News AMD Intros Zen 4 Ryzen 7000 CPUs and 600-Series Chipset: Up to 5.5 GHz, 15%+ Performance, RDNA 2 iGPUs, PCIe 5, DDR5

[KyaraM]

AMD used 142W for total package power before and now it is going up to 170W. The Intel chip with the closest clocks is the 12900KS that running no PL (standard for most Intel motherboards) pulls 265W. Worst case vs worst case the 12900KS is still pulling >50% more power than the AMD chip despite having 8P + 8E cores. Overall Intel's Core uArch is not very power efficient especially in higher core and frequencies.

If you run a CPU without power limits, that's on you. Intel specifies the CPU for 150W sustained draw, with 241W being a short duration maximum. Also, other high-end Intel chips like the 12700k are specified for up to 190W short duration and 125W sustained. Amd that is only under full load. They draw close to AMD niveau in games, while usually having more FPS, equalling things out. Personally, I have never seen my 12700k go over 70W. It's draw is actually quite similar to my old 7600k with far better performance.

Plus, as someone stated above, it might be higher based on previous board advertisement. We will have to see.

 

[TerryLaze]

If you run a CPU without power limits, that's on you. Intel specifies the CPU for 150W sustained draw, with 241W being a short duration maximum.

Intel does allow for PL2 to be sustained for ever without calling it overclocking , which is what is causing a lot of confusion.
People think that because intel allows it it automatically has to be the default setting which is just nonsense.
Both settings are valid and allowed and the end user can choose between them.

 

[tennis2]

AMD officially caught up to Intel somewhere between Ryzen 3000 and 5000. Not it seems they're running into the same issues. >>> power to increase clock speeds due to lackluster IPC gains.

Good to see the IGP though. Doesn't have to be a low-end dGPU replacement. Just needs to work for general office use.

 

[jeremyj_83]

The way that it's worded in the article the 170W could be the equivalent of the old 105W and not the 142W.
Current boards are advertised as taking CPUs up to 105W.

In any case, it's way too early and we will have to wait and see.

Later in the article is states "To that effect, AMD also boosted the maximum power delivery of the AM5 socket that will house the Ryzen 7000 chips to 170W, a 28W increase over the previous-gen AM4 socket's 142W peak." That to me sounds like the Total Package Power for AM5 is 170W.

 

[jeremyj_83]

If you run a CPU without power limits, that's on you.

If you are building your own system, Intel motherboards default to infinite Tau and you have to go into the BIOS to change that. Most people are never going to go into the BIOS to change that setting. Heck most people avoid the BIOS entirely unless it is to set XMP on first install. You cannot say "that's on you" when the default behavior is the opposite.

 

[TerryLaze]

Later in the article is states "To that effect, AMD also boosted the maximum power delivery of the AM5 socket that will house the Ryzen 7000 chips to 170W, a 28W increase over the previous-gen AM4 socket's 142W peak." That to me sounds like the Total Package Power for AM5 is 170W.

Yes but that's toms interpretation which doesn't have to be correct, the wording is not clear and that's my point in the first place.

 

[javiindo]

AMD used 142W for total package power before and now it is going up to 170W. The Intel chip with the closest clocks is the 12900KS that running no PL (standard for most Intel motherboards) pulls 265W. Worst case vs worst case the 12900KS is still pulling >50% more power than the AMD chip despite having 8P + 8E cores. Overall Intel's Core uArch is not very power efficient especially in higher core and frequencies.

They are not using the same production node (5 vs 10). And Intel has an hibrid architecture that it should help.

 

[-Fran-]

After mulling over the announcements and reading between the lines, Intel should make sure Meteor Lake does not hold any punches. Zen 4 is going to deliver in more fronts than just better CPU performance.

