Intel Core i9-14900K

[FreezingPC]

With AMD you can turn on eco mode with only a small performance hit. Just not sure how it affects idle power and temps compared to Intel.

ECOmode doesn't seem to change something on that departement, at least, i didn't saw some major difference on my 5700x...
Not like you would need a high-end, high on watts, CPU to watch some YT, websurf or ms office, i3-12100 should already be good, unless im missing something ?

 

[lexluthermiester]

I'm trying to decide between AMD and Intel, and from what I've been reading so far, it seems Intel runs cooler / draws less power at lower loads and idle than does AMD. But AMD runs cooler and draws less power than Intel at higher loads - (for similar comparably spec'd CPUs).

With AMD you can turn on eco mode with only a small performance hit. Just not sure how it affects idle power and temps compared to Intel.

Which platform to go for, as always, will depend on what the primary purpose for your PC is, what's important to you and what your budget is. If money is no object, the 14900k is the best consumer level CPU and will be for a while. However, if you need to be more careful about money, you'll need to make a more careful choice.

AMD draws less power at higher loads because they have a lower power limit from the factory. Nothing stops you from lowering power limit on intel as well.

That's a good point for people who care. Not everyone does.

 

[rv8000]

A 7950x does not draw 125w but 250w according to nexus. So why would you buy that if you care about efficienfy?

But anyways, that's irrelevant. I don't care what you want to buy, buy whatever the heck you want. I'm saying a 13/14900k is insanely efficient if you don't run at 4096 watts power limits and thus cripple performance by 20-30%, aka highly inefficient.


30 to 40

FTFY

How many goal posts are you gonna move? The review itself has efficiency ratings in Cinebench for MT that you used for comparison, and quite frankly the 13900/14900k results are downright embarrassing in comparison.

But you can keep dreaming otherwise.

 

[JustBenching]

FTFY

How many goal posts are you gonna move? The review itself has efficiency ratings in Cinebench for MT that you used for comparison, and quite frankly the 13900/14900k results are downright embarrassing in comparison.

But you can keep dreaming otherwise.

What goal post am I moving? I don't understand what your point is. The 13/14900k is the 2nd most efficient cpu in ISO wattage. Calling it inefficient is just stupid, since everything else that exists is even more inefficient. How can you not get this is beyond me.

 

[HBSound]

Is there any post about when the I9-1400KS will get released?

 

[Tek-Check]

In measuring performance of a PC part or aspect of operation, if you START at 50% and go to 100%, you might be doubling the performance, but you're still only increasing by 50%. If you go from 100% to 150% you are not doubling the performance, but you are still only increasing by 50%.

No. It is not correct to calculate it this way. The other members were correct.

- up-comparison (the point of view of new CPU): say an old CPU has 100% of performance, e.g. 100fps. If a new CPU gets 150fps, the new CPU is 50% faster because 150/100= value 1.5. So, the new one has performance of 150% relative to what 100% represents for the old one.
- down-comparison (the point of view of old CPU): however, if you change the perspective and treat the same new CPU with 150 fps as 100%, the old one with 100 fps has 67% of performance because 100/150 = 0.67. But, is it 33% slower? No, the older one is 50% slower and new one is still 50% faster.

So, having a certain percentage of performance (67%) below 100% does not mean that you can simply add the missing part (33%) and say new CPU is 33% faster or old is 33% slower. It does not work like that. You need to do the work with raw performance and ratio numbers first and then derive percentages from those.

Lets's take different numbers.
up-comparison:
- current CPU has 100fps and new CPU has 125 fps
- current CPU is 100% and new CPU is 125% because 125 fps/100 fps gives value 1.25, so the new one is 25% faster
down-comparion:
current CPU has 100 fps and new CPU has 75 fps
- current CPU is 100% and new CPU is NOT 25% slower, but 25 fps slower. Big difference!
- let's check it: 75 fps/100 fps gives value 0.75, so the new CPU has 75% of performance of 100% of the old one
- this does not mean it is 25% slower. It is more slower than that, as down-comparisons are calculated differently than up-comparisons
- let's reason: to get from 75% to 100%, is it enough to add 25% of new CPU performance to it? It seems obvious, but the answer is no.
- let's double check: 75 fps + 19 fps (~25%) = 94 fps. So, it does not reach 100 fps (100%).
- you need to add 25 fps to reach 100 fps (100%), which is more than 25% (19 fps)
- you need to add 33% of performance (25 fps) to 75 fps to reach 100 fps
- therefore, new CPU is 33% slower (75 fps), and the old CPU is 33% faster (100 fps)

 

[TheNightLynx]

Because it's the same price and performance as the 13900K and AMD haven't made anything new yet either.
It's literally a drop-in replacement for the 13900K with zero downsides, so it gets the same award as the 13900K.

