News Intel Core i5-12600K Test Suggests 50% More Threaded Performance Than 11600K

Status
Not open for further replies.
Oh, this looks amazing if true.

Come on AMD, drop prices so I can upgrade my 3800XT to a 5950X for about $400 :D

Regards.
Due to the very high platform price + ddr5 it's very doubtful that AMD will have to drop prices very soon.
If and when they will have to drop prices it's not going to be by a lot, they are not going to lose money just to keep selling them, they can't afford to even if they wanted to do that.
 
  • Like
Reactions: artk2219
Due to the very high platform price + ddr5 it's very doubtful that AMD will have to drop prices very soon.
If and when they will have to drop prices it's not going to be by a lot, they are not going to lose money just to keep selling them, they can't afford to even if they wanted to do that.
Not to mention that so far we have only seen synthetic benchmarks LEAKS. Before Rocket Lake came out there were tons of leaks saying it was going to be crazy fast. Then it is released and was nothing but a letdown. All those leads in synthetics equated to nothing in real world performance.
 
  • Like
Reactions: artk2219
Not to mention that so far we have only seen synthetic benchmarks LEAKS. Before Rocket Lake came out there were tons of leaks saying it was going to be crazy fast. Then it is released and was nothing but a letdown. All those leads in synthetics equated to nothing in real world performance.
Do you mean for laptop/mobile maybe?! Because all the leaks for desktop where like so-so they showed improvement but nothing crazy.
 
  • Like
Reactions: artk2219

InvalidError

Titan
Moderator
"Three separate validations have been submitted to CPU-Z's database, painting a relatively consistent performance expectation for Intel's historically mid-tier CPU. "

The i5-12xxx is still a mid-tier CPU and the only reason a 50% improvement looks good today is because we've had nearly 10 years of relative stagnation eroding people's expectations. 20 years ago, 50% year-on-year performance improvements were normal or even disappointing, not anything to be impressed about.
 

Neilbob

Distinguished
Mar 31, 2014
195
226
19,620
"Three separate validations have been submitted to CPU-Z's database, painting a relatively consistent performance expectation for Intel's historically mid-tier CPU. "

The i5-12xxx is still a mid-tier CPU and the only reason a 50% improvement looks good today is because we've had nearly 10 years of relative stagnation eroding people's expectations. 20 years ago, 50% year-on-year performance improvements were normal or even disappointing, not anything to be impressed about.

Though to be fair, those improvements were attained in large part thanks to clock speeds that were still very rapidly increasing on a yearly basis. That seems rather less likely these days.
 
Though to be fair, those improvements were attained in large part thanks to clock speeds that were still very rapidly increasing on a yearly basis. That seems rather less likely these days.
Ding ding ding ding.

When you're not only getting 200Mhz gen over gen for your clocks, you gotta get creative in other ways.

EDIT: One note worth mentioning I completely forgot. It should be said in % of previous gen speeds instead of actual Hz, as it's a bit clearer what the point is about.

Regards.
 
Sep 25, 2021
2
0
10
Its not exactly impressive when you realize the 12600k is now a 10-Core CPU vs the 11600k’s 6 cores. It better show a lot more threaded performance, it has 4 more cores. More cores=higher multithreaded performance. It’s not magic, it’s just math.

Your title should be “Captain Obvious Reporting in: 10-Core CPU 50% more threaded performance than 6-Core CPU. News at 11”.
 
Last edited:

InvalidError

Titan
Moderator
Though to be fair, those improvements were attained in large part thanks to clock speeds that were still very rapidly increasing on a yearly basis. That seems rather less likely these days.
Clock frequencies were leveling off quite a bit: Intel launched the 2GHz Willamette P4 back in 2001, Northwood maxed out at 3.4GHz in 2004 and we barely got another 1GHz more on stock clocks from there to modern day. Clocks haven't been a major performance scaling factor at about +30% in 17 years.

Architecture, extra cores, new instructions, IO integration, etc. on the other hand have increased performance by 10-20X since then.
 
  • Like
Reactions: Memnarchon
That sound incredible.

A few questions from the pictures in the arcticle:

CPUZ is able to use those 4x Efficiency core when runing the MT benchmark, right? Cause having 4 extra cores (threads) would of course prove to be handy vs the old core i5 11400.

