[UPDATE - 220W CPUs? Get yo chiller] All Your + Are Belong To Us - Comet Lake S performance review leaks AND Power Consumption/Heat Output outted

[GDRRiley]

3 minutes ago, mr moose said:

Let's be realistic though, how many people buy a new CPU and not a new Motherboard when they upgrade?  The reality is very, very, very few do it, so few it probably doesn't even show up in sales analysis.  

I wanted to but my upgrade path on z170 was limited.

[5x5]

3 hours ago, mr moose said:

It doesn't mean that absolutely, except there is evidence it can and does happen.   If we look at the last time a platform was supported for 6 years,  by the time you got to the end you could buy a new CPU only but it meant running with 6 year old features (3 year old features if the board supported DDR3 which came in half way through).   No point in buying a new CPU only to gimp it with shit VRM, lower rams speeds etc.

 

 

Let's be realistic though, how many people buy a new CPU and not a new Motherboard when they upgrade?  The reality is very, very, very few do it, so few it probably doesn't even show up in sales analysis.  

Actually, when you provide support, people tend to upgrade quite readily as evidenced by the fact that the 3600 and 3700X both outsold all Intel CPUs combined in Germany.

 

https://www.notebookcheck.net/AMD-s-Ryzen-5-3600-sold-almost-three-times-as-many-units-than-all-of-Intel-s-chips-put-together-in-latest-Mindfactory-CPU-sales-data.459514.0.html

[GoldenLag]

with DDR5 right around the corner. isnt this a single generation socket? i mean they could hold off with DDR5 untill 2022, and see some success due to lower ram prices. but right now seems like a bad time to release a new socket, memmory wise. 

[mr moose]

3 hours ago, GDRRiley said:

I wanted to but my upgrade path on z170 was limited.

My upgrade path was limited too, I had a h77 mobo.

17 minutes ago, 5x5 said:

Actually, when you provide support, people tend to upgrade quite readily as evidenced by the fact that the 3600 and 3700X both outsold all Intel CPUs combined in Germany.

 

https://www.notebookcheck.net/AMD-s-Ryzen-5-3600-sold-almost-three-times-as-many-units-than-all-of-Intel-s-chips-put-together-in-latest-Mindfactory-CPU-sales-data.459514.0.html

Not sure what sales figures has to do with the average upgrade cycle.   The average upgrade cycle for most enthusiasts is anywhere between 4 and 8 years (as reported on various threads on forums).     Regardless which CPU you have they are nearly always a new Motherboard due to features if nothing else.

 

 

Of course there are hard core enthusiast who do it every year, but they are not the majority. 

[porina]

31 minutes ago, 5x5 said:

Actually, when you provide support, people tend to upgrade quite readily as evidenced by the fact that the 3600 and 3700X both outsold all Intel CPUs combined in Germany.

At one store in Germany. It by no means represents the wider market, but leaning heavily towards the enthusiast self-build market.

 

8 minutes ago, GoldenLag said:

with DDR5 right around the corner. isnt this a single generation socket? i mean they could hold off with DDR5 untill 2022, and see some success due to lower ram prices. but right now seems like a bad time to release a new socket, memmory wise. 

There was a separate rumour thread recently this socket will also do Rocket Lake, which appears to be a backport to 14nm of the stuff that was supposed to be on 10nm. It would be the first beyond-Skylake architecture on desktop and I think is also the one after Ice Lake, so expect a good IPC boost. If they can still maintain clock on 14nm is the interesting question.

 

I'd really hope for DDR5 to become mainstream sooner than later, but I really don't expect it to happen.

[mr moose]

3 minutes ago, porina said:

At one store in Germany. It by no means represents the wider market, but leaning heavily towards the enthusiast self-build market.

 

 

Even then, if total sales did not increase (which it strongly looks like but those numbers aren't given) then people weren't upgrading more frequently or out of the normal upgrade cycle.   It just means in that market people started building ryzen systems instead of Intel. 

[porina]

30 minutes ago, mr moose said:

Even then, if total sales did not increase (which it strongly looks like but those numbers aren't given) then people weren't upgrading more frequently or out of the normal upgrade cycle.   It just means in that market people started building ryzen systems instead of Intel. 

I haven't looked closely at the Mindfactory data since some months after the Zen 2 desktop launch. I did find it interesting that the Zen 2 sales, at the time, were incremental. Intel sales were largely unchanged. That was then, I haven't looked again to see if it holds today.

