Discussion Intel current and future Lakes & Rapids thread

[H433x0n]

Ian reported a new round of marketing lay offs in the past 24 hrs (yesterday). I hope it doesn't affect engineers in this 'round'. Not that I hate the people who have jobs in marketing, I mean it much suck getting laid off, but like Intel needs their engineers more than ever.

That would make sense. It appears as if LNL's uncore changes are the main star of the show, so even if LNC is a bit lackluster, LNL as a whole could still be really impressive.

A bit unrelated, but I wonder if both AMD and Intel kinda shot themselves in the foot recently with their all core frequency boosts. The 7950x boosts up to 5.4 GHz all core, and the 7600x literally boosts the same speed in ST as it does in MT (5.5GHz).
The 13900k meanwhile has its P cores go up to 5.5GHz and its E-cores at 4.3GHz, which again, seem really close to the P core peak ST frequency of 5.8Ghz.
Usually, when a company moves to a newer node, they are able to boost the all core frequency as a large part of their MT gains- but it appears for ARL and Zen 5, at maxed out power draws, the only MT perf gain they are going to get is from IPC increases, since the all core frequency doesn't have anywhere to go.
Obv this doesn't mean that there aren't any benefits, since efficiency and perf at lower power draws is going to be improved, but in terms of total performance uplift, without a care for power, the uplift could look small. For example, the 7950x vs 5950x was nearly a 50% perf uplift in MT CB23, largely due to the massive all core frequency uplift, but for the 8950x vs 7950x, I would expect the MT uplift to be at most 35% since it's going to have to be essentially all IPC, since the 7950x all core clocks are already so high.
I think for this reason alone it should make sense for PTL/Zen 6 to increase core counts (8+32 for PTL maybe, 24 cores for Zen 6?). Those shouldn't have large IPC increases, and unless we see Max frequency core clocks hitting like 7GHz, we won't see large gains in all core frequency either. Obv we could also see Zen 6/PTL also only increase MT perf marginally (kinda like zen 3) too, but I hope with increased competition between AMD and Intel we could see more (though feeding all those cores bandwitdh is a problem that is also going to have to be solved, obviously, as well).

It certainly doesn’t seem like there’s anything in the pipeline that will clock like RPL (let alone the refresh).

It’s more than a little concerning with respect to where ARL will end up. At this point the performance is almost entirely dependent on the performance characteristics of N3* because as of now it seems to be full steam ahead with this product.

 

[Geddagod]

It certainly doesn’t seem like there’s anything in the pipeline that will clock like RPL (let alone the refresh).

It’s more than a little concerning with respect to where ARL will end up. At this point the performance is almost entirely dependent on the performance characteristics of N3B because as of now it seems to be full steam ahead with this product.

Apparently it's not N3B, the 20A version isn't much better ¯\_(ツ)_/¯

 

[H433x0n]

Apparently it's not N3B, the 20A version isn't much better ¯\_(ツ)_/¯

I had figured it was N3B, I didn’t know for sure. The only thing I know for sure is that internally there doesn’t seem to be any change of plans and it’s going to be ARL as the next client product. So I’m guessing this info wasn’t that concerning? I’m really unsure.

I couldn’t get any info on how the 20A version was clocking so that’s news to me.

 

[Abwx]

Apparently it's not N3B, the 20A version isn't much better ¯\_(ツ)_/¯

It s neither better nor worse, 20A simply doesnt exist elsewhere than in marketing slides.

 

[Geddagod]

It s neither better nor worse, 20A simply doesnt exist elsewhere than in marketing slides.

It was originally planned, and it appears it has been cancelled pretty recently, if it has been at all. While it definitely won't have reached ES 1 or ES 2 phases, it would almost certainly had initial silicon back from the labs.

I had figured it was N3B, I didn’t know for sure. The only thing I know for sure is that internally there doesn’t seem to be any change of plans and it’s going to be ARL as the next client product. So I’m guessing this info wasn’t that concerning? I’m really unsure.

I couldn’t get any info on how the 20A version was clocking so that’s news to me.

Even if it was extremely concerning , not much Intel could have done. If they want to launch a product in 2H 2024, the best they could do is porting MTL over to Intel 3, or continue any canned MTL-S programs. Only problem is that I'm pretty sure neither of those options is going to be better than ARL, and I don't think Intel wants anything eating up GNR/SRF wafers, unless they are external customers to legitimize IFS.

 

[Geddagod]

Also I want to expand on the Intel 20A ARL not being much better than TSMC 3nm ARL rumor;
It's a bit hard to believe tbh. Intel 20A and ARL are massively different, node wise, and concerning design rules. If both designs are messed up, that should mean that LNC has to be flawed on a fundamental architectural level, or on the high level organization of the core, I'm guessing. And if it's that 'high level', I would be a bit surprised that simulations haven't caught it.
I also do want to point out, Intel should have had a lot of time to clean up LNC, considering how delayed ARL as a whole is.
There is of course, the off chance that LNC isn't as bad as the projections say, but based on how other leakers have been agreeing with it, the chances are looking slim...

