Question Discussion over P-core and E-core vs AMDs regular vs C-core

[Thunder 57]

Skymont and Raptor Cove are about equal in floating point and massively better than Gracemont.

Based just on memory I am not so sure about that part. I would have to look in to it though. The others look right though.

 

[FlameTail]

A classification of cores based on size and performance.

Tried to use the latest core IP from each vendor.

Cortex X925 and A720 as deployed in Dimensity 9400.

Lion Cove/Skymont and Zen5/Zen5C as deployed in Lunar Lake and Strix Point.

Only L1 cache included into core area. Private L2 is also excluded.

Y-axis = size
X-axis = performance

So the cores to the bottom-left are the most area efficient.


Just for fun

Edit: Added core sizes, and fixed some errors.

 

[MoogleW]

The processor overall is as you outlined, but a single core of Zen5 is much higher performance in MT than a single Skymont or Lion Cove core. This is very important in DC.

I think so, but I don't believe IPC is as comprehensive as performance per core. In MT for DC I believe this is the most important metric.

In laptop and desktop where die cost is very important, performance per area seems like a better metric.

I am not sure how important IPC is compared to these metrics.

are you accounting for hyperthreading?

 

[511]

So you believe Arrow Lake would be okay without lion cove?

Yes absolutely on Desktop 24/32 core Skymont would be better cause if they are regressing ST might as well annihilate in MT no quirks like we have seen only if Intel Bios allows booting without P core for more accurate comparison

 

[511]

A classification of cores based on size and performance.

Tried to use the latest core IP from each vendor.

Cortex X925 and A720 as deployed in Dimensity 9400.

Lion Cove/Skymont and Zen5/Zen5C as deployed in Lunar Lake and Strix Point.

Only L1 cache included into core area. Private L2 is also excluded.

Y-axis = size
X-axis = performance

So the cores to the bottom-left are the most area efficient.

View attachment 111575
Just for fun

Edit: Added core sizes, and fixed some errors.

All is good but there is the biggest unknown factor Library used we need someone to share SEM images for the core to know what library they use

 

[OneEng2]

Yes absolutely on Desktop 24/32 core Skymont would be better cause if they are regressing ST might as well annihilate in MT no quirks like we have seen only if Intel Bios allows booting without P core for more accurate comparison

Is it possible to make it run with only one core type or the other? I haven't seen anything yet that shows it.

 

[511]

Is it possible to make it run with only one core type or the other? I haven't seen anything yet that shows it.

Sadly No 1P is required that is an idiocity considering you can use without E and Intel sent engineer sample with 16E cores only

 

[FlameTail]

There's a surprise amount of similarity between Oryon-M and Skymont LPE.

Both clock very high (3.5 GHz and 3.7 GHz).

Both score about 5 points in SPEC2017 INT, and the power consumption also appears to be close.

Skymont is 1.1 mm² and Oryon-M is 0.85 mm². But the latter has double the L2 per core (4 LPE + 4 MB sL2 vs 6M + 12 MB sL2). If we account for that, Skymont and Oryon-M are pretty closely sized.

 

[511]

View attachment 111786
There's a surprise amount of similarity between Oryon-M and Skymont LPE.

Both clock very high (3.5 GHz and 3.7 GHz).

Both score about 5 points in SPEC2017 INT, and the power consumption also appears to be close.
View attachment 111787View attachment 111788
Skymont is 1.1 mm² and Oryon-M is 0.85 mm². But the latter has double the L2 per core (4 LPE + 4 MB sL2 vs 6M + 12 MB sL2). If we account for that, Skymont and Oryon-M are pretty closely sized.

This is a gimped skymont as well doesn't have RingBus Access it will be even higher but yes Skymont is impressive even in a weak state and apparently over on reddit there is a claim of 20% st of next generation mobile soc reaching 4000ST but fun fact 10% gains are from SME

Just for reference how SME is boosting scores in GB

https://twitter.com/x/status/1857310612969398332

https://twitter.com/x/status/1855838039563759633

 

[OneEng2]

View attachment 111786
There's a surprise amount of similarity between Oryon-M and Skymont LPE.

Both clock very high (3.5 GHz and 3.7 GHz).

Both score about 5 points in SPEC2017 INT, and the power consumption also appears to be close.
View attachment 111787View attachment 111788
Skymont is 1.1 mm² and Oryon-M is 0.85 mm². But the latter has double the L2 per core (4 LPE + 4 MB sL2 vs 6M + 12 MB sL2). If we account for that, Skymont and Oryon-M are pretty closely sized.

Would be interesting to know Zen 5c size on N3E (Turin Dense). Still likely much larger than Skymont by a good margin ..... but on MT workloads, I still wonder if a combination of SMT, AVX512, and other enhancements that Zen 5c has, how well Skymont fares against Zen 5c without these advancements.

Nice chart!

 

[DZero]

View attachment 111786
There's a surprise amount of similarity between Oryon-M and Skymont LPE.

Both clock very high (3.5 GHz and 3.7 GHz).

Both score about 5 points in SPEC2017 INT, and the power consumption also appears to be close.
View attachment 111787View attachment 111788
Skymont is 1.1 mm² and Oryon-M is 0.85 mm². But the latter has double the L2 per core (4 LPE + 4 MB sL2 vs 6M + 12 MB sL2). If we account for that, Skymont and Oryon-M are pretty closely sized.

Who cares on A5XX? Is too weak that even the olde Intel Core era processor can defeat it. The disadvantages of being an In Order core are real. Now bussiness will start to switch to all Out of Order cores configs.

 

[DavidC1]

Would be interesting to know Zen 5c size on N3E (Turin Dense).

It's ~1.9mm2.