[mr moose]

6 minutes ago, porina said:

I haven't looked closely at the Mindfactory data since some months after the Zen 2 desktop launch. I did find it interesting that the Zen 2 sales, at the time, were incremental. Intel sales were largely unchanged. That was then, I haven't looked again to see if it holds today.

 

From about 2017, AMD sales go up and Intel sales fall, over all a marginal increase in sales which looks like normal growth.    It doesn't look like people were upgrade any sooner than the normal upgrade cycles, it's just that from the 2000 series Ryzen was the CPU of choice for upgrades.  Which makes perfect sense when all things are considered.

 

EDIT: maybe even a bit earlier than the 2000 series.  Seeing as they were 2018.

[thorhammerz]

10 hours ago, mr moose said:

Of course there are hard core enthusiast who do it every year, but they are not the majority. 

Though they are certainly vocal about how their own preferences should dictate the broader product stack .

[5x5]

Just now, comander said:

Remember Ryzen 1000/2000 were around 10% behind on IPC...

 

So Ryzen 3000 only ends up around 5% ahead on IPC. It's around 5-15% behind on clock speed though... But at half the power draw per core. 

 

At the end of the day, AMD can cram 64 cores into a socket while maintaining solid lightly threaded performance and Intel can't. 

Actually - no. That's not remotely correct. Look at LTT's video on the G14 and you'll also see them confirm - Zen 1 was 10% behind. Zen+ matched. Zen 2 took a 15% lead. Hence why Intel's single threaded performance is pretty much the same as AMD's at the moment.
 

 

[5x5]

36 minutes ago, comander said:

Zen and Zen+ have basically the same IPC, within single digit percentages. 

 

There's a reason why a 6C/12T Intel CPU was generally faster than the 2700x. 

 

https://www.techspot.com/amp/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/page2.html

 

 

If you're referring to Renoir, then it's very possible that AMD has around a 15% edge. AMD is claiming that Renoir is 25% faster than it's predecessor at the same clocks. Compare with desktop Zen2 which is "only" 15% faster. 

Dude - Renoir is the same architecture as Matisse - it's Zen 2. And no, Zen+ has superior IPC over Zen 1 due to cache improvements, clockspeed enhancements and BIOS microcode updates. Please, as you can CLEARLY see. The 8700K @ 4GHz is the SAME as the 2600X at 4GHz. SAME single-core results and SAME IPC. Intel HAD an advantage (past tense) in gaming due to optimization and OS scheduler issues that have long since been resolved. Currently, gaming on a 4.4GHz 3700X is the same affar as gaming on a 5GHz 9700K

[SpaceGhostC2C]

18 hours ago, GDRRiley said:

the lie is needing a new chipset and socket. the old socket and higher end chipsets would have been fine but intel decided to change them.

I'm not even sure it's a bad thing: anyone with a motherboard that "could have supported" the new chips is probably better off skipping them anyway, and anyone upgrading from something older or starting anew would have to buy a motherboard regardless.

[5x5]

3 minutes ago, SpaceGhostC2C said:

I'm not even sure it's a bad thing: anyone with a motherboard that "could have supported" the new chips is probably better off skipping them anyway, and anyone upgrading from something older or starting anew would have to buy a motherboard regardless.

Well, B350 working fine with 3000 series CPUs proved that backwards compatibility is not bad at all. Just have to design products with forward thinking, tis all.

[mr moose]

1 hour ago, SpaceGhostC2C said:

I'm not even sure it's a bad thing: anyone with a motherboard that "could have supported" the new chips is probably better off skipping them anyway, and anyone upgrading from something older or starting anew would have to buy a motherboard regardless.

I can understand the reason people would want to be able to do a CPU only upgrade, but it seems most of the reasoning people give lacks any depth or acceptance of the reality of features and support.

 

 

57 minutes ago, 5x5 said:

Well, B350 working fine with 3000 series CPUs proved that backwards compatibility is not bad at all. Just have to design products with forward thinking, tis all.

 

As I pointed out before, you can run a 3000 series on a B350 (I am right now), however if you do that in most cases you will lack higher memory speed support and the cheaper boards that have less power capabilities will reduce your overall performance.   This is one thing that Intel are trying to avoid.   It's one thing to have the option, but do you really want that option if it means people will market a CPU to you that will only perform at 90% because of your motherboards VRM and memory support?  There is already less than 10% improvement between most Intel CPU iterations as it is. 

