Discussion Intel current and future Lakes & Rapids thread

[lobz]

How does that compare to this?

View attachment 64324

what the... ??? 😂🤣🤣

 

[repoman27]

The first article you quote was about the completion of a new factory, while the second is from well into 2020. And Q3 volume production might still be enough to squeak out a Mac chip for late in the year. TSMC has been very consistent about their public communications, which is commendable, but that doesn't mean their partners don't have their own expectations.

Sorry to nitpick, but the first article was actually about the *beginning* of construction for TSMC Fab 18 phase 4-6, which is relevant because that is where N3 high volume manufacturing will take place. Fab 12B, the development fab that handled the early risk starts, is the only other TSMC site with N3 capabilities. Fab 15 has some EUV systems, but only enough for N7+ and N6. Fab 18 phases 1-3 are full tilt with N5 to the extent that phase 4, which was originally slated to be the first N3 fab, is being used for N5/N4 expansion instead. Phases 5 and 6 will be for N3, and TSMC has already broken ground across the street on phases 7 and 8 for future N3 expansion. It looks like phase 4 is already operational, 5 is pretty much good to go, and 6 is probably close but construction was still ongoing earlier this year. This is exactly what we'd expect given what TSMC has disclosed about the N3 timeline.

Anyway, the point was that the quote directly contradicts what you said, and the IThome piece it was sourced from was picked up by dozens of other media outlets around that time in 2019. Also, going from, "Around 2019, the assumption for pretty much everyone was that N3 would be ready for volume production around mid-2022, and would show up in the A16." to, "Q3 volume production might still be enough to squeak out a Mac chip for late in the year." is a mighty shift of the goal posts, especially since Apple didn't publicly announce their silicon ambitions for the Mac until June 22, 2020.

 

[pakotlar]

I haven't seen it myself, but I know someone that works at Lenovo and according to what they say from their perspective it has been anything but consistent. So please don't make up nonsense.

Ah yes, your uncle at Lenovo

 

[Exist50]

Sorry to nitpick, but the first article was actually about the *beginning* of construction for TSMC Fab 18 phase 4-6, which is relevant because that is where N3 high volume manufacturing will take place. Fab 12B, the development fab that handled the early risk starts, is the only other TSMC site with N3 capabilities. Fab 15 has some EUV systems, but only enough for N7+ and N6. Fab 18 phases 1-3 are full tilt with N5 to the extent that phase 4, which was originally slated to be the first N3 fab, is being used for N5/N4 expansion instead. Phases 5 and 6 will be for N3, and TSMC has already broken ground across the street on phases 7 and 8 for future N3 expansion. It looks like phase 4 is already operational, 5 is pretty much good to go, and 6 is probably close but construction was still ongoing earlier this year. This is exactly what we'd expect given what TSMC has disclosed about the N3 timeline.

Anyway, the point was that the quote directly contradicts what you said, and the IThome piece it was sourced from was picked up by dozens of other media outlets around that time in 2019. Also, going from, "Around 2019, the assumption for pretty much everyone was that N3 would be ready for volume production around mid-2022, and would show up in the A16." to, "Q3 volume production might still be enough to squeak out a Mac chip for late in the year." is a mighty shift of the goal posts, especially since Apple didn't publicly announce their silicon ambitions for the Mac until June 22, 2020.

I'll point out that that article is from quite late '19. Moreover, it sources Digitimes, and while I can't find the full text of the original, they talk about accelerated plans in this article that seems to be the reference. So was the 2023 bit a prediction, or editorialization? https://www.digitimes.com/news/a20191024PD202.html

Additionally, yes, Apple didn't announce their plans till June '20, but rumors proceeded them by quite a while. Additionally, the M1 is basically an A14X. Substitute "Mac chip" with "iPad chip" if you want. Doesn't change the point.

 

[Saylick]

I haven't seen it myself, but I know someone that works at Lenovo and according to what they say from their perspective it has been anything but consistent. So please don't make up nonsense.

