Discussion Intel current and future Lakes & Rapids thread

[DrMrLordX]

I have a feeling it's going to be the best at some things (like those requiring low latency memory) for a very, very long time.

Intel hasn't had a really good memory controller since Comet Lake.

 

[Exist50]

Arrow Lake comes with Intel 20A says Intel, Meteor Lake is the only Intel 4 client core CPU as far as we know unless the plan changed. Maybe we will get an update at Intel innovation. I know about the lite and full Royal core, to me lite doesn't sound major and Arrow Lake in the reddit leak didn't sound like a major architecture uplift.

Arrow Lake will also be N3, but that aside, this whole "Royal lite" thing, along with pretty much everything MLID says about the program, should really just be thrown out. He even still seems to think Lion Cove is related to Royal, which shows how little information he's actually working with.

Anyway, as I think I've said before, I'm expecting Royal to have at least double Golden Cove IPC in the first incarnation we see, and if I'm being honest expect the real value to be closer to 2x Lion Cove, though with significant error margin both plus and minus. I'd then hope for a Sandy Bridge-like followup.

Edit: Though I'm curious how Panther Lake (and Panther Cove?) fit in. I'm not expecting much, but perhaps we'll see something interesting to avoid piling everything on Royal + Nova Lake.

 

[Tarkin77]

Igor has new information

Never Ending Story: Intel’s Sapphire Rapids (maybe) comes in the 12th Stepping - Planned shipping, a growing bug list and possible availability | Exclusive | igor´sLAB

Intel’s new Xeons, i.e. the scalable server processors of the 4th generation, are delayed again. Once again, one might say, and the reasons are very diverse. The problems and errors are collected…

www.igorslab.de

"Intel has now announced the “launch window” for Sapphire Rapids (SPR) for calendar week 6 to 9 (Feb. 6, 2023 to March 3, 2023), while the first shipment to selected recipients is still scheduled for 2022 in two waves. Calendar week 42 for the smallest models (2S) and calendar week 45 for the larger models (4 and 8S) are being speculated."

By Q1 2023, SPR will face Genoa X and probably Bergamo and maybe Mi300?

 

[mikk]

Arrow Lake will also be N3, but that aside, this whole "Royal lite" thing, along with pretty much everything MLID says about the program, should really just be thrown out. He even still seems to think Lion Cove is related to Royal, which shows how little information he's actually working with.

Anyway, as I think I've said before, I'm expecting Royal to have at least double Golden Cove IPC in the first incarnation we see, and if I'm being honest expect the real value to be closer to 2x Lion Cove, though with significant error margin both plus and minus. I'd then hope for a Sandy Bridge-like followup.

Edit: Though I'm curious how Panther Lake (and Panther Cove?) fit in. I'm not expecting much, but perhaps we'll see something interesting to avoid piling everything on Royal + Nova Lake.

You mean TSMC 3nm?

And of course I won't go with MLID; the reddit leak seems more legit. Lion Cove is more likely a typical tock increase in the range of 10-20%. The Mont gains could be higher, historically Intel made bigger gains there. I guess Nova Lake isn't ready after Arrow Lake and they introduced Panther Lake which wasn't in the reddit leak, whatever it is. I hope it's more than a simple Arrow Lake refresh. Unless they renamed Nova Lake into Panther Lake.

 

[pakotlar]

Igor has new information

Never Ending Story: Intel’s Sapphire Rapids (maybe) comes in the 12th Stepping - Planned shipping, a growing bug list and possible availability | Exclusive | igor´sLAB

Intel’s new Xeons, i.e. the scalable server processors of the 4th generation, are delayed again. Once again, one might say, and the reasons are very diverse. The problems and errors are collected…

www.igorslab.de

"Intel has now announced the “launch window” for Sapphire Rapids (SPR) for calendar week 6 to 9 (Feb. 6, 2023 to March 3, 2023), while the first shipment to selected recipients is still scheduled for 2022 in two waves. Calendar week 42 for the smallest models (2S) and calendar week 45 for the larger models (4 and 8S) are being speculated."

By Q1 2023, SPR will face Genoa X and probably Bergamo and maybe Mi300?

Biiiig ouch.

 

[Exist50]

You mean TSMC 3nm?

Yes. Presumably N3B.

Lion Cove is more likely a typical tock increase in the range of 10-20%.

I think Lion Cove is somewhat atypical, but that sounds like a reasonable IPC range.

I hope it's more than a simple Arrow Lake refresh.

Porque no los dos?

