Core i9-9900K Power & Thermals, Did Linus (and OC3D TV) Get it Wrong?

[Carclis]

23 minutes ago, mr moose said:

But that's just it.   Intel have tested the chip and know it can get as high as X Freq (calling it the boost freq), but they cannot guarantee it will stay there because they have no control over the cooling solution, the load type or how clean the power is and so on.   That is why TDP is rated at base clock.  You buy a processor that runs at base, It is advertised to be able to boost to X freq. for short periods of heavy use,  but if you are going to force the chip to run in boost all the time then it is expected you know what you are doing.  It's like a car, the manufacturer guarantees the engine to run within it's RPM limit, if you remove the limiter and over rev the engine, is it now false advertising because the engine uses more petrol? is it now the manufactures fault that your tank is too small for longer trips?

That's kinda the issue I have though. I know it's a similar case with other products but I'd rather keep it within PC hardware. With older Intel parts such as the 7700k or even the 1080ti you could expect that both the TDP and boost clocks would be adhered to ie the boost/turbo frequencies were implemented with some level of sanity and respect for the TDP of the product. That's why the 1080ti (and all Pascal cards) would boost past the advertised boost frequency and meet the power target at the same time. If we look at the 9900k in particular you could have all of the things you mentioned under control, but the CPU would still throttle because the quoted TDP is unrealistic for the turbo frequency on the spec sheet.

I'd personally prefer that Intel just made 150w the spec, allowing the CPU's to perform higher and the product to be more transparent. It would also mean that Z390 boards would be of a higher standard and we'd be far less likely to see 4 phase boards and VRM throttling. At the end of the day this just looks worse for Intel because as is the case with Linus's benchmarks, it showed the CPU to be performing worse, as it did in all of the DX12 tests.

[mr moose]

2 minutes ago, Carclis said:

I'd personally prefer that Intel just made 150w the spec,

How would engineers work out the required cooling solution if the spec was just a high number?  They would have to do extensive testing themselves or laptop, custom and SFF coolers would all be best guest and might result in macbook like throttling issues.

 

TDP is carefully measured to give engineers the data they need to implement the product.  If Intel just make it 150W then PC builders would be putting in 150Watt coolers unnecessarily increasing the price of the product to avoid being liable.

[Carclis]

2 minutes ago, mr moose said:

How would engineers work out the required cooling solution if the spec was just a high number?  They would have to do extensive testing themselves or laptop, custom and SFF coolers would all be best guest and might result in macbook like throttling issues.

Well at this point it's a lottery draw whether or not your board defaults to a more "realistic" TDP or throttles as is the case with Linus. So people have no way of knowing whether they will throttle anyway. If you're running a 150w in SFF or a laptop you'd have to be kinda crazy. That's why more often than not those builds have mobile parts instead. The Macbook example you mentioned is just outright poor design.

7 minutes ago, mr moose said:

TDP is carefully measured to give engineers the data they need to implement the product.  If Intel just make it 150W then PC builders would be putting in 150Watt coolers unnecessarily increasing the price of the product to avoid being liable.

Yes, exactly. Because right now people are putting 95 watt coolers on a part that the default bios often opens up to 150W, or some number far above 95W. The cost increase would be a shame but lets be honest here. The 9900k is by most measures not a competitive product. It's already unreasonably expensive and bumping up the TDP (it is an i9 after all) would simply ensure a more consistent experience.

[mr moose]

3 minutes ago, Carclis said:

Well at this point it's a lottery draw whether or not your board defaults to a more "realistic" TDP or throttles as is the case with Linus. So people have no way of knowing whether they will throttle anyway. If you're running a 150w in SFF or a laptop you'd have to be kinda crazy. That's why more often than not those builds have mobile parts instead. The Macbook example you mentioned is just outright poor design.

Yes, exactly. Because right now people are putting 95 watt coolers on a part that the default bios often opens up to 150W, or some number far above 95W. The cost increase would be a shame but lets be honest here. The 9900k is by most measures not a competitive product. It's already unreasonably expensive and bumping up the TDP (it is an i9 after all) would simply ensure a more consistent experience.