There's a few interesting bits, which I'll just directly call out as the big elephants in the room and or the in-between-the-lines information:
1- RDNA2 iGPU (notice they're not classed as APUs), which brings parity to the lack of "iGPU to diagnose". Those RDNA2 iGPUs will most likely be like 8 CUs at best and just have the very basic of elements to get video out and most of the silicon will be used for ML/AI stuff alongside acceleration. Also, DP2.0 and up to 4 connections may imply they can drive up to 4 monitors from the iGPU; that is actually not trivial for productivity stuff.
2- Then you have the no-mention of V-Cache, which is undoubtedly the thing Intel should keep in mind when pricing and segmenting Meteor Lake. They already know AMD can blow their socks if they want to by just getting a "Zen4-3D" out there, so even if Ry7K does not match Alder Lake in every aspect, or flat out beat it, then they can still use (and will if needed) the V-Cache card.
3- Zen4 performance. This is a really interesting one and as others in other places have mentioned: AMD is playing poker here. They just let everyone know Zen4 is at least able to beat and/or keep up with Alder Lake in several areas where they could just edge it out or flat out lose. The 15% "uplift" is poking the bear. Why? Because it is a baseline. What you should be reading from there is this: "the slowest Zen4-based CPU will be at least 15% better than any Zen3-based CPU". Or at least, that is how I interpret it. And last in this bit is the 31% better time for Blender. As I read somewhere else with a bit of napkin math: "Actually, the 31% uplift in CineBench over the 12900k is an understatement. Greymon55 pointed out that the Zen 4 CPU completed the CB rendering in 204 seconds, whereas the 12900k CPU completed it in 297 seconds, if you look at the footnotes. That is a 31% reduction in time required to render, but that is actually equivalent to a 45.6% increase in processing speed. To reduce overall rendering time by 31% you need 46% faster speed". So, you can say that the new 16c/32t Zen4 CPU is about 45% better than the 12900K in MP. That being said, Blender does favour AMD a tad, so even the 5950X pulls ahead of the 12900K here, but marginally. You can extrapolate more from this and Intel should, as I said before, keep this in mind.
4- They blow it out of the water in connectivity (for mainstream, that is); plain and simple. I mean 24 PCIe5 multipurpose lanes (the "multipurpose" is important here) is huge; keep in mind this is equivalent to 48 PCIe4 lanes in a mainstream platform. That is HUGE, period. This also makes me believe they are making sure the first iteration of the 600-series chipsets don't have the same problem as 300 and 400-series. Or at least, they suffer much less from it in the eye of the multitude of people decrying "boot muh PCIe5!". USB4 being a rare omission though, but I'm sure this will be implemented before the 700-series chipsets roll out. Any OEM/AIB can just add USB4 and just use some of the PCIe5 lanes and still have plenty for more stuff.
5- They were really vague on B650 and no mention of an A-series chipset is also weird and interesting. I'm still not entirely sure what to make of it, but I think they're also keeping something there under wraps. I just hope they don't gimp B650 too much.

Overall, I would have liked to see explicit mentions of more things, but I can't blame AMD for playing it safe-ish and not giving Intel too much information. Same with the lack of RDNA3 information. Which, to be honest, was the biggest glaring omission for me. I was expecting AMD to slap Intel with something, but they were dead silent and this also means they do not want anyone know how RDNA3 is cooking, which is odd to not even mention it.

Regards.

 

[jeremyj_83]

They are not using the same production node (5 vs 10). And Intel has an hibrid architecture that it should help.

Then compare that with Zen 3. Zen 3 has a 142W max, unless PBO is used, and the Intel in AVX2 with 16c/24t uses 272W. Those are on more or less the same process node (Intel 10nm = TSMC 7nm). The 8e cores us 48W total and just the 8p cores use 239W. This is with the 12900K which will have slightly lower power consumption since the clock speed is lower than the KS. https://www.anandtech.com/show/1704...hybrid-performance-brings-hybrid-complexity/4 Any way you look at it the big Intel cores are power hungry. They are not power efficient beyond 4 cores and like 4GHz clock speed.

 

[KyaraM]

If you are building your own system, Intel motherboards default to infinite Tau and you have to go into the BIOS to change that. Most people are never going to go into the BIOS to change that setting. Heck most people avoid the BIOS entirely unless it is to set XMP on first install. You cannot say "that's on you" when the default behavior is the opposite.

My mainboard asked me for the preferred power setting when I had to reset the BIOS to optimized defaults due to XMP issues after Windows install. Since I had the system built by someone else due to time constraints, I can't say how it was before, but I would assume the same. Also, "most people" aren't doing much more than gaming, either, which, again, is far more frugal than rendering something, for example. And if you want to use that computer commercially, you either got IT who hopefully knows what they do, or better read up stuff yourself.

 

[neblogai]

1- RDNA2 iGPU (notice they're not classed as APUs), which brings parity to the lack of "iGPU to diagnose".

'iGPU to diagnose' is nice to have. But where it matters- is that now OEMs will no longer have to include a dGPU when building business PCs, and that lowers cost.
Another important niche is gaming and productivity laptops- we can expect these chips to be used next year for top gaming/productivity laptops. You have to have an iGPU for low power use when idling, otherwise battery life of the laptop will be junk.

