Discussion Intel current and future Lakes & Rapids thread

[ondma]

Neither is personal criticism direct at other posters.

 

[Thunder 57]

Neither is personal criticism direct at other posters.

Perhaps you're right. I won't bring up your name anymore.

 

[trivik12]

Tigerlake-U to support LPDDR5 ram.

Intel 10nm Tiger Lake-U Mobile CPUs Rumored To Be DDR5 RAM Enabled

A next generation Intel 'Tiger Lake-U' processor submission for the Eurasian Economic Commission (EEC) suggests the upcoming CPU series will support both LPDDR4X and LPDDR5 memory when it arrives next year.

hothardware.com

 

[DrMrLordX]

Looks like IceLake-SP is going to be a late 2020 CPU, and Sapphire Rapids will be a late 2021 CPU:

Intel’s 2021 Exascale Vision in Aurora: Two Sapphire Rapids CPUs with Six Ponte Vecchio GPUs

www.anandtech.com

Those CPUs will be Sapphire Rapids CPUs, Intel’s second generation of 10nm server processors coming after the Ice Lake Xeons. The announcement today reaffirmed that Sapphire Rapids is a 2021 processor; and likely a late 2021 processor, as the company also confirmed that Ice Lake will have its volume ramp through late 2020.

Cooper Lake will have a lot of time on market all by itself.

 

[IntelUser2000]

Looks like IceLake-SP is going to be a late 2020 CPU, and Sapphire Rapids will be a late 2021 CPU:

Cooper Lake will have a lot of time on market all by itself.

Yes, just as expected. The more they say it'll change, the more it stays the same. You need a 1 year cycle for the previous gen to sell. So unless they wanted to keep the bifurcation, Sapphire Rapids would only arrive a year after the latest arrival date of the predecessor. And in this case its Icelake-SP.

Bifurcation isn't something they desired, but forced to because their roadmap was in disarray.

This is why I basically ignore release dates that seem earlier than the historical trend. You'll always be disappointed. You think companies are going to lose their chance at maximizing profits?

 

[DrMrLordX]

@IntelUser2000

I read this more as, "IceLake-SP will never be available in significant quantity". Intel doesn't expect IceLake-SP to sell well, and they probably don't expect to yield many of those 38c dice anyway. If IceLake-SP's volume ramp is "through late 2020" then it will have too short of a market window for it to be a serious product. Realistically-speaking, Cooper Lake will dominate the Xeon landscape until Intel can deliver Sapphire Rapids.

 

[IntelUser2000]

If you look at how Icelake is doing compared to Comet Lake, its doing pretty well. It's available in quite a few systems. Just browsing through the Dell site I can find roughly equal number of systems using both.

Icelake-SP has advantages over Cooperlake. It has PCIe Gen 4, so it has double I/O bandwidth. It has native 8-channel memory support, while Cooper Lake will have to split it over two dies. Being the latest core also means it'll have some unknown changes in it too. Cooper Lake is largely known, but we don't know what changes the -SP Icelake will have compared to the client version.

Cooperlake will have higher core counts with dual-dies meaning it can go in the systems requiring dense configurations and in absolute performance.

Icelake with higher perf/clock can be used for high single thread performance SKUs and in lower end ones and those requiring better I/O such as large storage systems. Not needing MCM might mean it can also go into lower power setups.

Neither is ideal, but they'll work to find places for both.

 

[JoeRambo]

I am looking forward to Icelake-SP. The yields and margins should be left to corporations and beancounters, but obviuosly with AMD giving such great competition, I think Intel will be forced to properly price and release ASAP.

Just because there is 64C Epyc, does not mean there is suddenly no market for Icelake-SP with 24-38 cores, I think we have already discussed in one of these threads that not every CPU sold is 28C or 64C monster, and most sales happen in "smart" range. And big playas are buying off counter and off SKU-list.

So combining the two, all we need is properly priced product for its core count (read performance) range. I don't care how Intel prices 38C, they might ask $66666, but I care how they price for example 24C-32C range.

8 chans of faster DDR4, 3+Ghz with all cores active with non-AVX2 load, PCIE4 Optane once available? Count me in.

 

[DrMrLordX]

he yields and margins should be left to corporations and beancounters,

So what happens when IceLake-SP becomes the next Cannonlake? IceLake-U seems to be doing okay-ish (though clocks are not great). Producing a die at least six times larger on 10nm is a whole 'nother ballgame.

