Uh?
The i5-10600 is rated at 65W ('TDP') at all-core base frequency.
So was the i3-10100.
Uh ... where did that i3-10100 come from? I responded to your out-of-the-blue mention of the i5-10600, which, in case you forgot, you brought up as a response to me saying the 12600 seems kind of redundant. I honestly don't see what you're getting at at all here. Have I complained about the TDP of this chip? No, I've complained that it doesn't make sense at this point in the product stack, doesn't fit the 12600 name (in light of the spec differences vs. the K), and that this just follows Intel's habit of having way too many SKUs. Whether previous i5s and i3s shared a TDP is entirely irrelevant to that point.
This number is essentially a fiction with fiddled frequencies to meet the target.
You see that this sentence is contradicting itself, right? Yes, TDP is very different from actual power targets. I've never said anything even relating to that. But those "fiddled frequencies" are precisely why they can give chips different TDPs, and why at any reasonably low TDP a higher core count chip will have a lower base clock. Which is a meaningful distinction, as how much you're able to cool is highly variable between different PCs.
Therefore the i3 went as high as 4x3.9 GHz, whereas the i5 had a base of 6x3.3 GHz (or less).
Alder Lake has up to 3.5 GHz for the 4 core TDP now reduced to 60W, and 3.3 GHz for the 6 core.
Again, apparently I have to remind you here: I made a statement about the 12600 seeming rather redundant. You responded to that with a list of SKUs and frequencies, which ... I assume was supposed to contradict that somehow? But which also for some reason left out some of the major differentiating factors between those SKUs, making the list rather useless overall - and still it didn't bring any clarity to why there needs to be 3 6c12t i5 SKUs, or why this SKU shares a name with a 10c16t chip for some reason.
These numbers are not very meaningful in that if someone, let's say Asrock, created a board, that was incapable of dealing with the arbitrary 'TDP' of a K chip (125W), then people scream that it is terrible.
That TDP is e.g. 8 x 3.6 GHz + 4 x 2.7 GHz, as on the 12700K
.... relevance? If a bad motherboard has a bad VRM, does that affect whether or not a CPU SKU is unnecessary?
Intel now provide a second more useful 'TDP', which for the non-K i5s is 117W. This is therefore 6 x 4.8 GHz.
Again: Yes, but relevance? I only brought up TDPs as you failed to account for that as a differentiating factor between the two chips
you brought up in order to contradict
my point.
Meanwhile the 12600K is up to 8 x 4.9 + 4 x 3.6 = 150W
The only problem with Intel's numbers is that they aren't SKU-specific.
I believe the 12600 uses every bit of that 117W, whereas the 12400F can run all day at 4.4 GHz on 75W or less.
I don't know what river you're paddling up currently, but it bears no relation to what I was saying, nor my response to you. Why are we discussing TDPs? My point was about
product segmentation. TDPs play into that, but they are one of many variables, and discussing TDP alone gets us nowhere. Also, do I need to remind you that my first post here said that "Intel are the absolute masters of incremental product segmentation" - does that somehow imply that this hasn't been true up until now? It should really be plenty clear that this isn't new - I'm simply pointing out that this is a particularly egregious example of it.
For Comet Lake and Rocket Lake you had non-K SKUs with low base frequency and a low TDP.
This has literally been how Intel SKUs have worked since Skylake, though arguably since Sandy Bridge: for any model number, a K SKU is unlocked, higher clocked, and might have a higher TDP than the non-K SKU, but they were the same hardware and were configured very similarly outside of this. They've now broken with this system, for no good reason beyond making three near-identical i5 SKUs that perform within a few % of each other. I find that worthy of pointing out.
Since most people just buy a board which ignores the TDP, we now have two TDPs, the same one from Comet/Rocket Lake, and one at full turbo, maxed chip.
Yes, that's a good thing. But also entirely irrelevant to this discussion. Whether previous TDP figures were nonsense or not (they mostly were),
I'm talking about product segmentation.
The i5-12600k should be the i6 or something, and it is quite confusing.
No, the i5-12600 non-K should have been a 10c16t chip or not have existed at all. Then they could have had a couple of 6c12t i5s below that and this would have been a lot less messy.
I assumed it was obvious that the i3 and i5 are different C/T so didnt list that.
So what was the point of the list? To say that different chips have clock frequencies a few % apart from each other? I don't see how that in any way refutes my point about this being a redundant SKU.
Again, clearly there is more justification for having multiple i5 chips in the past in that:
12500 vs 12400 is a totally different GPU (32 vs 24 EUs, 2 encoding engines vs 1)
12600k vs 12600 is a totally different CPU (6+4 cores vs 6)
"totally different GPU" - but they're both uselessly slow, so ... who cares? Sorry, but that's not a meaningful differentiator for anyone otuside of perhaps a few digital signage OEMs. Nobody in the world cares if their desktop CPUs comes with a 24 or 32-core Intel iGPU. Xe is better than their previous stuff, and can compete with Vega when the drivers work, but they generally don't, so that point is moot. This just underscores my point of Intel producing a ton of useless SKUs for no good reason.
whereas in the past it was literally only down to clock speeds and the TDP you cite could be done with BIOS settings anyway so not significant.