[mr moose]

I find the whole IPC debate to be a bit erroneous at times.  If they turned off all cores bar one and disabled HT/SMT would the above results be the same?  If you are testing across multiple cores then you have to account for core scaling not just frequency.  Without doing that it's not longer IPC that you are testing but also how efficient HT/SMT is and how well different workloads scale across multiple processors. Now I get that with everything on you are still measuring instructions per clock, however if someone is talking about IPC as being a single thread, then comparing multiple cores does not really address the initial claim.  It just clouds the debate and distracts from which ever is the most important metric to each user.

[thechinchinsong]

20 hours ago, thorhammerz said:

Though they are certainly vocal about how their own preferences should dictate the broader product stack .

Whats wrong with being vocal?

[thechinchinsong]

16 hours ago, mr moose said:

It's one thing to have the option, but do you really want that option if it means people will market a CPU to you that will only perform at 90% because of your motherboards VRM and memory support?  There is already less than 10% improvement between most Intel CPU iterations as it is. 

I've talked with you about this before, but more options is always better than less options. What are the drawbacks? You always have the choice to still buy newer boards. The 99% of people who want a new motherboard every generation can get one , while the 1% of people who want to run a new processors on older motherboards are also satisfied. Yes, TDP limitation and memory/IO support will be worse, but if you're part of the 99% who buy new motherboards every CPU upgrade, then it literally doesn't matter for you since you can just buy a new motherboard. For the 1% who want motherboard compatibility, they'll get it, albeit with only 90% of CPU performance. Better to have the option than not have the option. (Percentages taken out of my ass)

 

Yes, people always say, "iT dOeSn'T mAkE sEnSe FiNaNcIaLlY" for Intel, but how does it hurt anyone to be vocal about it or at least keep asking. Yes, AMD's 300 through 500 series motherboard compatibility isn't the greatest implementation, but since Intel has so much more money and so much more leverage with OEMs/motherboard makers, shouldn't they be able to deliver a superior implementation of this compatibility? I'm sick of people implying others as being spoiled or ignorant or entitled, simply for asking for features that could have been added.

 

I say could have been added, since it looks like in the next couple of generations, Intel might actually be changing architectures significantly enough to warrant new motherboards and not provide backward compatibility. Moving to Willow Cove is a big change, and I'm not expecting nor calling for backward compatibility with previous motherboard generations. Nobody thought when AMD was moving from Bulldozer to Zen "golly gee I wish AM3+ was compatible with Zen" (although it would've been interesting if it was). But when you move from socket 1151 -> 1151 -> 1151-> 1151 all the while keeping the same architecture and don't provide at least partial backward compatibility, it gets some people agitated and rightfully so. 

[MageTank]

18 hours ago, 5x5 said:

Well, B350 working fine with 3000 series CPUs proved that backwards compatibility is not bad at all. Just have to design products with forward thinking, tis all.

Backwards compatibility is fine as long as they don't overdo it. Looking back at AM3, where AMD allowed AM3 processors to work in AM2+ motherboards and their desire to maintain backwards compatibility ended up holding back on their ability to innovate outside of that platform restriction. I hope they don't repeat that mistake with Zen. The consumers also need to understand that they can't have backwards compatibility forever, and that a company needs to move on to bigger, better platforms eventually. I still remember people that were angry when they learned AM4 CPU's wouldn't fit in their AM3 boards, despite AMD delivering a major uplift in performance and platform features with Zen.

 

I am just annoyed that Intel artificially locked us out of using Coffee Lake on Z170 boards, despite the boards clearly being capable of supporting the processors. There are people using 8700k's on Z170 OC Formula boards still pushing absurdly high clock speeds and memory configurations without any real issues.

[5x5]

31 minutes ago, MageTank said:

Backwards compatibility is fine as long as they don't overdo it. Looking back at AM3, where AMD allowed AM3 processors to work in AM2+ motherboards and their desire to maintain backwards compatibility ended up holding back on their ability to innovate outside of that platform restriction. I hope they don't repeat that mistake with Zen. The consumers also need to understand that they can't have backwards compatibility forever, and that a company needs to move on to bigger, better platforms eventually. I still remember people that were angry when they learned AM4 CPU's wouldn't fit in their AM3 boards, despite AMD delivering a major uplift in performance and platform features with Zen.

 

I am just annoyed that Intel artificially locked us out of using Coffee Lake on Z170 boards, despite the boards clearly being capable of supporting the processors. There are people using 8700k's on Z170 OC Formula boards still pushing absurdly high clock speeds and memory configurations without any real issues.

