Flagship Model: Core Ultra 9 285K24 cores (8 P-cores + 16 E-cores) and 24 threadsLion Cove architecture for P-cores and Skymont for E-cores76 MB total cache (36 MB L3 + 40 MB L2)Base clocks: 3.7 GHz (P-cores) / 3.2 GHz (E-cores)Boost clocks: 5.7 GHz (P-cores) / 4.6 GHz (E-cores)TDP: 125 W (PL1) / 250 W (PL2)Mid-Range: Core Ultra 7 265K / 265KF20 cores (8 P-cores + 12 E-cores) and 20 threads66 MB total cache (30 MB L3 + 36 MB L2)Base clocks: 3.9 GHz (P-cores) / 3.3 GHz (E-cores)Boost clocks: 5.5 GHz (P-cores) / 4.6 GHz (E-cores)TDP: 125 W (PL1) / 250 W (PL2)Entry-Level: Core Ultra 5 245K / 245KF14 cores (6 P-cores + 8 E-cores) and 14 threads50 MB total cache (24 MB L3 + 26 MB L2)Base clocks: 4.2 GHz (P-cores) / 3.6 GHz (E-cores)Boost clocks: 5.2 GHz (P-cores) / 4.6 GHz (E-cores)TDP: 125 W (PL1) / 159 W (PL2)
>>102470449>E-coresThat's an easy skip for me
>cuck coresnot interested
it feels disingenuous to add all the cores together as if equal.
>>102470547but the cores are split
>>102470449I thought Arrow Lake would be screwed but Zen 5% gave them a chance.
>>102471230The same way Ivy -10% Bridge gave a chance to Bulldozer?
>>102471313Bulldozer was usually slower than Intel's offerings and more inefficient. Intel's CPUs are just more inefficient, but like Bulldozer can sometimes compete on value.The competition is much closer this time, and if the rumors are true that Arrow Lake is focusing more on efficiency things can get interesting.
If the new simplified scheduler actually prevents the cuck cores from getting loaded with heavy tasks then this could be a genuinely good processor, especially since it maintains last gens performance after deleting hyper threading
>>102470449>8 P-cores
>>102470449Are the embedded explosives and hardware backdoors included for free?
>>102470477>>102470481>>102470547>>102472832
>>102472832LMG
>>102470449>P-Cores>E-CoresWhy is this thing exist?
>>102473913Because big.LITTLE is the most popular CPU architecture on the planet?
>>102470449>ultra mega 9000>ecoreslmfao
>>102470449avx512 status?will it be disabled just like in raptor lake?
>>102470449Who is this for?
>>102473913e-cores are 1/4 the size of p-core but they only sacrificing 33% performance. its a really good way to achieve best price to multicore performance way better value than ccx that amd uses.
reminder that Intel-aviv has the ability to remotely detonate your cuck core laden cpu if you buy from them...
>>102475428>E-Cores>Explosive Coresdear God it's so obvious now
>>102475061but how does software choose what kind of core a particular thread goes to? seems like an unsolved problem they are hoping nobody notices
>>102470449These are going to mog the new ryzen lineup, but i like not having a space heater.
>>102470449Not falling for the estrogen cores again
>>102470449but what does this mean for Intel Arc?
>>102470449why are there E-cores in a flagshipbetter question, why is the ratio of these E-cores 2:1 to P-coresintel is dead
>>102478098>DEI corescost the same do half the work
>>102470449oh boy I can't wait for the super ultra cora AI spectacular 395K with 32 E-Cores. it's looking like the E-Cores is just a long con to eventually transition to 100% E-Cores and drop the nomenclature.
why have they completely changed the names
>>102473913Having a core that is inefficient in die area and power usage but is performant, alongside a core that is efficient in die area and power usage at the cost of performance, is an ideal mix so you can assign the right kinds of threads between them. AMD understands this as well, having neutered Zen 4c cores which fulfill the same role, but AMDrones still haven't updated their script from the Zen 3 era.
>>102475991You don't really.The problem with X86's software ecosystem is that any given task is usually unconstrainted by anything but CPU speed.You open a browser, it immediately gets processed on the highest performance core as the browser isn't waiting on something else like a network connection. Only when the browser is done loading will it move to an efficiency core as normal operation is now more likely waiting to be served web resources which is more time dependent.Android which has had heterogenous core groups for ages now has an API that lets applications recommended what type of cores it should be schedule to. Browsers and regular apps may ask to be put on a "little" core, more interactive apps and light games may request "mid" cores, full fat heavy games may request "big" cores.
>>102478523no actually it isnt, the reality is that intelaviv moved to ecores because the transistor density of intel 7 (10nm) is garbage, they can fit nearly 4 e cores in the same area as 1 p core, it is the direct result of not being able to hit 100mtr/mm^2 as originally intended. a 16 pcore design with avx 512 would perform better in 100% of workloads, vs the hybrid design.efficency wise there is little difference between a pcore and an ecore, as the pcores scale with power just as well as the ecores, they just have a higher frequency ceiling.
>>102479198>because the transistor density of intel 7 (10nm) is garbageTransistor density is irrelevant, the concept scales with any node density because it's just a good idea, which is why ARM started it, and both Intel and AMD adopted it for future products. A Heterogenous core design is seen on Meteor Lake's Intel 4, on Zen 4's N4, and on every node imaginable with ARM.>a 16 pcore design with avx 512 would perform better in 100% of workloadsNo, that depends on the workload. E cores are roughly 33-40% as performant as a P core, but you get 4 of them for slightly more than the size of a single P core. A program that could take advantage of the 64-odd E cores that would replace the 16 P cores is probably going to outperform the 16 P cores, and it might even do it more efficiently.>the pcores scale with power just as well as the ecores,At the top end, maybe. At the bottom, not at all.
