Intel Core i9-12900K Allegedly Beats AMD Ryzen 9 5950X at Cinebench R20

[Turmania]

Power consumption that is what I care about, if its rated 95w. I expect consumption to be around that plus or minus 10 percent. Rest is secondary for me.

 

[Auer]

Are these wafers coming out of Arizona plant?

Would be nice if they would be U.S. made.

U.S. should not have to depend on other countries for this.

 

[londiste]

Let's not forget that Intel's R&D budget is literally over 684% larger than AMD's and therefore Intel SHOULD be crushing AMD.

This is an ignorant argument. Technically correct, $13.56B vs $1.98B. However, within that R&D spending Intel is competing in a much MUCH wider area than AMD.

In addition to CPUs (and GPUs) that they are working on, similarly to AMD:
- Most notably, they are competing with semiconductor foundries like TSMC ($4B R&D in 2020 and increasing fast) and Samsung (no idea how much foundry spends but Samsung Electronics total is $18.9B).
- They are competing in storage with XPoint/Optane and last year still included NAND R&D before selling it to SK Hynix. For reference, both SK Hynix and Intel's old partner Micron spent $2.6B in 2020.
- Intel is also major player in FPGA. AMD joined in by buying Xilinx in 2020, Xilinx R&D spending was $0.7B a couple years ago. The other big player is Microchip Technologies with R&D spending above $0.8B in 2020.
- Intel is pretty big in NICs and some network solutions, trying to get into IoT, researching a bunch of new ideas, for example silicon photonics comes to mind which they should be one of the leaders in plus whole bunch of other smaller areas that do add up.

If that is true, there are 2 scenario, and for some reason, i think the worst one will be the one that happen.

Scenario 1. Intel release it, price it like the 5950x forcing AMD to reduce it's price
Scenario 2. Intel release it, price it 20-30% more expensive than the 5950x so people pay for the extra performance

Scenario 3. Intel continues with their current i9 lineup pricing - i9-9900K and i9-10900K were $500, i9-119000K is €550 - or does a modest increase and undercuts 5950X in price

Power consumption that is what I care about, if its rated 95w. I expect consumption to be around that plus or minus 10 percent. Rest is secondary for me.

Rumors say Alder Lake PL2 is at 228W.
Btw, for comparison 5950X TDP is 105W, power limit is at 140W. Latest rumors of the next lineup (either Zen3 refresh or Zen4) include 120W and 170W TDP (assuming power management continues the same way as Zen2 and Zen3 based desktop CPUs that would mean power limits at 160W and 230W respectively).

 

[RandallFlagg]

This is an ignorant argument. Technically correct, $13.56B vs $1.98B. However, within that R&D spending Intel is competing in a much MUCH wider area than AMD.

In addition to CPUs (and GPUs) that they are working on, similarly to AMD:
- Most notably, they are competing with semiconductor foundries like TSMC ($4B R&D in 2020 and increasing fast) and Samsung (no idea how much foundry spends but Samsung Electronics total is $18.9B).
- They are competing in storage with XPoint/Optane and last year still included NAND R&D before selling it to SK Hynix. For reference, both SK Hynix and Intel's old partner Micron spent $2.6B in 2020.
- Intel is also major player in FPGA. AMD joined in by buying Xilinx in 2020, Xilinx R&D spending was $0.7B a couple years ago. The other big player is Microchip Technologies with R&D spending above $0.8B in 2020.
- Intel is pretty big in NICs and some network solutions, trying to get into IoT, researching a bunch of new ideas, for example silicon photonics comes to mind which they should be one of the leaders in plus whole bunch of other smaller areas that do add up.

Don't forget 5G, things like Mobileye (autonomous driving), and Intel's relatively vast software suites including extensive system management packages.

I see that lazy argument all the time comparing R&D budgets, it's like comparing a grape to a fruit salad.

 

[Crackong]

Will be interesting to see how Intel tackles this problem, since most of the power should be going to the Golden Cove cores, of which there are only 8 to share the power budget. Being monolithic is a boon, but it changes little at 7nm/ESF density.

+1

Since the little Gracemont cores are Atom class,
Judging by current gen Atom, the TDP of those as a discrete CPU should be ~10W for 4 cores
Bump that to 8 cores ~20W
Double that amount for a desktop application ~40W
Now 228 - 40 = 188 , 188 / 8 = 23.5
So 23.5W per BIG core within a N7FF size die....

Dunno how Intel is gonna handle that heat.

 

[ThrashZone]

+1

Since the little Gracemont cores are Atom class,
Judging by current gen Atom, the TDP of those as a discrete CPU should be ~10W for 4 cores
Bump that to 8 cores ~20W
Double that amount for a desktop application ~40W
Now 228 - 40 = 188 , 188 / 8 = 23.5
So 23.5W per BIG core within a N7FF size die....

