• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Discussion Intel current and future Lakes & Rapids thread

Page 356 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Exist50

Member
Aug 18, 2016
193
241
116
Giving the half of the year instead of the quarter would make total sense with a planned September release. If it slips even a week or two over the next 8 months (extremely common), that could push it into Q4, so better to just give a broader window to avoid any "delayed" headlines.

And no, production is not planned down to the week this far in advance.
 
  • Like
Reactions: Zucker2k

lobz

Golden Member
Feb 10, 2017
1,565
1,982
136
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.
Foundry is not the only thing that can hinder a release, this past year the whole supply chain was suspect to serious constraints.
 

jpiniero

Diamond Member
Oct 1, 2010
8,854
1,617
126
Shipping date down to the week at tapeout? You've gone from wrong to outright laughable.
Unless Stuff Happens, the timeline is probably close to that.

I still think they are talking about the beginning of production, not the end; and it's just as simple as S starting in Q3 and P starting in Q4.
 

Exist50

Member
Aug 18, 2016
193
241
116
Assuming nothing pops up in the validation of the chips, you should be able to set your shipping week at tapeout.
"Assuming nothing pops up" is precisely the problem. If there were never any bugs discovered after tapeout, then there would be no need for over a year's worth of validation. Sure, you can set target dates down to the week, but with the expectation that those are allowed to slip within a certain window. And that's assuming nothing major that warrants a new stepping.
 

Hitman928

Diamond Member
Apr 15, 2012
3,400
3,554
136
"Assuming nothing pops up" is precisely the problem. If there were never any bugs discovered after tapeout, then there would be no need for over a year's worth of validation. Sure, you can set target dates down to the week, but with the expectation that those are allowed to slip within a certain window. And that's assuming nothing major that warrants a new stepping.
If the CPU doesn't pass validation, you're not hitting your H1/H2 window either, so it doesn't really anything in terms of being forced to push out the delivery schedule.
 

dmens

Platinum Member
Mar 18, 2005
2,079
473
136
"Assuming nothing pops up" is precisely the problem. If there were never any bugs discovered after tapeout, then there would be no need for over a year's worth of validation. Sure, you can set target dates down to the week, but with the expectation that those are allowed to slip within a certain window. And that's assuming nothing major that warrants a new stepping.
Hope you realize schedules account for validation cycles and time to fix bugs. Oh wait, you don't, because you have obviously never worked on a tapeout to product cycle.
 

Exist50

Member
Aug 18, 2016
193
241
116
Hope you realize schedules account for validation cycles and time to fix bugs. Oh wait, you don't, because you have obviously never worked on a tapeout to product cycle.
Yes, a real schedule does account for those things. Which is why a real schedule doesn't specify production down to the week over a year in advance.
 
  • Like
Reactions: mikk and Zucker2k

Hitman928

Diamond Member
Apr 15, 2012
3,400
3,554
136
Yes, a real schedule does account for those things. Which is why a real schedule doesn't specify production down to the week over a year in advance.
When dealing with semiconductor design, you only have certain windows that you have to target to bring everything together and wafer runs have to be purchased months and months in advance and then you need to make sure all of the following steps are in place in order to process those purchased wafers. If you don't know the week you can launch a product at time of tapeout (or really even before tapeout), you're doing something wrong and asking for major issues with your production schedule and a large cost increase in launching your product. There are of course buffers added to hit that scheduled week, but that just means if everything goes well you can have everything ready early and breath a sigh of relief and then you can put more focus on the next product. This is of course all internal scheduling, externally each company is different with how much info they share and to which partners/public.
 

Ajay

Diamond Member
Jan 8, 2001
8,159
3,106
136

dmens

Platinum Member
Mar 18, 2005
2,079
473
136
Yes, a real schedule does account for those things. Which is why a real schedule doesn't specify production down to the week over a year in advance.
Say what now? The whole point of the post-silicon schedule is a high confidence time plan based on prior experience to achieve qualification, which enables production. Which is precisely why it is down to the work-week. Again I ask, how many products have you worked on? This is 100% BS you are emitting.
 
  • Like
Reactions: Tlh97 and lobz

dmens

Platinum Member
Mar 18, 2005
2,079
473
136
I'm talking about public time frames, to prevent confusion. From AT: https://www.anandtech.com/show/16402/amd-to-launch-midrange-rdna-2-desktop-graphics-in-first-half-2021
Announcing midrange RDNA2 products availability for 1H21, on Jan 12, 2021!
Those are just scaled down cards of already qualified products on the market, that also happen to be supply constrained. Not quite the same thing as failing to specific a timeframe for the great big.little x86 savior with silicon already in the labs.
 