So domestic people are not doing their homework.  Remember it is a technical spec for engineers, you are an enthusiast who is moving out of the domestic realm and into the more technical engineering,  that means you must do your homework and use the TDP supplied and calculate what size PSU/cooler you will need based on what you are going to do with the CPU.  

 

I dare say many people on these forums (and similar forums) walk the grey line between domestic user (who doesn't need to know the TDP or what it means) and system builder who should take the time to understand it and do the math.   The reason we have so many debates about it is because there are too many people who think they understand it and take every opportunity to use it to flame up their personal ideologues that it becomes a narrative and new members read it and continue the misinformation. 

 

[leadeater]

3 hours ago, schwellmo92 said:

Generally GPU rated boost clocks are available within their stated TDP (usually even higher boost than advertised), that's the difference here.

 

I'm not saying Intel are wrong, they obviously have some clever marketing but it is definitely misleading.

True, at least with GPUs the board is rated at a TDP rather than the GPU die itself. You can raise and lower the board limit but the stated specs are all achievable within the stated TDP. Intel CPUs cannot, now we can all complain about that but the problem comes down to wanting the TDP to be something it's not. If you grab a 9900K and put it in a very basic OEM system it'll have a sustained power draw of 95W so that system OEM uses that stated TDP to pair the correct cooling solution. That same system OEM is free to increase either the cTDP or the Turbo Boost Window TDP, or both however you are knowingly doing that and will again have to pair the correct cooling solution for this non default setting.

 

Where it gets hard is us, PC gamers and builders, who buy a motherboard not knowing what the default cTDP is or honestly not really having to care in the past. This is a new product, with new behavior and now suddenly the TDP actually matters. How are we supposed to know this? The reviewers is the simple answer to that, and they have done their job thus far. We are now aware of it, we are now discussing it and we are learning new things.

 

This doesn't make Intel's TDP incorrect or misleading, under stock configuration the CPU will use 95W sustained as stated. The default Turbo Boost TDP is too low though, the default 120W as we have seen appears to be lower than what is required for the stated Turbo frequencies. 

 

Basically the TDP is not misleading, the Turbo frequencies are.

[Carclis]

55 minutes ago, mr moose said:

So domestic people are not doing their homework.  Remember it is a technical spec for engineers, you are an enthusiast who is moving out of the domestic realm and into the more technical engineering,  that means you must do your homework and use the TDP supplied and calculate what size PSU/cooler you will need based on what you are going to do with the CPU.  

 

I dare say many people on these forums (and similar forums) walk the grey line between domestic user (who doesn't need to know the TDP or what it means) and system builder who should take the time to understand it and do the math.   The reason we have so many debates about it is because there are too many people who think they understand it and take every opportunity to use it to flame up their personal ideologues that it becomes a narrative and new members read it and continue the misinformation.

If I'm completely honest, I agree with what you're saying. But I'm sure Intel is more than happy to be able to slap a 95W and 5Ghz sticker on it and be done with it, because at the end of the day motherboard manufacturers will most likely remove the power limits anyways. It means Intel can basically wash their hands of any responsibility and still get to claim 95W and 5Ghz. It's all just a bit silly but I don't expect it to change.

[mr moose]

8 minutes ago, Carclis said:

If I'm completely honest, I agree with what you're saying. But I'm sure Intel is more than happy to be able to slap a 95W and 5Ghz sticker on it and be done with it, because at the end of the day motherboard manufacturers will most likely remove the power limits anyways. It means Intel can basically wash their hands of any responsibility and still get to claim 95W and 5Ghz. It's all just a bit silly but I don't expect it to change.

I agree, Although I have a formal education in electronic design principals, so it doesn't really stand out to me as anything major.  It's really just another spec along with all the others when designing a circuit.  

