Discussion Intel current and future Lakes & Rapids thread

[RTX2080]

I doubt DDR5 availability more than Intel's 10nm yield now. IIRC when DDR4's first release, price seems to be not much expensive than DDR3, correct me if I'm wrong?

Some other things that is worth paying attention is the interconnection/system bus in big+little design, which might greatly affect the latency and windows' thread scheduling. Waiting to see what will Intel do with it.

 

[Ajay]

Not long ago many of these people called 10nm desktop dead and 10nm never will clock high and things like that.

10nm desktop is still dead ATM. Obviously, Intel is having trouble getting good yields even on Tigerlake mobile (it appears Tigerlake desktop is a no go). From the Analyst call transcript on AT, it looks like Intel is finally getting the cost of producing 10nm down. It was killing their margins. They put in a huge and expensive engineering effort the past couple of years to get 10nm to a real HVM level. Likely, they are yielding better an throwing fewer dice into the trash bin saving money their as well. I suspect that a lot of chips were junk because they couldn't hit target frequencies. if Intel isn't playing a bit fast and loose with their info, AdlerLake be a solid desktop CPU, even though the big.little approach is more optimal for mobile. At least ADL-S looks poised to beat Zen4 to the market.

As far as desperately hoping for an AdlerLake delay - I don't know who these people are; that sort of thinking is irrelevant in any case . Intel has spoken in the past of not reaching the goals vis-a-vis the competition till 5nm, which is looking like 2024 - and they will still be behind TSMC in density. Intel has a ton of work to do the get sorted out and it will take time, meanwhile the competition isn't slowing down. This is not a rosy picture for Intel; the facts just don't support that outlook at the present or near future.

 

[uzzi38]

I doubt DDR5 availability more than Intel's 10nm yield now. IIRC when DDR4's first release, price seems to be not much expensive than DDR3, correct me if I'm wrong?

Some other things that is worth paying attention is the interconnection/system bus in big+little design, which might greatly affect the latency and windows' thread scheduling. Waiting to see what will Intel do with it.

DDR5 pricing, availability and actual performance will all be rather suspect in 2021.

But Alder Lake has DDR4 support so no biggie.

 

[DrMrLordX]

You are correct but this is nothing some fans are hoping for.

"some fans" are welcome to wish for whatever they like. It doesn't change Intel's plight, nor does it change the fact that Intel left open the possibility of a Q4 volume ramp with broad availability in Q1 2022. The only takeaway here is that Alder Lake-S is likely to launch at some point, as opposed to a Charlie-esque prediction of 10nm desktop never happening in a million years. Which is (I suppose) quite exceptional when you consider that both IceLake-SP and Tiger Lake-H are both vaporware.

I mean many of them are so deluded they don't buy Intel products because it's Intel the "evil company"

Oh boo hoo. What people do with their money is their own business.

Some other things that is worth paying attention is the interconnection/system bus in big+little design, which might greatly affect the latency and windows' thread scheduling. Waiting to see what will Intel do with it.

Hopefully they learned from Lakefield.

At least ADL-S looks poised to beat Zen4 to the market.

Maybe. That would require speculation on Zen4's likely release date, and we have a thread for that already. This point is further complicated by the broad timeframe Intel left open for Alder Lake-S' launch.

 

[dmens]

”H2 20” is a huge red flag since product qualifications are planned down to the week. Even at Intel where qualification takes 4 or 5 quarters (which is longer than everywhere else), not even saying which quarter they plan on finishing qual means they expect delays.

Or you can just call me an AMD fanboy LOL.

 

[Abwx]

You are correct but this is nothing some fans are hoping for. I mean many of them are so deluded they don't buy Intel products because it's Intel the "evil company", you cannot expect from them to cheer for Intel, they are in hope for more delays.

Out of 8 PCs i once had 5 were Intel and 3 AMD, what about your own case, did you ever bought a single AMD one..?..

As for Intel s delays dunno what you call AMD "fans" have to do with it, truth is that Intel screwed things so badly as to make any return to perf crown problematic, and that s not a saying of AMD "fans"....

Tough times ahead for Intel: industry insiders spill the beans on subpar 10 nm yields, server CPU plus 7 nm delays, production outsourcing and more

Industry insiders reveal that Intel's current 10 nm yields are nowhere near viable for “full production” and even with the SuperFin tech, 10 nm production may never match that of the 14 nm nodes. Plus we get more details about the performance issues with the upcoming server-grade CPUs, even more...

www.notebookcheck.net

 

[jpiniero]

”H2 20” is a huge red flag since product qualifications are planned down to the week. Even at Intel where qualification takes 4 or 5 quarters (which is longer than everywhere else), not even saying which quarter they plan on finishing qual means they expect delays.

