Discussion Intel current and future Lakes & Rapids thread

Page 708 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

nicalandia

Platinum Member
Jan 10, 2019
2,956
4,558
136
That Genoa part has an issue while running the multithreaded AES-XTS benchmark, the score is lower than the single core score.
As to be expected from unreleased hardware and BIOS.

The same goes for the Sapphire Rapids-SP parts, those are only preliminary benchmarks, that paint a trend in performance that we are aware of.

We know that Zen4 at ISO Core and ISO Speed are either a match for Golden Cove or slightly ahead(Which will be problematic for Intel because AMD has numerical core advantage and process node advantage that will make EPYC 1S and 2S systems more efficient). The unknowns are what impact on performance will the 4 tile design have on Sapphire Rapids-SP, does the Mesh of half Rings being extended that far(across four tiles) and only 1.875 MiB per core has any impact on performance?

On general purpose apps we can almost be certain that AMD will be ahead.. On specific task that take advantage of HBM2e on package memory and AMX I believe Intel will produce impressive performance not seen before.
 
Last edited:
  • Like
Reactions: Geddagod

nicalandia

Platinum Member
Jan 10, 2019
2,956
4,558
136
Here is the same Intel Xeon Platinum 8480+ scoring 37,794 Points due to being a single CPU 56C/112T instead of a 2S system.

1665756895076.png

And a single 32C/64T
1665756984145.png
 

nicalandia

Platinum Member
Jan 10, 2019
2,956
4,558
136
MLID says Granite Rapid gets 5-10% more IPC on Redwood+ architecture and Intel 3 node. No Lion Cove?
Thats correct, Redwood Cove+ on a better node. Its going to need more than 10% on IPC to still be relevant against Turin Zen5. Sapphire Rapids and Emerald Rapids will have to endure such a difficult task against Zen4, Zen4X3D
 

naad

Member
May 31, 2022
63
175
66
Genoa single core gets similar score to a 4.4/4.5GHz Cezanne APU, quite impressive.
We've been fed 2.7-3.0GHz single core mainstream (non-F) server chips for years, finally improvements are in order.
 

eek2121

Platinum Member
Aug 2, 2005
2,295
3,002
136
Thats bad... Intel is behind

this an 96 core EYPC zen 4.

View attachment 69106
Intel plans on making up for it with more sockets. Compute density with Intel will actually be higher than with AMD. A 448 core Intel system will likely be ahead by a somewhat decent amount. Don’t ask about power consumption, however.
 

Exist50

Golden Member
Aug 18, 2016
1,715
1,818
136
MLID says Granite Rapid gets 5-10% more IPC on Redwood+ architecture and Intel 3 node. No Lion Cove?
Ignoring for a sec his IPC predictions (which I'll note, he's already more than halved vs his prior claims), if GNR is RWC-based ("+" or otherwise), then it will not realistically compete head to head with Turin. Should close much of the SPR/EMR-Genoa gap from the increased core count and node shrink, but they're not going to be competitive with that core and similar process nodes.
 
Last edited:

Exist50

Golden Member
Aug 18, 2016
1,715
1,818
136
Intel plans on making up for it with more sockets. Compute density with Intel will actually be higher than with AMD. A 448 core Intel system will likely be ahead by a somewhat decent amount. Don’t ask about power consumption, however.
If they need two blades for 8S, then they're not likely to have better or comparable compute density. And >2S is a small and dying market anyway. Long term, I imagine 1/2S systems will be the norm, interconnected with CXL for coherency.
 

Geddagod

Member
Dec 28, 2021
175
152
76
Intel's Own CEO Pat said that it would be an enhanced core on a smaller node.
Ik but the IPC figures? Tbh I don't expect a 5-10% IPC uplift from redwood cove+, maybe like >5 or ~5? because I doubt they could change the architecture too much without having to move the timeline again.
Plus I think @Exist50 thinks that you meant over 5-10% ipc over raptor cove, Tom didn't give any figures, he just said a bigger uplift from alder lake to raptor cove. So Tom expects slightly more than the 10-20% IPC uplift he cited before for redwood cove.
 

Exist50

Golden Member
Aug 18, 2016
1,715
1,818
136
Ik but the IPC figures? Tbh I don't expect a 5-10% IPC uplift from redwood cove+, maybe like >5 or ~5? because I doubt they could change the architecture too much without having to move the timeline again.
Plus I think @Exist50 thinks that you meant over 5-10% ipc over raptor cove, Tom didn't give any figures, he just said a bigger uplift from alder lake to raptor cove. So Tom expects slightly more than the 10-20% IPC uplift he cited before for redwood cove.
Yeah, I didn't actually watch the video. But if he's still claiming 10-20% or whatever, much less anything higher, well then get your popcorn ready, because this'll be a trainwreck you don't want to miss. Well, more for him than anyone else, but we all know how that goes.
 