There is something wrong in your motivations. Let's suppose Intel, from now on, releases a new series every 6 months and call it 15900, 17900, 18900 and so on, what do you do ? Assign an "Editor's Choice" every 6 months cause they have same performance of previous "Editor's Choice" ?

 

[KLMR]

There is something wrong in your motivations. Let's suppose Intel, from now on, releases a new series every 6 months and call it 15900, 17900, 18900 and so on, what do you do ? Assign an "Editor's Choice" every 6 months cause they have same performance of previous "Editor's Choice" ?

That was my point, when you have so many variables that makes you change the decision there is no clear choice to "endorse".

Imagine 14th gen had all or some of the following:
- New socket for 14th and 15th with compatible cooling.
- +16 pcie lanes 8 socket + 8 bridge.
- AV1 hardware encoding gen2 (2 or 3 hardware encoding streams).
- 16p+16e cores
- triple channel RAM or quad channel RAM
- ECC (real ecc)

Then you say, holly molly, 14th gen i9 I choose you. Would be anyone having any discussion about intel offering a new product? or about "efficiency" in a useless program? No. The benefits are clear.
Do you think that is so crazy to implement? Or their costs would ramp up? I mean they only need to work a little on one of their xeons...

 

[caroline!]

So the same as before but overclocked by 100MHz. Such wow. /S

 

[chrcoluk]

That was my point, when you have so many variables that makes you change the decision there is no clear choice to "endorse".

Imagine 14th gen had all or some of the following:
- New socket for 14th and 15th with compatible cooling.
- +16 pcie lanes 8 socket + 8 bridge.
- AV1 hardware encoding gen2 (2 or 3 hardware encoding streams).
- 16p+16e cores
- triple channel RAM or quad channel RAM
- ECC (real ecc)

Then you say, holly molly, 14th gen i9 I choose you. Would be anyone having any discussion about intel offering a new product? or about "efficiency" in a useless program? No. The benefits are clear.
Do you think that is so crazy to implement? Or their costs would ramp up? I mean they only need to work a little on one of their xeons...

Chip would be size of a threadripper, and need 500w, but no issues lol.

 

[HBSound]

That was my point, when you have so many variables that makes you change the decision there is no clear choice to "endorse".

Imagine 14th gen had all or some of the following:
- New socket for 14th and 15th with compatible cooling.
- +16 pcie lanes 8 socket + 8 bridge.
- AV1 hardware encoding gen2 (2 or 3 hardware encoding streams).
- 16p+16e cores
- triple channel RAM or quad channel RAM
- ECC (real ecc)

Then you say, holly molly, 14th gen i9 I choose you. Would be anyone having any discussion about intel offering a new product? or about "efficiency" in a useless program? No. The benefits are clear.
Do you think that is so crazy to implement? Or their costs would ramp up? I mean they only need to work a little on one of their xeons...

I agree!

I always wonder why the smaller chips cannot handle ECC Registered RAM. I am about to try an Intel I9 system but to support the system, I plan on maxing out the memory to 98GB with some of the best DD5 RAM I can find. It is my hope to use the RAM because it is quality RAM, and it provides and helps support a stable system in that area.

 

[Chrispy_]

There is something wrong in your motivations. Let's suppose Intel, from now on, releases a new series every 6 months and call it 15900, 17900, 18900 and so on, what do you do ? Assign an "Editor's Choice" every 6 months cause they have same performance of previous "Editor's Choice" ?

You're not understanding the concept, and I'm not sure why I have to explain this to you. The "Editor's Choice" is an industry standard thing that applies ONLY AT THE TIME OF REVIEW. There is, of course, discretion of the award by "The Editor", in this case W1zzard, but the timeframe for which the award is relevant is consistent across all journalists, and across all market sectors.

So, let me make this super, objectively, unambiguously clear to you:

At the time of this review, the 14900K is the fastest consumer CPU for rendering, encoding, productivity, simulation, compiling, browsing or emulating on the market.