 

[DrMrLordX]

technically off topic because we didn't anticipate forests (im kidding)

We didn't expect the Spanish Inquisition either, did we?

 

[dullard]

It s neither better nor worse, 20A simply doesnt exist elsewhere than in marketing slides.

Intel 20A has been planned to be a short-lived node for the past year and a half. Originally 20A was supposed to be a year-long bridge between 2024 and 2025. Now it is just a small stepping stone on the way to 18A, and 18A has been moved up a full year.

July 2021 Timeline


Apr 2022 Timeline with 20A basically a small blip


That said, your comment is disingenuous. Of course it isn't in products yet, the process isn't ready and Intel just finalized the design rules in spring of this year.

 

[Abwx]

Intel 20A has been planned to be a short-lived node for the past year and a half. Originally 20A was supposed to be a year-long bridge between 2024 and 2025. Now it is just a small stepping stone on the way to 18A, and 18A has been moved up a full year.

July 2021 Timeline
View attachment 83318

Apr 2022 Timeline with 20A basically a small blip
View attachment 83319

That said, your comment is disingenuous. Of course it isn't in products yet, the process isn't ready and Intel just finalized the design rules in spring of this year.

That s a way to dupe people, nothing else, they know that they are unable to have a legit 20A reliable process, so they are branding it as being a step stone, that s just a kool aid to get shareholders being less nervous.

If they could mass product using 20A in a forseeable future they would instantly do it as soon as possible, instead they are struggling to have a TSMC s N5 equivalent.

Currently the only process they have at hand, Intel 7, is more or less the equivalent of TSMC s N7, think about it, they are saying that they can go directly from this node to a N2 equivalent without needing to first master the equivalents of TSMC s N5 and N3, at some point one has to get realistic rather than taking those rosy predictions at face value.

 

[Geddagod]

That s a way to dupe people, nothing else, they know that they are unable to have a legit 20A reliable process, so they are branding it as just being a step stone, that s just a kool aid to get shareholders being less nervous.

If they could mass product using 20A in a forseeable future they would instantly do it as soon as possible, instead they are struggling to have a TSMC s N5 equivalent.

Currently the only process they have at at hand, Intel 7, is more or less the equivalent of TSMC s N7, think about it, they are saying that they can go directly from this node to a N2 equivalent without needing to first master the equivalents of TSMC s N5 and N3, at some point one has to get realistic rather than taking those rosy predictions at face value.

Are we just going to ignore Intel 4 and Intel 3?
I will say this though, with Intel 4, they are essentially skipping a '5nm' class node in terms of max theoretical transistor density iso fin count. It's kinda impressive. But the lacking in SRAM density (worse than N5 IIRC) and lacking HD libs make it more reasonable. Either way, whether it's going to be successful or not will be found out in 3-5 months, and is that really that long of a wait before calling Intel's future node roadmap a failure and a con?
Also N2 isn't a major shrink over N3. Check out TSMC's density gain estimates for N2. Similarly, I don't expect 20a/18a to be major jumps over Intel 3 density either.
Shifting back the convo to the original point though, I believe we will see at least one 20a Intel sku of ARL, even if it is later in 2025, just to prove that they could launch something on 20a. Maybe they won't if Intel 18A CLW is near mid 2025, but otherwise it's going to be a lot of doubt and skepticism for 18A until PTL in late 2025, even if GNR and MTL have good launches in 2024 and end of 2023 respectively.

 

[Doug S]

Having two nodes in a row, Intel 3 and Intel 20A, shown to last only 6 months each seems highly suspicious. If Intel 18A is that great and able to be pushed up that far great, then why waste your time with TWO short lived nodes ahead of it? In the real world you'd concentrate on one, presumably the one that comes first so you're able to at least get a year out of it.

So why not skip 20A entirely and use Intel 3 for a year? Perhaps Intel 3 is having issues (i.e. all the talk about use of TSMC N3 instead of Intel 3) and hasn't been removed from roadmaps to avoid panicking investors? Now sure 20A can make a good "stepping stone" to 18A as far as the first node to use the new transistors and PowerVIA, but if it is only going to last six months they might have what, one product able to use it? Which would presumably be replaced/updated quickly to use 18A?

