Discussion Intel current and future Lakes & Rapids thread

[JoeRambo]

Newer one than that non-working microcode update mentioned in that article?

Please stop spreading unintentional FUD. uCode ver 150 seems to fix the problem for most people. Too early to say if fix is complete and how much/if of performance it kills.

 

[jpiniero]

Newer one than that non-working microcode update mentioned in that article?

At least on Macs, seems the microcode was only updated when Apple released 10.15.6, which was Wednesday.

 

[Adonisds]

Why are people saying 7nm will slip? What's the recent news/rumor that made people say that?

 

[jpiniero]

Why are people saying 7nm will slip? What's the recent news/rumor that made people say that?

Charlie saying Keller's departure was related to him pushing for TSMC. That seems more realistic than any other reason people have given so far.

You can also look at it in that Icelake Server is now for all intents and purposes a 2021 product. You would think that if Granite Rapids was still realistic for 2022 that Intel would just cancel Icelake Server at this point and roll with Sapphire Rapids.There really isn't any reason for Icelake Server to exist at this point; Rome is so much better and if you badly want bfloat, it doesn't have that.

 

[lobz]

Please stop spreading unintentional FUD. uCode ver 150 seems to fix the problem for most people. Too early to say if fix is complete and how much/if of performance it kills.

How is politely asking a legit question spreading unintentional FUD? Tell the same thing Dr. Ian and see what you get as an answer. Even the word triggered doesn't cover the way you're acting right now.

 

[lobz]

Charlie saying Keller's departure was related to him pushing for TSMC. That seems more realistic than any other reason people have given so far.

You can also look at it in that Icelake Server is now for all intents and purposes a 2021 product. You would think that if Granite Rapids was still realistic for 2022 that Intel would just cancel Icelake Server at this point and roll with Sapphire Rapids.There really isn't any reason for Icelake Server to exist at this point; Rome is so much better and if you badly want bfloat, it doesn't have that.

When will a company that never officially cancelled Cannon Lake do something like that?

 

[ajc9988]

Ask in the relevant thread. If you followed the AMD news, you would know better. N7+ is confirmed. I don't feel like digging it up but it's in the Zen3/Ryzen 4000 speculation thread and possibly elsewhere. That data from March is obsolete and was never conclusive. I can't help you if you insist on clinging to fallacies.

@lobz

AMD is already milking the DiY market. Watch what happens in October . . .

And I can't help you if you ignore the edited part of my post, talking about wafers and discussing the use on consoles, which is 7nm.

Then there is the contractual aspect of continuing to output zen 2 on servers. Then there is the GN part that they are pushing Intel out of product review attacks like Nvidia does on graphics cards variations.

Hell, look at how AMD even recommended the 3700x for most people at launch. Or look how low consumer sales of the 3600X and 3800X are. You are assuming 1) consumers are STUPID and don't understand the product stack or how to read or watch reviews, and 2) that consumers for AMD act like those on intel's side, where hubbub was made on the KF chip and the K R0 stepping over the P0 stepping, so they will go out and purchase an XT chip from AMD now. Especially since generally the pricing is the same as the counterparts minus the cooler, which you'd likely buy another one anyways.

So you say stagnation over nothing when they reduce latency and increase IPC by 20%, you argue delayed although expected release is within their cadence window, you argue bilking (not just milking, look it up) when consumer sales suggest those products are low volume, unlike Intel's bilking the market including memory lock down, and then argue over wafers and process when refinement of process and wafer can matter because both old zen2 has to be made for replacement for servers for years and because you only said zen 3 core chiplets will have 7nm+ EUV, but ignore working with GPUs, console chips that are like APUs, upcoming APUs, etc.

You ignore all the factors to shoe horn your salty point.

Even Intel has not fully stagnated. They have great architectures they have been working on for years ready to go. The manufacturing process is what is killing them, to be fair. Those engineers on core design deserve credit. And bad management and trying to correct their mistakes and corporate culture isn't easy. But you keep being reductive.