It doesn't matter what came before it.
It doesn't matter what comes after it.
Right now, it's the best you can buy for those workloads I listed above (and more).

There are Editor's Choice awards for products from 1993 that you shouldn't buy today because they're obsolete.
There are going to be Editor's Choice awards for products released decades from now that you can't buy today because they don't exist yet.

Right now, and ONLY right now, the 14900K gets an Editor's choice award for being the best CPU at all the things it wins at, which happens to be quite a lot of what most people want a CPU to do.

Would I personally buy a 14900K?
No; It's too hot, too power-hungry, and unnecessary for me - but that doesn't change the fact that the 14900K is, for many people, the best and fastest CPU they'll have access to right now, at a price that is (relative to inflation, platform costs, and competitor prices) more reasonable than both it's 13900K predecessor, as well as the 7950X3D competition.

 

[lexluthermiester]

No. It is not correct to calculate it this way. The other members were correct.

Whatever. You do you. No one who calculates math PROPERLY cares about your opinion.

Not going to waste my time with the rest of the silliness you typed out..

 

[Tek-Check]

Whatever. You do you. No one who calculates math PROPERLY cares about your opinion.
Not going to waste my time with the rest of the silliness you typed out..

I gave you more examples of what other members had attempted to explain to you. It's really simple.

 

[lexluthermiester]

I gave you more examples of what other members had attempted to explain to you. It's really simple.

Your explanations are as superfluous as they are rhetorical. Your flaw in thinking is that what I stated was incorrect. It is not. Please do carry on.

 

[TheNightLynx]

It has nothing to do with architecture, it is the process Intel makes it on. If Intel redid these CPUs for TSMC process, there would be no problem with efficiency at all.

Maybe yes, maybe not, but will never know until Intel drop 10nm.

So, let me make this super, objectively, unambiguously clear to you:

At the time of this review, the 14900K is the fastest consumer CPU for rendering, encoding, productivity, simulation, compiling, browsing or emulating on the market.

This is not true, the review says that the fastest is 13900KS. This is what is written in the review : "Compared to the 13900K the performance uplift is 3%, the 13900KS is a tiny bit faster"

 

[LuxZg]

Don't want to rain on the party, but when top several processors are within 5%, I will choose based on perf per dolar. And that dolar includes cooling cost and power cost. And when AMD has 2-3x better power efficiency, I'll say that 14900K is a pointless release. Unless you live somewhere where outside never goes over 10C, so you welcome the heat and you'd spend money to heat your home or office anyway. Ryzen can be run without AIO without performance penalty, 14900K not so much. So calculation always needs to include all required components / expenses. That all, combined with naming it 14th gen when it's v.13.1 makes me hate on an otherwise acceptable product. Don't fight over 1-3% differences. Cheers all!

 

[JustBenching]

And when AMD has 2-3x better power efficiency

For god's sake, it doesn't. At this point this is considered trolling. Take a close look at the picture, it's from hardware unboxed's review. Ignore everything else and just focus on the 7950x vs the 7950x 3d. This is system power, CPU power draw is around 230-250w for the 7950x, and around 150 for the 7950x 3d. So, the 7950x draws 66% more power for 3% more performance compared to the 7950x 3d.

Let me repeat it, 66% more power for 3% more performance. Now do you realize that the same scaling applies to the Intel cpus as well? Dropping power draw to 200w instead of 400w will only reduce performance by like 5%. In which case AMD is 5-10% more efficient, not 200-300% like you are claiming. If you want to compare efficiency, you do it at ISO power. It's obvious that a CPU with lower power limits will be more efficient than a CPU with higher power limits. That's why eg. the 14900t will be the most efficient CPU, cause it has a very low power limit.

 

[ratirt]

For god's sake, it doesn't. At this point this is considered trolling. Take a close look at the picture, it's from hardware unboxed's review. Ignore everything else and just focus on the 7950x vs the 7950x 3d. This is system power, CPU power draw is around 230-250w for the 7950x, and around 150 for the 7950x 3d. So, the 7950x draws 66% more power for 3% more performance compared to the 7950x 3d.

Let me repeat it, 66% more power for 3% more performance. Now do you realize that the same scaling applies to the Intel cpus as well? Dropping power draw to 200w instead of 400w will only reduce performance by like 5%. In which case AMD is 5-10% more efficient, not 200-300% like you are claiming. If you want to compare efficiency, you do it at ISO power. It's obvious that a CPU with lower power limits will be more efficient than a CPU with higher power limits. That's why eg. the 14900t will be the most efficient CPU, cause it has a very low power limit.



