Discussion Intel current and future Lakes & Rapids thread

[maddie]

Intel is already one of the biggest TSMC customers. And didn't new CEO described their future earlier, there will be leadership products for both client and server environments from TSMC 2023? Intel doesn't have choice as their own abilities to produce are somehow limited.

True? Making what?

 

[maddie]

Yes, I'm sure they are. Man, gotta love those 5nm Zen 3s!

I guess any publication with "times" in the name has seen better days.

 

[scineram]

My desktop has over 2000 active threads and hundreds of active processes at any given point. Sure, many of them run in the background, but for many workloads, we will always benefit from more cores.

Alder Lake has quite a few unknowns that go beyond the 20% “big core” improvement. Those tiny little gracemont cores with no hyperthreading are going to sip power, so ask yourself why ADL-S currently has a PL1 of 125W and a PL2 of 228W (without AVX-512 no less). I am willing to bet it isn’t because Golden Cove is power hungry (well, it probably will be, but not like rocket lake). We still don’t know how high the final silicon will even boost. We know 10SF can hit higher clocks at a given voltage, and can also hit 5ghz, and 10ESF will offer further improvements beyond that. PCIE5 uncore was mentioned, but I believe DDR5 savings will offset that. Could we see Intel hit multicore 5.3 and a single core 5.4 or 5.5? I don’t know, but anything is possible.

Whatever they will be we can be sure Arm, especially Apple will continue to embarass, humiliate the clock chasers.

 

[naukkis]

True? Making what?

TSMC's Top Customers 2019-2021

Not a big surprise but it will be interesting to see how much of a bump Intel will see at 3nm. I'm also guessing Mediatek will jump up a bit as well.

semiwiki.com

Intel is a big company. Until now Intel isn't produced it's core products in TSMC, but their new CEO recently published that they will soon, on TSMC more advanced process.

Actually I wonder how most of people are missed that, it was public announcement and reported everywhere, also there in Anandtech.

Intel’s x86 Designs No Longer Limited to Intel on Intel: IP Blocks for Foundry, Cores on TSMC

www.anandtech.com

 

[Zucker2k]

Could we see Intel hit multicore 5.3 and a single core 5.4 or 5.5? I don’t know, but anything is possible.

I want some of what you're smoking. Between @IntelUser2000 and you, ADL-S should be somewhere in the middle. I mean, I hope you're right, though, I think your figures are unrealistic given what we know so far about the performance of Intel's 10nm SF node.

 

[maddie]

TSMC's Top Customers 2019-2021

Not a big surprise but it will be interesting to see how much of a bump Intel will see at 3nm. I'm also guessing Mediatek will jump up a bit as well.

semiwiki.com

Intel is a big company. Until now Intel isn't produced it's core products in TSMC, but their new CEO recently published that they will soon, on TSMC more advanced process.

Actually I wonder how most of people are missed that, it was public announcement and reported everywhere, also there in Anandtech.

Intel’s x86 Designs No Longer Limited to Intel on Intel: IP Blocks for Foundry, Cores on TSMC

www.anandtech.com

Wow. A 30% bigger TSMC customer than AMD in 2019.

 

[IntelUser2000]

I want some of what you're smoking. Between @IntelUser2000 and you, ADL-S should be somewhere in the middle. I mean, I hope you're right, though, I think your figures are unrealistic given what we know so far about the performance of Intel's 10nm SF node.

Alderlake will likely put Intel in a far better situation than today, but there's little chance 8+8 will be faster than 5950X in embarrassingly parallel scenarios.

In the most optimistic scenario 5950X simply has more compute. It's a mobile chip first and foremost.

The 2+8 config will likely end up faster than 4 core Tigerlake by a fair bit. That's real nice.

so ask yourself why ADL-S currently has a PL1 of 125W and a PL2 of 228W (without AVX-512 no less).

Or, Alderlake has power far better controlled, so with PL1 125W it can perform to the full, unlike Rocketlake and even Cometlake where you needed unlimited PL2, which in practice means the PL1 needed to be 200W+.

 

[jpiniero]

Or, Alderlake has power far better controlled, so with PL1 125W it can perform to the full, unlike Rocketlake and even Cometlake where you needed unlimited PL2, which in practice means the PL1 needed to be 200W+.

Not sure what you mean by need. It's very likely that the boost frequencies allowed at stock will burn 200+. It's gonna be exactly like Rocket Lake.

 

[IntelUser2000]

Not sure what you mean by need. It's very likely that the boost frequencies allowed at stock will burn 200+. It's gonna be exactly like Rocket Lake.

