambient air at 30C would have a 60C difference, which would dissipate more power than if the CPU was running at 60C with an ambient of 30C which results in a 30C difference.
So it's also extremely important to know the target temperatures you're intending to run at - both your case air temperature or water temperature if you're using liquid cooled, and the CPU temperature, of which it gets more complicated because the CPU temperature readout in software is the die temperature and not the metal casing temperature of the chip, so there's yet one more piece of information to source for.
The stable-safe temperature depends on what the ambient temperature is + heat sink characteristics. IMO all they have to do is to list the maximum actual electrical power draw when under 100% load without throttling. As long as the temperature is within limits, it doesn't depend on the actual temperature value.
In a very... very vague sense, the purpose of something like the TDP would be to design for a certain electrical power upper limit based on what temperature limits must be maintained BUT it's still useless without knowing the variables an seems unnecessarily complicated.
Long story short, I think the TDP is just a marketing number. The formula is just an attempt to give it some credibility if someone questions it. The idea is that you'll see the artificial complexity and give up examining further.
I've only been a viewer for a month or so but your content is awesome. Thanks for your efforts.
Welcome aboard! Always great to have new viewers.
0:20 "Voodoo mysticism" - Anyone remembered 3dfx at that moment? Would be a nice retro review regardless of what Voodoo card...
Bruh if you had a voodoo card it was no problem to run Terminal Velocity
When a "deep dive" on any other channel is standard fare on this channel, deep really does mean deep. good work :}
OMG that 8700k looks like its been in battle!!!!
8700k Battle Scarred
Absolutely thrilled with this level of detail! Very happy to already being a patreon supporter!
Now I just need to get a mod mat...
"4x2 - in Freedom Units" - priceless!
Hey GN, since you wanted a definition on "amortize", I have one for you. In Computer Science and Engineering, amortization is used as a term to essentially mean an approximation of the best case scenario, with the knowledge it's guaranteed to have infrequent worst case performance.
Example: an algorithm implementation of a set will have an amortized time complexity of O(1) when adding a new item to the set, but if enough items are being added, you'll get a couple O(n) insertions into the set.
"approximation of the best case scenario" is not quite right, it's stronger than that. When we say amortized O(1) we mean that on average it is O(1), or in other words that we can guarantee that there will be enough fast operations to compensate for the slow ones in some way. In principle the "best case scenario" doesn't have to be the same as the amortized value, e.g. you could have best-case O(1), worst-case O(n^2) and amortized O(n).
i swear steve's hair just grow larger and longer the more i watch this video. cant tell if its a reflection of his growing judgment, his power, or both.
It’s a reflection of his TDP, being annoyed by misleading information. In order to continue the video the cooler quality, or thermal conductivity has to increase. The cooler in this instance being his hair of course. 😁
Sounds like the T in TDP should stand for take-this-with-a-grain-of-salt.
it really should
Thanks for putting it in black and white. Cleared some misconceptions and provided a lot of facts.
Teachers my whole life: never use wiki as a source
Steve at GN: shuts down a multi billion dollar company and its scientists with wiki
Teachers still believe anyone can edit it and make it say made up shit
I'm really glad you guys do videos like this one, I really enjoy being informed of shady practices of the vendors we rely on to tell us information about their own products. It is not easy to find this information online with any reliability of the authenticity. I appreciate the fact that your team takes the time out to actually check the details of the information provided for your videos. Your channel is amazing and hope you keep it up long term, I don't know where I would go if you disappeared.
Given AMD's formula for TDP, it seems to me that it makes perfect sense with how it behaves under LN2.
deltaT/HSF means that the TDP increases linearly as the ambient temperature falls.
Awesome! Thanks to GN for taking the time. More please!
When in doubt, go overkill on the cooling.
Not "when in doubt". Always go overkill on the cooling.
Squeeze that third fan on the Fuma 2
Macho, Ninja, and DH15 are minimums.