 

[TerryLaze]

Then compare that with Zen 3. Zen 3 has a 142W max, unless PBO is used, and the Intel in AVX2 with 16c/24t uses 272W. Those are on more or less the same process node (Intel 10nm = TSMC 7nm). The 8e cores us 48W total and just the 8p cores use 239W. This is with the 12900K which will have slightly lower power consumption since the clock speed is lower than the KS. https://www.anandtech.com/show/1704...hybrid-performance-brings-hybrid-complexity/4 Any way you look at it the big Intel cores are power hungry. They are not power efficient beyond 4 cores and like 4GHz clock speed.

So you are basically comparing one platform that has a power limit enforced at 142W with a platform that has no power limit enforced. Of course the one with the power limit enforced will use less power, that has nothing to do with efficiency though.
The 12900k tops out at 241W, anything above that is overclocked and will void your warranty.

Here you can see the 12900k running at 125W, the 5950x at 142W is about 12% faster in cinebench 23, and 142W is also about 12% more power than 125W.
In the last table of the page they show the 12900k needing 160W to get 12% faster than the 125W setting, still extremely close to the efficiency of the 5950x and even with fewer cores available.
https://www.hardwareluxx.de/index.p...-desktop-cpus-alder-lake-im-test.html?start=8

 

[SunMaster]

They are not using the same production node (5 vs 10). And Intel has an hibrid architecture that it should help.

Why do you assume Intel's hybrid architecture "should help". Choosing a dual type of core was Intels solution to solve their problem vs AMDs massive performance lead. I've seen no proof/indication/evidence/insert_wording_here that such a setup, by design, has the performance upper hand.

You could turn it around and say that Intel wasn't capable of fitting the same number of high performing cores on the available die area. Then there is the power consumption issue of the performance cores, and how performance does not scale linearly with power consumption,

 

[jeremyj_83]

So you are basically comparing one platform that has a power limit enforced at 142W with a platform that has no power limit enforced. Of course the one with the power limit enforced will use less power, that has nothing to do with efficiency though.
The 12900k tops out at 241W, anything above that is overclocked and will void your warranty.

Here you can see the 12900k running at 125W, the 5950x at 142W is about 12% faster in cinebench 23, and 142W is also about 12% more power than 125W.
In the last table of the page they show the 12900k needing 160W to get 12% faster than the 125W setting, still extremely close to the efficiency of the 5950x and even with fewer cores available.
https://www.hardwareluxx.de/index.p...-desktop-cpus-alder-lake-im-test.html?start=8

Wow 12% more performance but needing 28% more power to do so...that really sounds efficient. FYI you are also reading the information for the 5950X incorrectly. They measured a total package power of 120.5W for the 5950X under the Cinebench r23 nT test. At 4.5W LESS (3.73%) than the 12900K @ 125W the 5950X performed 8.79% better. Using PBO generated 2.21% higher performance for 1.91% more power on the 5950X. That is efficiency. This is why when you look at the score/watt the 5950X is 12.84% better than the 125W Intel. The Intel needs 160W to be 3% faster but use 33% more power than the Ryzen. Calculated to their efficiency ratings (score per watt) the 12900k @ 160W gets 159.94 that puts it at 28.87% less efficient than the 5950X.

 

[2Be_or_Not2Be]

'iGPU to diagnose' is nice to have. But where it matters- is that now OEMs will no longer have to include a dGPU when building business PCs, and that lowers cost.
Another important niche is gaming and productivity laptops- we can expect these chips to be used next year for top gaming/productivity laptops. You have to have an iGPU for low power use when idling, otherwise battery life of the laptop will be junk.

Yeah, have a built-in GPU will certainly get AMD more OEM design wins. I think Intel CPUs were used a lot (Dell, HP, etc.) just for that reason alone. Business systems still usually have life-cycle replacements, so that should be some steady gains there. I'll be happy to have more choices when it comes to AMD-powered business PCs.

 

[2Be_or_Not2Be]

I'm a bit confused by the article saying:
"The RDNA 2 engine supports up to four display outputs, including DisplayPort 2 and HDMI 2.1 ports. "

...then later:
"Socket AM5 motherboards support up to four display outs via HDMI 2.1 Fixed Rate Link (FRL) and Displayport 1.4 High Bit Rate 3 (HBR3) outputs, powered by the RDNA 2 graphics engine onboard the 6nm I/O die in the Ryzen 7000 processors. "

So is it DisplayPort 2.0 or 1.4? I definitely want 2.0 across the board, so I hope this is just a typo in the article. I'd hate to think that the m/b makers were just continuing with 1.4 outputs, even if 2.0 capability is now available.