[schwellmo92]

1 hour ago, Carclis said:

If I'm completely honest, I agree with what you're saying. But I'm sure Intel is more than happy to be able to slap a 95W and 5Ghz sticker on it and be done with it, because at the end of the day motherboard manufacturers will most likely remove the power limits anyways. It means Intel can basically wash their hands of any responsibility and still get to claim 95W and 5Ghz. It's all just a bit silly but I don't expect it to change.

Why not just have a 1GHz base clock at 10w TDP while we’re at it?

[Carclis]

2 minutes ago, schwellmo92 said:

Why not just have a 1GHz base clock at 10w TDP while we’re at it?

Yeah, with a 5Ghz single core boost.

[mr moose]

19 minutes ago, Carclis said:

Yeah, with a 5Ghz single core boost.

and thrusters for private interstate travel.. 

[leadeater]

1 hour ago, schwellmo92 said:

Why not just have a 1GHz base clock at 10w TDP while we’re at it?

Because after 22* seconds the CPU would drop to 1Ghz and be drawing 10W which would make games unplayable, but I'm assuming that was more a joke .

 

Seems the correct boost window spec for 9th Gen is 22 seconds.

[asus killer]

4 hours ago, Carclis said:

I'd personally prefer that Intel just made 150w the spec, allowing the CPU's to perform higher and the product to be more transparent. It would also mean that Z390 boards would be of a higher standard and we'd be far less likely to see 4 phase boards and VRM throttling. At the end of the day this just looks worse for Intel because as is the case with Linus's benchmarks, it showed the CPU to be performing worse, as it did in all of the DX12 tests.

board manufacturers know what they are building for but they are also building board for several CPU's, not just one. Should all z390 boards be spec'd for 9900k? sometimes people just want pcie lanes and not the highest clocked CPU. 

People that want a 9900k and go for the cheapest board are morons. I really can't see a problem here.

 

And apart from stock speeds, TDP will vary according to how much you want and can push the GPU

[Carclis]

5 minutes ago, asus killer said:

board manufacturers know what they are building for but they are also building board for several CPU's, not just one. Should all z390 boards be spec'd for 9900k? sometimes people just want pcie lanes and not the highest clocked CPU.

Well if you want even the 9700k the board would have to be equally equipped. Only the 9600k would be fine with a lesser board and even then it's socket compatible with Z370 boards, many of which will be better than Z390 boards still. Mind you the 9600k is not much different from the 8600k amyways. Bios flashback is a pretty painless process making this a perfectly valid option. If you want more PCIe lanes you won't be buying into this platform as it doesn't offer anything over previous ones.

13 minutes ago, asus killer said:

People that want a 9900k and go for the cheapest board are morons. I really can't see a problem here.

Yes and no. Z390 boards were only made with 3 chips in mind and only one of them is a six core. It kinda makes sense that they should all be decent at the very least. I still don't think the buyer should skimp on the board however since the CPU is already absurdly expensive.

15 minutes ago, asus killer said:

And apart from stock speeds, TDP will vary according to how much you want and can push the GPU

Well... it is an overclocking platform

[asus killer]

21 minutes ago, Carclis said:

Well if you want even the 9700k the board would have to be equally equipped. Only the 9600k would be fine with a lesser board and even then it's socket compatible with Z370 boards, many of which will be better than Z390 boards still. Mind you the 9600k is not much different from the 8600k amyways. Bios flashback is a pretty painless process making this a perfectly valid option. If you want more PCIe lanes you won't be buying into this platform as it doesn't offer anything over previous ones.

Yes and no. Z390 boards were only made with 3 chips in mind and only one of them is a six core. It kinda makes sense that they should all be decent at the very least. I still don't think the buyer should skimp on the board however since the CPU is already absurdly expensive.

Well... it is an overclocking platform

you are considering everybody makes an ideal decision to pair the best cpu and gpu match, i fell like that is only sometimes the case. I follow some local stores on facebook, they share customers builds and oh boy you see the weirdest of choices. You can buy a Z390 and a i5 whatever non OC'd so later you can upgrade. Buy a 1080ti to play fortnite on a 60hz display. etc...