The beautiful part of modern processors is that you can't really melt them. If things get too hot, they just ratchet down the clocks and power draw (or just flat out turn off) to keep thermals under control. Therefore, your best bet is install a cooler, stress the CPU, and see if it throttles. If it doesn't throttle, your cooler is sufficient. If it does, get a bigger/better cooler, rinse and repeat. Sounds like a lot of work, but there's really not that much variation in overall cooler performance within product lines. As long as it's a reputable company, you can generally guess the model of cooler, based on the tier of the CPU. Something like a 3600 can use the budget model, 3700x would require mid tier, and 3900x would be top tier.
I'm AIO cooling my i7-6700T, if possible.
Well f-ing dissected! Thanks, GN for going this far down the rabbit hole in your research and eloquently sharing your results. Loads of work and we appreciate it!
"The CPU's electric consumption measured in watts is not the same as the CPU's dissipated heat measured in watts because our processors are so efficient that they convert most of the energy into calculations."
-- AMD
When an electron in our CPU finishes calculating, it happily ceases to exist, removing its energy from this universe and doesn't go on bouncing into the silicon like a bandit.
Had this on in the background... Man so much words that don't mean what they sound like... Conclusion: start with the anwser and adjust the formula accordingly. Will give it another watch tonight! Thanks for the indept pieces of content!!
I woke up in the middle of the night, sweating.... Tech Jesus must've released new content :D
you're doing huge job preparing each video, thank for that!
Is it bad that I can say it's a GN video just by glancing at the title in notidication?
What does "GN" mean? "notidication" This I don't think is a word?
@@Bass.Player Gamers Nexus and *in the notification
@@optiquest86 What are attempting to say? Try a complete sentence if you want me to understand...
@@Bass.Player This video was uploaded to the Gamers Nexus RUclips channel so we can most likely guess that 'GN' means Gamers Nexus. 'notidication' most likely means 'notification' as in a standard QWERTY keyboard layout the letter d is right next to the f key, meaning that he most likely pressed it by accident when typing.
@@JesusMeza3 Thanks
Excellent work guys. I think I would have just walked into busy traffic after spending so much time dealing with PR paperwork and Reddit.
A few words from the land of thermodynamics, to confuse anyone who isnt yet: Heat transfer is governed by thermal resistance (ie cooler and TIM performance) and temperature difference between heatsink and heat source. If a given CPU has a heat output of 100 Watts (For worst case look at the electrical power consumed, thats the cpus heat output.)
Now: There are two major heat transfers that you have to worry about: From the CPU to the Cooler (Via TIM), and from the cooler to the outside air.
Now comes the key part: The Heat flow (Measured in Watts) is proportional to temp difference. 20° Ambient, with the cooler being able to do 10W/°C, 100 Watts of Power into the cooler would mean a cooler temp of 30°C. Then, the CPU has to be hotter than the Cooler to put in 100 Watts into the cooler. Lets say, the TIM is awesome, with a conductance of 100W/°C (The actual formula works with resistance per Meter times the Length heat has to travel), Then the CPU has to be at 31°C to put out all heat.
If the ambient temp goes up, your temps rise linearly. If your Power Consumption rises, the cpu and cooler have to get hotter to transfer the new load.
You could push 500Watts through an intel stock cooler. But if the outside is 20°C, the cooler would be way hotter than cpu safe temps.
Last thing: watercooling allows to use bigger coolers than an air heatsink might allow (Space), and the advantage of dumping the heat to the outside air.
My cpu sits at 40°C under full cpu load. If i am gaming and my gpu increases case temp, the cpu goes to 60°C bc of more heat. Water can evade this problem. if done correctly unlike the 92mm tower aio going in the place of a tower heatsink. That thing is no better than a similar size air cooler.
It's ridiculous that when I work on my electronic projects I look into datasheets to find devices' Tjmax and junction to case thermal resistance, I look at thermal paste's thermal resistance and I calculate what heatsink I should get to achieve the target temperature at junction. But when it comes to building a PC all the engineering knowledge goes down the gutter because of plenty of misconceptions, misleading numbers some voodoo factors and marketing bullshit which confuses everyone. Can you imagine what if all electronics design companies had to measure and verify EVERYTHING on their own rather than just go to manufacturers' datasheets to obtain necessary information? That would be nuts.