This is why I've always been suspicious of Intel's roadmap, there are too many different nodes coming in too short of a time. It simply doesn't make sense to do it like that. Why would Intel's engineers want to target nodes that will be replaced so quickly with a better one? If their design schedule slips a bit suddenly it is "man, should have designed for 20A instead of Intel 3" or for 18A instead of 20A. Maybe that's part of the reason for the confusion over Intel's chip roadmap, and what is using what node. Or maybe that confusion stems from issues like Intel 3 having issues and engineers forced to retarget or use backup plans like TSMC N3 or pushing back the schedule and retargeting for a later Intel node and hoping it doesn't have issues too.

Intel's share price has been recovering a bit of late, but I'm not sure whether that's because investors are believing in them again or it is just part of the general stock market and tech market uplift. Personally I'm still skeptical, until I see some proof that things have changed. Because they've certainly retained their ability to write a good roadmap and talk up their future, something they had for years while 10nm didn't happen and then later when 10nm became all that happened.

 

[dullard]

This is why I've always been suspicious of Intel's roadmap, there are too many different nodes coming in too short of a time. It simply doesn't make sense to do it like that. Why would Intel's engineers want to target nodes that will be replaced so quickly with a better one? If their design schedule slips a bit suddenly it is "man, should have designed for 20A instead of Intel 3" or for 18A instead of 20A. Maybe that's part of the reason for the confusion over Intel's chip roadmap, and what is using what node.

Intel explained all of this though. After being burned with 10 nm, they are now designing for multiple nodes, keeping what works best for their needs, and cancelling the rest. There is no "man, should have designed for 20A instead of Intel 18A" since they are designing for both (or more) nodes and cancelling one (or more).

https://images.anandtech.com/doci/15217/20190916%20SPIE%20Photomask%20and%20EUVL%20Plenary%20-%20Phillips%20v23-VRL2%20distribute-page-019.jpg

 

[Abwx]

Intel explained all of this though. After being burned with 10 nm, they are now designing for multiple nodes, keeping what works best for their needs, and cancelling the rest. There is no "man, should have designed for 20A instead of Intel 18A" since they are designing for both (or more) nodes and cancelling one (or more).

https://images.anandtech.com/doci/15217/20190916%20SPIE%20Photomask%20and%20EUVL%20Plenary%20-%20Phillips%20v23-VRL2%20distribute-page-019.jpg

This slide is from may 2019, roughly four years old, all they did was to improve and rename their 10nm that started at this time to Intel 7, that must be the agressive scaling they re talking about in said slide...

 

[dullard]

This slide is from may 2019, roughly four years old, all they did was to improve and rename their 10nm that started at this time to Intel 7, that must be the agressive scaling they re talking about in said slide...

It covers the four paths Intel is pursuing.
1) Move to the next node.
2) Follow the Easiest Design Path of sticking with a node and iterating it.
3) Backport to a node.
4) Not well appreciated but still on the image: "Establish close partnerships" in case #1 to #3 don't work.

 

[eek2121]

Having two nodes in a row, Intel 3 and Intel 20A, shown to last only 6 months each seems highly suspicious. If Intel 18A is that great and able to be pushed up that far great, then why waste your time with TWO short lived nodes ahead of it? In the real world you'd concentrate on one, presumably the one that comes first so you're able to at least get a year out of it.

So why not skip 20A entirely and use Intel 3 for a year? Perhaps Intel 3 is having issues (i.e. all the talk about use of TSMC N3 instead of Intel 3) and hasn't been removed from roadmaps to avoid panicking investors? Now sure 20A can make a good "stepping stone" to 18A as far as the first node to use the new transistors and PowerVIA, but if it is only going to last six months they might have what, one product able to use it? Which would presumably be replaced/updated quickly to use 18A?

This is why I've always been suspicious of Intel's roadmap, there are too many different nodes coming in too short of a time. It simply doesn't make sense to do it like that. Why would Intel's engineers want to target nodes that will be replaced so quickly with a better one? If their design schedule slips a bit suddenly it is "man, should have designed for 20A instead of Intel 3" or for 18A instead of 20A. Maybe that's part of the reason for the confusion over Intel's chip roadmap, and what is using what node. Or maybe that confusion stems from issues like Intel 3 having issues and engineers forced to retarget or use backup plans like TSMC N3 or pushing back the schedule and retargeting for a later Intel node and hoping it doesn't have issues too.

Intel's share price has been recovering a bit of late, but I'm not sure whether that's because investors are believing in them again or it is just part of the general stock market and tech market uplift. Personally I'm still skeptical, until I see some proof that things have changed. Because they've certainly retained their ability to write a good roadmap and talk up their future, something they had for years while 10nm didn't happen and then later when 10nm became all that happened.

20A shrank due to 18A being ahead of schedule. You also can’t compare past Intel with present Intel, given that leadership has changed.

This slide is from may 2019, roughly four years old, all they did was to improve and rename their 10nm that started at this time to Intel 7, that must be the agressive scaling they re talking about in said slide...