The point is manufacturers and even retailers most of the time couldn't care less for customers weird choices, if someone makes a stupid choice and goes for the top mobo for no reason, nice for them.

Exactly because it's an OC platform that you make my point, you can go from someone slapping a pentium with a intel stock cooler, to a 5000$ custom loop or a chiller, or a nitrogen OC, so how to define TDP on a range like this?

[porina]

10 hours ago, leadeater said:

They do note below during testing that it actually uses 168W for all core turbo, wonder where that 210W is documented though. I imagine other motherboard makers set their cTDP to that value, need someone with one of these Z390 to confirm what it is set to.

Just to add some more confusion...

 

 

 

[XenosTech]

16 hours ago, The Benjamins said:

Honestly I don't like the disconnect of car people vs engineers on their stance on the CVT.

For performance and efficiency the CVT is the best type of transmission.

 

But Manual is the most fun.

Yeah manual is more fun for sure. I just don't like automatics because I'll fall asleep behind the wheel in traffic >.>

[Taf the Ghost]

Seen on the Twitters:

 

[Dylanc1500]

1 hour ago, Taf the Ghost said:

Seen on the Twitters:

 

We know who the true enthusiasts are. I'd like to know how frequent we are talking though. The people doing that are going a bit over board on the chiller. They could get away with a 1/13th horse chiller and still have capacity to spare (not to mention they can be purchased for about $350).

[leadeater]

7 hours ago, porina said:

Just to add some more confusion...

Oh FFS, knew I should have trusted the Intel XTU screenshot I actually saw showing 120W for the Turbo Boost Short Power Max.

 

Source: Hardware Unboxed Video

 

So maybe the 22 seconds someone else said is not correct either, or rather not Intel default. The above image is using Intel SVID so everything should be actually stock/to Intel spec.

[Taf the Ghost]

10 minutes ago, leadeater said:

Oh FFS, knew I should have trusted the Intel XTU screenshot I actually saw showing 120W for the Turbo Boost Time Window.

 

Source: Hardware Unboxed Video

 

So maybe the 22 seconds someone else said is not correct either, or rather not Intel default. The above image is using Intel SVID so everything you be actually stock/to Intel spec.

Highly varied motherboards appears to be the big driver in a lot of the power consumption figures. Beyond just the Stock vs some form of overclock.

[leadeater]

12 minutes ago, Taf the Ghost said:

Highly varied motherboards appears to be the big driver in a lot of the power consumption figures. Beyond just the Stock vs some form of overclock.

Yep, there really needs to be a consensus on what "stock" actually is because right now it's possible to go read 3 or 4 different reviews using different motherboards and "stock" CPU settings are different. If they are all optimizing the CPUs in different ways i.e. cTDP, Turbo TDP, Turbo Length, vCore then labeling results as "stock" it's a bit useless. Stock for the motherboard sure but it's not stock for the CPU and it's not going to match the Intel spec sheet, comparing reviews becomes harder.

 

That's why I agree with setting the SVID to Intel default, or anything else required, to make the CPU run exactly to Intel spec then every review will have an accurate representation of stock performance regardless of motherboard. After that pull out non stock/OC results because we need to know those as well.

[Trik'Stari]

On 10/22/2018 at 5:21 AM, Spotty said:

Response issued by LTT is pinned to the top of the comments for their video [on youtube].

Say what you will about Linus, but he responds to serious forms of criticism with transparency.

 

Gotta respect that.

[zassou]

on a slightly unreated note, the comment section on this video is quite pitchfork-y.

im not really into latest ltt videos but the way steve presented it is really drama-y and can draw unwanted hate.

[HashBrowns]

Still over here on my i7 2600 and in heaven. Not sure when I will  be incentivized to upgrade... 9900k is insanely expensive, so is ram...my lord

 

 

[MMKing]

1 hour ago, HashBrowns said:

Still over here on my i7 2600 and in heaven. Not sure when I will  be incentivized to upgrade... 9900k is insanely expensive, so is ram...my lord

 

 

Sitting on the 4770K myself. At this point, i'm probably gonna end up skipping DDR4 altogether.