News Intel Core i9-14900K, i7-14700K and i5-14600K Review: Raptor Lake Refresh XXX

[bit_user]

Why should they be noted as a suggestion? It's what the part is rated for which means it's what the part is guaranteed to work at.

Where do they say that? I think you simply imagined that part.

Then why does Nvidia bother putting power specs out for their parts?

Show me where they do. I'm not interested in debating your imagination. Let's look at the facts.

Here's a completely different example of specs/settings not matching: the N series processors only support 16GB according to spec,

Irrelevant. Using more memory doesn't cause the CPU to run hotter or the machine to run louder. It's harmless for them to specify a low value, which can be exceeded under certain (or even all) circumstances, because there's essentially no downside or tradeoff.

I don't know how I can possibly make this more clear

By quoting where I supposedly said what you're claiming I said. Either show me where I said it or drop the point.

I'm asking you if it's not okay to change the power settings to enforce PL2 (your stance this entire time) for a review why is it okay that they enable XMP/EXPO for one?

I already said that if the reviewer is testing "out-of-the-box" settings, then they shouldn't change anything.

If they're testing vendor-recommended settings, then they should change according to those & document what they tested.

As I've essentially been saying: we're in an era where you can no longer just test these CPUs as if the rest of the system doesn't matter. If you want to include "official Intel specifications" as one of those configurations, that's fine. Just don't pretend it represents some ground truth, as it's really almost as arbitrary as anything else you might test.

I've been extremely clear in what my position is:

Position and agenda are two different things. Agenda is essentially the outcome which motivates your position. If you're clear - at least, in your own mind - about the outcome you want, it's easier to find a good position. If you don't know what you want to achieve, or if you're trying to achieve something different than I am, we'll probably never agree on a position.

 

[thestryker]

Show me where they do. I'm not interested in debating your imagination. Let's look at the facts.

I'll get to the rest later, but this took me 5s to find: https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3080-3080ti/

shocker that was sure hard

 

[thestryker]

Irrelevant. Using more memory doesn't cause the CPU to run hotter or the machine to run louder. It's harmless for them to specify a low value, which can be exceeded under certain (or even all) circumstances, because there's essentially no downside or tradeoff.

So you're picking and choosing what matters based on an arbitrary metric you find relevant got it.

I already said that if the reviewer is testing "out-of-the-box" settings, then they shouldn't change anything.

Then none of the reviews that have come out other than Anandtech's meet your metric. That's totally fine and I at least understand where you're coming from even if I disagree.

By quoting where I supposedly said what you're claiming I said. Either show me where I said it or drop the point.

By changing nothing the CPUs are running with power limits removed. This is what you've been advocating the entire time. That's totally fine as we've established now that you want no settings changed, but it's still a fact that you're advocating for power limits removed testing since that's out-of-the-box.

As I've essentially been saying: we're in an era where you can no longer just test these CPUs as if the rest of the system doesn't matter. If you want to include "official Intel specifications" as one of those configurations, that's fine. Just don't pretend it represents some ground truth, as it's really almost as arbitrary as anything else you might test.

Then you do not want CPU tests you want pseudo real world system tests. The entire point behind testing a component is to isolate that component from everything else as well as you can hence why reviewers tend to test CPUs with the highest GPU they have available and standardize the other components across CPU tests as much as possible.

Position and agenda are two different things. Agenda is essentially the outcome which motivates your position. If you're clear - at least, in your own mind - about the outcome you want, it's easier to find a good position. If you don't know what you want to achieve, or if you're trying to achieve something different than I am, we'll probably never agree on a position.

I think it's fair to say neither one of us wants the same thing from a CPU review. I want a CPU review and you want a system review those two things will never be the same barring some massive shift in manufacturer behavior.

 

[bit_user]

I'll get to the rest later, but this took me 5s to find: https://www.nvidia.com/en-us/geforce/graphics-cards/30-series/rtx-3080-3080ti/

shocker that was sure hard

Except:

"Note: The above specifications represent this GPU as incorporated into NVIDIA's Founders Edition or reference graphics card design. Clock specifications apply while gaming with medium to full GPU utilization. Graphics card specifications may vary by add-in-card manufacturer. Please refer to the add-in-card manufacturers' website for actual shipping specifications."
​

Obviously, they were going to say that. Intel does not.

 

[bit_user]

So you're picking and choosing what matters based on an arbitrary metric you find relevant got it.

No, the burden is on you to show why it's analogous, though it's clearly not. Trying to blame that on me suggests that you're not posting in good faith.

