Press "F" For Graphics: Leak Suggests Intel "F" Lineup Returns For Alder Lake

Intel 12th Generation Alder Lake CPU
(Image credit: Intel)

Intel's Alder Lake lineup of processors was detailed last week at Intel's Architecture Day, providing us with further details about the first x86 hybrid architecture with big and little cores packed onto one die. The company mentioned that these CPUs will be accompanied by Xe GPUs. However, according to the well-known industry leaker Komachi Ensaka, we have come to know that Alder Lake will bring back the "F"  versions without integrated graphics.

Alder Lake is packing many types of processing elements onto one package. There are smaller, low-power, efficient cores called E-cores, and higher-performance P-cores. Alongside the usual I/O elements, Intel also embeds the Xe graphics in the die.

With the past few generations, Intel introduced a concept of the "F" series of processors, where the F model has integrated graphics disabled. 

According to Komachi (@KOMACHI_ENSAKA) on Twitter, the person who is a well-known hardware leaker in the hardware community, Intel has prepared a series of F models based on Alder Lake design. There are lots of F models listed there, starting from Core i5-12600KF, Core i7-12700KF, and Core i9-12900KF. All these F processors will come with the standard core configuration of the K model (Core i5-12600K, Core i7-12700K, and Core i9-12900K), just with the internal graphics being disabled and still present on the die. 

As we already know, the F models have their iGPU disabled, and that isn't necessarily a bad thing. As the integrated GPU produces heat, some users with overclocking intentions could benefit from the disabled iGPU as there is no heat output coming out of it. Intel could also introduce some price cuts compared to the regular models, allowing for the PC budget to go elsewhere.

  • Giroro
    I wonder how much of an upcharge Intel thinks they can get away with for their iGPU. It will probably be at least $100, this time.

    Intel is pushing the 8-core 12900k as "16 core" for a reason. My guess is because they are desperate to find any way to price it like a 5950X.
    Intel marketing is acting like they legitimately believe a mainstream audience is willing to pay over $700 for a "slightly better than i7" class CPU.
    There's such heavy use of desperate-feeling semi-deceptive smoke-and-mirrors in their marketing slides so far, that it's worrying.
    For example they still have made no claims to the real performance of their small E cores, nor even technically claimed that they can be used at the same time as the big cores. It's all been weasely implication. They shove the word "performance" repeatedly into their efficiency claims, apparently to lead people to overly-optimistic assumptions.
    It feels like they're trying to hide something. Maybe it's disappointing performance, maybe it's a big price hike. Possibly both.

    I think with sales dropping on their massive-margin scalable Xeon, they're going to try and squeeze out bigger margins wherever they can.
    Maybe they'll surprise me, but I'm not optimistic for the desktop Alder Lake processors, at all.
    Reply
  • TerryLaze
    Giroro said:
    I wonder how much of an upcharge Intel thinks they can get away with for their iGPU. It will probably be at least $100, this time.
    Based on the 11th gen CPUs with the exact same iGPU costing zero to negative amount of money more?
    Giroro said:
    Intel is pushing the 8-core 12900k as "16 core" for a reason. My guess is because they are desperate to find any way to price it like a 5950X.
    Intel is very clearly stating up to and is even telling you that it's 8+8
    Also they are all cores, they aren't all the same cores but they are all full cores, unlike the shared resource thing that AMD did intel can legally and even ethically call it that. The fact that they are clearing it up that they are different cores makes it even better.
    https://www.tomshardware.com/features/intel-architecture-day-2021-intel-unveils-alder-lake-golden-cove-and-gracemont-cores
    Giroro said:
    Intel marketing is acting like they legitimately believe a mainstream audience is willing to pay over $700 for a "slightly better than i7" class CPU.
    You have no amount of idea on how this will perform or how it will be priced, for all you know it might outperform the 5950x and still retain the max intel desktop price of about $550 that the 11900k holds.
    Giroro said:
    There's such heavy use of desperate-feeling semi-deceptive smoke-and-mirrors in their marketing slides so far, that it's worrying.
    For example they still have made no claims to the real performance of their small E cores, nor even technically claimed that they can be used at the same time as the big cores. It's all been weasely implication. They shove the word "performance" repeatedly into their efficiency claims, apparently to lead people to overly-optimistic assumptions.
    Well, what word should they shove into an efficiency claim? "beauty" ? It's how much performance you get for how much energy. How could they not use the word performance?
    Also they did show what kind of real world performance you can expect from them.
    Reply
  • Giroro
    TerryLaze said:
    Based on the 11th gen CPUs with the exact same iGPU costing zero to negative amount of money more?

    We all know pricing is highly variable right now, and street pricing is all over the place in ways that often makes no sense whatsoever. Usually when the F part is more expensive than the non-F, it's because the F part is being sold at some elevated price that nobody should pay, while the non-F is still near MSRP
    But, the MSRP of the non-F 11th gen processors is indeed $25 higher than the F parts.
    I believe Intel exec is going to look at a $150 GT 1030 and think "our iGPU is about that good, lets get in on that action".


    TerryLaze said:
    Intel is very clearly stating up to and is even telling you that it's 8+8
    Do you think that your parents are going to understand that difference when they see a giant "16 Core" sticker on a computer at best buy, or are they going to think that this new computer is exactly twice as good as an 8 core processor? We still don't even know that 8+8 is even an accurate statement, because there is only implication that all cores can run at once, not any kind of proof or confirmation.


