Intel's Comet Lake Absence at CES Reportedly Related to Power Consumption Wall

[Turmania]

We did not have this confusion before turbo boost days. To be honest, I prefer the old style no turbo boost. Then we can overclock or not at our own risk and or preference

 

[hat]

Turbo isn't really that complex. IMO, more than a performance enhancer for the end user, it's more of a gimmick for Intel (and now AMD) to advertise really high clock speeds, under certain conditions*(asterisk to the 10th power). Just like with manual overclocking in the old days, inferior cooling and power delivery can get in the way of Turbo being effective. In other words, you don't put a 9900k in a $50 motherboard under a $10 cooler... just like you wouldn't put, say, a Q6600 in a budget board with the stock cooler and try to get to 4GHz.

Of course, you can always adjust the Turbo multipliers and voltage yourself. That's what I did with my 2600k. What was once a 3.8GHz single core Turbo chip is now a 4.4GHz all core Turbo chip (with more room, if I had a better cooler). The disappointment comes when we're used to seeing huge overclocks starting with C2D all the way up to Sandy Bridge. Ivy Bridge was okay, but that's where the disappointment started to set in. Now, with Coffee Lake and above, and its AMD counterparts, there's not much room left to overclock yourself.

So yeah, it's mostly a marketing tool to put big, impressive numbers on the box. To reach those numbers you essentially have to build a system like you would in the old days if you intended to overclock. Decent board, good cooler. That's why the 9900k goes way above 95w... they overclocked the snot out of it, much like the popular i7 920 would draw copious amounts of power when it was pushed. Think of it like horsepower ratings... my Honda Civic is rated at 115 HP... at 6100RPM! I've never pushed it beyond 4500 or so, even in first gear.

 

[Valantar]

Turbo isn't really that complex. IMO, more than a performance enhancer for the end user, it's more of a gimmick for Intel (and now AMD) to advertise really high clock speeds, under certain conditions*(asterisk to the 10th power). Just like with manual overclocking in the old days, inferior cooling and power delivery can get in the way of Turbo being effective. In other words, you don't put a 9900k in a $50 motherboard under a $10 cooler... just like you wouldn't put, say, a Q6600 in a budget board with the stock cooler and try to get to 4GHz.

Of course, you can always adjust the Turbo multipliers and voltage yourself. That's what I did with my 2600k. What was once a 3.8GHz single core Turbo chip is now a 4.4GHz all core Turbo chip (with more room, if I had a better cooler). The disappointment comes when we're used to seeing huge overclocks starting with C2D all the way up to Sandy Bridge. Ivy Bridge was okay, but that's where the disappointment started to set in. Now, with Coffee Lake and above, and its AMD counterparts, there's not much room left to overclock yourself.

So yeah, it's mostly a marketing tool to put big, impressive numbers on the box. To reach those numbers you essentially have to build a system like you would in the old days if you intended to overclock. Decent board, good cooler. That's why the 9900k goes way above 95w... they overclocked the snot out of it, much like the popular i7 920 would draw copious amounts of power when it was pushed. Think of it like horsepower ratings... my Honda Civic is rated at 115 HP... at 6100RPM! I've never pushed it beyond 4500 or so, even in first gear.

That's a rather pessimistic view. Turbo/boost modes also improve real-world performance quite dramatically after all,especially in bursty and responsiveness-sensitive workloads like application loading. Pretty much all modern CPUs also run all-core turbo above base clocks even when strictly power limited to TDP and equipped with a shitty cooler. So without any form of turbo we would both se massive drops in responsiveness due to no short peak turbos, but also slower overall performance.

Of course SKUs would change due to this, but the changes would be to the worse - high end chips would be scarce top bins or have high TDPs to ensure sufficient chips meet the clock speed target within TDP. This would also likely lead to a lot more SKUs as the range from best to worst chips would grow, making lineups more messy (and top end chips likely more expensive). And even then you wouldn't come anywhere near modern boost speeds as no cores would clock beyond the highest all-core speed within TDP regardless of power or thermal headroom.


Of course this would make for improved overclocking (simply because the entire point of boosting is to leave less performance on the table) but it would still be a significant net loss to the vast majority of users. Let's please abandon this foolish nostalgia back to when our PC components were dumber, less dynamic and overall much more poorly managed than today.

 

[hat]

That's a rather pessimistic view. Turbo/boost modes also improve real-world performance quite dramatically after all,especially in bursty and responsiveness-sensitive workloads like application loading. Pretty much all modern CPUs also run all-core turbo above base clocks even when strictly power limited to TDP and equipped with a shitty cooler. So without any form of turbo we would both se massive drops in responsiveness due to no short peak turbos, but also slower overall performance.

