• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Intel Core i9-14900KS

I don't understand why you would test the CPU at 115ºC, this is 15ºC over the max tjunction.

What is the point, breaking the CPU?

115C is the actual junction max spec for Intel's K SKUs. Just about any high end board since 10th gen has the 115C option in BIOS and some vendors opt to enable it automatically if you install a KS chip. 135C critical temp usually as well, but it appears Intel may have nudged that up a bit for the KS if it was sustaining 118C without shutting down.

Min FPS is still better with 14900KS interestingly, I wonder why that is.

Only at high res, and literally two decimal places outside the 2% margin with all the power limits removed. I wouldn't necessarily be praising that considering the doubling in power over the already margin-for-error results of the 14900K to achieve that.
 
Bah, another 5GHz product overclocked by Intel and pushed to the very edge of stability, power consumption, silicon health, and cooler capabilities at vast cost heat/energy/noise/cooling/power to the end user.

Your average 13900 (non-K) is boosting to ~5.5GHz at one third the power draw and IIRC the 13900T is locked at a strict 106W max package power and still boosts to 5.3GHz, not any of this 353W+ nonsense. I really struggle to find any scenario outside of e-peen willy-waving and overclocking competitions where the K or KS ever makes much sense.

This is a product for gamers, mostly, and the 7800X3D is simply better at that.
What's left, single-threaded productivity workloads? I'm struggling to think of any of those where the extra 10% over a 13900 (or 15% over an i5-13400, for that matter) is really worth all the hassle.
 
That was referencing when tuned.

https://www.techpowerup.com/review/...ke-tested-at-power-limits-down-to-35-w/8.html TPU testing shows tuned 14900K indeed being the most efficient. Note this is a simple tune, power limit only. More involved tuning together with a per core overclock and specified voltages will offer good frequencies and in many cases better than stock performance, while also improving efficiency. View attachment 338998

Every one of those processors is also very efficient when heavily power limited so any claim of the 14900K being most efficient is likely inaccurate. Too bad those tests weren't included in the graph to actually substantiate that claim.
 
I can imagine somebody running a particular application at work, for research etc. and simply choosing this CPU because it performs the best.
Well that's what I'm asking - outside of gaming, are there any time-sensitive, single-threaded applications left? The companies I support do a variety of coding, compiling, encoding, rendering, simulation, and all those power users get 16C/32T Ryzen 9s. We've built a few Threadrippers for the renderfarm, but they're genuinely not a good workstation solution for us since the very few instances where more cores are needed are the exact same solutions that favour pushing the job out to in-house dedicated render farms or compute clusters.

We've been in the multicore era for over 20 years, and quad-core has been mainstream at affordable price points for about 15 years. If any applications written in the last 15 years needed performance, they've tapped extra cores.
 
Hi @W1zzard, if you eventually find the time, would you be interested in making a mini-review (something just quick with 1 or 2 games and 1 or 2 MT applications) at iso clocks and power limit between the i9-13900KS and i9-14900KS? This would be probably the absolute best way to see the binning progress Intel has made.

I'm quite surprised at how much higher the power draw is against the original 12900K... that's quite concerning, although I suppose the performance uplift is there after all. As a 13900KS owner this isn't a product meant for me, but glad to see Intel could improve on what's already a very refined and matured processor and drop the price a little at the same time.
 
For example, a lot of finacial stuff including simulations is done in Excel. You can read the review to see a lot of applications in which this CPU performs the best.
 
Why did you stop using joules per task for effieciency ? honestly point per watt are pretty confusing.
 
For example, a lot of finacial stuff including simulations is done in Excel. You can read the review to see a lot of applications in which this CPU performs the best.
Yes, that's why the accounts department gets Ryzen 9. Better Excel performance than a 13900KS at 1/3rd the power draw - so an air-cooled, quiet, compact workstation rather than something huge with liquid cooling.

1710434781683.png


In their experience, the time it takes massive xlsx files with vast pivot tables and scripts to run favours AMD the more demanding the scenario. By the time you're waiting minutes rather than seconds, it's AMD all the way.
 
Hot damn that's hungry. Also I shall always complain that the game tests don't show 1% lows, since chips like the x3d ones always stand out massively there.
 
Average of 2.5x the power of the 7950X3D for 8% more average performance over the application suite. FOR MORE MONEY!!!

Is this a joke? Are people really still defending Intel? Why?
I don't think anyone's defending Intel - it's a stunt processor that exists at a ridiculous premium just so that you can say "you have the fastest" in a specific situation for a specific period. It becomes obsolete shortly just like there's absolutely zero point in the 13900KS now.

It doesn't mean the 13900KS is bad in any way, but it's no longer the halo product of the moment, and people who "must own that halo product" are the target audience. If you want a fast, power-efficient part, just pick up a 14700K and set your motherboard to sensible PL1/PL2 limits. It'll still be plenty fast enough and you don't have to throw efficiency completely out of the window if you don't need that last 2% of bleeding edge performance.

