News Intel Core i9-14900KS Review: The Last Core i9 Hits Record 6.2 GHz at Stock Settings

[bit_user]

With the 8600G, we're talking about the lowest-end budget gaming systems that don't even use a GPU—there is a complete and total price sensitivity in that market that simply cannot tolerate expensive RAM.

Understood. But, if you are using its iGPU, then the performance benefits of DDR5 are really part of the equation! If you used it with DDR4, iGPU performance would definitely suffer.

You shouldn't assume that level of iGPU performance is available without DDR5. So, it should be seen as an intrinsic part of the deal, and not an adjunct.

 

[PaulAlcorn]

Understood. But, if you are using its iGPU, then the performance benefits of DDR5 are really part of the equation! If you used it with DDR4, iGPU performance would definitely suffer.

You shouldn't assume that level of iGPU performance is available without DDR5. So, it should be seen as an intrinsic part of the deal, and not an adjunct.

Yes, and as an intrinsic part of the deal for a budget-class chip, it is a con.

 

[thestryker]

That is incorrect. The power settings used are listed in the article in the first testing section (games benchmarks).

Apologies for missing the sentence regarding the settings in the review, but literally everything else I said is still 100% applicable as I made the assumption your were doing what you are indeed doing.

 

[usertests]

Most people would throw more juice at their processor for performance, if it could handle it.

I know I wouldn't. Modern CPUs are more than fast enough for most basic tasks and gaming. I did overclock my Raspberry Pi 4, because that Broadcom CPU needs all the help it can get.

Like you said, there's no reason to get mad about the 14900KS approach. It's just out there for those who want it. How many units are going to be sold, tens of thousands at most?

We are already being warned that processors in the future will require more electricity and better cooling solution with what appears to be little regard for energy savings. Unfortunately, we aren’t engineering architectures to run faster with less.

The APUs are where to look for efficiency, especially as they are laptop-centric products. Lower stock TDPs, greater use of low-power cores, less power consumption than CPU + dGPU. If we see mega APUs like Strix Halo pushed to e.g. 150 Watts (under full load), that still won't be very much most of the time and you could probably use a 300 Watt power supply.

 

[PaulAlcorn]

Apologies for missing the sentence regarding the settings in the review, but literally everything else I said is still 100% applicable as I made the assumption your were doing what you are indeed doing.

I had 36 hours to test the chip, so there was no time to test that as a secondary test point. I barely even got the chip tested in a standard config.

As noted in the text, this is the default setting for nearly every enthusiast motherboard now, and it remains warrantied. Yes, testing in the different power profiles is good data to have, but when shooting for the center mass of the intended market, which is folks hell-bent on getting every last drop of frequency they can get, this was the best option given the time limits.

 

[JamesJones44]

And in only about 12 months after that they will release yet another something...you have to pull the trigger at some point.
Also especially for gamers it doesn't even matter, if you buy a CPU with decent performance today it will last you for many years.

You are likely talking about a "tick" type change in those next 12 months which isn't worth waiting for. Arrow Lake should be a much bigger boost than the tiny 5% you got for the last 2 generations.

If you used this argument and bought the last Intel Pentium D instead of waiting 6 months for the Core 2 Duo you were kicking yourself hard for missing a massive leap in performance and reduced power draw.

 

[bit_user]

You are likely talking about a "tick" type change in those next 12 months which isn't worth waiting for. Arrow Lake should be a much bigger boost than the tiny 5% you got for the last 2 generations.

I understand, if you're talking strictly about single-threaded perf improvements. However, due to the extra E-cores, L2 cache, faster DDR5 speed, and other tweaks, Raptor Lake improved multithreaded performance (over Alder Lake) by a lot more than 5%!

Gen 14 was just a straight clockspeed increase over Gen 13, so you can do the math on that one. Not even 5%, I'd say.

If you used this argument and bought the last Intel Pentium D instead of waiting 6 months for the Core 2 Duo you were kicking yourself hard for missing a massive leap in performance and reduced power draw.

100% agree. However, I don't expect Arrow Lake to be nearly as big a jump as Alder Lake (Gen 12) was over Rocket Lake (Gen 11). And definitely not like Pentium D -> Core 2 big!