Exactly my point - I expect Zen 3 to be the end of AM4 and Zen 4 to be on AM5. Otherwise AM4 will, indeed, be a limiting factor. After all, DDR5 should be ready and in good supply by the end of 2021/start of 2022 when Zen 4 is supposedly due.

[thorhammerz]

2 hours ago, thechinchinsong said:

Whats wrong with being vocal?

People are free to say whatever the hell they want, as much as others are free to ridicule and laugh at whatever they want .

[thechinchinsong]

12 hours ago, thorhammerz said:

People are free to say whatever the hell they want, as much as others are free to ridicule and laugh at whatever they want .

Agreed. I'm just curious why you are ridiculing a made up demographic .

[mr moose]

5 hours ago, thechinchinsong said:

I've talked with you about this before, but more options is always better than less options. What are the drawbacks? You always have the choice to still buy newer boards. The 99% of people who want a new motherboard every generation can get one , while the 1% of people who want to run a new processors on older motherboards are also satisfied. Yes, TDP limitation and memory/IO support will be worse, but if you're part of the 99% who buy new motherboards every CPU upgrade, then it literally doesn't matter for you since you can just buy a new motherboard. For the 1% who want motherboard compatibility, they'll get it, albeit with only 90% of CPU performance. Better to have the option than not have the option. (Percentages taken out of my ass)

 

Yes, people always say, "iT dOeSn'T mAkE sEnSe FiNaNcIaLlY" for Intel, but how does it hurt anyone to be vocal about it or at least keep asking. Yes, AMD's 300 through 500 series motherboard compatibility isn't the greatest implementation, but since Intel has so much more money and so much more leverage with OEMs/motherboard makers, shouldn't they be able to deliver a superior implementation of this compatibility? I'm sick of people implying others as being spoiled or ignorant or entitled, simply for asking for features that could have been added.

 

I say could have been added, since it looks like in the next couple of generations, Intel might actually be changing architectures significantly enough to warrant new motherboards and not provide backward compatibility. Moving to Willow Cove is a big change, and I'm not expecting nor calling for backward compatibility with previous motherboard generations. Nobody thought when AMD was moving from Bulldozer to Zen "golly gee I wish AM3+ was compatible with Zen" (although it would've been interesting if it was). But when you move from socket 1151 -> 1151 -> 1151-> 1151 all the while keeping the same architecture and don't provide at least partial backward compatibility, it gets some people agitated and rightfully so. 

 

My problem is the confusion and less than adequate consumer education that becomes required when products have so many nuances coming from outside the actual product. CPU's do not come with a Motherboard performance guarantee and so they can only guarantee performance if they independently test every board (and who knows if some of the cheaper boards are even go to survive a new CPU being pushed on it).  I was one of the people who ended up paying for a processor that was not performing as reviewed/promised because I used it on an older motherboard, one that was compatible only but didn't support all it's features.  Had I spent another few weeks trying to decipher the multitudes of posts on forums working out which specific motherboard I need (a new one) I may have avoided that.

 

My second issue is not that people are complaining about it (I understand the desire to keep your motherboard and just upgrade the CPU), but that people are making assumptions about motive, and taking their probem well beyond reality in order to shoe horn the situation into their preconceived ideas. 

 

 

Having said that, my personal reasoning is it is like complaining you have to buy a whole new car rather than just the engine when you want to upgrade the performance.   The car (motherboard) is as much designed to support the engine (CPU) as much as the engine gives the car all it's power.    And while the manufacturer doe not test performance of new CPU's on older motherboards there is no guarantee your upgrade will be an upgrade,  just an expensive purchase for little gain.   And that to me is more anti consumer than only supporting each hip for 2 generations and forcing all upgrades to ensure that you get the performance you pay for.   At least doing it that way you know before you buy what you are getting and how much it will cost, the other way around the only thing you know is it will work, how well is a gamble without spending an inordinate amount of time looking for people who have already done it and actually tested it.  

 

Fortunately now days we have a few reviewers who have done basic testing of 3000 stuff on B350, but they only did one B350 board and it wasn't a low quality one so we still don't know if a cheaper B350 board will give a 3600 any chance of beating a 2600 (especially with memory limitations).  If reviewers start doing this in detail for All new CPU's on most older boards then the issue becomes much less of a problem, but I can't see that happening.