View attachment 318707

There is one thing i need to point out.
7950x and 7950x3d are not the same chips per se. They vary a lot. the x3d is clocked lower and has a different voltage due to vcache thus lower power consumption.
Now look at the 14900k power draw and then compare it to 13900k. for how many % of performance increase? Not to mention, these are literally the same chips and the difference is 200Mhz frequency bump which is the main factor in the power consumption increase. You can lower voltage and clocks and 14900k will be more efficient but the performance will drop as well and 7950x is not that far away in some tasks. Gaming is another story but the power consumption difference between these chips stays relatively the same.
Now if you compare 7950x to 14900k look a the power difference. 364 vs 533. That is huge and double the difference between 7950x and 7950x3d for not much of a performance bump to be fair.

Best comparison would be to set the power draw for all chips to same value that would not go above that value and measure the performance across the board. That would be a good thing to see. The question is what value? 50w 100w maybe 200w? or which one exactly? The results will vary among different values right? do you see the problem here? Do you know why reviewers use default settings and judge by that metric?

 

[JustBenching]

There is one thing i need to point out.
7950x and 7950x3d are not the same chips per se. They vary a lot. the x3d is clocked lower and has a different voltage due to vcache thus lower power consumption.
Now look at the 14900k power draw and then compare it to 13900k. for how many % of performance increase? Not to mention, these are literally the same chips and the difference is 200Mhz frequency bump which is the main factor in the power consumption increase. You can lower voltage and clocks and 14900k will be more efficient but the performance will drop as well and 7950x is not that far away in some tasks. Gaming is another story but the power consumption difference between these chips stays relatively the same.
Now if you compare 7950x to 14900k look a the power difference. 364 vs 533. That is huge and double the difference between 7950x and 7950x3d for not much of a performance bump to be fair.

Best comparison would be to set the power draw for all chips to same value that would not go above that value and measure the performance across the board. That would be a good thing to see. The question is what value? 50w 100w maybe 200w? or which one exactly? The results will vary among different values right? do you see the problem here? Do you know why reviewers use default settings and judge by that metric?

The 7950x and the 7950x 3d are the exact same chips. Besides gaming and a few other applications at similar watts performance is similar.

What value you test doesn't really matter. Yes there will be differences, but if the 7950x is 3x times more efficient than the 14900k at 200w then it will be more efficient at any wattage honestly. But of course that is not the case, the 7950x is nowhere near 2 or 3 times more efficient. It's close to 0.1 times more efficient.

 

[ratirt]

The 7950x and the 7950x 3d are the exact same chips. Besides gaming and a few other applications at similar watts performance is similar.

I disagree. They are same architecture but are not the same. x3d has vcache and the no x3d does not. I dont think you should judge processors being same strictly by performance.

What value you test doesn't really matter. Yes there will be differences, but if the 7950x is 3x times more efficient than the 14900k at 200w then it will be more efficient at any wattage honestly. But of course that is not the case, the 7950x is nowhere near 2 or 3 times more efficient. It's close to 0.1 times more efficient.

I disagree again. it wont because the increase in clock speed which comprises of increased voltage and wattage is not linear. And you have confirmed what is actually common. You may argue that AMD counterparts do not OC well (x3d for instance) but is there a need to OC those? They have automatic OC. Intel can OC but then it becomes even more inefficient due to higher voltage and wattage with very little change in performance. If you want to buy a fast CPU like 14900k which is the fastest in Intel's stack and you make it more efficient (lower voltage, wattage with reduced clocks) this makes the purchase pointless since it will drop in performance. AMD is a differnet CPU, you buy it and you use it and this is the performance and power usage. No need to OC but you can give it a go with a memory OC and that will boost performance exponetialy untill you hit the wall that is.
What I'm saying is, you cant call it the fastetst and constrain it with 100w power and call it the most efficient at the same time. Full power of the CPU is achieved when it is the least efficient.
AMD does not have that problem. The most performance it can produce is when it is fairly efficient in comparison to Intel's.
If you constrain Intels 14900K to whatever output AMD 7950x produces (364watts as per the graph) will Intel's 14900K still be on top? I dont know but something tells me it won't.

 

[JustBenching]

I disagree. They are same architecture but are not the same. x3d has vcache and the no x3d does not. I dont think you should judge processors being same strictly by performance.