 

[thestryker]

I had 36 hours to test the chip, so there was no time to test that as a secondary test point. I barely even got the chip tested in a standard config.

As noted in the text, this is the default setting for nearly every enthusiast motherboard now, and it remains warrantied. Yes, testing in the different power profiles is good data to have, but when shooting for the center mass of the intended market, which is folks hell-bent on getting every last drop of frequency they can get, this was the best option given the time limits.

Oh I understand the reasoning I just disagree with it given the way it causes voltage to spike and the very real problems it can cause. I don't think it's unreasonable to want the tech reviewer crowd to call out the behavior and also not reinforce it. This isn't like the original MCE days where we were talking minor changes because that's all the silicon could reliably do. I get that at the end of the day Intel could force them to stop, but they're not the only potential force for change here.

 

[-Fran-]

"Buht muh Intel CPU can undervolt!"

Well, so can AMD? Motherboard vendors love pushing high voltages because "reasons" (case in point: see Asus burning CPUs) and "bigger voltage better" (lel) it seems.

If you want an energy efficient Intel CPU, get a T SKU if you can find one... If you're buying a K/KS for undervolting... Oof, I don't know what to say XD

Intel needs all of the juice to make a difference in the charts, including power, but for the wrong reasons. Paul's point on the RAM is excellent and it very underrated. Almost ironic in fact, since it used to be Intel that could get away with slower RAM speeds and still perform very well, but how the turns have tabled!

Anyway, Intel crowd just accept Intel is just scraping the bottom of the barrel and hope for improvements to come soon. Alder Lake was a spark, but it was just a firework at best. Raptor Lake (vanilla and R), while not Rocket Lake bad, it was still pretty lame in comparison. Yes, this is a rant towards E cores. Overall, just deal with it.

Regards,

 

[JamesJones44]

100% agree. However, I don't expect Arrow Lake to be nearly as big a jump as Alder Lake (Gen 12) was over Rocket Lake (Gen 11). And definitely not like Pentium D -> Core 2 big!

Oh for sure, I didn't mean to imply it would be it would be another Core 2 moment (I'm very doubtful that it will be), just that the gains are likely to be more than what we've gotten from 12th to 13th and 13th to 14th.

 

[bit_user]

Oh for sure, I didn't mean to imply it would be it would be another Core 2 moment (I'm very doubtful that it will be), just that the gains are likely to be more than what we've gotten from 12th to 13th ...

I like your optimism, but if we return to the point about E-cores, it doesn't seem as if Arrow Lake is going beyond the current 8P + 16E formula. The extra E-cores were a big factor in Raptor Lake's multithreaded improvements, plus it seems like Arrow Lake might lack hyperthreading, which will nullify some of the advancements in IPC and made elsewhere.

Arrow Lake will be a bit of a tick and a tock, because the desktop market missed the microarchitectural improvements in Meteor Lake. So, I am hopeful. Just, multithreaded-wise, not expecting the same kinds of generational gains that either Alder Lake or Raptor Lake provided.

 

[bit_user]

If you want an energy efficient Intel CPU, get a T SKU if you can find one...

Are they really any different than the other variants, aside from power limits? If not, you could just buy a 65 W TDP model and dial in the power limits you want.

The plot I included in Post #39 shows the sort of power/performance tradeoff you get.

My take is that the AMD CPUs have a much sharper "knee". It really doesn't make much sense running a Ryzen 7600X or 7700X beyond 65W or 88W. A higher power limit doesn't make a lot of sense until you go up to the 7900 tier.

With Intel, the good and the bad is that you really do get more performance by pumping in more juice. The flipside is that you also give up a meaningful amount, as you turn down the taps.

 

[TerryLaze]

"Buht muh Intel CPU can undervolt!"

Well, so can AMD? Motherboard vendors love pushing high voltages because "reasons" (case in point: see Asus burning CPUs) and "bigger voltage better" (lel) it seems.

If you want an energy efficient Intel CPU, get a T SKU if you can find one... If you're buying a K/KS for undervolting... Oof, I don't know what to say XD

Intel needs all of the juice to make a difference in the charts, including power, but for the wrong reasons. Paul's point on the RAM is excellent and it very underrated. Almost ironic in fact, since it used to be Intel that could get away with slower RAM speeds and still perform very well, but how the turns have tabled!

