WOW…Intel Canada leaks more ‘Raptor Lake’ specifications. Personally I am bit tired of the leaks and overbearing hype on the subject matter. Lets just wait a bit longer until AMD and INTEL will be for real stepping up to the bar and having most of the tech-channels running the chips through the course, paces and benchmarking hoops, including our good friend Steve at Gamers Nexus. Besides I would also like to see a few hundred or even more of these touted halo CPU’s in the hands of actual buyers and see what they have to say!

"accidentally"

P cores for screaming fast serial execution + E cores for super-efficient parallel execution is a dream come true. It should be mathematically provable that you can't beat P+E design.

Noisiv:

P cores for screaming fast serial execution + E cores for super-efficient parallel execution is a dream come true. It should be mathematically provable that you can't beat P+E design.

Calm down there, shareholder, and better pull out your calculator because your math probably isn't lining up. The only reason P cores are screaming fast is because of the absurd boost clocks causing the fans to scream. While I haven't seen any direct IPC comparisons from Zen3 to AL, both Zen3 and AL have a substantial increase over TL, so they're likely close enough in IPC where there's no clear winner. The lines become further blurred since the E and P cores aren't likely to have the same IPC, especially once you account for HT threads. The E-cores, meanwhile, aren't doing much to compensate for how power hungry the P-cores are, when the 5950X offers overall better performance-per-watt than a 12900K, despite the 5950X having all "P-cores". This of course is under full load conditions, because when the E-cores are doing what they're meant to do (handle multitasking efficiently), they do take a lead.

schmidtbag:

Calm down there, shareholder, and better pull out your calculator because your math probably isn't lining up. The only reason P cores are screaming fast is because of the absurd boost clocks causing the fans to scream. While I haven't seen any direct IPC comparisons from Zen3 to AL, both Zen3 and AL have a substantial increase over TL, so they're likely close enough in IPC where there's no clear winner. The lines become further blurred since the E and P cores aren't likely to have the same IPC, especially once you account for HT threads. The E-cores, meanwhile, aren't doing much to compensate for how power hungry the P-cores are, when the 5950X offers overall better performance-per-watt than a 12900K, despite the 5950X having all "P-cores". This of course is under full load conditions, because when the E-cores are doing what they're meant to do (handle multitasking efficiently), they do take a lead.

In gaming 12900k is up to 40-50% more fps per watt than 5950x with fast memory. 😉 For Cinebench gaming, 5950x is nice. For real gaming, 12900k is actual very good in performance per watt, so is 5800x3d.

400W TDP this time, intel?

nizzen:

In gaming 12900k is up to 40-50% more fps per watt than 5950x with fast memory. 😉 For Cinebench gaming, 5950x is nice. For real gaming, 12900k is actual very good in performance per watt, so is 5800x3d.

So... the only applications that exist to you are CB and games played at 720p with low details... Ok then.

schmidtbag:

Calm down there, shareholder, and better pull out your calculator because your math probably isn't lining up. The only reason P cores are screaming fast is because of the absurd boost clocks causing the fans to scream. While I haven't seen any direct IPC comparisons from Zen3 to AL, both Zen3 and AL have a substantial increase over TL, so they're likely close enough in IPC where there's no clear winner. The lines become further blurred since the E and P cores aren't likely to have the same IPC, especially once you account for HT threads. The E-cores, meanwhile, aren't doing much to compensate for how power hungry the P-cores are, when the 5950X offers overall better performance-per-watt than a 12900K, despite the 5950X having all "P-cores". This of course is under full load conditions, because when the E-cores are doing what they're meant to do (handle multitasking efficiently), they do take a lead.

The Pcores really aren't that bad, they are actually quite good at powergating, so when the load is light they don't use that much power, They only start to draw tons of power when the clock speeds are pushed really high and you're using heavy SIMD instructions. zen3 cores really are no different other than the fact they don't clock as high on 7nm tsmc, the primary reason for the e-cores isn't power consumption actually, its performance per mm^2, they use a lot less transistors. so you can put more of them on the chip, thats why the upcoming chips have more of them, if they didn't the die sizes would balloon , amd doesn't have this problem because of their chiplet strategy, with vermeer they put 160mm^2 worth of cores on there + cheap io die, where as intel's alderlake is 210mm^2 , and a bit under half of that is basically graphics and SA/io , so if they wanted a 16 pcore version, it would have to ditch the graphics or make the die very huge, even as it stands now it costs intel more to make a 210mm^2 die than amd's chiplets. This comparison works because intel 10 and tsmc 7 are very similar density wise, if intel had the same chiplet strategy there is a good chance desktop parts wouldn't have the e-cores.