Dunno how Intel is gonna handle that heat.

Hi,
Intel will handle it the way they always do = cooler not included lol

 

[Crackong]

Hi,
Intel will handle it the way they always do = cooler not included lol

Hey it is Okay
Just use that chiller hidden under a table in the W-3175X event, it would be fine.

 

[Shatun_Bear]

Don't forget 5G, things like Mobileye (autonomous driving), and Intel's relatively vast software suites including extensive system management packages.

I see that lazy argument all the time comparing R&D budgets, it's like comparing a grape to a fruit salad.

What craziness is this, Intel's budget for CPUs is many times larger than AMD's, doesnt take a breakdown of financials to work this out. AMD also doesn't just funnel all investment into CPUs and GPUs.

 

[RandallFlagg]

What craziness is this, Intel's budget for CPUs is many times larger than AMD's, doesnt take a breakdown of financials to work this out. AMD also doesn't just funnel all investment into CPUs and GPUs.

AMD doesn't make CPUs, for starters, so even your comparison of CPU-only budget (do you even have such a number?) is invalid as you'd be including work done at TSMC (node development which AMD does not do) without accounting for that R&D. Then you've got all these areas that Intel is a major player in, like RAID controllers, ethernet controllers, wifi controllers, where AMD is not present at all. On top of that you've got things like Thunderbolt, PCI, and PCI-e that are Intel developed technologies - and you can bet they are getting paid for their IP on every AMD motherboard. What's more, most AMD motherboards have an Intel chip on it somewhere. Like I said, you are comparing a grape to a fruit salad. It is a false comparison.

 

[Shatun_Bear]

AMD doesn't make CPUs, for starters, so even your comparison of CPU-only budget (do you even have such a number?) is invalid as you'd be including work done at TSMC (node development which AMD does not do) without accounting for that R&D. Then you've got all these areas that Intel is a major player in, like RAID controllers, ethernet controllers, wifi controllers, where AMD is not present at all. On top of that you've got things like Thunderbolt, PCI, and PCI-e that are Intel developed technologies - and you can bet they are getting paid for their IP on every AMD motherboard. What's more, most AMD motherboards have an Intel chip on it somewhere. Like I said, you are comparing a grape to a fruit salad. It is a false comparison.

So you think Intel and AMD have a comparable R&D budget for their processor department? You heard it first here folks; minnows Intel hitting back against the behemoth that is AMD in the processor market

 

[Rahmat Sofyan]

Intel ? Really ?
Mehhh . . .

 

[RandallFlagg]

So you think Intel and AMD have a comparable R&D budget for their processor department? You heard it first here folks; minnows Intel hitting back against the behemoth that is AMD in the processor market

I didn't say that - I asked if you even had a breakdown of their R&D. Obviously you don't. More obviously you are just trolling.

 

[HenrySomeone]

Lost to little cores nonetheless. It’s game over with Raptor Lake.

Bingo! It's just like I've been saying for years now - Intel was only held back by its node problems, not the architecture; mind that this was pretty much ready to go 3 years ago and it's still slaughtering the best from team (d)red today - now imagine them going against it with puny Zen(+), lmao! Granted, this isn't that great as far as competitive market goes, because Intel will once again hold complete dominance in 2 years or even less and it would have been better if AMD could stay competitive a bit longer, but to anyone paying attention it was obvious that their Zen line was a couple years to late for that to happen and it will be nice to see their insufferable fanboys having seizures, hehe

 

[Crackong]

AMD doesn't make CPUs, for starters, so even your comparison of CPU-only budget (do you even have such a number?) is invalid as you'd be including work done at TSMC (node development which AMD does not do) without accounting for that R&D. Then you've got all these areas that Intel is a major player in, like RAID controllers, ethernet controllers, wifi controllers, where AMD is not present at all. On top of that you've got things like Thunderbolt, PCI, and PCI-e that are Intel developed technologies - and you can bet they are getting paid for their IP on every AMD motherboard. What's more, most AMD motherboards have an Intel chip on it somewhere. Like I said, you are comparing a grape to a fruit salad. It is a false comparison.

By you standards
Intel doesn't make CPUs either.

Intel does not make the capacitors used within their CPU package.
Intel did not mine their own silicon / copper from the ground, nor brew all the chemicals used in the fab.

Your statement is totally invalid (by your own standard) as you'd be including all the work done at the silicon wafer manufacturer and all the chemical plants which are, obviously NOT own by Intel.

 

[The red spirit]

I want a high end 5ghz CPU that's only drawing 20watts........oh wait we're gonna have to wait 20 years for that. /s

High end CPU's will have a high power draw, what do you people want?

No more than 150 watts for flagship. If it needs more, then it's a workstation chip. Over 200 watts means very heavy overengineering, poor value and near zero practical improvement in games. But sure, such machine will burn some legs.