 

[JoeRambo]

How is politely asking a legit question spreading unintentional FUD? Tell the same thing Dr. Ian and see what you get as an answer. Even the word triggered doesn't cover the way you're acting right now.

The microcode that fixes is out for like a week, and there is little if any reason atm to doubt it.

But You are right, about "triggered" part, I misunderstood the question, since I was watching original IntelliJ thread and thought it was about it. My apologies to the original asker @naukkis

 

[ondma]

Charlie saying Keller's departure was related to him pushing for TSMC. That seems more realistic than any other reason people have given so far.

You can also look at it in that Icelake Server is now for all intents and purposes a 2021 product. You would think that if Granite Rapids was still realistic for 2022 that Intel would just cancel Icelake Server at this point and roll with Sapphire Rapids.There really isn't any reason for Icelake Server to exist at this point; Rome is so much better and if you badly want bfloat, it doesn't have that.

Unless of course, one chose to believe the stated reason instead of innuendo from a source who is known to never miss a chance to spread FUD about Intel. (Charlie, not you).
Even then, would it be even remotely possible for TSMC to have enough capacity to satisfy Intel's needs on top of their current customers?

 

[NostaSeronx]

Intel's 7nm basics was finished in 2016.
Intel's 7nm design kit was finished in 2019.
Intel's has been leaking out 2021 as the year of 7nm EUV. I have verified 7nm Xeon, 7nm Atom, 7nm Graphics, 7nm ROM/RF/SRAM, 7nm IO/Analog/Misc(Scan, SerDes, Temp, etc) being done by now.

Arizona 7nm = 2020 HVM ramp
Oregon 7nm = 2021 HVM ramp
Ireland 7nm = 2022 HVM ramp
Israel 7nm = 2023 HVM ramp

 

[TheGiant]

Intel's 7nm basics was finished in 2016.
Intel's 7nm design kit was finished in 2019.
Intel's has been leaking out 2021 as the year of 7nm EUV. I have verified 7nm Xeon, 7nm Atom, 7nm Graphics, 7nm ROM/RF/SRAM, 7nm IO/Analog/Misc(Scan, SerDes, Temp, etc) being done by now.

Arizona 7nm = 2020 HVM ramp
Oregon 7nm = 2021 HVM ramp
Ireland 7nm = 2022 HVM ramp
Israel 7nm = 2023 HVM ramp

 

[lobz]

The microcode that fixes is out for like a week, and there is little if any reason atm to doubt it.

But You are right, about "triggered" part, I misunderstood the question, since I was watching original IntelliJ thread and thought it was about it. My apologies to the original asker @naukkis

I take it back too, in this case. Props to you

 

[lobz]

Intel's 7nm basics was finished in 2016.
Intel's 7nm design kit was finished in 2019.
Intel's has been leaking out 2021 as the year of 7nm EUV. I have verified 7nm Xeon, 7nm Atom, 7nm Graphics, 7nm ROM/RF/SRAM, 7nm IO/Analog/Misc(Scan, SerDes, Temp, etc) being done by now.

Arizona 7nm = 2020 HVM ramp
Oregon 7nm = 2021 HVM ramp
Ireland 7nm = 2022 HVM ramp
Israel 7nm = 2023 HVM ramp

Good then.

As long as you have verified.

 

[NostaSeronx]

As long as you have verified.

With an Intel person, no names, not like you can find her/him.

Product line of hell 14nm, 10nm, and 7nm;
All products on 10nm++ could also be on 7nm, all products on 10nm+++ could also be on 7nm+, etc.

International products = 10nm
USA-only products = 7nm

 

[dmens]

You got it backwards, 10nm designs could definitely be on 14nm, it's called backporting.

 

[jpiniero]

Even then, would it be even remotely possible for TSMC to have enough capacity to satisfy Intel's needs on top of their current customers?