Also, Did I read right?, This test were using Z690 mobos with quad channel memory support?, if thats right how much of an improvement would runing quad vs dual channel memory be on this test (vs the old i5 11400 in dual channel)?

Thanks!
 
Last edited:

Giroro

Splendid
Due to the very high platform price + ddr5 it's very doubtful that AMD will have to drop prices very soon.
If and when they will have to drop prices it's not going to be by a lot, they are not going to lose money just to keep selling them, they can't afford to even if they wanted to do that.
I don't think AMD would lose money if they sold a 5950X at $400. They would just make way less profit.
The cost to manufacture a CPU is a lot closer to $50 than it is to $500.

But, of course, they're always going to make the most money that they possibly can
 

Giroro

Splendid
That sound incredible.

A few questions from the pictures in the arcticle:

CPUZ is able to use those 4x Efficiency core when runing the MT benchmark, right? Cause having 4 extra cores (threads) would of course prove to be handy vs the old core i5 11400.

Also, Did I read right?, This test were using Z690 mobos with quad channel memory support?, if thats right how much of an improvement would runing quad vs dual channel memory be on this test (vs the old i5 11400 in dual channel)?

Thanks!
Intel has implied that efficiency cores can run at the same time as performance, but I don't think that they've ever confirmed it. They've been suspiciously quiet on Alder lake's real world performance.

Also, as far as I know, Intel is sticking to dual channel for Alder Lake-S desktop parts
 

TJ Hooker

Titan
Ambassador
A normal memory channel is 64 bits, but DDR5 splits this up further into two 32-bit 'sub' channels. So what we would normally think of as dual channel ends up looking like quad channel.

I'm guessing CPU-Z reports the memory bus width as 64 x # of channels (which is no longer correct with DDR5), which is why it lists 256 bit when it should be 128.
 
A normal memory channel is 64 bits, but DDR5 splits this up further into two 32-bit 'sub' channels. So what we would normally think of as dual channel ends up looking like quad channel.

I'm guessing CPU-Z reports the memory bus width as 64 x # of channels (which is no longer correct with DDR5), which is why it lists 256 bit when it should be 128.

That would make sense. Didn't knew that sub-channel splitting of DDR5, thank you, always something new to learn. Guess I will have to deep dive into DDR5 to get to know what else could be different.
 

anonymousdude

Distinguished
Intel has implied that efficiency cores can run at the same time as performance, but I don't think that they've ever confirmed it. They've been suspiciously quiet on Alder lake's real world performance.

Also, as far as I know, Intel is sticking to dual channel for Alder Lake-S desktop parts

I'm guessing they've been rather quiet on real world performance cause most current applications even with their new Thread Director isn't anything to write home about or flat out don't work properly. Basically your natural growing pains that will take a little bit to sort out.
 

blppt

Distinguished
Jun 6, 2008
568
89
19,060
"I don't think AMD would lose money if they sold a 5950X at $400. They would just make way less profit.
The cost to manufacture a CPU is a lot closer to $50 than it is to $500. "

Yes, but built into the price is the immense cost of R+D for the entire Zen architecture that needs to be made up somehow.
 

InvalidError

Titan
Moderator
"I don't think AMD would lose money if they sold a 5950X at $400. They would just make way less profit.
The cost to manufacture a CPU is a lot closer to $50 than it is to $500. "
For a 3-chips package, it was likely somewhere in the $100-150 range pre-covid, possibly $150-200 now with all the raw materials, processed materials, substrates, testing facilities, etc. shortages.
 

InvalidError

Titan
Moderator
Who thinks the 4 extra E cores will contribute about 2/3rds of this 50% increase in score?
The E-cores are narrower, have a shorter re-order queue, smaller caches, no SMT and don't clock as high as the P-cores. Core for core, I'd expect E-cores to be maybe 2/3 as fast as P-cores in multi-threaded workloads and with E-cores counting for 4/10 of cores, they might contribute 25-30% to multi-threaded score, or about half of the 50%.

HWUB benchmarked 10th-gen series to compare how performance scales between the i5, i7 and i9 with only six cores enabled when L3$ size is the only net difference and found that the i5's 12MB L3$ leaves quite a bit of performance on the table. 15-20% of Alder Lake's performance gains at the i5 tier likely come straight from having 20MB of L3$ as a baseline,
 
Status
Not open for further replies.