To be fair, Intel have been pretty consistent over the last few years.




Consistently inconsistent.

 

[DrMrLordX]

So 7% ST improvement

Actually there are some 12900Ks listed on GB5's site with a roughly-equivalent ST score as that sample. In fact, I'd say there are quite a few:

core i9-12900k - Geekbench 5 CPU Search - Geekbench

browser.geekbench.com

 

[deasd]

Actually there are some 12900Ks listed on GB5's site with a roughly-equivalent ST score as that sample. In fact, I'd say there are quite a few:

core i9-12900k - Geekbench 5 CPU Search - Geekbench

browser.geekbench.com

ST score might be a hint that looks like GoldenCove @10nm hits wall hard above 5.5Ghz and throttling, not a good sign at all. KS model is doubtful as some people suggest.

https://twitter.com/x/status/1546533557820088322

 

[igor_kavinski]

What are the chances of Intel porting RPL to TSMC N5 for getting higher clockspeeds, to deal with MTL delays?

 

[DrMrLordX]

What are the chances of Intel porting RPL to TSMC N5 for getting higher clockspeeds, to deal with MTL delays?

Slim to none.

 

[igor_kavinski]

Then the only thing left for Intel is to start a price war. Good for us.

 

[Henry swagger]

https://twitter.com/x/status/1546242241584173056

intel the innovator.. amd and apple updated it 🙂

 

[biostud]

Obviously the focus is on the top processors, 7950X vs 13900k which is fine if you have the need. But how much is RPL going to improve in P cores compared to ADL? Especially in the 13700 and 13600 segment.

 

[poke01]

https://twitter.com/x/status/1546242241584173056

intel the innovator.. amd and apple updated it 🙂

Still cannot believe Apple made the first consumer MCM GPU. If you had told me that 10 years ago I would have laughed at the possibility.

Sadly Intel is playing catch up nowdays in implementing new tech. Once Intel was the king. Now just another chip maker but Intel does hold 74% of the desktop chip market.

RPL looks okay a ADL+ so to say. MTL is where things get interesting.

 

[jpiniero]

But how much is RPL going to improve in P cores compared to ADL? Especially in the 13700 and 13600 segment.

I imagine it'd be the same deal. More cache and higher boost clocks.

 

[igor_kavinski]

I wish they would release P-core only 10 and 12 core parts or a 2P+32E part. I think they still have a high defect rate for 10nm that is preventing them from being too ambitious with having different die configurations.

 

[pakotlar]

What are the chances of Intel porting RPL to TSMC N5 for getting higher clockspeeds, to deal with MTL delays?

0. Think about how long it took to port Ice Lake to 14nm, and think about the fact that Intel 4 is in volume production this year.

 

[dullard]

Obviously the focus is on the top processors, 7950X vs 13900k which is fine if you have the need. But how much is RPL going to improve in P cores compared to ADL? Especially in the 13700 and 13600 segment.

I think the 13700 and 13600 will get a very necessary boost in avoiding the currently problematic situation of 4 E cores. As it is, unless you are careful, you can easily get your CPU intensive work to be processed in the background on just your E cores. Just 4 E cores causes that work to slow to a crawl. The rumor of going from 6P+4E to 6P+8E will significantly help that problem. All background multithreaded work will have double the number of cores and will almost double in speed. So, if you start one big task and then switch over to say browsing the internet, you will have a much better time with Raptor Lake than Alder Lake.

Even if you keep it in the foreground, just having 4 more E cores will have a small but noticeable increase in speed. For example, the 12700H (6P+8E) and 12650H (6P+4E) are nearly identical except for the E cores (the only other change is the 12700H has to power more EUs in the iGPU so the CPU gets a little less power to work with). And the 12700H does 7% to 16% better on multithreaded tasks in the foreground. https://nanoreview.net/en/cpu-compare/intel-core-i7-12700h-vs-intel-core-i7-12650h

For just the P cores? I wouldn't expect much change. Maybe just a small frequency boost, giving maybe a ~5% improvement.

