Intel Core i5-14600K and Core i7-14700K Show up in the Wild

[Space Lynx]

OC is dead ... days with Q6600 are gone ... btw back in the days i was overclocking to run stable 60 fps .. now if its 200 or 250 it doesnt matter

Ya, I'm just bored today is all

 

[ir_cow]

DDR4-4600 at 1.624v. I've pushed 1.5v on DDR4 in the past, but I've never seen a kit built to take more than 1.6v. That's DDR3 levels of voltage.

DDR4-5066 and 5333 kits exist. They use 1.6V. But in Gear 2, pretty much anything in Gear 1 will beat it.

Does current gen Intel stuff still base DDR_VREF at the system agent voltage for the IMC? Could this be pushing 1.4v+ at the IMC to achieve these speeds?

Good question. It's split now after 10th Gen. For DDR5 you have, SA, VDDQ_TX and VDD2 (IMC) rails, but DDR4 has VDDQ_TX removed (thinking its only DDR5 relevant).

 

[kondamin]

might as well just launch the things at this point.

 

[Dyatlov A]

OC is dead ... days with Q6600 are gone ... btw back in the days i was overclocking to run stable 60 fps .. now if its 200 or 250 it doesnt matter

Nope, that is why need to buy non k, 12400 and overclocking is awesome.

 

[MaMoo]

That's a weird behavior. You mean that your P cores are just sitting there doing nothing while they wait for the e-core to finish their task ?

Yes exactly.

 

[dyonoctis]

Yes exactly.

That's really odd, It's like the app is dividing the task on exactly 24 threads for a 13700k...
meanwhile content creation apps just give the core a new task when it's done

 

[MaMoo]

That's really odd, It's like the app is dividing the task on exactly 24 threads for a 13700k...
meanwhile content creation apps just give the core a new task when it's done

Yeah. I don't like the behaviour because it is not a most efficient use of dissimilar cores. But I get it - the multiprocessing model is based on identical core performance. The type of workload needs to be executed on equal-sized work chunks, because aggregate statistics are derived from it after all chunks are complete. May be there would be a better way eventually.

 

[MaMoo]

Try 6w per E core

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

for something with 8 e cores, that's 48w. Not exactly a small number. More importantly, turning them off turns that silicon into an extended die for the P cores to sink heat into, and into the IHS for cooling, helping to somewhat alleviate the issue of tiny transistors shedding heat.

As if the P cores cannot handle that. AMD manages to compete with intel without using such E cores.

I finally managed to test out your idea of turning off ecores to make more thermal or power headroom for the pcores. You are right. After I disabled all my ecores, my Linpack performance went UP exactly 11%. The temps are also cooler on a Noctua DH15. This is much better than with the ecores on. At least for my workload the ecores were not useful anyway. Also the performance variance is much smaller between runs. It used to vary about 5 to 10 % from run to run but now is within less than 1 %.