This was a really great comprehensive video. The GN tours of cooler factories were awesome as well. Thanks!
The kilogram (mass) of steel is heavier (weight) than the kilogram (mass) of feathers, because steel is denser (density) than features, so buoyancy will make the features lighter (weight).
Exactly, Archimedes should not be forgotten.
38min! We need a TL;DW for this. But love the work you put into man.
Basically, Intel, AMD and CPU cooler manufactures all calculate TDP differently so Steve is saying look at CPU cooler manufacture websites for compatible processors for their coolers and review websites.
@@DamienWhite you are a legend mate!
AMD and Intel: He's turned the weans against us!
Steve: YOU HAVE ALREADY DONE THAT YOURSELF!
weans
My case is probably the most extreme of all I can think of. I'm cooling an overcklocked R9 290X with a TDP of 290w with a modified first gen arctic cooling freezer 7 pro with a TDP of 65w!
It was mostly an experiment that turned out to work great so now I'm using it on a daily basis. :)
Gotta say; mad respect for AMD for releasing this information and collaborating with you guys. Also mad respect to GN and the whole team for putting out this content; i question if any of you sleep honestly.
AMD confirmed specs, but they did not *release* it to consumers. The cat was out of the bag so the gave in and provided correct information to prevent [more?] bad press.
Goddamn. My brain queued up the Limmy clip, then you guys played it. Kudos ♥️
I've never been this early, but I suspect this video will be good.
Screw it, I know this will be good.
Thank you GN crew for continuing to evolve your test methodologies(insert Jayz pronunciation) and your "subtle" humor and ambivalence.. With such a small team: you've gotten attention from some of the biggest corporations in the world.
I never doubt that you will be neutral, regardless of the sponsor.Cheers
8:07 anyone with any knowledge in basic physics would know that two variables can very well measure different things and be different despite having the same unit of measurement. For example work (W) and energy (E) are totally different measurements but they have the same unit joules (J). So just because two things have the same unit, it doesn't necessarily mean they are both the same.
Technical ally, buy the energy used in a processor to switch transistors, and that is converted to radiation, is very small. It will be far below any user's or pro reviewer's margin of error.
Sound like you don't have a basic knowledge of physics then. A watt is a unit of energy. Energy can exist in different forms (kinetic, potential, thermal, electrical, etc). When you convert that energy from one form to another, you're doing work. The amount of energy that you convert is the amount of work that you do. Work is just energy, energy converted. Saying work and energy are different things is as stupid as saying a metre of distance and a metre travelled are different things or a dollar in your pocket is different from a dollar you actually spend. Yes, work and energy are the same thing. You're wrong, Gamers Nexus is right.
And if you continue watching the video, he explains (correctly) how all the electrical energy consumed by the CPU gets converted into heat (there's no such thing as a watt of computer calculations), so in the long run the electrical energy consummed by the CPU is going to be equal to the thermal energy dissipated.
Energy = joule. Watt = joule/second. Thermodynamics, as energy is converted from 1 form to another, some energy is lost as heat. In this case, practically all of it. Sure entropy also comes into the mix here somewhere reference 2nd law of thermodynamics
My head exploded about mid way through but I do like the kind of technical content you guys produce.
ACTUALLY, I needed this content...
32:15: nice save, Steve: I almost got the you were talking about Noctua... :-)
My guess is Noctua
AND TECH JESUS HAS BLESS US WITH HIS KNOWLEDGE AGAIN.
As someone who has been building and supporting pcs for 28 years, that's a useful video, almost to the level of adored. Thanks for clearing up some of the confusion on something that's not that obvious.
This is what happens when the marketing department roleplays as engineers
This is one of the best videos I have ever seen in my life about ANYTHING. I love you.
W doesn't stand for electricity, it stands for joules per second.
