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

14900k - Tuned for efficiency - Gaming power draw

Not a lie technically because they had to remove support for a lot of older gen CPU's, remember x370 also supported *Dozer offshoots!

What was a lie however is that they were "committed" to a third gen(?) AM4 (zen)CPU support ~ which is basically marketing speak for we'll do it only if we have to :slap:

 
I pretty much only game on my PC. what an idiot I am for buying the best available cpu for gaming then tweaking it for gaming while playing games that don't utilize HT..... LOL

Why would you turn off HT on a 7800x3D ?
 
Last edited:
Not a lie technically because they had to remove support for a lot of older gen CPU's, remember x370 also supported *Dozer offshoots!

What was a lie however is that they were "committed" to a third gen(?) AM4 (zen)CPU support ~ which is basically marketing speak for we'll do it only if we have to :slap:

Hi,
Indeed Intel priority is to sell new chips
Fortunate for them intel fanboy's will do just that follow the carrot on the stick every release just about :laugh:

Guess they're just happy the socket has lasted this long.
 
Not a lie technically because they had to remove support for a lot of older gen CPU's, remember x370 also supported *Dozer offshoots!

What was a lie however is that they were "committed" to a third gen(?) AM4 (zen)CPU support ~ which is basically marketing speak for we'll do it only if we have to :slap:


I mean, it was a lie for the express intent of not allowing them to upgrade, I'm sure that everyone who had a high end X370 such as the Crosshair 6, myself included, truly and deeply cared for Bristol Ridge support when we were precisely intending to head the other way with a Zen 3 CPU... thankfully Alder Lake changed their minds fast

Hi,
Indeed Intel priority is to sell new chips
Fortunate for them intel fanboy's will do just that follow the carrot on the stick every release just about :laugh:

Guess they're just happy the socket has lasted this long.

The only involvement of Intel in this particular subject is that the i5-12400F existing made it basically impossible for AMD to continue to charge $300 for a 5600X, and that they also needed to ensure that people would retain their AM4 systems instead of getting an Intel build entirely, hence their fast backtracking on that
 
Because it gives you better FPS?
I think it was supposed to be a joke, because in the original post the guy was talking about a 14900k and called it the best gaming cpu.
 
1) it's not "HALF" - it's just HT - which gives you 30% more utilization of P-cores, in the best case scenario, with increased latency and temperatures.
2) Even if you disable the HT cores on a intel i7 or i9 you still get 40-50% more MT performance over an 7700x/7800X3D
3) it gives you an extra 100-200Mhz with the same power and temps, or gives you better power and temps.

Is Hyper-Threading Useless For Gaming Now? 40 Game Benchmark with i9 13900K (youtube.com)

4) Intel knows this and future cpus will not have HT anyways.

View attachment 325655

I just tried disabling HT and increased clocks by 200mhz (before: 5.4ghz Pcores, now 5.6ghz Pcores), the result is repeatable 7% increase in 1% Low FPS in PUBG, power consumption only increases by 20W

Totally worth disabling HT on Intel 13/14th if there is a competitive game that benefit from doing so.

17-12-2023, 11:01:49 TslGame.exe benchmark completed, 26824 frames rendered in 120.344 s
Average framerate : 222.8 FPS
Minimum framerate : 137.5 FPS
Maximum framerate : 240.0 FPS
1% low framerate : 138.2 FPS
0.1% low framerate : 104.7 FPS
17-12-2023, 11:34:11 TslGame.exe benchmark completed, 27313 frames rendered in 120.359 s
Average framerate : 226.9 FPS
Minimum framerate : 148.0 FPS
Maximum framerate : 240.0 FPS
1% low framerate : 148.1 FPS
0.1% low framerate : 125.5 FPS
 
Must define "saving grace" please.

Because the OP tuned his rig, I never saw him say it saved the house from fire or some other outlandish statement.

The guy tuned his rig and supplied a result.

What's with everyone's crap posting?

If you under volted your 7800, was that no different than under-volting a cpu? He's just tuning it and lowered some power draw.

Should be no big deal??
Don't get me wrong, it is an achievement, and as a tech junkie, I totally appreciate it!

By "saving grace" I mean, 253 W by default is way too much for a consumer-grade CPU, and the ability to fine tune it doesn't make it a better product in my eyes. One BIOS reset or upgrade, and you can start again.

I understand marketing wasn't OP's intention, and that's cool.

Like I said, having a non-K chip that is power limited by default, then raising those limits with a capable motherboard and cooling is a much better option, in my opinion. Should you have to reset your BIOS, you're back with the safer limits, not with ridiculously high ones.
 
Don't get me wrong, it is an achievement, and as a tech junkie, I totally appreciate it!

By "saving grace" I mean, 253 W by default is way too much for a consumer-grade CPU, and the ability to fine tune it doesn't make it a better product in my eyes. One BIOS reset or upgrade, and you can start again.

I understand marketing wasn't OP's intention, and that's cool.

Like I said, having a non-K chip that is power limited by default, then raising those limits with a capable motherboard and cooling is a much better option, in my opinion. Should you have to reset your BIOS, you're back with the safer limits, not with ridiculously high ones.