I disagree again. it wont because the increase in clock speed which comprises of increased voltage and wattage is not linear. And you have confirmed what is actually common. You may argue that AMD counterparts do not OC well (x3d for instance) but is there a need to OC those? They have automatic OC. Intel can OC but then it becomes even more inefficient due to higher voltage and wattage with very little change in performance. If you want to buy a fast CPU like 14900k which is the fastest in Intel's stack and you make it more efficient (lower voltage, wattage with reduced clocks) this makes the purchase pointless since it will drop in performance. AMD is a differnet CPU, you buy it and you use it and this is the performance and power usage. No need to OC but you can give it a go with a memory OC and that will boost performance exponetialy untill you hit the wall that is.
What I'm saying is, you cant call it the fastetst and constrain it with 100w power and call it the most efficient at the same time. Full power of the CPU is achieved when it is the least efficient.
AMD does not have that problem. The most performance it can produce is when it is fairly efficient in comparison to Intel's.
If you constrain Intels 14900K to whatever output AMD 7950x produces (364watts as per the graph) will Intel's 14900K still be on top? I dont know but something tells me it won't.

But I did not call the 14900k the fastest. All I'm saying is at similar power limits it offers same performance and efficiency with amds best in Mt workloads. It is not the fastest in MT workloads, it loses to the 7950x, but not by 2-3 times like the guy above me argued. The difference is 5 to 20% depending on what exactly you are testing.

 

[Tek-Check]

Your explanations are as superfluous as they are rhetorical. Your flaw in thinking is that what I stated was incorrect. It is not. Please do carry on.

There is no flow in thinking, but there are different narratives, one of which was used to show something different from what the other members were explaining, hence a glitch. This is what you wrote earlier:
"...if you START at 50% and go to 100%, you might be doubling the performance, but you're still only increasing by 50%"
Can you start anything at, say, 50%, in vacuum, without relation to a reference point? You have not explained the logic of this reasoning to other members to show them the difference of taking this perspective in that context. It appeared as incorrect. Below is why.

Let's put what you wrote in context and follow that logic of reasoning to try to find out when we could really "start at 50%" with anything in computing.
- let's assume that a starting point for best GPU1 last year was 50 fps. This is represented as 100% because GPU1 is on the top of performance chart
- can this be a starting point "at 50"%? It can't, because GPU1 is 100% at the moment and one data point. Ok, all good.
- new GPU2 this year doubles the performance and scores 100 fps. It is 100% faster and this is new 100%, so GPU1 is not anymore 100%
- now we have two data points. Is this the moment we can introduce the idea of "if you start at 50%..."?
- possibly, but would it be meaningful? Let's check it out.
- GPU1 now has 50% of performance of GPU2, but GPU2 is still 100% faster than GPU1
- has the performance of GPU2 increased by 50%, from "GPU1 50%" (50 fps) to "GPU2 100%" (100 fps)? The answer is yes, according to your proposition.
- GPU2 is still 100% faster than GPU1, not 50% faster, despite the wording "you're still only increasing by 50%"
- there are two narratives here, indvidual GPU performance improverment and wider, gen-to-gen performance improvement, which you usuallly need 3 data points for as "start at 50%" also comes from somewhere
- I get what you mean; what was "100%" yesterday becomes "50%" today and there is new "100%" now, therefore generational performance increases by 50%
- such narrative needs more context rather than being abstracted from two GPU data points only
- GPU3 brings 150 fps and becomes new "100%". It's 50% faster than GPU2 and 300% faster than GPU1
- at this point, we can say that top SKU in each generation brings 50% more performance overall than the previous one
- you are correct to use generic narrative to show gen-to-gen percentage increase, but you need to mention reference point and give it context
- other members were focused on individual component difference
- we just need to be careful in which context we use those narratives, as they serve to show different things

 

[Chrispy_]

Maybe yes, maybe not, but will never know until Intel drop 10nm.


This is not true, the review says that the fastest is 13900KS. This is what is written in the review : "Compared to the 13900K the performance uplift is 3%, the 13900KS is a tiny bit faster"

The KS isn't the same price, so it's out of the discussion; You're also arguing the logic of the 14900K's Editor's Choice award directly against the 13900K's Editor's Choice award.

Don't.
Move.
The.
Goalposts.

 

[firejohn]

And where does APO come from? Windows Update, extra software, directly integrated in games, drivers?