This can be any form of energy not just electrons.
Yeah. We said that. The reference to "electrical watts" was because of AMD's phrasing.
@@anasbinmuslim Would you like to point out how the hell a CPU is getting anything other than electrical power? And how it's outputting anything other than heat? Would you like to show us all how your CPU is turning electricity into light and kinetic energy?
The CPU takes electricity and only electricity. It outputs heat and only heat. We are able to get use out of *how* it changes electrical energy into heat energy because it's going through a CPU, but that doesn't change the fact that it only inputs electricity and outputs heat.
And if you want to get technical, sure, there could be some energy that turns into light or is dissipated in other ways, but it's extremely minor. The exact values of energy input to energy output may vary by an extremely small amount. That amount isn't large enough to truly make a difference though.
Hi Steve,
Thank you for making this video. As an electronics hobbyist I never understood why the computer world uses different parameters than the electronics world. In the electronics world we calculate the worst case power dissipation ( what would be the real TDP), and from the datasheet of the part you get max T junction en junction to (CPU)case parameters in C or K per watt. Then you set/decide your maximum ambient (inside the case) temperature and the maximum junction temperature you are comfortable with. From these paramaters you can calculate your heatsink or other cooling solution (also in Kelvin or Celcius per Watt). All these values can be found in normal component datasheets and heatsink suppliers, even for 10 cent parts. But these parameters are never found for CPU's or coolers. It's not that difficult.
PS. For above calcultions I would have no problem using the electrical power draw and forget about parasitic (pins etc.) conductivity unless I'm in a really tight spot.
Thanks, Jeroen
This is the comment I've been looking for! I myself used the formula countless times when picking component types, but was quite shocked to see it misused in such a way here.
Was not expecting a GN video to help me verify answers from my energy systems analysis test
The most interesting thing is that AMD assumes a case that maintains 42 C ambient in presumably a 30 C room. Time to start cranking up the heater for your AMD CPU reviews GN.
That's actually not a bad idea. Steve should do this to see how the two major companies (AMD/Intel) deal with less than ideal situations. The code that defines their performance characteristics in different conditions are written differently and therefore should yield interesting results.
AMD does not assume it. AMD's rated TDP is calculated at these measurements. If your case has a X temperature, ambient is Y temperature, and your heatsink is able to dissipate Z watts then the CPU will consume the amount stated in the TDP.
Needless to say, transistors in the silicon die change characteristics according to temperature and use different amounts of energy at different temperature. Also, the CPU might throttle according to your heatsink's performance at dissipating thermal energy.
You can't have a universal measurement of CPU power usage regardless of the environment it's tested in.
Given they've defined case in this as the cpu's heat spreader I'm assuming 'ambient' in this case is the temperature of the air around the heat sink inside your case, not the air in your bedroom outside the case. Which means for a lot of cases without decent airflow 42 C is pretty ambitious.
@@coinexchange1761 Heat transfer only depends on the difference in temperature between the heatsink and the air, not the absolute numbers
@@Mandragara I haven't said otherwise? Did you read what I wrote?
Thanks for making this video. In addition to TDP we have thermal capacitance to think of. We really need to have 2 separate value for HSF (heat capacitance and heat transfer per temperature delta). In CPU we also need 2 numbers: burst energy consumption and how long (power x time), and continuous power consumption.
I'm going to need AMD to provide me an amortization schedule before I'm willing to purchase one of their CPUs.
So from an electrical engineering point of view:
The formula is TDP = delta temperature / thermal resistance. This is very similar to equations electrical engineers use every day to specify hearsinks. You would design for a whole chain of thermal interfaces from the heat source (the die) to the ultimate heat sink (the air). Thermal resistances of each layer add up (die to IHS to paste to heatsink to air). The outcome of the formula in electrical engineering is a sentence like "the combined thermal resistance of the paste and heatsink must be less than X degrees C per Watt".