I agree, it is a pessimistic view, but that's how I see it. Of course you can get an all-core turbo "above" base clock, but it may not be the max turbo allowed without the rest of the system to support it. Of course, this is an extreme example, but take the 9900KS: it's specifically advertised at 5GHz all cores. You won't see this without an adequate board and cooler. Without a beefy infrastructure to support it, that high turbo goes away... and the same is true for any chip equipped with Turbo. See any laptop for an example. Of course, (most) laptops aren't designed to be powerful machines like their desktop counterparts, but it does illustrate the fact that you won't always (sometimes rarely, even) see those turbo speeds without the rest of the system being good enough to support it.

Of course SKUs would change due to this, but the changes would be to the worse - high end chips would be scarce top bins or have high TDPs to ensure sufficient chips meet the clock speed target within TDP. This would also likely lead to a lot more SKUs as the range from best to worst chips would grow, making lineups more messy (and top end chips likely more expensive). And even then you wouldn't come anywhere near modern boost speeds as no cores would clock beyond the highest all-core speed within TDP regardless of power or thermal headroom.

TDP is defined in firmware: the same software that tells my system my CPU is a 2600k, and tells said 2600k what speeds to run at and when. In other words, it's just a number. No additional binning would be required to say the 9900k is a 200w chip. Or, rather, they could come out and define two TDP values: one for base clock, and one for Turbo. As long as the Turbo TDP is defined correctly, there's no reason a chip like that would not be able to contend with current chips that lack such a definition. In fact, at least in the case of the 9th generation, a Turbo TDP does actually exist, it's just not advertised:

In this case, for the new 9th Generation Core processors, Intel has set the PL2 value to 210W. This is essentially the power required to hit the peak turbo on all cores, such as 4.7 GHz on the eight-core Core i9-9900K. So users can completely forget the 95W TDP when it comes to cooling. If a user wants those peak frequencies, it’s time to invest in something capable and serious.

-Anandtech review

Of course this would make for improved overclocking (simply because the entire point of boosting is to leave less performance on the table) but it would still be a significant net loss to the vast majority of users. Let's please abandon this foolish nostalgia back to when our PC components were dumber, less dynamic and overall much more poorly managed than today.

Given my responses above, no additional SKUs nor binning are required. Without a supporting system in place designed to take maximum advantage of Turbo, such as a system one may have built in the old days for manual overclocking before Turbo existed, Turbo remains to be a marketing tool to advertise big, impressive numbers that aren't really obtainable. And, to be honest, I find the argument that I can launch light programs or load webpages faster by being at 5GHz for a split second (because that's all my inadequate motherboard and/or cooler could manage, as is the case in many prebuilt systems and laptops) to be a fallacy. Not only does it not refute my original argument, but it also doesn't carry much weight on its own. Who cares if the 9900k can load this TPU page .1 seconds faster than my old i7 920 (even at stock) if the performance isn't there when the time comes to run a heavier application when the performance would actually be appreciated?

Finally, to wrap up my argument, I'm not trying to say that Turbo, in itself, is bad. I'm just saying it's badly advertised. To me personally, it makes about as much sense as a commercial that might advertise the newest Mustang, going nice and fast with impressive handling around sharp curves or other potentially hazardous conditions... with the disclaimer professional driver on a closed course etc etc... it doesn't take into account that a normal person can't actually drive their car like that unless they are also a skilled driver on a closed course. Otherwise, you'll probably crash into something, or you'll run into other trouble when you are running from the police. Or you'll just pull over right away and get a nice fine, or worse.

 

[Arc1t3ct]

...just like you wouldn't put, say, a Q6600 in a budget board with the stock cooler and try to get to 4GHz.

I remember the time when intel shills roamed this site (and others) making false claims like that. Of course the Q6600 could never go that high on air ...not even close!

 

[TheGuruStud]

It's 2020 and people believe intel is honest with tdp vs real world (no matter how you believe it's calculated) lol

Shoot me. I'm done.

 

[Vayra86]

Finally, to wrap up my argument, I'm not trying to say that Turbo, in itself, is bad. I'm just saying it's badly advertised. To me personally, it makes about as much sense as a commercial that might advertise the newest Mustang, going nice and fast with impressive handling around sharp curves or other potentially hazardous conditions... with the disclaimer professional driver on a closed course etc etc... it doesn't take into account that a normal person can't actually drive their car like that unless they are also a skilled driver on a closed course. Otherwise, you'll probably crash into something, or you'll run into other trouble when you are running from the police. Or you'll just pull over right away and get a nice fine, or worse.

