I really don't care that much about power consumption but when your CPU is using as much power as your GPU that's insane and even then that would be more palatable if it actually gave you a tangible improvement.
Even if you have no concerns over energy use consider power consumption usually ties into heat so it means its harder to keep cool.
On my 13700k, on spice wars I switch to my soft p-states power scheme and it saves 50w on the CPU, I save another 50w or so by switching GPU to throttled clocks. With zero loss of game performance.
Usually my 13700k isnt crazy on power in games, but spice wars definitely was showing it up on my power scheme thats similar to the stock balanced hardware p-states profile.
I did expose Joules in the past, people didn't understand it, which is why I'm now listing "Cinebench Points per Watt" for efficiency
Core i9-14900K is Intel's new flagship with clock speeds of up to 6 GHz. It's actually clocked even higher than the 13900KS, thanks to an extra 100 MHz when more than two cores are active. Our review confirms that Raptor Lake Refresh is amazing for both applications and gaming, if you can live...
www.techpowerup.com
That graph is interesting, it seems to indicate when you factor in the productivity gains (without your manual OC) its not terribly far behind AMD on points per watt.
So the graphs that look pretty bad seem to be lighter load stuff which is perhaps highlighting the over aggressive behaviour of how rapid voltages and clocks rise on these chips.
Playing youtube e.g. unless I tame the chip via custom power scheme, my 13700k clocks to 5.4ghz and is at 1.32v whilst the video is playing, consuming about twice as much as my 9900k did to do the same thing and almost 3x as much as my 5600G (with XFR disabled).
Combination of undervolting and putting a leash on the settings that control the clock speed ramp up's can have a huge impact, but it is actually quite a challenge to make a profile that eg. makes youtube power efficient whilst also not giving noticeable interactive experience regressions. Things that consume constant CPU steadily its better to tame the chip, but things that require a very short burst of performance like displaying a UAC prompt, will appear slower when chip ramp up is tamed.
Since these chips are hybrid though one can e.g. keep the e-core clocks down, and use them for things like twitch and youtube, to keep power usage down, whilst still allowing p-cores to flex for interactive stuff. I expect Intel with all their software stuff they working on to do things like this in coming years.
I was expecting more saltiness from
@Wizzard for making him do an entire review cycle for what is basically just the 13900KS he's already reviewed
Its content and revenue for him.
He touched upon this on the unboxing thread.
This option has existed for several years now. At least since 12th gen. Maybe not all motherboards expose it.
It makes my life much easier because I don't have to worry so much about thermal throttling around 100°C, which would invalidate test results just because my cooling is not good enough
I think I would prefer if you did tests at the specified temperature limit, I wouldnt concern yourself if Intel produce a poor product that requires insane levels of cooling. Also not sure why you think thermal throttling would invalidate the test, thermal throttling is what end users will have to deal with on these chips in certain situations so will be a fair test.
As I do consider 115C performance figures as potentially misleading. Especially if its already on a workbench and not in a case which already is helping the review.
I dont think your job is to get as good a score as possible.
Maybe some clarification on how much this affected the results? Looking at the temp graph it seems in blender your chip at stock was below 100C anyway, so maybe only affecting the OC and power limit removed results?
Thats my small opinion on this.
On my 690 board it is exposed, and was actually defaulted to 110C by ASRock on their initial bios. (later bios corrected it to 100C).