The Intel 7 that current products are being made on is very different from the original Intel 10nm.

 

[dullard]

20A shrank due to 18A being ahead of schedule. You also can’t compare past Intel with present Intel, given that leadership has changed.


The Intel 7 that current products are being made on is very different from the original Intel 10nm.

Exactly. Plus, with 18A getting the Department of Defense behind it and just this week Boeing and Northrup Grumman signing up to be the first 18A customers, we know that 18A will be Intel's focus. That doesn't necessarily mean 20A is good or bad. It just means that 20A will be a short-lived node.

 

[clemsyn]

Exactly. Plus, with 18A getting the Department of Defense behind it and just this week Boeing and Northrup Grumman signing up to be the first 18A customers, we know that 18A will be Intel's focus. That doesn't necessarily mean 20A is good or bad. It just means that 20A will be a short-lived node.

This is interesting. Looks like the foundry is actually doing good. Hopefully Intel will announce more customers soon.

 

[LightningZ71]

I don't know why anyone is surprised over the 20A/18A movements. They were fairly explicit that 20A was an "early" version of the node with limited, Intel internal focused libraries, and that 18A would have MANY more libraries to make it more attractive to external customers. I only ever expected to see maybe a couple of Intel chips on 20A prior to 18A maturing to have those products either ported or abandoned.

 

[dullard]

This is interesting. Looks like the foundry is actually doing good. Hopefully Intel will announce more customers soon.

It makes national security sense. The military and military contractors can send their secret designs to TSMC (China seems to want to take over Taiwan), Samsung (South Korea), GlobalFoundaries (UAE), or Intel (USA). I don't think for a second that any foundry other than Intel would be chosen for obvious reasons.

 

[dr1337]

The Intel 7 that current products are being made on is very different from the original Intel 10nm.

At that rate Intel should rebrand 9th and 11th gen 14nm+++++ chips as Intel 12 and Intel 10 respectively.

Despite the reassuring marketing label, for some reason I doubt the 10nm used in Tiger Lake is functionally any different from what's shipping as intel 7 today.

 

[H433x0n]

At that rate Intel should rebrand 9th and 11th gen 14nm+++++ chips as Intel 12 and Intel 10 respectively.

Despite the reassuring marketing label, for some reason I doubt the 10nm used in Tiger Lake is functionally any different from what's shipping as intel 7 today.

It’s pretty different. The version of Intel 7 that will ship with RPL-R is almost a full node improvement in efficiency (not density) over Alder Lake.

If you were to run ADL-S against RPL-R at iso clocks, RPL-R would use 20-25% less power. This will also be apparent when EMR launches since it’s significantly more efficient than SPR, although in the case of EMR it’s not just the node improvements but also optimizations with cache and tile layout.

This was also true with 14nm+++++ with the final version of 14nm outperforming the first iterations of 10nm.

This is one of the things Intel is pretty good at - for better or worse (usually the latter).

 

[Hitman928]

It makes national security sense. The military and military contractors can send their secret designs to TSMC (China seems to want to take over Taiwan), Samsung (South Korea), GlobalFoundaries (UAE), or Intel (USA). I don't think for a second that any foundry other than Intel would be chosen for obvious reasons.

Intel is definitely in play as a foundry for national security interests. The problem is that the volumes needed by this interest are very small compared to the commercial sector. It's nothing that can sustain Intel in trying to be a leading, or even a close 2nd, foundry.

 

[Abwx]

It’s pretty different. The version of Intel 7 that will ship with RPL-R is almost a full node improvement in efficiency (not density) over Alder Lake.

The efficency improvement already happened in regular RPL compratively to ADL, at some point there will be nothing more to extract.

Prove is that RPL-R is still 253W and has negligible perfs improvements according to recent leaks, so much for "almost a full node improvement" that only bring a few percents uplift..

 

[H433x0n]

The efficency improvement already happened in regular RPL compratively to ADL, at some point there will be nothing more to extract.

Prove is that RPL-R is still 253W and has negligible perfs improvements according to recent leaks, so much for "almost a full node improvement" that only bring a few percents uplift..

When I said full node improvement I meant the cumulative effect of RPL + RPL-R node changes. I don't know the exact performance characteristics of the latest version, but it is a new version of P1274.

 

[Abwx]

When I said full node improvement I meant the cumulative effect of RPL + RPL-R node changes. I don't know the exact performance characteristics of the latest version, but it is a new version of P1274.

There was a significant gain but not up to a full node in perf/watt improvement, otherwise RPL wouldnt be lagging AMD s RHL perf/watt wise that much at equal configuration, that is 32T 8 + 16 vs 16C/32T 7950X, it s the same with lower parts, a 24T 8 + 8 RPL lag a 12C/24T 7900X.