Performance has always come with heat. This is nothing new.

Optimizing a system is a thing. Doesn't matter what they bought, it fit their budget and they tweak within the 253w power envelope. Or 65w envelope.

So in order for a 12400F to run the same frequency as a 12/3/46/7/900K, there's a whole lot more settings to optimize. Because BCLK only. Cause 12400F is max multi 40x. And some people want a bunch of e-cores.

Would it be something if Intel did like AMD and have all their Chips unlocked OC?

Or should Intel be more clear and start naming them Extreme Edition again? Why? Does AMD not boost to it's throttle point? There's no difference. Those get under-volted like it was meant to happen.

All the same just a little different. Just like some gutter wenches from back in the day.
 
Performance has always come with heat. This is nothing new.

Optimizing a system is a thing. Doesn't matter what they bought, it fit their budget and they tweak within the 253w power envelope. Or 65w envelope.

So in order for a 12400F to run the same frequency as a 12/3/46/7/900K, there's a whole lot more settings to optimize. Because BCLK only. Cause 12400F is max multi 40x. And some people want a bunch of e-cores.

Would it be something if Intel did like AMD and have all their Chips unlocked OC?

Or should Intel be more clear and start naming them Extreme Edition again? Why? Does AMD not boost to it's throttle point? There's no difference. Those get under-volted like it was meant to happen.

All the same just a little different. Just like some gutter wenches from back in the day.
That's way too much theoretical thinking for me.

All I'm saying is, the 13900 non-K exists, and the 14900 non-K is coming soon. Those are much safer options, in my opinion, because then, your CPU runs cool by default even if your cooling can't handle 253 W, but you can always raise your power limits if it does. I don't like the idea of default settings that push my CPU to Tjmax, and having to lower it manually to be safe(r), that's all.

Every CPU boosts to throttle point by default if your cooling can't handle it. That doesn't mean it should. The difference is, some CPUs will never reach throttle point with a normal air cooler by default, while others will.
 
Intel just needs to make another 9700k.
 
That's way too much theoretical thinking for me.

All I'm saying is, the 13900 non-K exists, and the 14900 non-K is coming soon. Those are much safer options, in my opinion, because then, your CPU runs cool by default even if your cooling can't handle 253 W, but you can always raise your power limits if it does. I don't like the idea of default settings that push my CPU to Tjmax, and having to lower it manually to be safe(r), that's all.
Set it and forget it. With a performance loss against the K parts.

I found the 13700K to OC slightly better with more cores than a 13600KF. Cause I've played with both. 12400F, 12600K and G6900 processors are all different creatures too.

Pretty sure a 13900 non K part will be noticeably slower in performance to save a little heat.

If the CPU is designed to run its TJmax with a MTBF of say 100,000 hours, you only have to find a way to dissipate the heat better. That's all.
 
Set it and forget it.
Until you reset or update your BIOS.

I found the 13700K to OC slightly better with more cores than a 13600KF. Cause I've played with both. 12400F, 12600K and G6900 processors are all different creatures too.

Pretty sure a 13900 non K part will be noticeably slower in performance to save a little heat.

If the CPU is designed to run its TJmax with a MTBF of say 100,000 hours, you only have to find a way to dissipate the heat better. That's all.
I have an 11700 non-K that is limited to 65 W by default, but performs near-identical to the 11900K with power limits removed. The difference is 200 MHz maximum. I'm happy to sacrifice this much for the thought that my system is always safe.

If you intend on custom looping from the get-go, a K CPU makes sense. But with any everyday cooler, I'd much rather get a non-K and raise its power target until it's still within Tjmax.

Naturally, you don't have to agree.
 
Until you reset or update your BIOS.


I have an 11700 non-K that is limited to 65 W by default, but performs near-identical to the 11900K with power limits removed. The difference is 200 MHz maximum. I'm happy to sacrifice this much for the thought that my system is always safe.

If you intend on custom looping from the get-go, a K CPU makes sense. But with any everyday cooler, I'd much rather get a non-K and raise its power target until it's still within Tjmax.

Naturally, you don't have to agree.

Here's the Direct Difference between a K and non K model.

This is cooled via Wraith Prism, not clamp, diamond nano paste.

50w 6ghz anyone?

Note the date time. That was just a couple minutes ago..... And I just typed this out to you on it.
Edit: Central US time.
 

Attachments

  • RX 6800 GPUPI 3.3 32B.png
    RX 6800 GPUPI 3.3 32B.png
    244.8 KB · Views: 159
Here's the Direct Difference between a K and non K model.

This is cooled via Wraith Prism, not clamp, diamond nano paste.

50w 6ghz anyone?

Note the date time. That was just a couple minutes ago..... And I just typed this out to you on it.
Edit: Central US time.
This is a GPU test on a 6800 with a 13700K. Or am I missing something?
 
This is a GPU test on a 6800 with a 13700K. Or am I missing something?
I had to look close to that CPUz.. and the core count..
 