 

 

[thechinchinsong]

6 hours ago, mr moose said:

 

My problem is the confusion and less than adequate consumer education that becomes required when products have so many nuances coming from outside the actual product. CPU's do not come with a Motherboard performance guarantee and so they can only guarantee performance if they independently test every board (and who knows if some of the cheaper boards are even go to survive a new CPU being pushed on it).  I was one of the people who ended up paying for a processor that was not performing as reviewed/promised because I used it on an older motherboard, one that was compatible only but didn't support all it's features.  Had I spent another few weeks trying to decipher the multitudes of posts on forums working out which specific motherboard I need (a new one) I may have avoided that.

 

My second issue is not that people are complaining about it (I understand the desire to keep your motherboard and just upgrade the CPU), but that people are making assumptions about motive, and taking their probem well beyond reality in order to shoe horn the situation into their preconceived ideas. 

 

Having said that, my personal reasoning is it is like complaining you have to buy a whole new car rather than just the engine when you want to upgrade the performance.   The car (motherboard) is as much designed to support the engine (CPU) as much as the engine gives the car all it's power.    And while the manufacturer doe not test performance of new CPU's on older motherboards there is no guarantee your upgrade will be an upgrade,  just an expensive purchase for little gain.   And that to me is more anti consumer than only supporting each hip for 2 generations and forcing all upgrades to ensure that you get the performance you pay for.   At least doing it that way you know before you buy what you are getting and how much it will cost, the other way around the only thing you know is it will work, how well is a gamble without spending an inordinate amount of time looking for people who have already done it and actually tested it.  

 

Fortunately now days we have a few reviewers who have done basic testing of 3000 stuff on B350, but they only did one B350 board and it wasn't a low quality one so we still don't know if a cheaper B350 board will give a 3600 any chance of beating a 2600 (especially with memory limitations).  If reviewers start doing this in detail for All new CPU's on most older boards then the issue becomes much less of a problem, but I can't see that happening.

 

 

So ultimately, you don't have a problem with motherboard compatibility, you have a problem with the way its implemented. Again, the 99.9% of people who don't care about compatibility will simply buy a new motherboard along with the new CPU, while anyone who wants to only buy a new CPU will have to make the decision on whether to also upgrade the motherboard or not (based on the information that they can acquire). Obviously if there are drawbacks (in AMD's case) for using an older motherboard with a newer CPU, it will be up to the small niche to acquire that consumer education, again, I don't understand why having the choice is worse than not having the choice. If you can't get that reliable reviewer information, then simply ignore the option, and buy a new motherboard, but for those that scour the forums and  "spent another few weeks trying to decipher the multitudes of posts on forums working out which specific motherboard I need", give them the option to do so.

 

If assumptions are the issue, then stop disagreeing with the motive, but disagree with the assumptions. One doesn't discount a factual motive held by a group of people simply because there are some amount of people who share that same motive, but have the completely wrong assumptions. 

 

I don't see how its anti-consumer, since its not taking away the option for the consumer to simply purchase a whole new car, its giving those auto tinkerers the option to do what they want. Those auto tinkerers are going to be very happy that they at least have the option to put in new engine into an old car, and everybody else is none the wiser (everyone else literally doesn't even have to know if it somehow makes others unhappy).

 

Intel and AMD have both done backward compatibility before in the past. Haswell was on z87 and z97. AM3+ was compatible with so many refreshes and new archs. AM4 has had the option for backward/forward compatibility even if its not perfect, but IMO, the trade-offs are still well documented if you take care. In fact socket-1151 compatibility has been literally done with custom BIOSes to work between sky/kaby/coffee-lake processors. If unofficial sources can get it working, my assumption is that Intel could make it official and make it a lot better. Hell, Colorful has a h310 motherboard that works with every ****-lake Intel has (https://www.tomshardware.com/news/colorful-h310m-e-v20-mothboard-intel-cpu-support). 

[mr moose]

3 minutes ago, thechinchinsong said:

So ultimately, you don't have a problem with motherboard compatibility, you have a problem with the way its implemented. Again, the 99.9% of people who don't care about compatibility will simply buy a new motherboard along with the new CPU, while anyone who wants to only buy a new CPU will have to make the decision on whether to also upgrade the motherboard or not (based on the information that they can acquire). Obviously if there are drawbacks (in AMD's case) for using an older motherboard with a newer CPU, it will be up to the small niche to acquire that consumer education, again, I don't understand why having the choice is worse than not having the choice. If you can't get that reliable reviewer information, then simply ignore the option, and buy a new motherboard, but for those that scour the forums and  "spent another few weeks trying to decipher the multitudes of posts on forums working out which specific motherboard I need", give them the option to do so.