 

[rluker5]

After mulling over the announcements and reading between the lines, Intel should make sure Meteor Lake does not hold any punches. Zen 4 is going to deliver in more fronts than just better CPU performance.

There's a few interesting bits, which I'll just directly call out as the big elephants in the room and or the in-between-the-lines information:
1- RDNA2 iGPU (notice they're not classed as APUs), which brings parity to the lack of "iGPU to diagnose". Those RDNA2 iGPUs will most likely be like 8 CUs at best and just have the very basic of elements to get video out and most of the silicon will be used for ML/AI stuff alongside acceleration. Also, DP2.0 and up to 4 connections may imply they can drive up to 4 monitors from the iGPU; that is actually not trivial for productivity stuff.

4- They blow it out of the water in connectivity (for mainstream, that is); plain and simple. I mean 24 PCIe5 multipurpose lanes (the "multipurpose" is important here) is huge; keep in mind this is equivalent to 48 PCIe4 lanes in a mainstream platform. That is HUGE, period. This also makes me believe they are making sure the first iteration of the 600-series chipsets don't have the same problem as 300 and 400-series. Or at least, they suffer much less from it in the eye of the multitude of people decrying "boot muh PCIe5!". USB4 being a rare omission though, but I'm sure this will be implemented before the 700-series chipsets roll out. Any OEM/AIB can just add USB4 and just use some of the PCIe5 lanes and still have plenty for more stuff.

1- I completely agree on the IGPU. They are very handy to have if, like most people, you don't play video games. They just have to put up a good desktop and play whatever videos you want them to.

4- I disagree here. AMD is just catching up and I don't think their added connectivity will make much difference.
My pc -as an example- is using a low end z690, Ausus Prime -P. I have a 3080, 4 Optane drives, soundcard, 2 SSDs, an HDD and am not using a pcie4x4, pcie 3x4, sata plug, 2-usb2 plugs, a usb3 plug, a usb c plug, 2 ram slots, or my thunderbolt header.
I've heard Raptor will have a pcie5 m.2 in addition to the gen5 gpu. How much connectivity can you put on a motherboard and even use? I think I'm already over the top and still have a bunch left over on the cheapest z690 I could find at the time I got it. I can't even think of a decent use for my leftover ports. Is mgpu coming back? Do people need the back of the mobo full of m.2 slots? I think the point of diminishing returns has already been passed by Intel and nobody would notice in real life if Zen5 came with 64 pcie6 lanes. The Zen4 mobos I've seen seem exactly on par with the year older Intel ones with the exception of gen5 m.2s. Hopefully they cost less.

 

[-Fran-]

1- I completely agree on the IGPU. They are very handy to have if, like most people, you don't play video games. They just have to put up a good desktop and play whatever videos you want them to.

4- I disagree here. AMD is just catching up and I don't think their added connectivity will make much difference.
My pc -as an example- is using a low end z690, Ausus Prime -P. I have a 3080, 4 Optane drives, soundcard, 2 SSDs, an HDD and am not using a pcie4x4, pcie 3x4, sata plug, 2-usb2 plugs, a usb3 plug, a usb c plug, 2 ram slots, or my thunderbolt header.
I've heard Raptor will have a pcie5 m.2 in addition to the gen5 gpu. How much connectivity can you put on a motherboard and even use? I think I'm already over the top and still have a bunch left over on the cheapest z690 I could find at the time I got it. I can't even think of a decent use for my leftover ports. Is mgpu coming back? Do people need the back of the mobo full of m.2 slots? I think the point of diminishing returns has already been passed by Intel and nobody would notice in real life if Zen5 came with 64 pcie6 lanes. The Zen4 mobos I've seen seem exactly on par with the year older Intel ones with the exception of gen5 m.2s. Hopefully they cost less.

Wait, so you're one of the people that after AMD actually announces a plethora of extra connectivity over Z690 it's just going to shrug it off and say "meh, it's enough at Z690". Am I reading your comment correctly?