 

[Manoa]

I would like to remind everyone that might have missed some important information, this intel "10nm" manufacturing is completely different to anything that exists, the physical size is (slightly) smaller than what TSMC has that it calls "7nm" and unlike it intel are now the first and only in the world to use cobalt as conductor for manufacturing material, whatever everyone said about intel being behind for the amount of time that they were, it was because they were making a huge revolutionary step in the development of a completely new manufacturing technology: cobalt. now that they are out of that corner they might start picking up speed because they are now the sole owners of this. they use cobalt to provide more power at smaller spaces than what copper can and this can turn into a big advantage. the reason I mention to you this is we don't know how cobalt behaves when it is transferring the power at such small spaces, we don't know the extent of electron jumps across silicon borders, don't know how much it turns into heat, which is we don't know how much resistance it has per distance compare to copper.... in short I wanna say that we can't compare anything that is known about copper technology to cobalt technology: temperatures cannot be compared, because the materials are different, e.g stability limit for copper circuit may not be the same as it is for cobalt, tolerance may not be the same, everything we know about the problems with copper technology may not apply anymore, you can compare power use I suppose but it means nothing because we don't know the relation of how that translate to temperature, and even if we did we don't know to what extent that temperature matters, people say "10nm" but nobody explain what is behind it, I hope I did that but anyway we all will know soon enough

 

[HenrySomeone]

No more than 150 watts for flagship. If it needs more, then it's a workstation chip. Over 200 watts means very heavy overengineering, poor value and near zero practical improvement in games. But sure, such machine will burn some legs.

Well, no current cpu uses over 200 watts while gaming, so I'm not sure what you're trying to say. Well, there was one (ore two), that did though, at least in some games - FX9590 (and 9370), hehe

 

[The red spirit]

Well, no current cpu uses over 200 watts while gaming, so I'm not sure what you're trying to say. Well, there was one (ore two), that did though, at least in some games - FX9590 (and 9370), hehe

Xeons, FX 8320 and some other chips do indeed pull more than 150 watts. Even overclocked FX 4300 could pull more than that. Anyway, Intel has PLs and they are adjustable. PL1 was at 125 watts and that's the long term power limit. PL2 is short term burst power limit, so it almost doesn't matter. Intel chips, including Rocket Lake are very reasonably efficient, if Intel spec isn't violated. Many boards violate it and then we have people believing that Intel chips are furnaces. I'm tired of this crap. I personally don't like PL2 and I wish that it will be removed, there's no point to have it, when PL1 works perfectly fine.

 

[Crackong]

I would like to remind everyone that might have missed some important information, this intel "10nm" manufacturing is completely different to anything that exists, the physical size is (slightly) smaller than what TSMC has that it calls "7nm" and unlike it intel are now the first and only in the world to use cobalt as conductor for manufacturing material, whatever everyone said about intel being behind for the amount of time that they were, it was because they were making a huge revolutionary step in the development of a completely new manufacturing technology: cobalt. now that they are out of that corner they might start picking up speed because they are now the sole owners of this. they use cobalt to provide more power at smaller spaces than what copper can and this can turn into a big advantage. the reason I mention to you this is we don't know how cobalt behaves when it is transferring the power at such small spaces, we don't know the extent of electron jumps across silicon borders, don't know how much it turns into heat, which is we don't know how much resistance it has per distance compare to copper.... in short I wanna say that we can't compare anything that is known about copper technology to cobalt technology: temperatures cannot be compared, because the materials are different, e.g stability limit for copper circuit may not be the same as it is for cobalt, tolerance may not be the same, everything we know about the problems with copper technology may not apply anymore, you can compare power use I suppose but it means nothing because we don't know the relation of how that translate to temperature, and even if we did we don't know to what extent that temperature matters, people say "10nm" but nobody explain what is behind it, I hope I did that but anyway we all will know soon enough

Before you deep dive to some magical "cobalt technology"

Please understand under some basic thermodynamics
A CPU has no physical work done
so 99.99% of energy going in will turn into heat (rest of it will turn into some form of radiation, which also turn into heat eventually).

So 228W goes in
227.99999W of heat goes out.

Maybe your magical "cobalt technology" could somehow send the excessive energy into the 5th dimension ?
How are you "Don't know how much it turns into heat" ?