 

[msj10]

להגות את הדבר הגדול הבא - ראיון עם האחראי על תהליך יצירת ליבות עיבוד עוצמתיות במעבדי אינטל • HWzone

בראיון זה ישבנו אחד על אחד עם עדי יועז, הארכיטקט הראשי של ליבת הביצועים של אינטל וזה שצוותו עומד מאחורי Golden Cove העוצמתית של מעבדי Alder Lake

hwzone.co.il

interview with the lead architect of golden cove. he is saying after raptor lake and meteor lake we are going to see "bigger and bigger jumps".
it's in Hebrew so use google translate.

 

[Markfw]

להגות את הדבר הגדול הבא - ראיון עם האחראי על תהליך יצירת ליבות עיבוד עוצמתיות במעבדי אינטל • HWzone

בראיון זה ישבנו אחד על אחד עם עדי יועז, הארכיטקט הראשי של ליבת הביצועים של אינטל וזה שצוותו עומד מאחורי Golden Cove העוצמתית של מעבדי Alder Lake

hwzone.co.il

interview with the lead architect of golden cove. he is saying after raptor lake and meteor lake we are going to see "bigger and bigger jumps".
it's in Hebrew so use google translate.

OK, not to derail, but this is funny IMO. The first page of that link, the first advertisement is for a Ryzen 5950x !

 

[mikk]

להגות את הדבר הגדול הבא - ראיון עם האחראי על תהליך יצירת ליבות עיבוד עוצמתיות במעבדי אינטל • HWzone

בראיון זה ישבנו אחד על אחד עם עדי יועז, הארכיטקט הראשי של ליבת הביצועים של אינטל וזה שצוותו עומד מאחורי Golden Cove העוצמתית של מעבדי Alder Lake

hwzone.co.il

interview with the lead architect of golden cove. he is saying after raptor lake and meteor lake we are going to see "bigger and bigger jumps".
it's in Hebrew so use google translate.

Here is the google translation

LM: Want to tell a little about the future of the P Cores ? The Golden Cove core sets a very high performance standard in the market, and is, at least in my opinion, the highest performance jump between two generations of processors for a processing core since the transition to Sandy Bridge , with all due respect to Skylake on the way

EI: From what I can tell. Among our plans for continued processor improvements in the coming generations, we're working hard on the Next Big Thing, things you'll see in the future. Meanwhile in the coming years you will see Raptor Lake and Meteor Lake and then you will see bigger and bigger jumps.

 

[igor_kavinski]

So the guy is basically admitting that Meteor Lake won't be that special.

 

[mikk]

So the guy is basically admitting that Meteor Lake won't be that special.

Meteor Lake is more about power efficiency (which can lead to massive gains in low power mobile SKUs), however an IPC incease of around 10% is not out of range given the size increase from Redwood Cove over Golden Cove in some areas (in particular the Integer execution block). Lion Cove should bring much more performance gains, I think this is hardly unexpected. And after Lion Cove the performance rumor about Nova Lake is much higher. What he says is basically a confimation.

 

[Exist50]

however an IPC incease of around 10% is not out of range

In this case, it is. I'd start managing expectations with half that.

 

[poke01]

Nova Lake is much higher.

Nova lake is said to completely redesign the Core arch. The biggest since the first Core arch launched. Intel may have an undisputed pref/w leadership with Nova Lake.

 

[poke01]

Raptor Lake is meh. Meteor Lake is only interesting because of new node, a better iGPU and will be good for mobile.

But it's Arrow Lake, Lunar Lake and Nova Lake is where Intel will shine. Those E cores are really powerful by the time it's arrow lake and there might be 32 E cores on the flagship i9. I cannot wait for late 2024/25.

 

[Timmah!]

להגות את הדבר הגדול הבא - ראיון עם האחראי על תהליך יצירת ליבות עיבוד עוצמתיות במעבדי אינטל • HWzone

בראיון זה ישבנו אחד על אחד עם עדי יועז, הארכיטקט הראשי של ליבת הביצועים של אינטל וזה שצוותו עומד מאחורי Golden Cove העוצמתית של מעבדי Alder Lake

hwzone.co.il

interview with the lead architect of golden cove. he is saying after raptor lake and meteor lake we are going to see "bigger and bigger jumps".
it's in Hebrew so use google translate.

No “bigger jumps” if marketing departments have any word in it.
If they can do bigger jump, they will divide it into 3 smaller ones to keep milking you over 3 years.

 

[NTMBK]

Nova lake is said to completely redesign the Core arch. The biggest since the first Core arch launched. Intel may have an undisputed pref/w leadership with Nova Lake.

Don't forget the Bulldozer was a complete redesign, too.

 

[NTMBK]

No “bigger jumps” if marketing departments have any word in it.
If they can do bigger jump, they will divide it into 3 smaller ones to keep milking you over 3 years.

Marketing want a product that will actually compete...

 

[NostaSeronx]

Don't forget the Bulldozer was a complete redesign, too.

Bulldozer was a new design, not a redesign.