I think we are going to find out... eventually. Clearly chiplets would have to be used with multiple nodes.

 

[NostaSeronx]

I think we are going to find out... eventually.

sooner than later;


Also, Thundar something IoT group is an Intel 7nm(intel 7nm in the US/internal 7nm) project. So, Ponte isn't the first and might actually be the last 7nm product in the pipe.

 

[DrMrLordX]

And I can't help you if you ignore the edited part of my post

Orthogonal to the (off-topic) conversation, which is Zen3. N7+ is confirmed. Post in the relevant thread if you need more information, that is all. I don't know what else you're going on about other than to argue for arguement's sake. What does any of that have to do with AMD dragging their heels on Zen3?

So you say stagnation over nothing when they reduce latency and increase IPC by 20%, you argue delayed although expected release is within their cadence window

Nonsense. AMD is only launching two SKUs in October. Staying "on cadence" would have been August/September 2020.

March 2017 - Zen1
April 2018 - Zen+
July 2019 - Zen2
August/Sept 2020 - Nothing!

We won't get the full stack of Vermeer parts until Nov/Dec. AMD can afford to slack because Intel is flopping about uselessly, offering no competition.

If you think that's "okay" then go crow about how great it is that AMD's cadence went from 13 months to 15 months to 17 months for the full stack of products (down to 6c parts anyway; AM4 4c parts have been on delayed launch since 2017).

If Intel were able to launch Golden Cove this year, none of this would be happening. Zen3 would be out sooner, the full stack would be here sooner, the XT launch would never have happened, and we'd get the full Zen3 stack instead of just two high-end SKUs. And the prices on Zen3 would probably mirror those for Zen2 - $499 for the 12c part and $749 for the 16c part. If not less if Intel really pushed them on price. Intel is screwing everyone with their delayed launches - not just Intel's customers.

Even Intel has not fully stagnated. They have great architectures they have been working on for years ready to go.

Wake me when they decide to actually sell products based on those architectures.

Sadly, this is getting a bit off-topic, so if you really want to argue that AMD is just knocking it out of the park and everything's perfectly fine, again, take it to the Zen3/Ryzen 4000 thread and address me there. Thank you.

You got it backwards, 10nm designs could definitely be on 14nm, it's called backporting.

I hope they don't plan on any more backport products aside from Rocket Lake. Golden Cove on 14nm would be a bit of a mess, wouldn't it?

 

[ajc9988]

Orthogonal to the (off-topic) conversation, which is Zen3. N7+ is confirmed. Post in the relevant thread if you need more information, that is all. I don't know what else you're going on about other than to argue for arguement's sake. What does any of that have to do with AMD dragging their heels on Zen3?



Nonsense. AMD is only launching two SKUs in October. Staying "on cadence" would have been August/September 2020.

March 2017 - Zen1
April 2018 - Zen+
July 2019 - Zen2
August/Sept 2020 - Nothing!

We won't get the full stack of Vermeer parts until Nov/Dec. AMD can afford to slack because Intel is flopping about uselessly, offering no competition.

If you think that's "okay" then go crow about how great it is that AMD's cadence went from 13 months to 15 months to 17 months for the full stack of products (down to 6c parts anyway; AM4 4c parts have been on delayed launch since 2017).

If Intel were able to launch Golden Cove this year, none of this would be happening. Zen3 would be out sooner, the full stack would be here sooner, the XT launch would never have happened, and we'd get the full Zen3 stack instead of just two high-end SKUs. And the prices on Zen3 would probably mirror those for Zen2 - $499 for the 12c part and $749 for the 16c part. If not less if Intel really pushed them on price. Intel is screwing everyone with their delayed launches - not just Intel's customers.



Wake me when they decide to actually sell products based on those architectures.

Sadly, this is getting a bit off-topic, so if you really want to argue that AMD is just knocking it out of the park and everything's perfectly fine, again, take it to the Zen3/Ryzen 4000 thread and address me there. Thank you.