The problem comes as power is an integral over time rather than an instantaneous thing. ( i.e. Joules per second) and depending on the timespan (1 second? 1 hour? 1 year?) the power will be different because of things like turbo. So AMD and intel have to make a design decision as to what power rating to use in this specification, and this single estimation is called TDP. In EE, the delta of temperature from die to ambient would typically be conservative and use the maximum safe die temperature and approximate the worst case ambient air temperature. (As performance of a CPU changes dramatically with temperature, they could substitude "maximum safe temperature" with "optimum temperature for peak performance"). I can see why they'd specifiy a different ambient air temperature for server parts for instance. Theta of die to IHS is controlled and know by AMD/intel. Theta of IHS to air is not and therefore they must make this recommendation to a heatsink manufacturer.
I think the whole thing would be much simpler if AMD and intel therefore simply specified a maximum theta_ca for 3rd party heatsinks and kept the estimated "thermal" power used in the equation to themselves. Or better still, specify a range of theta_ca (based on a range of ambient temperature and a fixed optimum die temperature) which the heatsink+fan should completely cover by way of its variable fan speed.
Oh, there's that Pentium II Klamath CPU trying to impersonate a heatsink!
The small cpu heatsink at 29:16 is a similar one to which I placed on to the backplate of my gpu with tim and the fan plugged into mobo sys fan 3 socket, and another bigger heatsink on another part of gpu backplate with fan into mobo sys fan 2. This heatsink warms up and with its small fan, reduces gpu temps a few degrees, through the backplate that would usually get quite hot. This not only makes use of left over stock cpu heatsinks, but also helps to increase gpu oc or quieter operation with fans on low speed
And yet power supply manufacturer's all reference TDP when recommending you what size PSU to buy....
This is why my rule of thumb is to always at least double the TDP of the components when buying a PSU. That, and it runs cooler, quieter, and more efficient.
@@MrStealYoBeef right, but if you look at for example Seasonic one of the biggest PSU makes, on their website they recommend to use 90% of TDP - which sounds completely wrong based on Steve's investigation.
@@mfdsuk Seasonic PSUs have been known to be insanely overbuilt though. My SS-X650KM3 can push upwards of 800w despite being a 650w PSU. Of course I never load it that hard, but it is interesting to know.
That was useless even back when AMD and Intel used realistic TDPs.
@@MrStealYoBeef PSUs have efficiency curves, they are most efficient at around 50% load, and are way less efficient when considerably underloaded.
Great research and expose, Gamers Nexus! TY for your diligence!
Kinda hoped that there going to be a chart with results of different cpus and cooler solutions.
I wanted a giant cooler beatdown
Different video
I'm sure they are trying to figure out how to properly review CPU coolers. Maybe something that doesn't require using it in a system to see how good it works.
Really great article!
I try to see the power consumption of parts before picking coolers!
Making a build perform at the level you want is so complicated its not only about tdp and cpu cooler, its about noise, ventilation ,clocks and volts!!!!
8:24. "what's heavier a kilogram of steel or a kilogram of feathers". the answer was steel? lol lol lol what is that video from? Limmy's show, which? hahahaha
ruclips.net/video/Qwr6_015ROo/видео.html
Great video guys! Tons of great info in here!
I could listen to Steve talk four hours.
GN: thank you so much for this work, today I’ve really learnt something as I always thought you could compare TDP between AMD and Intel, oh how wrong as a consumer I was, and this is the fundamental problem average consumers wouldn’t have a clue, this is seriously good work 👍
You're never going to get an exact match for TDP when the CPU is using that for its own internal operation for frequency variables whilst coolers are looking at how well the cooler can dissipate the heat from its base at a specific set of numbers.
Thermal dissipation of a heat sink is variable depending on conditions and the delta between the base and ambient temps required. Higher cfm of air flow for dissipation is not linear and has diminishing returns, so any given heatsink's TDP changes depending on the fan attached and the positive air pressure at the fan intake which changes between open air and a cases air flow and ambient (hot boxes vs HAF case).
