Intel will release Alder Lake to PC enthusiasts first, says report

[HEXUS]

Mainstream chips and chipsets are expected to debut at CES 2022 in January.

Read more.

[kalniel]

built on the Intel 7 node

Come on Hexus, this is where we need you - don't just parrot marketing, tell us what the actual process is. I presume this is 10nm?

[HW90]

built on the Intel 7 node

Come on Hexus, this is where we need you - don't just parrot marketing, tell us what the actual process is. I presume this is 10nm?

The marketing Intel is using with Intel 7 etc is actually the more reasonable term to use nowadays because it's matching how TSMC does their marketing, with the number of nanometers marketed being representative of the transistor density rather than transistor size.

So if you want them to call Intel 7 as 10nm, they also need to start changing what they call chips based on TSMC too and that would get confusing as hell given they actually put nm in the name of their processes.

[kompukare]

Strange the article says:

Intel's Z690 chipset is expected to support both DDR4 and DDR5 memory,

From the headline I was going to say:
"So it's a DDR5-only depute, then. As only cash cows, sorry enthusiasts, will be willing to stump up lots extra for brand new DDR5 whose speeds aren't likely to be great and like DDR4-2133 will quickly be succeeded by DDR5 which might be an actual meaningful upgrade."

Okay, Intel are hedging their bets with DDR4 then.

[John_Amstrad]

Old technology is tried to be sold with old marketing techniques

[DanceswithUnix]

So does this mean that Intel are expecting such poor production quantities that they only have enough for the enthusiast retail sector that they don't do very well in?

[CAT-THE-FIFTH]

So does this mean that Intel are expecting such poor production quantities that they only have enough for the enthusiast retail sector that they don't do very well in?

Yep:
https://www.bbc.co.uk/news/technology-57996908

Intel chief warns of two-year chip shortage

[bae85]

Node 7 or in other words - 10nm+++

Nothing like dodgy marketing to try and trick people into thinking they're at the same level as the competition.

[Xlucine]

So does this mean that Intel are expecting such poor production quantities that they only have enough for the enthusiast retail sector that they don't do very well in?

It's a vote of confidence in the die areas and high frequency scaling at least

Node 7 or in other words - 10nm+++

Nothing like dodgy marketing to try and trick people into thinking they're at the same level as the competition.

You do know that that intel process is about the same as TSMC 7nm, right? It's not a trick, tech journalists have been explaining for years now that intel node names are a bit larger than the equivalent performing TSMC et al branding

[DanceswithUnix]

You do know that that intel process is about the same as TSMC 7nm, right?

Same size, but Intel have been really struggling to get frequencies up at 10nm. If they really have fixed that, then to be fair to Intel that is like a whole new node.

[kalniel]

Same size, but Intel have been really struggling to get frequencies up at 10nm. If they really have fixed that, then to be fair to Intel that is like a whole new node.

At which point they should brand it with some kind of name, rather than a number. Then they can just change name when they have an improvement and there's no association with a (contentious) specification like size.

[edmundhonda]

Okay, Intel are hedging their bets with DDR4 then.

This is the weirdest part to me. The chipset supporting both is nothing, the memory controllers are on-CPU anyway, but these are still single-die processors, right? So either there's a whole lot of effectively wasted die space and interconnects supporting both, or there's different models that support either DDR4 or 5, but *not* both. Unless they're close enough that they could have a single memory controller design that can switch standards without causing problems?

[DanceswithUnix]

Unless they're close enough that they could have a single memory controller design that can switch standards without causing problems?

Most of the silicon area is the power transistors to drive the high speed signals across the motherboard PCB, and things like write posting buffers to improve efficiency and those have to be there regardless of which memory interface is used.

Driving modern ram is actually pretty difficult, almost like a processor in itself. So it isn't a big leap to have two different state machines driving the interface so it can switch between standards, once you realise you need a state machine to start with.

There are plenty of examples of CPUs and GPUs that have had the ability to do different ram standards, including fairly cheap stuff like Nvidia's 1030 that can be had in DDR4 and GDDR5 versions (I think GDDR5 has narrower memory channels like DDR5 adopts making it a similar situation, but fact check me on that).

[CAT-THE-FIFTH]

Same size, but Intel have been really struggling to get frequencies up at 10nm. If they really have fixed that, then to be fair to Intel that is like a whole new node.

Well if they are saying there are chip shortages for 2 years,it makes me thing their new processes don't have enough volume yet to fully replace 14NM. Laptops definitely need 10NM/7NM and desktop is going to be less of a priority for power savings for the immediate future.

It makes me wonder if Intel will just keep RKL going as a cheaper platform,and ADL as some kind of HEDT one between higher end AM4 and AMD TR.

AMD probably does not mind either!

[Xlucine]

This is the weirdest part to me. The chipset supporting both is nothing, the memory controllers are on-CPU anyway, but these are still single-die processors, right? So either there's a whole lot of effectively wasted die space and interconnects supporting both, or there's different models that support either DDR4 or 5, but *not* both. Unless they're close enough that they could have a single memory controller design that can switch standards without causing problems?

6th gen intel core supported DDR3 and DDR4 on the same die, and AMD's bristol ridge also supported both. Dual support is very common