By changing nothing the CPUs are running with power limits removed. This is what you've been advocating the entire time.

What I've been advocating for is testing that reflects actual usage. If "power limits removed" is what these boards actually default to, then find a board which does that and test it at the defaults.

Then you do not want CPU tests you want pseudo real world system tests. The entire point behind testing a component is to isolate that component from everything else

What I object to is testing products in a way that doesn't reflect actual usage, because it paints a false impression of how they behave. In that regard, if the CPU's performance cannot be accurately characterized in isolation, then reviews shouldn't pretend otherwise.

What I also object to is Intel not clearly stating that motherboards and systems can exceed the specified power limits on their unlocked CPUs. How many people do you think know that?

I want a CPU review and you want a system review

What you want is something largely fictional, but I don't know why. What I want is something realistic, and for Intel not to engage in misleading advertising.

 

[thestryker]

Except:

​

"Note: The above specifications represent this GPU as incorporated into NVIDIA's Founders Edition or reference graphics card design. Clock specifications apply while gaming with medium to full GPU utilization. Graphics card specifications may vary by add-in-card manufacturer. Please refer to the add-in-card manufacturers' website for actual shipping specifications."​

​

Obviously, they were going to say that. Intel does not.

Fair enough I totally missed that when scrolling on my phone.

 

[thestryker]

No, the burden is on you to show why it's analogous, though it's clearly not. Trying to blame that on me suggests that you're not posting in good faith.

You're saying that an OEM circumventing an Intel spec doesn't matter in this case because it doesn't impact what you care about. If you think Intel has to be the one doing the disclosures then you should want them to add one to their memory spec as well. That's all I'm saying here.

What I also object to is Intel not clearly stating that motherboards and systems can exceed the specified power limits on their unlocked CPUs. How many people do you think know that?

Hopefully anyone who does any research regarding what they buy, and I do agree it should be more clear, but I think that needs to be on the motherboard/OEM side. Even if Intel has to force them to do it like Nvidia did when they had laptop OEMs not disclosing GPU wattages. A statement/warning like "this CPU can use above maximum turbo power depending on your system setup" is effectively useless compared to specific disclosure like "xxx uses thermal limits instead of power limits for the CPU".

To call back to your correction regarding nvidia specs a warning like that is only useful if the AIBs disclose any differences. It seems like that must be voluntary on desktop cards as EVGA, Asus and Gigabyte don't cite power consumption, but MSI does (I didn't look for others). The first disclosure without the second makes it useless as you still have to rely on reviews if you want to know before buying. This is exactly how I feel about Intel's CPU power situation as well.

What you want is something largely fictional, but I don't know why. What I want is something realistic, and for Intel not to engage in misleading advertising.

It's not realistic though aside from "no power limits with good cooling = high power usage" which you shouldn't need a review to know. A Deepcool dual tower cooler will easily cool a 14900K, but with power limits removed compare it to their 360mm AIO and you end up with dramatically different power consumption (and varying amount of performance uplift depending on bin). I'm not sure why you would think results in this configuration by themselves can be useful.

It's obviously something we fundamentally won't agree on and that is what it is.

Reviewers should absolutely warn readers about the default settings no matter what since there's currently no other way of knowing (I checked a couple of motherboard manuals and don't even see the default settings listed).

 

[Order 66]

This is 13.5 gen at best. I'll be sticking with AMD until Intel releases an actual generational improvement that is significantly better than my 7700x.

 

[King_V]

Hey, everyone, I just realized something.

There is ONE . . and only ONE . . thing that Intel did right with this.

The 14th-Gen chips are backward compatible with existing 600- and 700-series motherboards

They have, with this compatibility, and outside of the leaked information that bit_user posted, admitted to us that it's not a new generation.

If this were truly a new generation, then Intel would have, for the first time in at least a decade, allowed more than 2 generations on the same chipset. That limit, I believe, is practically etched in stone.

Therefore, again, assuming we didn't have access to the leaked info, the fact that this new generation works on existing motherboards is basically Intel's confession that it's not new at all.

 

[dalauder]

I agree. Intel 13th and "14th" gen CPUs are essentially "Liquid Cooling Required" CPUs if you want to avoid thermal throttling.

And they've gotta be good liquid coolers at that. You can buy the wrong $100 liquid cooler and it's worse than a great $55 air cooler.

 

[dalauder]

Hey, everyone, I just realized something.

There is ONE . . and only ONE . . thing that Intel did right with this.