Or they don't want to admit it. You could still do some supply of DIY for an October release and the rest of the products are released in 2022.

 

[dmens]

Or they don't want to admit it. You could still do some supply of DIY for an October release and the rest of the products are released in 2022.

They have no reason to play subterfuge given their current market position.

 

[mikk]

”H2 20” is a huge red flag since product qualifications are planned down to the week. Even at Intel where qualification takes 4 or 5 quarters (which is longer than everywhere else), not even saying which quarter they plan on finishing qual means they expect delays.

Or you can just call me an AMD fanboy LOL.

Intel nor AMD won't share exact schedules to the public especially not like 3 quarters in advance, this is confidential and H1 or H2 is a typical announcement for an upcoming product. And you are completely wrong, three quarters before a (potential) production could start it isn't planned down to one week. There is a production and launch window from several weeks, you should check some old roadmaps. This windows can be narrowed down over time. For RKL-S there was a 4 weeks production start window 2 months before the production could have started.

 

[dmens]

Intel nor AMD won't share exact schedules to the public especially not like 3 quarters in advance, this is confidential and H1 or H2 is a typical announcement for an upcoming product. And you are completely wrong, three quarters before a (potential) production could start it isn't planned down to one week. There is a production and launch window from several weeks, you should check some old roadmaps. This windows can be narrowed down over time. For RKL-S there was a 4 weeks production start window 2 months before the production could have started.

Do you even know what qualification is? It is called PRQ in Intel lingo, go figure out what that is before you dig any deeper.

 

[mikk]

Do you even know what qualification is? It is called PRQ in Intel lingo, go figure out what that is before you dig any deeper.

I'm really experienced with Intel roadmaps, not sure about you. I told you check out some old roadmaps, preferably ILU roadmaps with the exact schedule. You will see there is a window of several weeks. Usually it's an 8 weeks window which they narrow down to 4 weeks shortly before it could start.

 

[dmens]

I'm really experienced with Intel roadmaps, not sure about you. I told you check out some old roadmaps, preferably ILU roadmaps with the exact schedule. You will see there is a window of several weeks. Usually it's an 8 weeks window which they narrow down to 4 weeks shortly before it could start.

And I am experienced in actually taking a chip from first tapeout to market. Not sure about you. Your shifting window for fab time (which is much smaller than a quarter, never mind half a year) is entirely dependent on the chip getting the sign-off for production. No qual, no mass production. That is the real long pole on release dates.

Where do you even think the roadmaps come from?

 

[mikk]

And I am experienced in actually taking a chip from first tapeout to market. Not sure about you. Your shifting window for fab time (which is much smaller than a quarter, never mind half a year) is entirely dependent on the chip getting the sign-off for production. No qual, no mass production. That is the real long pole on release dates.

Where do you even think the roadmaps come from?

You should know better then. Your initially post looked like you are clueless not aware of the basic things. And I'm not talking about the fab time lol, it really looks like you have no clue despite claiming you are the big expert. You also should know that they don't share confidential schedules to the public even if they could do, this is basic knowledge!

 

[dmens]

You should know better then. Your initially post looked like you are clueless not aware of the basic things. And I'm not talking about the fab time lol, it really looks like you have no clue despite claiming you are the big expert. You also should know that they don't share confidential schedules to the public even if they could do.

Yeah sure, keep reading your roadmap tea leaves and guessing how chip validation works LOL. Whatever you are talking about is likely your own fantasy... don't expect me to guess what that may be. By the way, I never said they can release the internal schedule, I said they can do better than give a half year window.

 

[lobz]

Intel nor AMD won't share exact schedules to the public especially not like 3 quarters in advance, this is confidential and H1 or H2 is a typical announcement for an upcoming product. And you are completely wrong, three quarters before a (potential) production could start it isn't planned down to one week. There is a production and launch window from several weeks, you should check some old roadmaps. This windows can be narrowed down over time. For RKL-S there was a 4 weeks production start window 2 months before the production could have started.

Among many others on this forum, you're falling for the same children's tale quarter after quarter, like a goldfish. So who are you calling fans and god forbid, why?