Hitman928

Diamond Member
Apr 15, 2012
3,400
3,554
136
Those are just scaled down cards of already qualified products on the market, that also happen to be supply constrained. Not quite the same thing as failing to specific a timeframe for the great big.little x86 savior with silicon already in the labs.
I think this is more of a PR thing. When AMD was fighting for its life against NV and Intel, they have more specific dates for products they were trying make people excited for because they needed to build the hype and get people to plan to switch from their competitors. Now that they are extremely competitive (to say the least for CPUs), they have also become less willing to share future product details and timelines. I think due in large part now to switching from wanting to prevent people from buying your competitor's chips to not wanting to hurt your own current generation sales. I don't know how Intel management operates outside of conversations I've had with some former employees, but I wouldn't be surprised if we don't really start hearing about Alderlake until Rocketlake launches and has some time in the market, more for PR reasons against its own current products than anything else. I think this is especially true if they are launching so soon after Rocketlake, though I find that doubtful (who knows though with what's happening at Intel right now).
 

dullard

Elite Member
May 21, 2001
22,606
900
126
Show examples from AMD/NV? My two cents, a half year window is acceptable when a product is in early to mid pre-silicon development. Not so much when first silicon has already arrived back.
It really isn't very hard to search for "first half AMD" or "2H NVidia" or similar.

"AMD's next-gen Arcturus GPU teased, could be here in 1H 2020"

"NVIDIA RTX 30 Series Mobile GPUs Coming in 1H 2020"

"Aiming 2020 Launch"

"scant on the details about either GPU besides a 1H 2020 release date."

"AMD Confirms Mainstream Radeon RX 6000 Desktop & High-Performance Notebook GPUs Coming in First Half of 2021"
 

dmens

Platinum Member
Mar 18, 2005
2,079
473
136
It really isn't very hard to search for "first half AMD" or "2H NVidia" or similar.

"AMD's next-gen Arcturus GPU teased, could be here in 1H 2020"

"NVIDIA RTX 30 Series Mobile GPUs Coming in 1H 2020"

"Aiming 2020 Launch"

"scant on the details about either GPU besides a 1H 2020 release date."

"AMD Confirms Mainstream Radeon RX 6000 Desktop & High-Performance Notebook GPUs Coming in First Half of 2021"
Clicked on one link and saw this: "The higher-end RTX 3080 and 3070 mobile will be released at CES 2021 in January," So not quite "1H 2021". Think your links just have shoddy reporting. AMD announced Q2/19 launch for 3rd gen ryzen back at CES 2019 in January.
 

Ajay

Diamond Member
Jan 8, 2001
8,159
3,106
136
Those are just scaled down cards of already qualified products on the market, that also happen to be supply constrained. Not quite the same thing as failing to specific a timeframe for the great big.little x86 savior with silicon already in the labs.
Who said ADL is going to be some sort of savior?
 

blckgrffn

Diamond Member
May 1, 2003
7,500
708
126
www.teamjuchems.com
Uh, I'm not sure if we have a thread up for Intel's graphics, I feel like we've been using this thread, so I figured I'd post this here:


Current thread for Intel Graphics.

That said - that has to be a qualification restriction, right? As they work it out? I mean, there is an exceptional opportunity to sell functional GPUs from $80 to $200 right now. Intel should figure out how to take advantage of it!
 

uzzi38

Golden Member
Oct 16, 2019
1,363
2,529
96
TGL-H35 is an absolute bust as far as I care.


Laptop here comes with a 3070 Max -Q configured to 80W compared against last gen 90W Max-Q designs. GPU is not the bottleneck here.

The CPU just absolutely sucks for gaming laptops with frametimes in some games being awful, and in terms of power/clock, sucks even compared to Renoir sitting at the very top end of it's V/f curve.

AMD comparison here with Renoir:

Beast mode on this system is probably most comparable for a power/clock comparison against TGL-H35's performance mode here. The former has 8 cores pulling an absolute peak of ~75W (with average likely closer to 60W or lower) for 4.1GHz sustained. The latter pulls 45-50W sustained for 3.9-4GHz.

The reason why I'm suggesting such a comparison is because Cezanne is the real competitor for TGL-H35, but sadly we don't have comparable numbers yet in AutoDesk, so have to make do with these.

Although in Prime95 the 5980HS sustains 4GHz in 42W if you still want a Cezanne comparison even still.


Overall... I honestly think were better off just launching TGL-H and not bothering with TGL-H35 at all. I can't speak for everyone, but IMO this isn't good enough for an i7 in the -H market today.
 
  • Like
Reactions: Tlh97

ASK THE COMMUNITY