I can't qoute the 40-50% extra fps for intel because i block noisy users, but that is bullshit. The FPS for games are pretty much on par, with usual 5-10% variations. 720p gaming in 2022 with rtx3080 is jsut cinebench with a different name

My 5900X system draws considerably more power than my 9900K system running World of Warcraft at same fps, using the same GPU. I also do blame the mainboard, though (I think its VRM is a pig at idle power).

user1:

The Pcores really aren't that bad, they are actually quite good at powergating, so when the load is light they don't use that much power, They only start to draw tons of power when the clock speeds are pushed really high and you're using heavy SIMD instructions. zen3 cores really are no different other than the fact they don't clock as high on 7nm tsmc, the primary reason for the e-cores isn't power consumption actually, its performance per mm^2, they use a lot less transistors. so you can put more of them on the chip, thats why the upcoming chips have more of them, if they didn't the die sizes would balloon , amd doesn't have this problem because of their chiplet strategy, with vermeer they put 160mm^2 worth of cores on there + cheap io die, where as intel's alderlake is 210mm^2 , and a bit under half of that is basically graphics and SA/io , so if they wanted a 16 pcore version, it would have to ditch the graphics or make the die very huge, even as it stands now it costs intel more to make a 210mm^2 die than amd's chiplets. This comparison works because intel 10 and tsmc 7 are very similar density wise, if intel had the same chiplet strategy there is a good chance desktop parts wouldn't have the e-cores.

I mostly agree; I've said for years that CPU architectures work a lot like an internal combustion engine, where each architecture has a specific "sweet spot" for efficiency. There is a point where the performance-per-watt suffers whether you clock too high or too low. Intel pushes their P cores to very inefficient levels, and back when they used to try to compete with ARM, they underclocked to inefficient levels too (I'm guessing that's still going to be the case considering they're still not really trying to compete with ARM). They can be much more competitive with AMD in performance-per-watt, though from what I recall, Intel models with more reasonable clock speeds are still overall less efficient. Don't get me wrong about the E-cores; unlike a lot of people here, I believe they're the right way forward and they're necessary, even for AMD. I've said in another thread that I would rather have 3 E-cores than a 1c/2t P core, since the 3 E-cores take up less die space yet yield overall better performance and (from what I recall) efficiency. The problem is the naming, because while they are efficient, their efficiency is rather underwhelming. Calling them E-cores artificially limits how high they can clock too, because push them too much higher and Intel would be undermining their namesake. If they were just referred to as "small" cores, that would be less of a turnoff to enthusiasts. A lot of enthusiasts don't want to compromise performance, and calling them E-cores implies it will do that. They may literally be slower and more efficient but they are by no means slow. Anyway, even if Intel went with AMD's chiplet approach, they would still have to make the E-cores. From what I understand (and perhaps I'm wrong), Intel's architecture is still larger per-P-core, a lot of which is thanks to AVX512. Other than some shame from the "glue" hypocrisy (which frankly, they've already done with Core2 Quad), there's nothing really preventing them from also adopting the chiplet design. But, they know doing this will result in worse latency, and that's pretty much the only thing that AMD has no chance of defeating. Intel is already pushing these absurd clock speeds just to keep a performance edge above AMD, and part of me feels like the main driver for these E-cores is so the overall performance-per-watt doesn't look so bad. A P-core will boost to 5GHz+ just to run windows update or watch a video , and since that's pushing the core past it's "efficiency sweet spot", that makes the overall stats look worse. But when an E-core does it, suddenly the performance-per-watt looks way better, because nobody cares if a CPU boosts to perform tasks it isn't even a bottleneck in. Even if you change the power profile to something more efficient, E-cores are still going to be more efficient.

interesting, not sure i want to know what kind TDP they are with those kind of clocks, Last time i actual looked the 13700 like most intresting thing to me 8p+8e and 65 TDP which would be considerable uprade to my 6700k and less TDP too, if i were to upgrade now or had too

TDP shouldn't increase compared to alders.