I hope they don't plan on any more backport products aside from Rocket Lake. Golden Cove on 14nm would be a bit of a mess, wouldn't it?

OK, normally this is where I'd inundate you with articles throughout the entire tech press, but I've gotten warnings whenever I do that and it is not from Anandtech, so I will try to make this easy for you to understand:

1) Intel deciding not to backport until 2019, while also not recognizing the report of the engineer that told them to but was ignored, then saying their architecture means nothing because it cannot be released due to not having the manufacturing process ready, along with the reports that Keller may have left due to saying just get the products out the door using TSMC until it gets straightened out, belittles their innovation during the time by other groups. You just care do you have a product in hand. You do not care about the sausage making. You just want the sausage to shove in your mouth once finished being made.

2) You assume that unless competition is happening all the time, especially when AMD finally takes the final crowns on performance, that this absolutely is stagnation. How? Because even though one firm overtook another firm, it must all be standing still, except standing still would leave relative positioning, but don't think too deeply on that.

3) YOU are who artificially limited all discussion to only be Zen 3. But even with that, let's assume they will not have a separate I/O die on 7nm for any product. Let's not discuss that the APUs lag by a generation. Let's not discuss AMD not owning their own fab and having to balance their fab time for console products, GPUs to compete with Nvidia releasing in September, and CPUs all in the same quarter. Let's ignore all these converging situations.

Not only that, let's ignore that for Zen, only the 8 core dropped in March and the rest of the SKUs released in April and finished in May. Let's ignore similar happened with Zen + with the April release and the bulk dropping in May. Let's ignore that supplies were limited for Zen 2 in July, or that the higher core count SKUs, the true high end mainstream SKUs, didn't arrive UNTIL NOVEMBER. https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40

Let's ignore all of these things that show only a couple SKUs drop at release and the rest of the lineup is fleshed out later, something Intel also does, and the dropping of multiple other products around the same time, all to shoehorn in the belief of stagnation.

16 months, for the record, is getting something out of the door around November. Sure, they have been better at the 14 month cadence, which would be a September product, but they also announced their graphics card is dropping first, which is needed to deal with the larger threat of Nvidia at this moment. But you don't want to mention that.

Further, you are believing rumors on Zen 3 pricing without any proof of a price increase. Even AMD is saying if you have Zen 2, do not buy these lines of new releases (edit: meaning XT). Think about that. A company saying don't buy a product. It almost fully matches GN's prediction this was to take up space on performance graphs.

But you do you, boo.

 

[beginner99]

Further, you are believing rumors on Zen 3 pricing without any proof of a price increase.

True there is no proof but it is one of the few explanations that make sense to even release these new skus. They reset the price to launch level eg. so you have to expect to pay $500 for a 12-core etc.

 

[DrMrLordX]

1) Intel deciding not to backport until 2019, while also not recognizing the report of the engineer that told them to but was ignored, then saying their architecture means nothing because it cannot be released due to not having the manufacturing process ready, along with the reports that Keller may have left due to saying just get the products out the door using TSMC until it gets straightened out, belittles their innovation during the time by other groups. You just care do you have a product in hand. You do not care about the sausage making. You just want the sausage to shove in your mouth once finished being made.

The reports about Keller are unsubstantiated, and what has this got to do with anything?

2) You assume that unless competition is happening all the time, especially when AMD finally takes the final crowns on performance, that this absolutely is stagnation.

Intel already showed us what happens when their is no competition. Observe Sandy - > Kaby and tell me that wasn't stagnation. Intel didn't even bother upping core counts until Zen1 showed up. Had AMD failed and gone out of business, where do you supposed we'd be now? Are you forgetting the people in this very forum that argued that competition wasn't necessary, with a straight face? Do you agree with them?

3) YOU are who artificially limited all discussion to only be Zen 3.