 

[lobz]

I am looking forward to Icelake-SP. The yields and margins should be left to corporations and beancounters, but obviuosly with AMD giving such great competition, I think Intel will be forced to properly price and release ASAP.

Just because there is 64C Epyc, does not mean there is suddenly no market for Icelake-SP with 24-38 cores, I think we have already discussed in one of these threads that not every CPU sold is 28C or 64C monster, and most sales happen in "smart" range. And big playas are buying off counter and off SKU-list.

So combining the two, all we need is properly priced product for its core count (read performance) range. I don't care how Intel prices 38C, they might ask $66666, but I care how they price for example 24C-32C range.

8 chans of faster DDR4, 3+Ghz with all cores active with non-AVX2 load, PCIE4 Optane once available? Count me in.

I really wouldn't expect Ice Lake SP to be in the smartly priced category.

 

[lobz]

IceLake-U seems to be doing okay-ish.

I believe that when I see a proper report of how many of those CPUs / laptops were actually produced, until then I believe Charlie (S/A), because ca. 98% of what he said about Intel's 10nm in the past 2 years was correct. According to him, it's better than Cannon Lake, but still not financially viable and the quantities are way below what one should call mass production.

 

[DrMrLordX]

I believe that when I see a proper report of how many of those CPUs / laptops were actually produced, until then I believe Charlie (S/A), because ca. 98% of what he said about Intel's 10nm in the past 2 years was correct. According to him, it's better than Cannon Lake, but still not financially viable and the quantities are way below what one should call mass production.

At least you can buy Icelake-U. Cannonlake was dumped in small quantity on the Chinese market.

 

[mikk]

I believe that when I see a proper report of how many of those CPUs / laptops were actually produced, until then I believe Charlie (S/A), because ca. 98% of what he said about Intel's 10nm in the past 2 years was correct.

Charlie told 1 year ago that Intel killed their 10nm entirely, looks like he was dead wrong. Icelake-U availability is better than expected and next year we will see several new 10nm products like Tremont, Lakefield, Tigerlake, dedicated graphics DG1...

 

[yeshua]

Charlie told 1 year ago that Intel killed their 10nm entirely, looks like he was dead wrong. Icelake-U availability is better than expected and next year we will see several new 10nm products like Tremont, Lakefield, Tigerlake, dedicated graphics DG1...

Charlie said they had killed the first iteration of 10nm because they set unrealistic expectations (and looks like he's right considering the anemic volumes of Cannon Lake CPUs parts of which Intel couldn't even bin enough of to enable their iGPU). Ice Lake U CPUs are produced using a "relaxed" 10nm+ node.

 

[lobz]

Charlie told 1 year ago that Intel killed their 10nm entirely, looks like he was dead wrong. Icelake-U availability is better than expected and next year we will see several new 10nm products like Tremont, Lakefield, Tigerlake, dedicated graphics DG1...

You didn't read what he said and you're just repeating something you've heard from someone else, or you may just haven't read it thoroughly enough.

Edit: what I mean with that -> the node Intel announced many years ago, doesn't exist at all.

 

[ondma]

Here is a direct quote from Charlie admitting that he was wrong: "A few months ago SemiAccurate claimed Intel killed their 10nm process, we were wrong. Instead what looks to be happening is that the 10nm process is severely deprecated because it makes no economic sense to carry on with."

 

[krumme]

Here is a direct quote from Charlie admitting that he was wrong: "A few months ago SemiAccurate claimed Intel killed their 10nm process, we were wrong. Instead what looks to be happening is that the 10nm process is severely deprecated because it makes no economic sense to carry on with."

Lol. That's Charlie's way of saying he is right.

 

[uzzi38]

Charlie told 1 year ago that Intel killed their 10nm entirely, looks like he was dead wrong. Icelake-U availability is better than expected and next year we will see several new 10nm products like Tremont, Lakefield, Tigerlake, dedicated graphics DG1...

Well, >0 is definitely better than expected, but let's not kid ourselves and say ICL-U is anything even close to high volume. The full ICL-U design has but a handful of devices it's in, and a very large number of ICL devices have only half the GPU enabled and pretty heavily cut down clocks. They can also cap out clock-wise at 3.8GHz (1068G7 is non-existent thus far).

If that wasn't enough, Lakefield is late, Tiger Lake will probably be nice but CML-U's existence says enough about volume and DG1 is mostly low end stuff and is releasing at the end of the year (judging by first working silicon just last week).

Let's not pretend 10nm is even remotely close to a good state and/or will be in a good state by the end of next year.