AMD are guilty of the exact same sins, to a lesser extent. I'm doing plenty of 7950X builds this year and they default to a 230W PPT which is silly because you can get 92-95% of that performance at 80 Watts less, and still around 85% of the performance at 100W or so.
 
Hot damn that's hungry. Also I shall always complain that the game tests don't show 1% lows, since chips like the x3d ones always stand out massively there.
 
Low quality post by gridracedriver
I remember someone in techpowerup saying that the 14900KS would be the most efficient processor in the market.*
Oh, it will be.
Once you restrict and disable everything that makes it a KS.
That's tuned to 35W against other CPUs at their default performance-focused 230W, 250W, or 353W stock PPTs settings?

7950X at ~50W would likely be competitive, if not better since that's TSMC's manufacturing which is at least one full generation of nodes and node refinements ahead, though at very low (sub-65W TDPs) the IO die and chiplet design hurt AMD's efficiency more than Intel's monolithic offerings, as they raise the baseline. A 7950X has a 'parasitic' 15W baseline from the chiplet design adding fabric overhead and of course the IO die itself is on a less efficient, older node rivalling Intel's 10nm uh, I mean '7'nm current-gen process.
 
For example, a lot of finacial stuff including simulations is done in Excel. You can read the review to see a lot of applications in which this CPU performs the best.
Hi,
And probably using a macbook :laugh:
 
That's tuned to 35W against other CPUs at their default performance-focused 230W, 250W, or 353W stock PPTs settings?

7950X at ~50W would likely be competitive, if not better since that's TSMC's manufacturing which is at least one full generation of nodes and node refinements ahead, though at very low (sub-65W TDPs) the IO die and chiplet design hurt AMD's efficiency more than Intel's monolithic offerings, as they raise the baseline. A 7950X has a 'parasitic' 15W baseline from the chiplet design adding fabric overhead and of course the IO die itself is on a less efficient, older node rivalling Intel's 10nm uh, I mean '7'nm current-gen process.
Not necessarily even 35W, 65W might be interesting too. I am making an assumption here that the KS is the creme-de-la-creme of silicon that RL and Intels current process can provide and as such should scale somewhat better than the regular 14900K. If not... RIP, I guess. At least if we are talking MT loads. I don't think its catching the 7800X3D in gaming efficiency any time soon, the gap is too big.
 
in other words: buy amd already.
 
It is it what it is here but using the same exact memory on Intel as AMD isn't really quite fair.
Why go and use AMDs preferred ram speed but just consistently ignore Intel like they don't have 8000MHz+ kits around.
Most people buying a KS aren't putting in two sticks of single rank 6000MHz.
 
Not necessarily even 35W, 65W might be interesting too. I am making an assumption here that the KS is the creme-de-la-creme of silicon that RL and Intels current process can provide and as such should scale somewhat better than the regular 14900K. If not... RIP, I guess. At least if we are talking MT loads. I don't think its catching the 7800X3D in gaming efficiency any time soon, the gap is too big.
It would be an interesting test to see for sure, though slightly academic as I don't think the target audience of the KS chips are buying them for efficiency.

However, this is older data and it shows that AMD still has a vast advantage - Intel's running much closer to it's sweet spot at 162W and basically matching the woefully inefficient, balls-to-the-wall 230W configuration of the 7950X.

1710437115718.png
souce: ars

I'm sure Intel would win at some TDP because of the efficiency advantages of a monolithic design at very low power levels, the only question is whether those power levels are remotely relevant outside of EPYC/Xeon Platinum ultra-dense datacenter compute servers.
 
Min FPS is still better with 14900KS interestingly, I wonder why that is.

Ring bus, monolithic die...

I don't think the power consumption and temperature pages leave room for any comment here.

Yup. AMD stands no competition whatsoever. AMD for the win - Ryzen is the highest performing and the most energy efficient gaming processor.

I wonder why TPU doesn't move to the Ryzen 7 7800X3D in its graphics cards reviews? :D

I also wonder how this insanely overclocked Intel junk succeeds in keeping these clocks and unlimited power consumption stable without BSOD and crashes?
How fast does this degrade because of this insane overclock?
 
Lucky for me my premium hardware enthusiasm is based around gaming and i don't get out of bed for 1%.

The enthusiasm is also built around efficiency, whisper quiet noise levels and lower elec bills with energy prices on the rise. With the KS power requirements and price... come on, it's just a collectors item.
 
I'm still happy with my 5950x and it holds up well in many of the charts and it runs happy with NH-L12S for smaller cases at stock if you don't mind the moderate fan noise at full load.
Sadly 3950x is missing from the charts I just thought it would be interesting to see it there for comparison too.
Amazing how well Intel does with emulation.
Also amazing how much power you can push into an Intel chip.
 
Last edited:
Back
Top