If so, them I'm sorry to say, but that's BS of the highest caliber. Z690 is a step up over X570, but not a massive one if you read the fine print. Sure, it supports PCIe 5 GPU (only) and allows for a split for a PCIe NVMe card to support PCIe 5 NVMe at the expense of the GPU link, but AMD's X670(E) won't do that: it's going to on top of giving you full PCIe5 for the GPU, it'll also give it for the NVMe slot (maybe 2 or just 1; dunno as it wasn't explicit) and the rest is going to be using PCIe5 for everything. Z690 uses PCIe4 equivalent for the DMI link and everything else. Whether you want to admit or not, X670(E) is a proper upgrade over X570 in all accounts and all regards that matter for any consumer. If you don't need all that connectivity, then B650 is for you, which if you read the fine print as well, it'll be slightly below the Z690 in terms of connectivity by trading the PCIe5 for the GPU for the NVMe. Again, and this is IMPORTANT: if Z690 wants to use a PCIe5 NVMe, the motherboard maker will need to include an additional chip to bridge the CPU link that splits the PCIe5 of the GPU in two (dual X8) so it can accommodate it.

Do I agree with the overall picture of "most users don't need this?". Yes, kind of. Have most outlets and people made it a HUGE FRIGGEN DEAL that X570 doesn't have PCIe5 (with a huge asterisk, again)? Absolutely they have.

Regards.

 

[Awev]

In the past it had been observed that Intel spent so much time on the 14nm die that it we where running out of space for all of those "+"s, yet all of the experience they gained meant that it ran at about the equivalency of a 10nm die. Just imagine if they do develop in-house 3nm dies in the next two years, and then spend a couple more years perfecting it, and allowing it for desktop CPUs, not just server or HPC uses.

I am glad to see that AMD has not conceded much to Intel, and is jumping back in with both feet. It was sad to see Intel take the lead with PCIe 5, yet they are not the only game in town - just not enough people have taken advantage of PCIe 4 yet to justify the move to the next gen unless you are in a high-end business that needs it.

I welcome the ARM, Apple, and RISC-V chips, it means that chip development is moving forward, and if you are not careful, you can be left behind in the dust. Not everything is based on the x86 instruction set, which helps make for good competition.

One of my cats spilled water into my computer earlier this year, so I ended up having to replace so much that more than half of it looks new. There goes any budget I may of had for a new computer this year (came early as a rebuilt unit). Well, I was planning to wait for at least 13th Gen Intel or Zen 4/5 8000 series before building a new PC, and seems as that might be a good idea. Things are progressing nicely, even if it is not a 15 - 20% IPC over the prior Zen factor. I do like the fact that it will now have RNA 2, and not a reheated Vega chipset (which has been better than Intel's iGPU) for graphics - my biggest complaint with the current APUs.

 

[rluker5]

Wait, so you're one of the people that after AMD actually announces a plethora of extra connectivity over Z690 it's just going to shrug it off and say "meh, it's enough at Z690". Am I reading your comment correctly?

If so, them I'm sorry to say, but that's BS of the highest caliber. Z690 is a step up over X570, but not a massive one if you read the fine print. Sure, it supports PCIe 5 GPU (only) and allows for a split for a PCIe NVMe card to support PCIe 5 NVMe at the expense of the GPU link, but AMD's X670(E) won't do that: it's going to on top of giving you full PCIe5 for the GPU, it'll also give it for the NVMe slot (maybe 2 or just 1; dunno as it wasn't explicit) and the rest is going to be using PCIe5 for everything. Z690 uses PCIe4 equivalent for the DMI link and everything else. Whether you want to admit or not, X670(E) is a proper upgrade over X570 in all accounts and all regards that matter for any consumer. If you don't need all that connectivity, then B650 is for you, which if you read the fine print as well, it'll be slightly below the Z690 in terms of connectivity by trading the PCIe5 for the GPU for the NVMe. Again, and this is IMPORTANT: if Z690 wants to use a PCIe5 NVMe, the motherboard maker will need to include an additional chip to bridge the CPU link that splits the PCIe5 of the GPU in two (dual X8) so it can accommodate it.

Do I agree with the overall picture of "most users don't need this?". Yes, kind of. Have most outlets and people made it a HUGE FRIGGEN DEAL that X570 doesn't have PCIe5 (with a huge asterisk, again)? Absolutely they have.

Regards.

X670E will compete with the next chipset for Raptor that will have your precious gen5 ssd slot off of the cpu.
Hopefully by then there will be a pcie gen5 ssd made for you to use so you can get close to the real world performance of a gen3 optane drive so long as you use active cooling for your ssd. But you want several? Guess you will need some custom watercooling for them because air cooling won't fit when you have a gpu installed. Want more pcie slots? You will have to install them yourself because the motherboard manufacturers are leaving lanes unused because you are the only one that thinks this makes a massive difference.