>>102470449why won't they give us pure p-core and pure e-core offerings e-cores are great at value and multi-core, far more than what the vast majority of the market needs, people who want p-core performance will gladly pay for it, price and electricity wise
>>102479703they released pure p-core sapphire rapids and enterprise said they'd like some ecores
>>102475061>e-cores are 1/4 the size of p-core but they only sacrificing 33% performanceits always been about 33% the size if you look at die shots. skymont got even bigger so its more like 40% the size now
>>102477992the only reason i got one was to get avx512. e-cores weren't even part of the consideration lol
>>102470449e-cores... pfff fff ff... haha... ah no thanks
>>102470449It's very obvious how e-cores are a way for intel to justify charging the same prices for a worse product to appear competitive to their shareholders against AMD.
>>102479596transistor scaling is irrelevant? you litterally cant fit as many cores because of it, as for scaling its pretty clear cut, you take a product like the 12400f , scores 12500 points in cinebench r23 without power limits 12500/4.4 ghz /6 cores × 5.2 × 16 = 39,393 that basically puts it over a 14900k, without avx 512 being used and using the smaller caches that alderlake has at 5.2ghz all core. and yeah they do scale to low power, at the very low end there are some advantages to the ecores, but we're talking sub 15w tdp, which is irrelevant to dekstop chips and anything above that. not to mention that the ecores are clocked so far outside their peak efficiency range on dekstop, that the point is moot.
>>102474267Investors
>>102470449>ultra giga niggadon't care, did they bring back avx512 or am i stuck with 11th gen for another cycle?inb4 all the amdjeets, i'm not buying a 4-way smt cuckoldry
>>102471519Intel is hedging on Arrow Lake to hold the netbook/laptop market their last remaining stronghold in x86. They are banking on efficiency and their long-standing OEM relationships to weather the storm.I can't see Arrow Lake contesting Zen4/Zen5 at desktop environment or above if the rumored clockspeeds are true.
>>102473913It is a short-term cope by Intel to keep-up with AMD in the core-count race until they fully get into chiplets. Monolithic dies cannot scale-up on core-count purely because of economics.
>estrogen coresinto the trash it goes.
You must be obsessive gamers if you hate E cores.The Intel E cores are the reason Intel cpus idle at around 8 watts when you are just using your pc for browsing the web, while AMD cpus idle between 25-40 watts
>>102481106factually incorrect
>>102481106>Intel cpus idle at around 8 watts
>>102481106My 10900k idles at the same 8 watts while having 10 real cores zoomer
>>102481571>>102481711Overkill gaymer boards and piss poor efficiency of most PSUs at idle inflate the numbers quite a bitIn an actually well-designed system, Intel CPUs take dick for power at idle
>>102481571not really
>>102480397alder lake has avx512see >>102479900
>>102479870Yeah but Skymont is Raptor Lake-tier IPC and is also much more efficient than RPL
>>102481947not talking about amd, ecores make no difference to intel's idle power as seen with the 12400f which has no ecores
>>102481926retard they are all using the same psu, and the ecores make 0% difference to intels idle power
>>102480397no avx 512 enabled on consumer chips, early microcode allows you to enable it on some 12000 series with ecores disabled, but later models have it fused off physically. if you want avx 512 , amd or xeon workstation are your only options for an upgrade with proper support.
>>102470449Why are you faggots complaining about the E-cores? They're way better both performance-wise and security-wise. I have yet to see a vulnerability cause by E-cores. Hyperthreading on the other causes not just one, but lots of unneeded vulnerabilities that had to be fixed by slowing down the computer part you bought.Don't get me wrong I don't like intel as a company, but I can respect their decision to not longer use hyperthreading.
>>102482538because they are cuck cores, they are intel atom cores and they suck. hyperthreading isnt going anywhere btw, server will keep it, seems the reason for ditching hyperthreading is mainly power consumption.
>>102475061>>102482538shalom
>nigga doesn't know about Skymont's performance
>>102482538current gen e-cores are only somewhat decent at integer workloads. in fp or latency sensitive loads, they fall flat on their face which is where all the e-core scheduling memes come from
>>102479596It's good on phones because they run off a battery and have problems getting rid of heat. It makes no sense on desktops, maybe on servers but honestly it's so Intel can call a chip 24 core even though it's not 24 of the same core so it's intentionally misleading.
>>102482771it makes sense on desktop (for intel) since intel can ship a chip that uses 12-13 p-cores worth of area that performs like a 16 p-core one in cinebenchthe only good thing that came out of this is that their i5 and i7 cpus have a much larger percentage of the i9's performance without costing more
>>102470449so same shit since 12th gen?what node are they built in?
>>102482751show me how much of that gain is moving from intel 4 to tsmc n3b and then i'll be impressed.
>>102482798I'll be skipping all Intel chips with e-cores from here on out. No thanks.
Arrow Lake will be the best selling CPU of all time. All PC users who have waited for a long time for a great upgrade will finally have to wait no longer. Normies only buy intel and 97% of people are normies. It's so over for the competition.
>>102481106Am I meant to care that my computer is using 20 fucking watts more than it would if I had an intel?Electricity is not that expensive.