 

[jpiniero]

What are the chances of Intel porting RPL to TSMC N5 for getting higher clockspeeds, to deal with MTL delays?

I had thought about that too. But don't you think porting the CPU tile to TSMC would be easier?

My guess is the alternative to TSMC is a Raptor Lake Refresh.

 

[JasonLD]

What are the chances of Intel porting RPL to TSMC N5 for getting higher clockspeeds, to deal with MTL delays?

There were leaks about Rocket Lake 2 years before its actual release. If there were another product between Raptor and Meteor, most likely we would have known about it already. So I would say none at this point.

 

[jpiniero]

There were leaks about Rocket Lake 2 years before its actual release. If there were another product between Raptor and Meteor, most likely we would have known about it already. So I would say none at this point.

That's why I mentioned the Refresh. You could do it with the same die. Almost like a stealth price cut including more E cores at the lower tiers. Not ideal of course but better than nothing. My preference would still be to do the TSMC port, even if it means a delay.

What I want to know if Raptor Lake's Small die is going to be 6+0 or 6+8. Note that Meteor Lake Mobile isn't increasing the core count for either die.

 

[JasonLD]

That's why I mentioned the Refresh. You could do it with the same die. Almost like a stealth price cut including more E cores at the lower tiers. Not ideal of course but better than nothing. My preference would still be to do the TSMC port, even if it means a delay.

What I want to know if Raptor Lake's Small die is going to be 6+0 or 6+8. Note that Meteor Lake Mobile isn't increasing the core count for either die.

Another Refresh would have gotten another Lake name designation. Regardless of being it another 10nm refresh or 5nm TSMC port. I don't think there will be another product between Raptor and Meteor. Meteor Being late 2023~early 2024 product still falls on normal Intel product cycle.

 

[jpiniero]

Another Refresh would have gotten another Lake name designation.

They did a full lineup refresh with Coffee Lake and Comet got a refresh at the low end. Same name in both instances.

 

[Doug S]

Still cannot believe Apple made the first consumer MCM GPU. If you had told me that 10 years ago I would have laughed at the possibility.

Sadly Intel is playing catch up nowdays in implementing new tech. Once Intel was the king. Now just another chip maker but Intel does hold 74% of the desktop chip market.

RPL looks okay a ADL+ so to say. MTL is where things get interesting.

A lot of times who is "first" to some new technology is more about need than capability. Apple being first with an MCM GPU doesn't mean they have better designers than Nvidia, AMD and Intel.

Apple simply had no other choice if they wanted to maintain the same cache coherent GPU design they'd been using for years in their SoCs, which requires the CPU and GPU on the same die to share the SLC path to main memory. This was pointed out more than a few times by more than a few people, but some insisted on believing Apple would do a discrete GPU because a chiplet based GPU was somehow "impossible" since no one else had done it.

Nvidia and AMD have been basically making the largest dies they could that 1) yielded sufficiently and 2) along with the VRAM stayed within the PCIe power limit. The only reason they will be forced to go chiplet are 1) removes yield as an obstacle 2) PCIe is getting a higher power limit so even a reticle sized die might not max out the power 3) reticle size will be halved soon so high NA EUV processes serve as a hard deadline for going chiplet.

 

[JasonLD]

They did a full lineup refresh with Coffee Lake and Comet got a refresh at the low end. Same name in both instances.

If we don't hear about Raptor Lake refresh by next month or two, most likely they don't exist, since we should heard about it already. Coffee Lake and Comet Lake refresh got leaked 1.5 years before launch.

 

[Abwx]

https://twitter.com/x/status/1546242241584173056

intel the innovator.. amd and apple updated it 🙂

Lol...

Athlon 64 in September 23, 2003
The first Athlon 64 X2 CPUs were released in May 2005

https://en.wikipedia.org/wiki/Athlon_64

May 25, 2005 for the Pentium D...

https://en.wikipedia.org/wiki/Pentium_D