They have, with this compatibility, and outside of the leaked information that bit_user posted, admitted to us that it's not a new generation.

If this were truly a new generation, then Intel would have, for the first time in at least a decade, allowed more than 2 generations on the same chipset. That limit, I believe, is practically etched in stone.

Therefore, again, assuming we didn't have access to the leaked info, the fact that this new generation works on existing motherboards is basically Intel's confession that it's not new at all.

They used to just throw a "5" on the end of the model numbers to replace a "0". But marketing says that telling everyone it's NEW when you literally just move the clocks and Watts up a notch is better.

Couldn't everyone with thermal headroom already have overclocked and achieved nearly identical benchmark numbers? ...Not that Intel leaves anyone thermal headroom.

 

[helper800]

They used to just throw a "5" on the end of the model numbers to replace a "0". But marketing says that telling everyone it's NEW when you literally just move the clocks and Watts up a notch is better.

Couldn't everyone with thermal headroom already have overclocked and achieved nearly identical benchmark numbers? ...Not that Intel leaves anyone thermal headroom.

If the 14000 CPUs are higher binned 13000 silicon, then no, on average you could not just OC and get similar benchmarking. Whether that is true or not, only Intel knows. Either way, the performance differences are not large enough to matter to 99% of people anyways.

 

[bit_user]

If the 14000 CPUs are higher binned 13000 silicon, then no, on average you could not just OC and get similar benchmarking. Whether that is true or not, only Intel knows.

Actually, we do know they're not new designs. It doesn't even make sense for them to be binned. Most likely, the new clockspeed limits are simply due to the maturity of their Intel 7 manufacturing process.

Although this article is purported about no-K Gen 14 CPUs, they actually shared the die specs on all Gen 12 - 14 models. There's no such thing as a Gen 14 die. They're all Gen 12 or Gen 13 dies & steppings:

​

Vendors Share Raptor Lake Refresh Non-K CPU Specs

Specs of 35W and 65W Raptor Lake Refresh processors get published.

www.tomshardware.com

​

 

[helper800]

Actually, we do know they're not new designs. It doesn't even make sense for them to be binned. Most likely, the new clockspeed limits are simply due to the maturity of their Intel 7 manufacturing process.

Although this article is purported about no-K Gen 14 CPUs, they actually shared the die specs on all Gen 12 - 14 models. There's no such thing as a Gen 14 die. They're all Gen 12 or Gen 13 dies & steppings:

​

Vendors Share Raptor Lake Refresh Non-K CPU Specs

Specs of 35W and 65W Raptor Lake Refresh processors get published.

www.tomshardware.com

​

I was speculating by saying "If" that either they are higher binned or they are not higher binned, not whether or not they are the same chips, I knew that.

 

[bit_user]

I was speculating by saying "If" that either they are higher binned or they are not higher binned, not whether or not they are the same chips, I knew that.

I don't see how it can be a matter of binning, given that you'd expect them to mostly switch over from making & selling Gen 13 to Gen 14 CPUs.

For the most part, the people still buying Gen 13 CPUs & systems that aren't merely excess inventory "in the channel" that's being discounted are those who have qualified Gen 13 CPUs & systems for a specific application. The rest of the market should generally move over to Gen 14.

Given that, they won't have a way to move large numbers of chips that fail to fit any Gen 14 bin.

 

[helper800]

I don't see how it can be a matter of binning, given that you'd expect them to mostly switch over from making & selling Gen 13 to Gen 14 CPUs.

For the most part, the people still buying Gen 13 CPUs & systems that aren't merely excess inventory "in the channel" that's being discounted are those who have qualified Gen 13 CPUs & systems for a specific application. The rest of the market should generally move over to Gen 14.

Given that, they won't have a way to move large numbers of chips that fail to bit any Gen 14 bin.

The reason we can speculate binning is that we don't know, only Intel does and the 14000 have increased clock speeds. If I am not mistaken, they are still making 13000 chips and obviously 14000 chips. It would make sense to take chips that cannot reach 14000 speeds, even if they are a small amount faster, and make them one of the various 13000 monikers.

Price also makes a reasonable argument to get the 13700k vs the 14700k and so on.

 

[Avro Arrow]

And they've gotta be good liquid coolers at that. You can buy the wrong $100 liquid cooler and it's worse than a great $55 air cooler.

They used to just throw a "5" on the end of the model numbers to replace a "0". But marketing says that telling everyone it's NEW when you literally just move the clocks and Watts up a notch is better.