Please
Give us an answer other than 100% (or 99.99% if you wanna be precise)

 

[Manoa]

I can't give you this answer becuase thare are no cobalt CPU yet :x it will be first of it kind, soon we will find out, but I suppose you right it's always 100%....except you will have no way of knowing how much of that heat is from the conduits, how much from the transistors, resistors or the capacitors, which will also be diffrent :x that's always the problem with big technological jumps like this - when the conduit material changes so may the components :x I can't give you a good answer, all I know for sure is that it allow intel to increase the amount of power the conduits transfer to the destination, how much of that in % turns into heat from the conduits themselfs (at whatever conditions) only intel knows :x like I say, it is a first of it kind, nobody outside know nothing about cobalt technology :x it's a different material that never been used in CPU before

but I suppose if you whant to find out you can take a copper wire 20nm in diameter and transfer power through it, then do the same with cobalt and run some tests

 

[Crackong]

I can't give you this answer becuase thare are no cobalt CPU yet :x it will be first of it kind, soon we will find out, but I suppose you right it's always 100%....except you will have no way of knowing how much of that heat is from the conduits, how much from the transistors, resistors or the capacitors, which will also be diffrent :x that's always the problem with big technological jumps like this - when the conduit material changes so may the components :x I can't give you a good answer, all I know for sure is that it allow intel to increase the amount of power the conduits transfer to the destination, how much of that in % turns into heat from the conduits themselfs (at whatever conditions) only intel knows :x like I say, it is a first of it kind, nobody outside know nothing about cobalt technology :x it's a different material that never been used in CPU before

but I suppose if you whant to find out you can take a copper wire 20nm in diameter and transfer power through it, then do the same with cobalt and run some tests

It seems you don't understand.
No matter you CPU is made of copper / gold / cobalt , or harry potter magic.
As long as it does not have any physical work done during the process.
All of the energy it "consumes" are turning to heat , 99.99%

You don't have to count how many % of it is made of copper or anything.

When it consumes 228W, it gives 227.99999W of heat.

If you can't get the basics right, rest of it are meaningless.

 

[Shatun_Bear]

Bingo! It's just like I've been saying for years now - Intel was only held back by its node problems, not the architecture; mind that this was pretty much ready to go 3 years ago and it's still slaughtering the best from team (d)red today - now imagine them going against it with puny Zen(+), lmao! Granted, this isn't that great as far as competitive market goes, because Intel will once again hold complete dominance in 2 years or even less and it would have been better if AMD could stay competitive a bit longer, but to anyone paying attention it was obvious that their Zen line was a couple years to late for that to happen and it will be nice to see their insufferable fanboys having seizures, hehe

You are setting yourself up for crashing disappointment if you think this 'slaughters' the competition based entirely on two Cinebench numbers. Because, first of all, the guy that leaked this also shot down someone that said this will destroy Ryzen 5000 with 'no, they're comparable'. Oof. I guess that doesnt fit with the fantasy so carry on.

Secondly, we dont know if this is legit or not, and we certainly do not know all of the context in which these numbers were achieved.

Alder Lake absolutely should be faster than a year old Zen 3 and more efficient to boot, its on a denser, more advanced node, anything less would be a disaster.

 

[Melvis]

Bingo! It's just like I've been saying for years now - Intel was only held back by its node problems, not the architecture; mind that this was pretty much ready to go 3 years ago and it's still slaughtering the best from team (d)red today

Link to the review please?

Ill wait

 

[GURU7OF9]

Logically thinking even with +25% single thread performance I can't see how it could possibly be double rocket lake in multi core performance!
Most likely it will be some kind of intel trickery whether that is avx512 or higher or something else I just can't see it being consistently double rocket lake !
But we shall wait and see !

 

[londiste]

Maybe your magical "cobalt technology" could somehow send the excessive energy into the 5th dimension ?

At the scale the (lower) metal layers of 7nm/10nm processes operate at cobalt becomes a better conductor than copper. Cobalt is already used as barrier layer but Intel was experimenting with having only cobalt and no copper in lower metal layers. This does not really play directly into chip power consumption but better conductor means less waste heat emitted.

Whether Intel actually ended up doing that is not clear. They seemed to be pretty determined but that was at the earlier stages of the 10nm disaster. The suspicion that this cobalt switch may have played a significant part in the failure is not far fetched either - cobalt has not been widely used in this application and the rumors about early Intel 10nm failures were talking about both cobalt and brittle metal lines, sometimes making the connection between the two.

As long as it does not have any physical work done during the process.
All of the energy it "consumes" are turning to heat , 99.99%

Yes. But when you make "wires" out of a material that is a better conductor, you can get by with running less voltage through it.

Logically thinking even with +25% single thread performance I can't see how it could possibly be double rocket lake in multi core performance!

Rocket Lake has only 8 cores. Alder Lake has 8 cores that are 25-30% faster plus 8 cores that are small but pretty unknown in what they are able to do in Cinebench. The second thing is power. Rocket Lake throttles like crazy when power limit is set to TDP and does not run at all that high of a frequency with the 228W PL2. If Alder lake on that 10nm can keep the power something like 30% lower, that gives a bunch of headroom to run the chip faster.