... which is kind of my point, no? That Intel is creating an ever-increasing number of undifferentiated SKUs that have no meaningful differences? I mean, you're actually here making excuses for them ("this one has a marginally faster iGPU!"). You're also acting as if this segmentation isn't
entirely by choice. There's nothing forcing Intel to have three non-K i5 SKUs whatsoever - and again, nobody cares about that iGPU. Nobody. If this 12600 was called the 12500 and the 12500 didn't exist, things would look a lot more sensible.
It annoys me that the i5-12600 doesn't have the same core configuration as the i5-12600K. IMO, they should always have the same core configuration if they have the same model number +/- the K. The clock speed can be slightly different, but the core configuration should be the same.
Yep. Breaking this system is beyond stupid.
That isn't true at all. The E cores are designed completely different than the P cores.
Again, entirely true. No idea what
@ThrashZone is on about here, but E cores are an entirely different architecture than P cores.
It's a solution to a problem that does actually currently exists. Governments, in their never ending fight to reduce energy usage, are implementing regulations that computers now have a power limit when they are sitting idle or under light load, even desktops. So the E-cores allow more powerful P-Cores(and more of them) while still keeping the computer under the power limits. It's stupid on the government's side, but that's an entirely different discussion. And the new laws have already made some PC manufacturers pull certain high end models of their computers from the markets those laws cover.
Sorry, but this is pure nonsense. You seem to have bought into some of the
sensationalism and misinformation that got tossed around a while ago when some new environmental regulations (that OEMs had known about for
years) came into effect, causing non-compliant OEMs to halt sales of certain models. Failure to comply with these regulations is
only the responsibility of said OEMs, as compliant components were plentiful and they had several years' notice. The only computers pulled from the market were also pulled due to using low efficiency PSUs, and not because of the power consumption of any of their other components.
As for the goal of this being to bring down idle power consumption: that's likely partly true, but given that current Intel mobile CPUs idle in the mW range, the differences from adding E cores is relatively minor overall - especially as the differences between desktops and mobile in this regard comes down to larger boards and more AICs requiring more power, not the CPUs themselves (as well as high powered desktop PSUs generally being very inefficient at low loads). There is no way in which E cores affect any of this meaningfully, so presenting that as the reasoning just doesn't add up. Intel's main motivation for adding E cores is to compete with AMD's
massive MT efficiency lead, as well as Apple, as it's clear their P architecture just can't deliver the necessary combination of efficiency and speed.
In and ideal world, Intel would have a processor that is all P-cores, as many of them as they could fit on the die space. I figure if they replaced the E-cores with P-cores, they could have a 10 or 12 core 12th gen CPU with all P-cores.
The die space used by the 4-core E clusters is widely documented, and is roughly the same as a single P core, so they would top out at 10 in the same die area, but you'd then also have lower clocks and increased thermal density, while getting fewer threads for your trouble. Most likely, a 10-core Golden Cove chip would be quite underwhelming due to thermal limitations. In most MT heavy applications, 4 E cores deliver more performance than two more P cores would, at least after the scheduler was updated to keep track of them. There are still applications that don't manage to make use of them, but those are growing increasingly rare.
Maybe call this beast the i9-12950K. This CPU would only be available in OEM systems outside those jurisdictions with the strict energy laws. But there's also nothing stopping someone in those areas from upgrading their computer themselves with this processor. Of course this won't happen because it doesn't make a lot of business sense for Intel. It's another die they have to design and product test. And they probably wouldn't make much money off of it.
Such a chip would likely meet those idle power requirements just fine - just like an all-P core ADL i5 does, after all. Cores can be power and clock gated after all, so why would a 10P CPU consume more power at idle than a 6P one? And those 6P CPUs are sold in those jurisdictions. So, sorry, but your reasoning here doesn't add up. You're giving environmental regulations the blame for something they have literally zero effect on - the architectural traits of Intel's P cores and how many of them can be packed into a CPU package and made to perform well. Intel isn't being stopped by regulations, they're stopped by their inability to put more than 8 of these cores in a single package and have them clock high enough to run well.
Me? I trust Intel's engineers to know what they're doing with the resources available to them. They made E cores for a reason, and that reason isn't because the gub'mint forced them to do so. For what they do, E cores are
good, and smarter and more dynamic PCs is a net benefit and a necessity if we want faster PCs going forward - the era of brute-forcing your way to victory is coming to an end. And that's also a good thing. And lastly, I don't hold to scapegoating governmental regulations (which are
only beneficial in this scenario) in lieu of attributing blame where it actually lies: with those engineers, and/or with the product designers at various OEMs. Intel has a high performance architecture that has efficiency and area issues. E cores are a solution to that. And OEMs failing to design compliant products when given several years' notice have nobody but themselves to blame.
I mean, it is what it is. I've been feeling the higher electric bills in the past year or so, so a lower power idling computer that still has tons of horsepower for gaming and multi-threaded work is A-OK with me. But I'd be nice if the option was there to just go full on beastly CPU if you wanted.