As you point out,  it will be up to those people to work out,  If a product is advertised to work at X Ghz then it should work at that and never be up to the consumer to workout if there are any unwritten limitations.     My point is that by giving others the option you open the market to even worse advertising and marketing standards than we already have.  Remember the boost clock issue AMD had with ryzen 3000?

 

https://techreport.com/news/3465389/ryzen-3000-boost-clock-amd-statement/

 

There are so many reasons for Intel to tie products to platforms they can control, not the least of which is stuff that prevents their products from doing what they advertise..

 

3 minutes ago, thechinchinsong said:

If assumptions are the issue, then stop disagreeing with the motive, but disagree with the assumptions. One doesn't discount a factual motive held by a group of people simply because there are some amount of people who share that same motive, but have the completely wrong assumptions. 

The motive is the assumption, I am disagreeing with the assumption that finances are the motive for this move .  There is nothing factual about the accusation that Intel are just doing it to sell more chipsets.   If evidence arises that suggest that I will change my stance naturally, but until then it is an assumption.

 

3 minutes ago, thechinchinsong said:

I don't see how its anti-consumer, since its not taking away the option for the consumer to simply purchase a whole new car, its giving those auto tinkerers the option to do what they want. Those auto tinkerers are going to be very happy that they at least have the option to put in new engine into an old car, and everybody else is none the wiser (everyone else literally doesn't even have to know if it somehow makes others unhappy).

Talk to the all the people who were threatening legal action over 100Mhz when ryzen 3000 didn't reach boost.  I don't think that was AMD's fault, fortunately it was an issue they were able to fix with an update.  Imagine if it was a hardware issue due solely to the hardware on the motherboard?  Remember there are issues with bios being too small to take updates and fixes for certain issues.  https://www.techpowerup.com/257240/msi-scampers-to-launch-new-amd-400-series-motherboards-with-256mb-bios-chips
 

Now I personally think most of these are not major issues and can be worked around or fixed,  but the fact they exist and can leave a consumer with a CPU that won't function or function only as good as the last CPU is enough reason to either have complete validation or restrict new products to known working solutions.  

 

3 minutes ago, thechinchinsong said:

Intel and AMD have both done backward compatibility before in the past. Haswell was on z87 and z97. AM3+ was compatible with so many refreshes and new archs. AM4 has had the option for backward/forward compatibility even if its not perfect, but IMO, the trade-offs are still well documented if you take care.

I don't think Intel have ever supported a Processor for more than 2 generations on any one chipset, and I have a lengthy post somewhere on this forum concerning the issues AMD consumers faced with AM3.  It seems I need to bookmark it because every time Intel release a new chipset this discussion starts all over again. 

 

3 minutes ago, thechinchinsong said:

 

In fact socket-1151 compatibility has been literally done with custom BIOSes to work between sky/kaby/coffee-lake processors. If unofficial sources can get it working, my assumption is that Intel could make it official and make it a lot better. Hell, Colorful has a h310 motherboard that works with every ****-lake Intel has (https://www.tomshardware.com/news/colorful-h310m-e-v20-mothboard-intel-cpu-support). 

Yes, they have done that, but getting it working on the cheapest boards and getting it to clock and boost remain within spec for the expected lifetime of the motherboard/CPU is a different story altogether.   Those guys bodging around with it don't have to honor the warranty, they can make it work however they want/need.  The life of all silicon is proportionate to heat, that is why the CPU throttles back when it gets too hot,  if you put a CPU that draws more power through the board or places other components under heavier load than they were designed for it shortens the life of those parts (caps in particular, but silicon has a limit also)  If the VRM isn't properly implemented (like on much cheaper boards) then the load may well cause it to break or slow down.   And this is before we get into feature limitations like ram speed, OC, etc.

 

 

Again, I am not arguing that these are all enough to argue they should never be backwards compatible, but they are enough reason to see why it is a thing and certainly enough reason to understand it is more than just money motive.

 

[5x5]

Update - we have power boys
https://www.notebookcheck.net/Comet-Lake-S-may-require-hefty-CPU-coolers-Core-i9-10900F-shown-to-have-a-170-W-PL1-TDP-and-224-W-PL2-for-a-4-6-GHz-all-core-boost.460196.0.html