    TerryLaze said:
    You have no amount of idea on how this will perform or how it will be priced, for all you know it might outperform the 5950x and still retain the max intel desktop price of about $550 that the 11900k holds.

    Obviously all I'm doing is making an educated guess. If you don't like analysis and speculation, then maybe don't click on pre-release media for anything ever.


    TerryLaze said:
    Also they did show what kind of real world performance you can expect from them.

    That chart is 100% my point when I call the Intel presentation misleading smoke and mirrors. This is one of the charts that bothers me deeply.
    That is not a performance claim, it is an efficiency curve. They put the word performance on the chart 5 times to make you think it is about performance, but it is not about performance. If you think this chart is a performance claim, then they have succeeded in misleading you.

    They want to lead you into thinking that the small cores will either be 80% more performant than Skylake, or that the cores will have the same performance as maxed-out Skylake while using far less power (Really, intel probably wants you to think the cores will be both at the same time).

    But think about what skylake (assuming 6th gen) actually was. An i5-6600k was a 95W 4c/4t part. So you're looking at around 20+ Watts per core at the high side of that curve. If Intel were running 8 E cores at 20 Watts per core, then that would be a part that runs in the rough ballpark of ~180 Watts before even considering the 8 full P cores. Of course, Intel is actually comparing it to their 2C/4T parts, which were closer to 25W per core in desktop.

    So, obviously, Alder Lake cannot possibly be running their E cores at anywhere close to the high end of that performance chart. They are implying a configuration that simply is not possible.

    Even if they are running at 20% of that power for "same" performance, than you're talking on the scale of ~36 Watts which is still a very significant proportion of the power budget to take away from the big P cores.

    Realistically, the E cores are mobile Atom cores, which are going to be allocated low Atom-like levels of power. It is likely going to vary a lot between different parts, but it has to be on the scale of 1-2W per core. Definitely not 20W, and likely not even 5W. So the overall performance must be somewhere in the extreme bottom-left of that efficiency curve, not the top-right.

    Here's another way to think about it: They can fit 4 E cores in the space of 1 P core. If those 4 cores combined were more performant at multithreading than a single P core, then Alder lake would have been configured with far more E cores (Giving Intel the benefit of the doubt that all cores can be used simultaneously). We would possibly be seeing a 4P/24E/32T configuration to meet thread parity with a 5950X, for example. Since configurations emphasizing more E cores don't exist, it can be assumed that a single E core will have less than 25% of the performance of a P core.
    So, for an E core to be as performant as Skylake, than a single P core will need to have 4x the performance of a single Skylake core. Nobody is expecting that to happen. So far it seems like the Alder Lake P cores will only incrementally improve single-core performance over Rocket Lake, which would put it at somewhere in the range of 1.6x-2x the single-core performance of 6th gen skylake.
    This would put the upper limit of expected performance from an Alder Lake E core at around half the max performance of a Skylake core. Which is, again, the extreme lower-left of the efficiency curve (a little less than 3 lines up from the bottom of the chart, assuming this unlabeled graph starts at 0 and has a linear scale).

    But again, this is all only an educated guess, based on an analysis of what intel is actually saying, as opposed to what they are implying.
    Reply
  • TerryLaze
    Giroro said:
    We all know pricing is highly variable right now, and street pricing is all over the place in ways that often makes no sense whatsoever. Usually when the F part is more expensive than the non-F, it's because the F part is being sold at some elevated price that nobody should pay, while the non-F is still near MSRP
    But, the MSRP of the non-F 11th gen processors is indeed $25 higher than the F parts.
    I believe Intel exec is going to look at a $150 GT 1030 and think "our iGPU is about that good, lets get in on that action".
    They showed no sign in the last two years of doing that, while every other company increased prices. It could happen I guess but all the signs are against it. They even kept prices while lowering margins in the last years to compensate for higher taxes and increased shipping prices due to human malware.
    Giroro said:
    Do you think that your parents are going to understand that difference when they see a giant "16 Core" sticker on a computer at best buy, or are they going to think that this new computer is exactly twice as good as an 8 core processor? We still don't even know that 8+8 is even an accurate statement, because there is only implication that all cores can run at once, not any kind of proof or confirmation.
    Are they going to think that an AMD 16 core is twice as good at email than an intel 8core one?
    This is irrelevant, computers are a specialty market and you have to do your homework to know what's what.

    AMD made cores from different chiplets (basically different CPUs) work together why would you think there would be any issues in making cores on the same CPU work together?
    Giroro said:
    Realistically, the E cores are mobile Atom cores, which are going to be allocated low Atom-like levels of power. It is likely going to vary a lot between different parts, but it has to be on the scale of 1-2W per core. Definitely not 20W, and likely not even 5W. So the overall performance must be somewhere in the extreme bottom-left of that efficiency curve, not the top-right.
    Well they can't guarantee you a fixed power number since the whole idea of CPUs for the last 10 years or so is to use as little power as the situation calls for.
    Did we ever see any atom device with high power draw? That would be the only way to know if they can handle high power.
    Sure it would be nice if they showed us the max power the e cores can scale to but then all the headlines would be -OMG intel at rock bottom e cores to use x (high number) power- , now we have to wait for reviews I guess.
    Reply