Couldn't everyone with thermal headroom already have overclocked and achieved nearly identical benchmark numbers? ...Not that Intel leaves anyone thermal headroom.

If you really want to see something funny, watch Steve Walton compare "12th-gen", "13th-gen" and "14th-gen", only to realise that there's nothing different apart from clock speed:

 

[bit_user]

If I am not mistaken, they are still making 13000 chips and obviously 14000 chips.

They're still selling them, but the key questions are how much volume and how much of it is out of existing inventory?

It would make sense to take chips that cannot reach 14000 speeds, even if they are a small amount faster, and make them one of the various 13000 monikers.

Depends on how many chips fail to fit any Gen 14 bin. If it's like 50%, there's no way they have enough Gen 13 buyers for that. They'd have to sell them at a steep discount, which I'm guessing they're loath to do. If Gen 14 yield were that bad, they wouldn't have made the specs so aggressive.

Price also makes a reasonable argument to get the 13700k vs the 14700k and so on.

Only around launch time. They didn't formally reduce the price of Gen 13. Any $ you're saving is by time-limited discounts, which are probably done just to clear old product out of the sales channel.

 

[bit_user]

If you really want to see something funny, watch Steve Walton compare "12th-gen", "13th-gen" and "14th-gen", only to realise that there's nothing different apart from clock speed:

Depends on which specific models. The chart in that article I linked shows that some Gen 13-branded CPUs are actually using Gen 12 dies. So, those will have no increase in L2 cache, for instance.

Also, it's hardly news that there were no changes to the microarchitecture since Gen 12. So, it's basically just a matter L2 cache, core count, and clockspeed which differ between the models & generations.

 

[thestryker]

They're still selling them, but the key questions are how much volume and how much of it is out of existing inventory?

There's no way Intel would be wasting enough warehouse space to just not be actively making 13th gen parts still, but I'm sure the rate is greatly reduced.

Depends on how many chips fail to fit any Gen 14 bin. If it's like 50%, there's no way they have enough Gen 13 buyers for that. They'd have to sell them at a steep discount, which I'm guessing they're loath to do. If Gen 14 yield were that bad, they wouldn't have made the specs so aggressive.

Yeah there's no way they're taking die that don't reach 14th gen clocks and making 13th gen parts out of them. They'd use them down stack for lower SKUs first, and I bet at this point any 13th gen is just at a set manufacturing rate.

Most likely, the new clockspeed limits are simply due to the maturity of their Intel 7 manufacturing process.

On the 14900/14700 this is mostly likely the reason they exist in the form that they do.

 

[helper800]

They're still selling them, but the key questions are how much volume and how much of it is out of existing inventory?


Depends on how many chips fail to fit any Gen 14 bin. If it's like 50%, there's no way they have enough Gen 13 buyers for that. They'd have to sell them at a steep discount, which I'm guessing they're loath to do. If Gen 14 yield were that bad, they wouldn't have made the specs so aggressive.


Only around launch time. They didn't formally reduce the price of Gen 13. Any $ you're saving is by time-limited discounts, which are probably done just to clear old product out of the sales channel.

There's no way Intel would be wasting enough warehouse space to just not be actively making 13th gen parts still, but I'm sure the rate is greatly reduced.

Yeah there's no way they're taking die that don't reach 14th gen clocks and making 13th gen parts out of them. They'd use them down stack for lower SKUs first, and I bet at this point any 13th gen is just at a set manufacturing rate.

On the 14900/14700 this is mostly likely the reason they exist in the form that they do.

It seems I was on the right track maybe with some of it.

 

[thestryker]

It seems I was on the right track maybe with some of it.

What this indicates is that the differences are likely all in manufacturing process refinement. If Intel is able to sell "lower" quality silicon in 13900K form in 2023 than 2022 that's a big indicator. The other one is that the 14900K doesn't need to be "better" quality than the 13900KS to hit the same base clocks with less wattage or higher all core/E-core with the same.

 

[dalauder]

If you really want to see something funny, watch Steve Walton compare "12th-gen", "13th-gen" and "14th-gen", only to realise that there's nothing different apart from clock speed:

Hahaha! That's hilarious! Thanks for sharing.

 

[Avro Arrow]

Hahaha! That's hilarious! Thanks for sharing.

You're welcome! As usual though, Steve Burke was even funnier than Steve Walton:

He is the true Jedi Master of passive-aggressive sarcasm!

"which may not seem like much.... BUT......"
- Steve Burke, 2023