I had to look close to that CPUz.. and the core count..
I didn't even notice the core count. 3 cores with no HT? Seriously? Am I supposed to be impressed? :laugh:
 
I didn't even notice the core count. 3 cores with no HT? Seriously? Am I supposed to be impressed? :laugh:

Considered it's a 13700K, and that it actually got to 6 GHz, I suppose so! Not sure about the stability though
 
That's amazing. Now I really want to know if this has to do with CPU scheduling because HT has always been a half-solid way of extracting that last little bit of performance from CPU cores. Maybe not back in P4 times, but modern HT usually isn't half bad.
In all the workloads HT shines, the reasoning is sound that e-cores shine in the same scenario. E-cores beat a extra logical core though as even though they slower than P-cores, they still beat a HT core. I plan to do my own experimentation on this.

On my old pfSense unit it has a i5-5250U, This chip is a 2 core 4 threaded chip, the unit at most might load up a single thread briefly when loading a UI page in the web UI, or running a script, but it was never doing heavy multi threaded load. Yet when I disabled HT on it, the temps dropped significantly. So as such I ran it day to day in a 2/2 configuration, with no performance loss from its shipping state.

My only concern with disabling HT, is if I play any games that are designed around 8/16 chips, and for whatever reason those games dont use E-cores. So for me its something I plan to experiment with

Something of interest I have observed as well. If Windows in its default state is allowed to park p-cores, it will always unpark physical cores first, and then only start unparking logical cores, when all 8 of those physical cores are being utilised, this is optimal allocation of cores. However if I run with all p-cores unparked configuration, then I see windows allocating threads to logical cores on the preferred cores ahead of to available unutilised other physical p-cores. This is suboptimal. Disabling HT would fix this and probably yield more performance.

HT made much more sense in the quad core no e-core days.
 
Perhaps the thing to do is just have a few profiles for different scenario's on a few hot keys to change CPU profiles on command be it efficiency or performance or different types of targeted performance uplift. It's certainly worth exploring.
 
Considered it's a 13700K, and that it actually got to 6 GHz, I suppose so! Not sure about the stability though
6 GHz is just an arbitrary number, which doesn't mean anything in scenarios where it matters, that is, doing actual work, or gaming. Especially with only 3 cores running with no HT. So no, I'm not impressed.
 
In all the workloads HT shines, the reasoning is sound that e-cores shine in the same scenario. E-cores beat a extra logical core though as even though they slower than P-cores, they still beat a HT core. I plan to do my own experimentation on this.

On my old pfSense unit it has a i5-5250U, This chip is a 2 core 4 threaded chip, the unit at most might load up a single thread briefly when loading a UI page in the web UI, or running a script, but it was never doing heavy multi threaded load. Yet when I disabled HT on it, the temps dropped significantly. So as such I ran it day to day in a 2/2 configuration, with no performance loss from its shipping state.

My only concern with disabling HT, is if I play any games that are designed around 8/16 chips, and for whatever reason those games dont use E-cores. So for me its something I plan to experiment with

Something of interest I have observed as well. If Windows in its default state is allowed to park p-cores, it will always unpark physical cores first, and then only start unparking logical cores, when all 8 of those physical cores are being utilised, this is optimal allocation of cores. However if I run with all p-cores unparked configuration, then I see windows allocating threads to logical cores on the preferred cores ahead of to available unutilised other physical p-cores. This is suboptimal. Disabling HT would fix this and probably yield more performance.

HT made much more sense in the quad core no e-core days.

Yeah I think that might be part of the reason hyperthreading is on its way out (thats what I heard anyway - that they were starting to get in the way of future development with intel). You really don't need all those extra threads when you have so many cores already. I still leave mine enabled though. too lazy to research which games they help and which ones they don't.
 
Ok I have already done some quick testing. This is with a youtube video playing in firefox (I moved firefox back to p-cores for this testing).

I set min unparked cores to 50% in the power schema, which basically parks all the HT cores (2nd logical core for each p-core) unless a all core load in which case they unpark, I would have to set max unparked to 50% to force them parked under all loads. I then watched the load how it was allocated, I am posting two screenshots, the first is with HT cores parked, the second is with all unparked.

Logical core number starts at 0. 0,2,4,6,8,10,12,14 are the first of each p-core, 16 through to 23 are the e-cores.

8 and 10 are the preferred p-cores (highest clockers).

Result is interesting, different to what I thought was happening, it seems the load pushed on to the HT cores, is only happening on the preferred p-cores and the load pushed onto cores 0,2,4 is the same with HT on or off. (a all core cpuz bench correctly unparked all cores)

HTtest.png

What I plan to do.

Cinebench run with HT cores force parked, to see % impact on score as well as max power draw difference.
Test various games to see if stuttering issues are worse with HT on or off. The gaming I have already done on this CPU with HT on in its default state is much better than it was on my 9900k, but I am curious if it is as performant with HT disabled.

This is different to disabling in the bios which makes the HT cores totally not available, I prefer OS side control, as it allows switching behaviour live on the system.

This test was done with the windows scheduler set to "prefer performant cores".
 
Back
Top