Anyway, Intel crowd just accept Intel is just scraping the bottom of the barrel and hope for improvements to come soon. Alder Lake was a spark, but it was just a firework at best. Raptor Lake (vanilla and R), while not Rocket Lake bad, it was still pretty lame in comparison. Yes, this is a rant towards E cores. Overall, just deal with it.

Regards,

If reviewers would do their job properly instead of trolling us around for clicks then the 7950x would be 4% faster than the 14900k but also use 33% more power to do that.
Base power for the 14900k is 125W, you can look it up on ark.
Intel does not force you to run a CPU at 250-300 or 400W , it is under warranty to run it at base power.

 

[TheHerald]

"Buht muh Intel CPU can undervolt!"

Well, so can AMD? Motherboard vendors love pushing high voltages because "reasons" (case in point: see Asus burning CPUs) and "bigger voltage better" (lel) it seems.

If you want an energy efficient Intel CPU, get a T SKU if you can find one... If you're buying a K/KS for undervolting... Oof, I don't know what to say XD

Intel needs all of the juice to make a difference in the charts, including power, but for the wrong reasons. Paul's point on the RAM is excellent and it very underrated. Almost ironic in fact, since it used to be Intel that could get away with slower RAM speeds and still perform very well, but how the turns have tabled!

Anyway, Intel crowd just accept Intel is just scraping the bottom of the barrel and hope for improvements to come soon. Alder Lake was a spark, but it was just a firework at best. Raptor Lake (vanilla and R), while not Rocket Lake bad, it was still pretty lame in comparison. Yes, this is a rant towards E cores. Overall, just deal with it.

Regards,

Undervolt is irrelevant, every CPU can undervolt. But what Intel cpus can do is, power limits. Yes, of course AMD cpus can also be power limited, but they already are out of the box so that's a moot point. An efficiency comparison should be done at ISO wattage, else you are not comparing the chip's efficiency but the out of the box settings. Which, especially for a KS cpu, are really really irrelevant.

 

[TheHerald]

If reviewers would do their job properly instead of trolling us around for clicks then the 7950x would be 4% faster than the 14900k but also use 33% more power to do that.
Base power for the 14900k is 125W, you can look it up on ark.
Intel does not force you to run a CPU at 250-300 or 400W , it is under warranty to run it at base power.

As much as I agree that it's silly to compare efficiency with the power limits removed, it's partly Intel's fault. Yes, I think Intel cpus are insanely good, but the out of the box settings are not. A 14900k is already almost unsusable out of the box, and the 14900ks will be even worse. Again, yes, if you are willing to go into the bios, I think Intel has much better offerings than amd, that's why I have an Intel cpu, but Intel isn't doing anyone any favors by allowing motherboard vendors to go to 500 amps and 4095 power limits. Same thing happened with AMD, if you remember how some mobos were basically frying the chips due to excessive voltage, but AMD stepped in and addressed it. Intel is sort of allowing this to happen.

Again, for me, it doesn't really matter since I'm never going to use a CPU out of the box, but still it's a valid criticism. But your graph and gamersnexus review paints a clear picture, AMD cpus aren't particularly efficient, they are just limited to a very low power draw.

 

[bit_user]

An efficiency comparison should be done at ISO wattage, else you are not comparing the chip's efficiency but the out of the box settings.

That's what the plot I posted in Post #39 shows!

Spoiler

AMD cpus aren't particularly efficient, they are just limited to a very low power draw.

That graph shows AMD providing options with better performance at the same power, from 65 W until about 230 W.

The main reason Zen 4 CPUs appear to have relatively poor efficiency is that their boost power limits are too far beyond the point of diminishing returns. If you restrict them to run in their sweet spot, the superior efficiency of the 7900 and 7950 is quite clear.

 

[TheHerald]

That's what the plot I posted in Post #39 shows!

​

​

Spoiler

​

​

​

That graph shows AMD providing options with better performance at the same power, from 65 W until about 230 W.

The main reason Zen 4 CPUs appear to have relatively poor efficiency is that their boost power limits are too far beyond the point of diminishing returns. If you restrict them to run in their sweet spot, the superior efficiency of the 7900 and 7950 is quite clear.