Not if it ends up at 5GHz or something noticeably lower.

And I hope it does because what they are doing now is insane. At frequencies above 5GHz we enter the area of nonstandard cooling setups. In a way we're already there with watercooling setups being common and air heatsink+fan setups weighing over 1.5kg.

Ironically the chip that started Intel on an efficient path eventually ended up being the ancestor to a chip that used even more power and runs hotter than Netburst. They also achieved their dreams of reaching 5GHz right?

 

[jpiniero]

Not if it ends up at 5GHz or something noticeably lower.

5 Ghz on all 8 big cores plus the small clusters is going to draw a ton.

 

[insertcarehere]

5 Ghz on all 8 big cores plus the small clusters is going to draw a ton.

This, the whole point of the small clusters should be to to enable somewhat competitive MT performance without pushing all the big cores well past optimal V/Fs. Alder Lake should have lower full MT boosts than RKL/CML.

 

[Dave2150]

Not if it ends up at 5GHz or something noticeably lower.

And I hope it does because what they are doing now is insane. At frequencies above 5GHz we enter the area of nonstandard cooling setups. In a way we're already there with watercooling setups being common and air heatsink+fan setups weighing over 1.5kg.

Ironically the chip that started Intel on an efficient path eventually ended up being the ancestor to a chip that used even more power and runs hotter than Netburst. They also achieved their dreams of reaching 5GHz right?

Performance matters more IMO. Not many know/care if their cooling solution is 500G, 1KG or more. They just care if it's faster and saves them time (money).

Been running my 11900k since launch, mild OC to 5.4Ghz on 3 cores, 5.1Ghz on other cores, on a Noctua NH-D15S. This is an air cooler, hardly an 'extreme' cooler by any means. My temperatures are actually lower than on my 6700k - though that's mostly due to solder vs TIM. Of course the 11900k pumps out more heat into my room, though still nothing compared to my 320W 3080.

Most importantly - it's crazy fast, has no stability issues/bugs/crashes. System responsiveness/transfer speeds on my 980Pro (pcie4) is truly awesome in my workloads.

 

[beginner99]

on a Noctua NH-D15S. This is an air cooler, hardly an 'extreme' cooler by any means.

As a long-term NH-D14 owner I disagree. It's pretty extreme. But it is a metal block don't break so it will last as long as new mounting brackets are made. The one big advantage over watercooling what you probably consider extreme.

 

[IntelUser2000]

Performance matters more IMO. Not many know/care if their cooling solution is 500G, 1KG or more. They just care if it's faster and saves them time (money).

Ok, if 11900K actually had good performance than power use could be excused.

Except it performs quite horribly. Zen 3 based chips are a far better deal. In scenarios where it would actually matter to save money and time is where Zen 3 whips Rocketlake completely.

Intel is forced to push it out of bounds because they are that far behind.

Zen 3 desktop chips have double the cores of Zen 3 mobile parts. Alderlake-S should have been at least 8+12, or even 8+16, and mobile getting less.

I can see Alderlake mobile potentially being very good. Desktop? Better than today is all I have to say.

 

[jpiniero]

This, the whole point of the small clusters should be to to enable somewhat competitive MT performance without pushing all the big cores well past optimal V/Fs. Alder Lake should have lower full MT boosts than RKL/CML.

Intel's not going to do that. Maybe they could make it easier to configure the board to run at those optimal curves... but it'd be slower or even a lot slower like that HW Unboxed video shows with Rocket Lake when Rocket Lake is run at base clocks.

Main point of the small core clusters is getting better idle while getting better MT performance while at load.

 

[DrMrLordX]

Will be interesting to see whether Intel will be a first mover on N3 alongside Apple or join somewhat later.

I'm guessing "no" on that one. It would not be in TSMC's best interest to put Intel ahead of their existing customers, for numerous reasons.

Intel is a big company. Until now Intel isn't produced it's core products in TSMC, but their new CEO recently published that they will soon, on TSMC more advanced process.

