• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

i9-12900K (QS) > Ryzen 9 5950X in ST, MT

Soldato
Joined
28 May 2007
Posts
18,190
Nope, needed a new CPU for the games I play, wasn't prepared to wait.

That said, I have more disposible income thanks to a new job, so will likely jump on a DDR5 platform to play around sooner than later :)

The only worthwhile parts coming from Intel pre 2025 will be graphics cards. Should be with anytime now so you won’t have to wait long to open your purse to Intel again.
 
Associate
Joined
27 Apr 2007
Posts
963
I don't believe the MT benchmark.
Surely it has too few BIG cores and threads in general to beat the 5950X?
If they can beat it, even by consuming maybe 250W, it's a step in the right direction at least.
 
Soldato
Joined
6 Feb 2019
Posts
17,463
Will this 12900k hybrid be a little problematic when it comes to all core overclocking, considering some are small cores and some are big?

There will just be an option in the BIOS to overclock both chiplet's separately, just like you can do on an AMD cpu
 
Caporegime
Joined
8 Jan 2004
Posts
31,993
Location
Rutland
I buy Intel for the ease of overclocking all cores, I don’t want to have to overclock the same chip twice.

I'm not sure I get your logic. You can do an all core overclock on both with the same ease. If you want there's the settings on AMD to do lots more if you really want to tinker or at the other end of the spectrum you can just turn on PBO and leave it at that.
 
Caporegime
Joined
8 Jan 2004
Posts
31,993
Location
Rutland
If the 12900k is a hybrid containing both big and small cores, then it would be more time consuming and problematic to overclock. AMD are not as good for overclocking.

Overclocking is dying anyway. Leaving significant performance untapped isn't happening in an era of complex per core boost algorithms and power limits.

Part of the reason Intel have more headroom is their older process and architecture let's them chuck silly amounts of power around for the gains available.

AMDs 7nm process is much more efficient, doesn’t respond as well to more voltage/power and their boost algorithm does a pretty decent job at getting what's possible out of their chips.

I suspect future CPUs will move towards overclocking being largely pointless (Zen 3 isn't far from this) much like on GPUs these days.
 
Soldato
Joined
6 Feb 2019
Posts
17,463
Overclocking is dying anyway. Leaving significant performance untapped isn't happening in an era of complex per core boost algorithms and power limits.

Part of the reason Intel have more headroom is their older process and architecture let's them chuck silly amounts of power around for the gains available.

AMDs 7nm process is much more efficient, doesn’t respond as well to more voltage/power and their boost algorithm does a pretty decent job at getting what's possible out of their chips.

I suspect future CPUs will move towards overclocking being largely pointless (Zen 3 isn't far from this) much like on GPUs these days.

my 5950x still gains 25% extra performance from an overclock.
 
Caporegime
Joined
17 Mar 2012
Posts
47,379
Location
ARC-L1, Stanton System
I can still get 200 to 300Mhz out of my 5800X to get it to 5Ghz or higher, its not great but its already tuned to run at high Mhz out of the box and i don't think that's any worse than Intel CPU's from the same generation, in fact they already run at 5.2Ghz out of the box and so far as i can tell they don't even run so much as 100Mhz higher than that reliably overclocked unless you get a rare sample.

Also AMD's 14nm chips were twice as efficient as Intel's, these 10nm chips (supposedly better than TSMC 7nm, according to Intel) still run at well over 200 Watts. Again twice the power consumption of AMD's chips.

Is it just the process node or does architecture also play its part?
 
Last edited:
Soldato
Joined
28 May 2007
Posts
18,190
I can still get 200 to 300Mhz out of my 5800X to get it to 5Ghz or higher, its not great but its already tuned to run at high Mhz out of the box and i don't think that's any worse than Intel CPU's from the same generation, in fact they already run at 5.2Ghz out of the box and so far as i can tell they don't even run so much as 100Mhz higher than that reliably overclocked unless you get a rare sample.

Also AMD's 14nm chips were twice as efficient as Intel's, these 10nm chips (supposedly better than TSMC 7nm, according to Intel) still run at well over 200 Watts. Again twice the power consumption of AMD's chips.

Is it just the process node or does architecture also play its part?

Seeing as Intel’s fabs are open for business and Intel want to buy GloFlo I doubt Intel’s production is better.
 
Caporegime
Joined
17 Mar 2012
Posts
47,379
Location
ARC-L1, Stanton System
Seeing as Intel’s fabs are open for business and Intel want to buy GloFlo I doubt Intel’s production is better.

Well they can't have it both ways, they can't have a "10nm Node that is better than TSMC 7nm" and have X86 CPU's on it that use twice as much power as those made on TSMC's 7nm, either their 10nm is junk or their X86 architecture is junk.

AMD's 14nm CPU's also only used half as much power as Intel's 14nm CPU's.

4.8 <> 4.9Ghz Cinebench MT load, 109 Watts total package power. 5800X, 8 cores 16 threads, Just saying...
 
Last edited:
Man of Honour
Joined
30 Oct 2003
Posts
13,227
Location
Essex
How confident are we in the numbers? I just looked at the benchmarks etc and it looks like a monumental effort if they pull out the scores they are claiming. Fag packet maths tells me its possible with the core config etc assuming 8/16t on the big and 8 on the small but I just wonder if they can turn around their MT game just like that... That's basically 70% scaling on HT cores.

Mind you if my numbers are roughly in the ball park the smaller cores are genuinely total junk.
 
Man of Honour
Joined
30 Oct 2003
Posts
13,227
Location
Essex
Sceptical here to say the least

I mean the numbers work but call for some almost perfect scaling/scheduling....

Big -
810 x 8 = 6480 (Big physical) & 6480/100 * 60 = 3888 (Big HT) = 10368

Little -
11600 - 10368 = 1232 (1232/8) = 154 (which actually puts it inline with the Atom x5-z8350) if it was running gracemont clocks (ish).

I mean this looks to be roughly how it pans out... Thoughts?
 
Last edited:
Back
Top Bottom