Discussion Intel current and future Lakes & Rapids thread

Page 940 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

511

Diamond Member
Jul 12, 2024
5,482
4,896
106
Shipping 300mm² of N2 on a consumer product is certainly a choice
Depends on how much they can charge tbh if they can get $1000+ for that sku it's fine.
Seriously what the heck is Intel physdes doing?
Unless it is better than expected and/or Z6 is underwhelming, there is nothing nice to say.
Uhh it's bLLC extra cache is costly in area what did you expect. In the end for bLLC gaming performance will decide it's fate cause people don't use their brains when buying SKU look at X3D.
 
  • Like
Reactions: MoogleW

Geddagod

Golden Member
Dec 28, 2021
1,683
1,720
136

511

Diamond Member
Jul 12, 2024
5,482
4,896
106
110mm2 sounds pretty large, but if ARL-S's compute tile is ~115mm2, and whatever node shrink benefits they get are eaten up by increasing area for IPC on the P-cores, and enabling AVX-512 on the E-cores...
Still seems a bit large, but not unbelievable that the compute tile area is actually around that range IMO. At least for the base 8+16 tile.
Well now that you think ARL is 117mm2 and if the core has 30-40% resourcing increase this size is reasonable you have AVX-512/APX taking area
 

adroc_thurston

Diamond Member
Jul 2, 2023
8,614
11,327
106
Depends on how much they can charge tbh if they can get $1000+ for that sku it's fine.
You're gonna give that guy expecting cores for cheap an aneurysm.
Uhh it's bLLC extra cache is costly in area what did you expect. In the end for bLLC gaming performance will decide it's fate cause people don't use their brains when buying SKU look at X3D.
Point is AMD wins at a lot less area.
 

511

Diamond Member
Jul 12, 2024
5,482
4,896
106
You're gonna give that guy expecting cores for cheap an aneurysm.
20$ a core sounds reasonable
I didn't even realise ARL 8+16 was already that big.
Still a heterogeneous CPU is supposed to win on area, like that is the whole point.
Uhh if it was not heterogenous and Intel didn't have Atom they would be over like way more area than this considering ratio of 2.25: for P:E in ARL for Core+L1