And because Gelsinger wishes it to be so, it is so? Apple alone is the reason why Intel won't be getting leading-edge wafers. They can easily outspend Intel if there's a bidding war. TSMC is also going to be reluctant to do anything that helps Intel too much, given that Intel was until recently one of the few remaining rivals to TSMC's leading edge node hegemony. Intel's foundry business 2.0 is also a minor threat to TSMC. I just don't see them doing anything to keep Intel alive as a design house, at least to the point that Intel can afford to bag on their own fabs (which is apparently what they're doing, at least volume-wise). Also you may remember this article:

Rumor: TSMC Won't Build New Capacity for Intel, Views Orders as Temporary

TSMC may be fabbing chips for Intel, but the company isn't planning to put in new fabs or lines specifically for Chipzilla. The implication is that Intel is still planning to return to its own facilities as quickly as possible.

www.extremetech.com

 

[Dave2150]

As a long-term NH-D14 owner I disagree. It's pretty extreme. But it is a metal block don't break so it will last as long as new mounting brackets are made. The one big advantage over watercooling what you probably consider extreme.

I disagree entirely. Air has more than one advantage over watercooling:

1. Noise. Watercooling requires a pump. Even the best pumps are quite noisy at idle, whereas air cooling at idle (on a NH-D15 for example) is silent, as the fan can spin sub 500RPM)
2. Reliability as you mention. Less parts to fail
3. Maintenance - Water will evaporate over time, needing time consuming draining and refilling.
4. Cost - Air is obviously far cheaper

Also, watercooling isn't extreme - it's just a method of cooling some prefer for looks, and for the best cooling performance (note - AIO watercooling is bested/matched by the best air coolers, you need a custom loop for better than air performance).

 

[moinmoin]

I'm guessing "no" on that one. It would not be in TSMC's best interest to put Intel ahead of their existing customers, for numerous reasons.

That's exactly why this will be interesting to watch: In normal circumstances there is indeed no way TSMC would allow Intel ahead of other existing longer time leading edge customers.

The only way I see is that Intel aims for the same position as Apple as a first mover, that would mean: Essentially paying TSMC upfront to build up fab capacity, features and timescale specifically for it. For TSMC this would be in addition to Apple and other customers (so a big win), and it would also explain Intel's low order for EUV machines to ASML as well as the rumor that Intel relinquished on them specifically in favor of TSMC.

 

[andermans]

That's exactly why this will be interesting to watch: In normal circumstances there is indeed no way TSMC would allow Intel ahead of other existing longer time leading edge customers.

On the other hand, without a successful launch on TSMC Intel is never going to shut down their own fabs. So if there is a chance that they can onboard Intel and then get Intel to really turn down / deprioritize their own fabs there might be a benefit to some give and take here.

(of course that would be a significant bet)

 

[moinmoin]

On the other hand, without a successful launch on TSMC Intel is never going to shut down their own fabs. So if there is a chance that they can onboard Intel and then get Intel to really turn down / deprioritize their own fabs there might be a benefit to some give and take here.

(of course that would be a significant bet)

I honestly don't think TSMC is up to the multidimensional chess you lot seem to like to allude to. To TSMC in the current situation any expansion of leading edge capacity is a win. If Intel pays up for TSMC to achieve that it will happen, if not tough luck for Intel.

 

[DrMrLordX]

That's exactly why this will be interesting to watch: In normal circumstances there is indeed no way TSMC would allow Intel ahead of other existing longer time leading edge customers.

The only way I see is that Intel aims for the same position as Apple as a first mover, that would mean: Essentially paying TSMC upfront to build up fab capacity, features and timescale specifically for it. For TSMC this would be in addition to Apple and other customers (so a big win), and it would also explain Intel's low order for EUV machines to ASML as well as the rumor that Intel relinquished on them specifically in favor of TSMC.

Intel would have to foot a hefty bill for that to truly work in TSMC's favor over the long term.

On the other hand, without a successful launch on TSMC Intel is never going to shut down their own fabs.

No, but if Intel is stuck at 10nm ad infinitum with 7nm operating as a low-volume boutique node, they're not in a much-better position with their fabs than, say, Globalfoundries.

 

[naukkis]

And because Gelsinger wishes it to be so, it is so?

Gelsinger didn't specify manufacturing node other than it's advanced. Those 3nm rumours come sources close to TSMC.

And TSMC do business, it doesn't matter where the money comes. Intel other hand will use Samsung foundries if TSMC won't co-operate and sure TSMC doesn't want lose Intel money for their rival.

Other what Gelsinger said : Intel designs are free to use best manufacturing process available, from Intel itself or from the foundries. So if they aren't also trying to get best possible process as soon as possible they aren't executing properly.

 

[Zucker2k]

The TigerLake-H Review

 

[eek2121]

The TigerLake-H Review

Not bad at all. Hopefully they flush out the review a bit more and fix errors.

 

[eek2121]

I preordered the Razer Blade 15 Advanced. It isn't coming until the end of next month, but when it does, if there are any tests or anything you'd like to see, just let me know.