 

[Hulk]

You will be wrong most likely, they confirmed some days ago volume ramp will start in H2 2021. Also it's a desktop release initially this time and not mobile only which will accelerate it, a standalone CPU release has a faster market ramp. To me it sounds like you and other people are really hoping it will be delayed because in hope for a better AMD. Intel is the first with DDR5 and PCIe5 and the IPC crown will go back to Intel as well most likely, no new core for AMD this year on the other side. AMD supporters desperately hoping for ADL-S delays therefore. Not long ago many of these people called 10nm desktop dead and 10nm never will clock high and things like that.

I am sorry if my post came off as anti Intel and pro AMD. That was not my intention. As I have written many times before I have no dog in this fight.

I am a fan of both Intel and AMD. Intel has a long and quite extraordinary history of microprocessor (and other) technical achievements. During the last few years they seem to have stumbled a bit, allowing a very well-organized AMD to catch up. Intel has enormous resources and I believe the new leadership it going to put emphasis back on technological innovation, which I think is needed right now.

Finally, my basis for Alder Lake's release is just my best guess based on the problems Intel has been having getting new products and processes out the door in a timely manner. Specifically a competitive node across the product stack, and new cores. In addition, it just seems unlikely that they would release Rocket Lake and then savage that release a few months later with was will hopefully be a better performing and more efficient design.

If I am wrong then I will be happy for them (and for me) as I will buy it. Rocket Lake could be a stopgap measure until Alder Lake arrives.

I have been building PC's since the early '90's and have yet to own an AMD system. But I can't deny the technological achievement that is Zen 3.

 

[Zucker2k]

Let's not get it twisted, Intel's real fight is a process fight against TSMC.

 

[teejee]

I'm really experienced with Intel roadmaps, not sure about you. I told you check out some old roadmaps, preferably ILU roadmaps with the exact schedule. You will see there is a window of several weeks. Usually it's an 8 weeks window which they narrow down to 4 weeks shortly before it could start.

Roadmaps are never used for the detailed project management in my experience. The internal project time plan is almost certainly much more detailed. Have you seen those? (I'm pretty sure dmens has).

I work in automotive, the exact day for start of production of a car is often known about 3 years beforehand.
And changing that day is a huge task and only done if absolutely necessary.

 

[coercitiv]

Let's not get it twisted, Intel's real fight is a process fight against TSMC.

Let's not get it twisted, Intel's real fight is internal. TSMC did not break their 10nm node or delay their 7nm either.

The only thing I like about Intel lately is their leadership change, the pedigree and first speech of their new CEO align a lot better with what needs to be done to turn the ship around. I expect it won't take more than 6 months to observe what kind of mandate Gelsinger was given, since fixing Intel manufacturing requires fixing Intel management first.

 

[JoeRambo]

Let's not get it twisted, Intel's real fight is internal. TSMC did not break their 10nm node or delay their 7nm either.

Yeah, the process failure is completely on Intel. Everyone in the industry is using same tools, so this failure is completely on management. Even more so given how huge of a lead Intel had before.
TSMC roadmaps used to be laughing stock, it was 100% certainty that they would slip. In fact the gap between processes used to be so wide, that people on this very board were wishing for Nvidia cards manufactured on Intel process.

What was so wrong about Intel failures, is how damn long board tolerated them, absolutely ridiculous how long they bought management cool-aid about process "improvements".

 

[jpiniero]

Maybe the issue isn't S but P. There should be some decent amount of time gap between the two, if the graphics drivers specifically calling out S only is any indication.

 

[RTX2080]

500 series mobo seems to be starting pre-order in my country, just have a glance ASUS Z590-A costs 22% higher than Z490-A, and MAXIMUS XIII HERO also 22% higher than XII.
I don't understand the situation but it's better to have prepared for the price hike just like the name ROCKETlake.

 

[Ajay]

Do you even know what qualification is? It is called PRQ in Intel lingo, go figure out what that is before you dig any deeper.

Unfortunately for us, PRQ dates are only available to OEMs and other partners under an NDA. Now a days, the 'roadmaps' shown by semiconductor companies are really marketing slides. Kind of useless compared to when we used to get at least product availability down to the quarter.

 

[dmens]

Unfortunately for us, PRQ dates are only available to OEMs and other partners under an NDA. Now a days, the 'roadmaps' shown by semiconductor companies are really marketing slides. Kind of useless compared to when we used to get at least product availability down to the quarter.

That is correct, no one in public gets the PRQ target date. However, like I said, a half year window is just FUD and likely CYA.

 

[Ajay]

That is correct, no one in public gets the PRQ target date. However, like I said, a half year window is just FUD and likely CYA.

With Intel, no doubt. But of late, even Nvidia and AMD have been doing this - and they have far less serious issues with their foundry partners (they are just supply limited). Why? I don't know.