There was nothing artificial about that limitation. It's called not wandering off-topic.

Not only that, let's ignore that for Zen, only the 8 core dropped in March

Yeah the 1600 came out one month after. The 1800x, 1700x, and 1700 hit in March. Of course I can see where you're trying to go with this, but it seems that the math here isn't clicking, so I'll make it even simpler:

AMD said Zen1 for 2017, Zen+ for 2018, Zen2 for 2019, Zen3 for 2020, and Zen4 for 2021. If AMD sticks to their current cadence, Zen4 won't show up until 2022. They aren't hitting their targets.

There, all better now?

16 months, for the record, is getting something out of the door around November.

AMD has not taken 16 months to launch any of their AM4 products since Zen1 came out in 2017, so I still don't understand why they think that's "normal". The schedule was 2017/2018/2019/2020 and they're slacking.

Further, you are believing rumors on Zen 3 pricing without any proof of a price increase.

Hah! Okay. Wait and see then.

Even AMD is saying if you have Zen 2, do not buy these lines of new releases (edit: meaning XT).

First I've heard of that, but hey the reviews speak volumes about why someone should just buy last year's chips instead. AMD pulled an Intel and released the same thing twice.

 

[Zucker2k]

True there is no proof but it is one of the few explanations that make sense to even release these new skus. They reset the price to launch level eg. so you have to expect to pay $500 for a 12-core etc.

What I really don't understand is why everybody seems to be buying into this price-reset nonsense. The precedent was already set with the Zen2 releases, that is how this industry has always functioned. You don't need to release a lackluster batch of CPUs in order to remind any of your loyal fans and potential customers that you'd be charging $499+ for a 12-Core and higher Zen 3 sku? If the performance justifies the price, customers will jump on it; and if you believe some in this thread, AMD doesn't even have to worry about the competition since the competition is stagnant. Which all seems a bit weird that AMD should be dropping prices on the first iteration of Zen2 chips to the point where they now have to release another slew of chips to remind everyone where their minds should be with Zen3 pricing. The excuses never end.

 

[liahos2]

Milan is Q4 20 announcement, Q1 21 availability for everyone outside of a few select customers.

Whoever you contacted is either talking about when they're getting Milan or is bullshitting you.

Inspur CFO? They have no reason to *** a group of investors. They have no incentive to be pro intel or pro amd and they represent half the Chinese server market.Hard to see them getting Milan later than anyone else. But i guess because it's not super positive AMD news it must be wrong?


Profanity is not allowed in the technical boards.

AT Moderator ElFenix

 

[jpiniero]

sooner than later;
View attachment 26292

Also, Thundar something IoT group is an Intel 7nm(intel 7nm in the US/internal 7nm) project. So, Ponte isn't the first and might actually be the last 7nm product in the pipe.

There's been persistent rumors of them fabbing some of their secondary products on a bleeding edge foundry node. So far that hasn't happened. I'm talking about the key products - ie: Core and Xeon.

 

[beginner99]

What I really don't understand is why everybody seems to be buying into this price-reset nonsense. The precedent was already set with the Zen2 releases, that is how this industry has always functioned. You don't need to release a lackluster batch of CPUs in order to remind any of your loyal fans and potential customers that you'd be charging $499+ for a 12-Core and higher Zen 3 sku? If the performance justifies the price, customers will jump on it; and if you believe some in this thread, AMD doesn't even have to worry about the competition since the competition is stagnant. Which all seems a bit weird that AMD should be dropping prices on the first iteration of Zen2 chips to the point where they now have to release another slew of chips to remind everyone where their minds should be with Zen3 pricing. The excuses never end.

Can't say for sure if it is the local market price gouging or coming from AMD but the 3900XT is roughly $550 while the 3900X is going for roughly $470 so thats a $80 difference. If a 4900X now goes on the market for $550 it doesn't look like such a big price hike it really is.