Cooler manufacturers can make it easy by setting a standard for measuring thermal dissipation with a set of fixed variables. E.g. The ability to maintain a 30°C delta @20°C open air @ the fan @70% RPM. There once was a kind of defacto standard before CPU's started getting fancy with TDP variables when they only had rudimentary thermal sensors (edge temp only) with turbo fixed by guesstimates of thermal junction temps in relation to edge temps. Accordingly we've seen cooler manufacturers change cooler ratings over time which conflates the ability to know which cooler you should buy.
As CPU temps have been allowed to go higher, so has the delta increased which in turn changes what the thermal dissipation of the cooler actually is. A delta of 20°C (i.e. cooler base @40°C @20°C ambient) TDP is lower than a delta of 40°C (i.e. cooler base @60°C @20°C ambient). Thus a cooler once rated @120W TDP is now rated @160W TDP, though nothing has changed except the allowed delta between base and ambient air temps. There's also a delta variable between heat spreader temps and cooler base due to the thermal paste/interface medium used and its thickness. So where a cooler once was rated to maintain a CPU edge temp of ~45°C @120W TDP, they're now rated to maintain a 60°C edge temp @160W TDP. Same cooler new variables and a whole bunch of confusion.
Personally I just overkill the cooler for the conditions as I prefer lower temps and quieter fan speeds whilst giving headroom for a decent amount of manual overclocking which turns off the CPU's thermal and power algorithms leaving just tjmax cutoffs.
I love the longer format videos. I'm always looking for more depth in tech tuber videos.
TDP = the biggest CLC that will fit in my case.🤫
Same for me, but instead it's the biggest heatsink I can shove in a case.
@@rcradiator it's not about the tool, it's about the size. It's always about the size....
@@integer0verload948 its not about the size... Stick faster rotating fan on the cooler and yoy get higher tdp just by changing fan, ofc the heatsink itself must be same sized prior fan change. So no, 95 tdp can mean bigger cpu cooler but can mean that you can get smaller one but much more noisy fan.
@@fulldnbboy lol yeah just getting it blown on harder works also.
Great job steve even though occasionally you go over my head i always feel like i've walked away more informed you're doing the consumer i huge service.
Damn, going to have to watch some air cooler reviews. Sure wish a trustworthy source would do some scientific testing with all these noctua competitors coming out recently *cough*
Noctua has no competitors. Once they go RGB they will be unstoppable.
@@Ty-ri7dy
Dunno, man, you can buy a Chinese cooler for something like a quarter of the price that will keep your CPU only a couple of degrees warmer than a Noctua and that may be a couple of decibels louder. Your CPU won't give a shit and, unless you use your PC in a soundproof music studio, your ear won't even notice. I've got nothing against fanboyism -- and in the cooling world Noctua seems to be the equivalent of Apple -- but the majority of us don't need the absolute best and prefer to spend our money more wisely. Each to their own, I suppose.
I just keep buying Corsair liquid coolers and keep having a good time, just use a program like COREtemp to watch your C's
@@chesshooligan1282 arctic is p good
I really appreciate that GN brings in some basic physics and calculations. I like that people at GN understand physics and GN is actually one of few channels that I trust with formulas and calculations. Thank you very much for this video!
Steve: There are no RGB CUPs , Yet
ME: *FACE PALM*
Also Me: Dont give them any ideas
Great video, lots of mechanical engineering thought went into this. Looks like my next air cooler will be something your channel or the manufacturer recommends. Nicely done!!
This is a smart piece.
Thank you Steve that was a very educational piece for concerned, including the edification of the industry.
AMD: A watt is a wstt.
Intel: It is what it is, good luck.
"There's no RGB CPU's, yet" good God if that ever happens I will shoot myself.
promotions.newegg.com/nepro/19-0630/index.html
Newegg will like to have a word with you ;P
Oh it's coming lol. Actually I could see them add RGB to the edges of the CPU so it lights up from under the heat sink. With that will come yet ANOTHER standard so you need yet more special software. Intel and AMD will of course have their own standards too. I feel dirty just for thinking this.