 

[lobz]

Here is a direct quote from Charlie admitting that he was wrong: "A few months ago SemiAccurate claimed Intel killed their 10nm process, we were wrong. Instead what looks to be happening is that the 10nm process is severely deprecated because it makes no economic sense to carry on with."

Is it just me, or more and more people are really struggling to comprehend complex English sentences and to put them in the proper context? What you quoted means exactly what I wrote in my previous comment

 

[lobz]

Well, >0 is definitely better than expected, but let's not kid ourselves and say ICL-U is anything even close to high volume. The full ICL-U design has but a handful of devices it's in, and a very large number of ICL devices have only half the GPU enabled and pretty heavily cut down clocks. They can also cap out clock-wise at 3.8GHz (1068G7 is non-existant thus far).

If that wasn't enough, Lakefield is late, Tiger Lake will probably be nice but CML-U's existence says enough about volume and DG1 is mostly low end stuff and is releasing at the end of the year (judging by first working silicon just last week).

Let's not pretend 10nm is even remotely close to a good state and/or will be in a good state by the end of next year.

Just what I said. What is now being sold as Ice Lake has nothing to do with the 10nm process Intel originally announced - and kept announcing it every quarter.
Not to mention, it took AMD almost 1,5 years to release Vega (in very low volume) after the first prototype came back.

 

[uzzi38]

Just what I said. What is now being sold as Ice Lake has nothing to do with the 10nm process Intel originally announced - and kept announcing it every quarter.
Not to mention, it took AMD almost 1,5 years to release Vega (in very low volume) after the first prototype came back.

They've promised investors it's 2020 for first consumer graphics IIRC, they'll launch something even if it's a 128EU die.

 

[ondma]

Is it just me, or more and more people are really struggling to comprehend complex English sentences and to put them in the proper context? What you quoted means exactly what I wrote in my previous comment

Is it just me, or more and more people are really struggling to comprehend complex English sentences and to put them in the proper context? What you quoted means exactly what I wrote in my previous comment

Well, there certainly is a comprehension, and/or selective perception problem. Maybe you should examine your own posts. In an unusual moment of rationality and a lull in his usual anti-intel rants, Charlie himself admitted he was wrong, and I posted the direct quote. No ambiguity there.

 

[mikk]

Charlie said they had killed the first iteration of 10nm because they set unrealistic expectations (and looks like he's right considering the anemic volumes of Cannon Lake CPUs parts of which Intel couldn't even bin enough of to enable their iGPU). Ice Lake U CPUs are produced using a "relaxed" 10nm+ node.

Lol he even admitted a few months later that he was wrong. He initially told 10nm is dead, he didn't differentiate between 10 and 10+ etc. The killed off first iteration of 10nm would be no news because everyone knew Cannonlake never existed really. He was dead wrong and he knows it. Also the relaxed 10+ was denied by Intel, it was wrong as well.

 

[coercitiv]

He was dead wrong and he knows it. Also the relaxed 10+ was denied by Intel, it was wrong as well.

Intel denied 10nm+ is a relaxed process?That settles it!

Intel denies further delays to 10nm chips beyond 2017

INTEL has moved to quash speculation that its first 10nm chips could be pushed back even further than the second half of 2017, after already delaying them from this year.The chipmaker had planned to introduce the first family of 10nm processors sometime in late 2016. However, technical challenges encountered in shrinking transistors to ever smaller scales led to the launch being delayed until the second half of 2017.

Speculation about a further delay was sparked by a recent job vacancy posted on Intel's website. The listing was spotted by journalists at investor website The Motley Fool and implied that Intel will not begin mass production of 10nm components until two years after the posting date.

This would not seem to be much of a problem, but Intel has responded by contacting the site with a statement that the job listing was incorrect, and reaffirming that the first 10nm products are still scheduled for delivery in the second half of 2017.

Can we drop this subject please? Intel has lost any credibility it had with regard to their 10nm adventure, and the only valid proof for a healthy 10nm process is good availability for 10nm CPU products that perform inherently better than their 14nm counterparts (performance, efficiency, preferably both). Up until now we have yet to see this, hopefully upcoming server parts will break the ice.

 

[maddie]

Is it just me, or more and more people are really struggling to comprehend complex English sentences and to put them in the proper context? What you quoted means exactly what I wrote in my previous comment

This 10nm is alive and OK talk reminds me of the opposite of the Shakespearean quote. A rose by any other name would smell as sweet. Well, we have the same name, but it doesn't smell as sweet