Of three new designs it was the mid-performance design:
Performance Core => Monolithic SMT2 core with HP co-processor on 45nm partially depleted strained silicon on insulator ( https://www.catrene.org/web/downloads/profiles_medea/2T101_profile.pdf )
Value Core => Cluster-based Multithreading with HiPerf co-processor on low-power/low-cost 45nm partially depleted silicon on insulator (1st Bulldozer)
Pervasive Core => Single-cluster architecture with LowPow co-processor but on extra-low-cost 65nm partially depleted silicon on insulator. (1st Bobcat)

Even if it is a a new design, Intel shouldn't have the issue of five different chief architects hot-potatoing the architecture.

AMD - K5-architect for Pre-Bulldozer (single-threaded clustered microarchitecture)
AMD - New team for Pre-Bulldozer (Low-power multithreaded clustered micro-architecture)
AMD - Glew gets involved in Pre-Bulldozer (Low-power -> High-performance, switches cluster-interconnect to something more difficult)
AMD - Moore gets involved (Switches multi-cluster interconnect to something simpler) - (1st Bulldozer M-SPACE scaled up from 1st Bobcat)
AMD/Microsoft - Butler+Microsoft guy (Switches from clusters to cores, Cluster-based Multithreading -> Chip-level Multithreading) - (2nd Bulldozer - Production)

I don't think a complete redesign or new design for P-core at Intel given above. Would ever be as that complex or end in a design which arbitrarily halves ST-perf.

 

[Timmah!]

Marketing want a product that will actually compete...

with whom? There is only AMD and they will do exact same thing.

 

[igor_kavinski]

Zen 3 is old now so possibly AMD can let Intel license it and then Intel engineers can add their own improvements to that plus their own process enhancements to make it something better than Zen 4. Yes, it might take 3 years minimum but it would be better than pushing Arrow Lake to 100 degrees celsius for a narrow benchmark win and losing in power efficiency. Intel's own architectural enhancement ideas have gone stale. They need outside counsel in this matter.

 

[NTMBK]

with whom? There is only AMD and they will do exact same thing.

They need to compete with internally developed CPUs from the likes of Amazon. They need to compete with ARM server vendors like Ampere. And no, AMD will not let an opportunity like that pass them by- if Intel are sandbagging, then that's a great big opportunity to claim market share by offering a superior product.

 

[mikk]

In this case, it is. I'd start managing expectations with half that.

5% seems low for a core area reduction of 25% compared to previous tick releases.

32nm Sandy Bridge --> 22nm Ivy Bridge= 38.5% core area reduction
22nm Haswell --> 14nm Broadwell= 48% core area reduction
10nm Alder Lake --> 7nm Meteor Lake= 25% core area reduction

SB to IVB brought 3-5% higher IPC afaik, Haswell to Broadwell was just a shrink without real IPC changes.

 

[lobz]

5% seems low for a core area reduction of 25% compared to previous tick releases.

32nm Sandy Bridge --> 22nm Ivy Bridge= 38.5% core area reduction
22nm Haswell --> 14nm Broadwell= 48% core area reduction
10nm Alder Lake --> 7nm Meteor Lake= 25% core area reduction

SB to IVB brought 3-5% higher IPC afaik, Haswell to Broadwell was just a shrink without real IPC changes.

I'm also pretty sure that MTL will bring at least as big of an IPC improvement as Zen4 does. While not huge, definitely not 5%.

 

[Exist50]

5% seems low for a core area reduction of 25% compared to previous tick releases.

32nm Sandy Bridge --> 22nm Ivy Bridge= 38.5% core area reduction
22nm Haswell --> 14nm Broadwell= 48% core area reduction
10nm Alder Lake --> 7nm Meteor Lake= 25% core area reduction

SB to IVB brought 3-5% higher IPC afaik, Haswell to Broadwell was just a shrink without real IPC changes.

I'm aware of the auguries attempted by so-called "analysts", but don't be fooled into thinking they know what they're talking about. Redwood Cove is ultimately a refinement of Golden Cove, and a relatively low priority for IDC. It's not going to impress, at least in IPC.

I'm also pretty sure that MTL will bring at least as big of an IPC improvement as Zen4 does. While not huge, definitely not 5%.

Oh certainly not. And frankly I'm giving 5% as a median expected value, not a low. I've been calling myself a Meteor Lake pessimist for a reason. Maybe I'm way off base, but I don't think so...

 

[igor_kavinski]

Maybe I'm way off base, but I don't think so...

It would be a PR disaster for Intel if the median improvement is 5%. I'm sure Intel knows that and they will do their best to at least target 10%. There is a slight chance that Intel may not want to be too ambitious with Intel 4 and hence keep architectural enhancements to a minimum but it may turn out to be their funeral in the mindshare department among gamers/enthusiasts.