But that is what turbo does for the average user who can't OC Its that professional driver built-in, all you need to do is provide the closed course to make it go fast (ie proper cooling), and for that the average user USED TO be able to use TDP. Now Intel has changed their interpretation of it and how it relates to optimal performance... and thát has made it worse, really. Its not the fault of turbo per say but of its interpretation and explanation, or as you say advertising. That said, in non-K parts the professional driver can handle things just fine on the regular road, too. Will it always be optimal, no, but definitely faster.

Now what really made things worse the past few gens was that mobo manufacturers figured it was a good idea to put oil spills all over the race course to make things go even faster, never mind the fact even a professional driver would then quickly lose control

I love car analogies

 

[deu]

Long story short: Intel and AMD calculate TDP different, AMDs being the closest to actual TDP. And Intels older architecture on 14nm runs hotter than AMDs newer architecture on 7nm. Cool! We can now lay down the hatchet and start exchanging meaningful reflextions about hardware where it actually matters. I swear to god in 4000 years someone will be reading the second bit-ble about how the religion x was totally better than the competing religion due to glue-together-sh** and other dumb and dividing sh**. (And now we all hug around the AVX2-fire)

 

[hat]

But that is what turbo does for the average user who can't OC Its that professional driver built-in, all you need to do is provide the closed course to make it go fast (ie proper cooling), and for that the average user USED TO be able to use TDP. Now Intel has changed their interpretation of it and how it relates to optimal performance... and thát has made it worse, really. Its not the fault of turbo per say but of its interpretation and explanation, or as you say advertising. That said, in non-K parts the professional driver can handle things just fine on the regular road, too. Will it always be optimal, no, but definitely faster.

That's pretty much what I'm saying. The average user can load web pages a fraction of a second faster (whoopee). It isn't until you learn how to actually take advantage of Turbo that it really matters.

Now what really made things worse the past few gens was that mobo manufacturers figured it was a good idea to put oil spills all over the race course to make things go even faster, never mind the fact even a professional driver would then quickly lose control

I love car analogies

Probably another marketing tool by either the mobo manufacturers, or Intel, or both. I can see mobo manufacturers doing that to make their board appear better than the rest, or Intel going to board manufacturers to do that to make their processors seem even better than what Turbo already does by running it somewhat out of spec. I feel like this is kind of a moot point, though, because if you're looking at motherboards and other components individually, presumably to build a system yourself, you probably already know about this... or at least you have a much better idea than the average user.

 

[R0H1T]

That is burning a lot of power for little gain.

Definitely outside the efficiency curve

And now we know why they named it Comet Lake

I wonder if there's a Supernova or Inferno Lake in there somewhere!

 

[TheUn4seen]

It's 2020 and people believe intel is honest with tdp vs real world (no matter how you believe it's calculated) lol

Shoot me. I'm done.

It's 2020 and people believe any corporation is honest with anything. It's about time people realise that any corporation would happily send people to cut out your kidneys and leave you dying in a pool of blood if it made for a long-term gain. Finding ways to lie while remaining within constraints of the law is what corporations actually have whole departments for. If you believe any corporation is "honest" it just means this particular company has a better PR strategy at the moment. Hell, I worked in a corporate world for a while and met people who would cut someone's eyes out if it meant a better parking spot for a day. You think such people are above lying to consumers for money? Think again

 

[EarthDog]

Finally, to wrap up my argument, I'm not trying to say that Turbo, in itself, is bad. I'm just saying it's badly advertised. To me personally, it makes about as much sense as a commercial that might advertise the newest Mustang, going nice and fast with impressive handling around sharp curves or other potentially hazardous conditions... with the disclaimer professional driver on a closed course etc etc... it doesn't take into account that a normal person can't actually drive their car like that unless they are also a skilled driver on a closed course. Otherwise, you'll probably crash into something, or you'll run into other trouble when you are running from the police. Or you'll just pull over right away and get a nice fine, or worse.

interesting point. Though turbo doesnt take a professional to get the most out of it. Turbo is there so any dick, tom, and Jane gets it out of the box. And, if you are Intel, that is true. Ryzen seems to be hit or miss, even with high end hardware (though it has improved since release).

 

[hat]

I can agree to that much. It's easier to read the installation guide for an advanced cooler than it is to learn overclocking.