Geddagod

Member
Dec 28, 2021
175
152
76
Pat said 10-12% not 5-10%. Another time he said "10% plus". Again MLID has some sources but they are tainted by rampant speculations of his own.
I recall the 10% plus but not the 10-12%. Not that I doubt it won't be in that range.
I believe this was the exact quote-
"We did change the timing of Granite Rapids, and we had a big internal debate show on should we even keep the Granite Rapids name because it was the same platform, but it was a new core on a new process. So to some degree, it was a very different product. But some said, hey, you delayed Granite Rapids. Hey, I say I enhance Granite Rapids, with a much higher performance product, a much — 18% process, a major new core, that’s 10-plus percent in the core. So a much better product and aligned to the customers’ timing. And they said, hey, Sapphire Emerald Granite was too compressed. "
 

Exist50

Golden Member
Aug 18, 2016
1,715
1,818
136
I recall the 10% plus but not the 10-12%. Not that I doubt it won't be in that range.
I believe this was the exact quote-
"We did change the timing of Granite Rapids, and we had a big internal debate show on should we even keep the Granite Rapids name because it was the same platform, but it was a new core on a new process. So to some degree, it was a very different product. But some said, hey, you delayed Granite Rapids. Hey, I say I enhance Granite Rapids, with a much higher performance product, a much — 18% process, a major new core, that’s 10-plus percent in the core. So a much better product and aligned to the customers’ timing. And they said, hey, Sapphire Emerald Granite was too compressed. "
"A major new core" almost makes me wonder if it was intended to be upgraded to Lion Cove, but they were forced to backtrack sometime after Gelsinger's statement. Surely something they don't even deign to give a new name would not be worth calling that? Or MLID is simply wrong. Certainly won't discount that possibility. But if it is just a Redwood Cove derivative, that would be extremely disappointing.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,583
3,637
136
"A major new core" almost makes me wonder if it was intended to be upgraded to Lion Cove, but they were forced to backtrack sometime after Gelsinger's statement. Surely something they don't even deign to give a new name would not be worth calling that? Or MLID is simply wrong. Certainly won't discount that possibility. But if it is just a Redwood Cove derivative, that would be extremely disappointing.
We can't focus on the naming too much. Generally a new name suggests a big change, but it's not an absolute thing. They can do whatever they want, it's their project after all. Goldmont Plus, anyone?

Technically, it is a plus and an expansion of the predecessor, but the performance increase was pretty fantastic.

Also they sandbagged Raptorlake. "Up to double digit performance boost". So we now know they were talking about single thread and planned to "beat" that.

Not saying they are sandbagging GNR. It's that we don't know crap.

As for MLID:

2024: "Oh I was never sure about that. The source doesn't have a 100% track record."
 
Last edited:
  • Like
Reactions: coercitiv

Exist50

Golden Member
Aug 18, 2016
1,715
1,818
136
We can't focus on the naming too much. Generally a new name suggests a big change, but it's not an absolute thing. They can do whatever they want, it's their project after all. Goldmont Plus, anyone?
Yeah, this is just reading the tea leaves, but if they were willing to give "Raptor Cove" its own name for such meager gains, then I wouldn't expect anything more from "RWC+". I certainly doubt such an upgrade would count as a "major new core". Barely even "new"! But I've been quite open about my pessimism regard Redwood Cove.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,583
3,637
136
Yeah, this is just reading the tea leaves, but if they were willing to give "Raptor Cove" its own name for such meager gains, then I wouldn't expect anything more from "RWC+". I certainly doubt such an upgrade would count as a "major new core". Barely even "new"! But I've been quite open about my pessimism regard Redwood Cove.
Well 10-12% is at least of significance. The P core in Raptorlake is just the L2 cache and the prefetcher, which the E cores got too!
 

Exist50

Golden Member
Aug 18, 2016
1,715
1,818
136
Well 10-12% is at least of significance. The P core in Raptorlake is just the L2 cache and the prefetcher, which the E cores got too!
I'm not convinced we'll see 10-12% IPC over Raptor Cove, much less Redwood Cove. But I guess we'll have to wait and see.
 

DrMrLordX

Lifer
Apr 27, 2000
20,508
9,590
136
Very good Info on Sapphire Rapids and Emerald Rapids..
That may be good news for Intel that Sapphire Rapids' delays have essentially compressed the launch window for the product without pushing Emerald Rapids back. But yes it does make you wonder if that means that Sapphire Rapids will have even more delays.
 

lightisgood

Member
May 27, 2022
35
12
41
So either Emerald Rapids is Ramping Up Sooner than expected or...Sapphire Rapids is Ramping Up Later as in late 2023?
Rumor has it that EMRs might be scheduled in early 2023.
Now, I'm putting a simple interpretation on SPRs/EMRs.

SPRs : low-end ( 600~700sqmm die area?).
EMRs : mid-range ( ~800sqmm die area).
SPRs with EMIB: high-end ( ~1600sqmm die area ).
EMRs with EMIB: extream ( >=1600sqmm die area).
 

nicalandia

Platinum Member
Jan 10, 2019
2,956
4,558
136
In Any Event, Genoa's NDA ends in just a few days STH will be releasing Benchmarks on that day and poor Ice Lake will have to pay the price of Intel Delays and will be the Victim of Genoa's Might.
 

ASK THE COMMUNITY