Does it? The 14700k should be matching the 7900x while being faster in games. The 7800x 3d which is fast in games is lagging behind both the 13700k and the 14700k in MT efficiency. It's even lagging behind (although not by a lot)to the 13600k.

 

[bit_user]

Does it? The 14700k should be matching the 7900x while being faster in games. The 7800x 3d which is fast in games is lagging behind both the 13700k and the 14700k in MT efficiency.

I'll grant that neither of those are in the data set. The publication date of the article should explain quite well why they're excluded (Oct 2022). If they publish an update with Gen 14, I'll update my plot.

 

[TheHerald]

I'll grant that neither of those are in the data set. The publication date of the article should explain quite well why they're excluded (Oct 2022). If they publish an update with Gen 14, I'll update my plot.

It's fine, going by reason alone the 14700k should be right in between the 13900k and the 13700k, probably closer to the former due to probably better binning on 14th gen.

 

[icycool]

Paul really needs to be replaced.
Constantly reading Intel praise centric crap is not only unfair on competitors, it's a blight on Toms.
Will he mention lack of DDR4 support on Intel 15900K, not a chance.
He is always begrudging in his reviews where AMD wins and will do anything possible to shine a good light on Intel.
With such bias, it's difficult to trust anything he posts, including results.

 

[bit_user]

@icycool , fine to air your concerns, but let's please not talk about people being "replaced".

FWIW, I've read Paul's articles for many years and appreciate his depth and attention to detail. I know many have voiced the same concern you shared about his apparent proclivity towards DDR4-supporting platforms at a time when it seems to conveniently align with Intel.

I think Paul could do a lot either to defend this position or perhaps reconsider, if he would make a numerical argument by pricing out a comparably-performing Intel or AMD system (i.e. just motherboard + CPU + RAM). If the overall cost argument no longer goes in Intel's favor, then drop DDR4 as a pro/con. However, if DDR4 still makes the difference, then at least he can point to an amount of $ that it's costing people to go with the AM5-based option.

 

[icycool]

@bit_user you are trying to be fair, but are you really being fair?
Nobody would ever recommend anyone to buy DDR4 these days for a new system.

GN provides unbiased reviews.
Jared provides excellent GPU reviews.
Paul is just dead to me, literally not even worth scrolling to the bit where "Intel is the best!".

 

[35below0]

Nobody would ever recommend anyone to buy DDR4 these days for a new system.

This is simply not true. Not only because of reuse of existing parts but because of lower latency in DDR4.
Speeds are comparable between 4 and 5 except for really high OC like 7600 or 8000. Those have drawbacks as well. Memory training or compatibility issues can appear.

Some people also just want DDR4, or it is cheaper. Those asking for build advice here always have a budget in mind. Sometimes a very tiny budget.
By the end of the year DDR4 will probably begin to fade away but as long as DDR4 boards are available, DDR4 will still be recommended. Right now, it is a still an option.
Same is true for non ATX 3.0 PSUs.

LGA 1700 is a dead end, and at the same time it is current tech. If i were to follow this to the extreme i would have to recommend everyone avoid Intel gen 12, 13, and 14 motherboards and CPUs as worthless.

GN provides unbiased reviews.
Jared provides excellent GPU reviews.
Paul is just dead to me, literally not even worth scrolling to the bit where "Intel is the best!".

Maybe you are too close to it all. Such strong opinions, too strong for computer/tech news.
I sometimes get too involved or too emotional. I lose the forest for the trees.

Like most information online, you have to filter out much of it. Exaggeration creeps into articles almost without the author even noticing; It needs to be filtered out. But bias is in the head of the reader, not the writer. All the writer can do is choose from a collection of facts and experiences, same as we all do. And they are cherry-picked by default. The articles are never perfect, and they don't need to be.

You're fussing too much. I hope i was polite enough in saying that.

 

[TheHerald]

@bit_user
Nobody would ever recommend anyone to buy DDR4 these days for a new system.

New system, sure. But if you already have decent DDR4 kits why not? I know a lot of people that have 13900k / 14900k and still use DDR4 cause they already owned dual rank 32gb bdie kits. We are talking about 4000c14 kits etc. I wouldn't replace them if I had them.

 

[TheHerald]

@bit_user
GN provides unbiased reviews.

Are you sure about that? Cause I nitpick 500 different things he does that are really, really, really biased.