Store publishes benchmarks of the Core i9-12900KF, i7-12700KF and i5-12600KF

What game? Ah, nevermind, Ashes of the Singularity. Well, if you dig through the benchmarks, there are far better 5950X scores with these settings. This one seems unusually low. Still, certainly impressive.

Edit: Christ, I can't seem to read benchmark data today. I looked at a 3090 result rather than a 3080 for the 5950X. In that context this is even more impressive. I might actually upgrade in Q4 2022 after all.

 

Well this really looks like another Conroe for Intel. But let's not jump to conclusions just yet. AMD is not the same company it was 15 years ago. They might actually have an answer to it this time around. Or at least I hope they have...

btw. That would be an interesting upgrade going from the legendary 2700K to 12700K

 

Conroe was literally double performance at half the price. This is ~40% at the same price. Not quite there yet I'm afraid, but come late next year, we'll have both an indirect and a direct AMD response and an Intel refresh. I'm salivating at the thought.

 

So that's intel's "Zen" moment...

Looks really impressive but unlike 2006 i think AMD will be prepared this time as they might have other tricks before Zen 4.

 

i did try to point this out in the icelake thread,

the 11000 series was just a stepping stone to get intels new cache and pipeline out into the wild.

the increments of skylake to comit lake have been tiny, underreaching changes that meant little without the major changes (now in alders lake) because without 10nm going all in was not going to happen,

Rocket had one of the biggest of those tiny changes diverging from icelake with its l1 cache down to 3 cycles from 5 but being on 14nm restrained its full design intended implementation.

You lost some cores true, but these 10 cores were not design optimal for intel ringbus.

Word is that Alderlake is 7nm btw, not 10nm as previously thought.

Changes from xLake to Sunny (Icelake)

Key changes from Palm Cove/Skylake

• Performance
  • IPC uplift (Intel self-reported average 18-20% IPC across proxy benchmarks such as SPEC CPU2006/SPEC CPU2017)
• Front-end
  • 1.5x larger µOP cache (2.3K entries, up from 1536)
  • Smarter prefetchers
  • Improved branch predictor
  • ITLB
    • Double 2M page entries (16 entries, up from 8)
  • Larger IDQ (70 µOPs, up from 64)
  • LSD can detect up to 70 µOP loops (up from 64)
• Back-end
  • Wider allocation (5-way, up from 4-way)
  • 1.6x larger ROB (352, up from 224 entries)
  • Scheduler
    • Larger scheduler (160, up from 97 entries)
    • Larger dispatch (10-way, up from 8-way)
• Execution Engine
  • Execution ports rebalanced
  • 2x store data ports (up from 1)
  • 2x store address AGU (up from 1)
  • New paired store capabilities
  • Replaced 2 generic AGUs with two load AGUs
• Memory subsystem
  • LSU
    • 1.8x more inflight loads (128, up from 72 entries)
    • 1.3x more inflight stores (72, up from 56 entries)
  • 1.5x larger L1 data cache (48 KiB, up from 32 KiB)
  • 2x larger L2 cache (512 KiB, up from 256 KiB)
    • Larger STLBs
      • Larger 1G table (1024-entry, up from 16)
      • Larger 4k table (2048 entries, up from 1536)
      • New 1,024-entry 2M/4M table
  • 5-Level Paging
    • Large virtual address (57 bits, up from 48 bits)
    • Significantly large virtual address space (128 PiB, up from 256 TiB)

Willow Cove
Key changes from Sunny Cove

• Expanded L2 Cache (512KB 8-way → 1.25MB 20-way)
• 50% Expanded L3 Cache (8MB 16-way → 12MB 12-way)
• Memory Subsystem with more bandwidth and LPDDR5 support
• New Total Memory Encryption(TME) feature

New instructions
Willow Cove introduced a number of new instructions:

• Control-flow Enforcement Technology (CET) enhancements
• MOVDIR - Direct stores
• Additional AVX-512 extensions:
  • AVX512_VP2INTERSECT - AVX-512 Vector Intersection Instructions

Only on server parts (Sapphire Rapids):

• ENQCMD - Enqueue Stores
• Intel Advanced Matrix Extensions (Intel AMX)

Golden Cove
Key changes from Willow Cove

• Performance improvements
  • Strong IPC improvement (19%)
  • AI workload improvement (AMX)
  • Network/5G performance improvements
• New security features
• Front-End
  • Add 2 decoders from 4
  • x2.5 BTB at 12K entries
  • 2x pages 4k
  • Add more 256 and 32 pages of 2M and 4MB respectively
• Back-End
  • Increased ROB 512 (from 352 Sunny Cove)
  • Add 2 execution port for a total 12
• Execution Engine
  • Add one ALU and LEA for a total 5

 

I'm probably gonna stick with my 5800X until the next console generation (probably 2025 or 2026 I guess) so whatever Intel or AMD have out by then should be a real monster. Maybe at least 3-4x the single thread performance I get now.

 

If all this ends up to be real, then the bigger pity is that the CPU is not made with a proper process and it needs the "low power" cores to basically hide the power design deficits. This thing should have been a real 16-core.

 

I was thinking 12 P-cores + 4 E-cores would be a better combo, 4Ecores/4threads is plenty for office workload and 12 Pcores/24threads would make it better gaming CPU than 10900K

 

seems like 330eur will be the standard launch retail price for the midrangers since 5600X.
5600X was very highly demanded despite launching the price of prev gen Ryzen 7
4 extra small cores are of very little consolation to me,though I am a huge fan of energy efficient designs for desktop work.
if this beats 5800X in gaming just like 5600X beat 9700K people will line up,despite the fact it's a six core

anyone with an octa-core from r7 3000/intel i7 running at least 3600MHz ram series should not care about upgrading until late ddr5,or at least not bother with it because of lackluster performance.
people who stayed on ryzen 1st/2nd gen or an older gen six core like 8700K w.3000mhz ram will need to upgrade in early ddr5 days,which will mean lower performance at inflated costs of memory and boards.

 

They seem to have a huge power deficit vs AMD, which I would say leaves AMD a lot of playroom.

 

AMD have options up there sleeves they can pull out, however even as an avid anti intel user I hope they can force AMD into another step or we will be in danger of stagnation once again