The X670E motherboards are there to see. Compare them to the existing Z690. Show me a single X670E that blows any Z690 out of the water in connectivity. Raptor will have 20 pcie5 + 8pcie4 for dmi vs Zen4's 24 pcie5. If you account for Raptor's gen5 m.2 then not really blown out of the water.

Going crazy and using caps over some doesn't yet exist ssd that needs active cooling just to be second best in real world performance. I'm guessing you have had both sata and nvme drives. Notice a huge difference in everyday use?

 

[-Fran-]

X670E will compete with the next chipset for Raptor that will have your precious gen5 ssd slot off of the cpu.
Hopefully by then there will be a pcie gen5 ssd made for you to use so you can get close to the real world performance of a gen3 optane drive so long as you use active cooling for your ssd. But you want several? Guess you will need some custom watercooling for them because air cooling won't fit when you have a gpu installed. Want more pcie slots? You will have to install them yourself because the motherboard manufacturers are leaving lanes unused because you are the only one that thinks this makes a massive difference.

The X670E motherboards are there to see. Compare them to the existing Z690. Show me a single X670E that blows any Z690 out of the water in connectivity. Raptor will have 20 pcie5 + 8pcie4 for dmi vs Zen4's 24 pcie5. If you account for Raptor's gen5 m.2 then not really blown out of the water.

Going crazy and using caps over some doesn't yet exist ssd that needs active cooling just to be second best in real world performance. I'm guessing you have had both sata and nvme drives. Notice a huge difference in everyday use?

You missed the point by like 300 nautical miles. I'll stop here.

Regards.

 

[escksu]

I mixed up cinebench and blender but the point still stands.12900k = 5950x in blender. Wheres the 30% performance coming from? 3% IPC? LOL.

the cinebench test could have easily been done at the same clock speeds. They've made ipc comparisons before doing clock for clock tests

Clockspeed is the answer.

In blender, all cores are utilised. Under stock condition, clock speed for all cores are much lower than running 1-2 cores. 5950x has just 105w tdp. This obviously limit its performance when all cores are utilised.

Now zen4 will have higher tdp and under all cores situation, it will be able to run at much higher clockspeeds.

Note thay 5950x has 4.9ghz boost but base is just 3.4ghz... the new zen likely have 5.5ghz boost.

 

[escksu]

Looks like AMD's marketing slides gotten many pple confused. But cant blame you all, those slides are rather misleading.

One thing for sure is that we wont see much IPC increase for zen4. Most of it will be clocks. If there is big increase in ipc, amd will put it in their slides, if not, nope... Not there (their usual marketing strategy for decades). Just like how amd posts huge performance/watt numbers but actual performance numbers are much smaller. Many are confused between this and actual performance.

Also, single thread performance arent the same as all cores. The TDP will limit how fast these cores can run when they are all utilised.

 

[escksu]

If you run a CPU without power limits, that's on you. Intel specifies the CPU for 150W sustained draw, with 241W being a short duration maximum. Also, other high-end Intel chips like the 12700k are specified for up to 190W short duration and 125W sustained. Amd that is only under full load. They draw close to AMD niveau in games, while usually having more FPS, equalling things out. Personally, I have never seen my 12700k go over 70W. It's draw is actually quite similar to my old 7600k with far better performance.

Plus, as someone stated above, it might be higher based on previous board advertisement. We will have to see.

That was prior to alderlake. With Alderlake, Intel nows allow full PL1 (241W) power draw indefinitely as long as the board can support and the heatsinks are able to handle it.

Btw, 241W (190W for 12700K) is not the power consumption, its just the power headroom. So, what this means is that your 12700K can use up to 190W but it doesnt have to do it all the time. Also, certain conditions must be met.

 

[escksu]

AMD used 142W for total package power before and now it is going up to 170W. The Intel chip with the closest clocks is the 12900KS that running no PL (standard for most Intel motherboards) pulls 265W. Worst case vs worst case the 12900KS is still pulling >50% more power than the AMD chip despite having 8P + 8E cores. Overall Intel's Core uArch is not very power efficient especially in higher core and frequencies.

265W is basically "PL" mode. The difference is that now Intel allows CPU to run in this PL mode indefinitely as long as board and cooling solutions are capable of sustaining it.

As for 50% more power.... Well, it depends. Its not like people are running full load all the time.