The industry needs a consistent power consumption measurement. Whether TPD or a new acronym. Great work guys 👍
Was that guy serious about the steel and the feathers
No he's a scottish comedian called limmy.
It's from a funny video. The dry delivery is great!
YES, THANKYOU
Heat generated in watts = Electricity consumed in watts - a negligible amount of work done
and to keep it att the wanted temperature you will need a cooler that:
Dissipates "Heat generated in watts" @ Desired max temperature and ambient temperature
"A kilogram of feathers in Watts!!!"
10/10 would watch again. You guys are awesome, well worded and super interesting. Thank you!
The second definition you gave from AMD is from engineering guidelines. That's not some super secret definition to trick consumers, it's telling the engineers what the target transfer rate should be for designing AMD coolers. In other words, they test the thermal output under heavy load at stock voltage, and then design a cooler to handle those conditions.
The first definition is for the consumer, as guidance for choosing a cooling solution.
Nothing nefarious going on there.
so you're telling me that a 12 core Ryzen 9 3900x and an 8 core Ryzen 7 3800x running at roughly the same frequency (4.5-4.6 ghz) actually output the same amount of heat? Because that's what their TDP is essentially showing. 105w TDP for both of those processors. One of these processors clearly has 50% more cores, and frequency is roughly the same as well as core architecture. Are you saying that the number of cores has nothing to do with heat output?
And you're telling me that there's nothing nefarious going on here?
@@MrStealYoBeef One versus two chiplets, and binning.
The silicon itself is one of the biggest bottlenecks in cooling, if you spread out the cores across two dies instead of one, even if you have 50% more cores, you will likely have lower temperatures simply because the thermal conduction is a function of both the thermal conductivity of a material and it's crossection. Double the crossection means double the conduction. The temperature at the thermal interface and the temperature in the die are wildly different, so it is possible that even with a higher temperature at the heatsink IHS junction, you will still see a lower temperature inside the die itself.
Is it kinda bullshit? Yes, absolutely. Is it probably accurate for what kind of cooling solution you will need? Yes, probably.
I do wish TDP was strictly defined as Watts at a specific dT over ambient. This would be far more useful, since the specifics of how a cooler will interact with each CPU is not at all modeled via the numbers.
In conclusion: we still don't know what TDP is, but it's pretty much useless on the cpu vendor side, better trust cpu cooler manufactures, they also don't know what TDP is, but have decades of experience preventing your cpu from turning into a frying pan.
Loved the video, thanks for making it.
Absurdly fantastic video! I would LOVE to see more like this!! Thank you all so much for putting in 3 WEEKS to do all the research and make this wonderful video!
I think that part of the problem is that we're not seeing guaranteed all-core boost clocks printed on the box, because that is the only way you could tie a specific "stock settings maximum power draw" number to a SKU. They purposely don't want to specify a guaranteed per-core-count minimum boosting behaviour for the SKU(they want to list only the higher numbered single-core boosting behaviour because it looks better on the box), so a stock maximum all-core power draw can't be established. The root of the problem is that they should be listing guaranteed all-core clocks on the box, not just guaranteed single-core clocks.
Because vids like these have Huge Respect for this channel
As an accountant, the best part about this video is watching Steve stumble over the word “amortize” and turning to Wikipedia for a definition. Also, his frustration with the fact that a formula can be solved for any of the variables within it. Great vid 👍
We have 50% viewership outside of English-speaking (primarily) countries. It is actually important to define. It wasn't for me. I use accountants too, man. You must know that running a business would have you running into accounting terminology quite frequently.
A simple way to put it would be "Power draw at steady state and maximum CPU utilisation".
We could also have single, quad, and manycore TDP ratings. It'd be great.
Truly excellent stuff... I cant believe its the same GN that botched a lapping test.
When developing products that have plastic injection molded parts the tooling cost ($5k-30k) is often amortized into the part cost. I learned the term "amortize" when I started developing products.
After about 2 mins into this video I lost any clue what anything really meant that was talked about but kept watching.