Easy to say, but there were no incentives for anyone to do so. Board manufacturers wanted to have the "fastest" boards, and Intel were quite happy to let them do so if it gave them an edge over AMD. Similarly, the media (let's face it, primarily youtubers) didn't have time to test with "Intel stock settings" vs the shipped settings as they have to feed the youtube algorithm - i.e. publish first or perish.
Journalistic integrity counts for nothing?
As I said, "publish first or perish", that's the reality of the modern tech media landscape. And before our resident Intel booster jumps in, yes, this absolutely applies to AMD products also.
Would it not make more sense to stop using that media as a reference or review? They can publish first, but if what they publish is suspect, it is not useful.
It has nothing to do with youtube vs text. Or put another way, which website tested with Intel's recommended power limits as opposed to the motherboard defaults? Anandtech? Tomshardware? Igorslab? PCWorld? Tomsguide? Techpowerup?
Picking a site I use and recall covers this, I see a fair amount of discussion on power limits on ananadtech in reviews over the last few years. They do not seem to ignore this, for example.
Exactly, despite being aware of how this could skew their results, they went ahead and tested with the motherboard defaults anyway. Just like the youtubers.Picking a site I use and recall covers this, I see a fair amount of discussion on power limits on ananadtech in reviews over the last few years. They do not seem to ignore this, for example.
The Intel Core i9-13900KS Review: Taking Intel's Raptor Lake to 6 GHz
www.anandtech.com
They do appear to have decided that motherboard defaults for power limits were most appropriate for intel systems in reviews
Background/explainers are not under the same deadline pressure as reviews. GN have also published similar videos.I also see an entire article on this from a while ago:
Yep++;
If everyone explains it, and until last generation it was fine and even encouraged by the manufacturer, I am not sure that is a problem for reviewers at all.
wccftech.com
9th and 10th gen are both derivatives of Skylake.
The only difference between a -K and a non-K CPU on boards that apply no limits is that the non-K CPUs have significantly lower base clocks, and lower boost clocks. Since the boost clocks are what push the CPU far beyond the sweet spot of the voltage/frequency curve, the non-K parts are less likely to burn themselves out, even when unlimited.
That is what is so interesting about the vendors who would have seen a lot of reports carefully saying it only mostly affects K CPUs. The locked chips should not be capable of the kind of aggressive settings you see in the K parts, so when those fail (even at a much lower rate), it brings up some additional questions.9th and 10th gen are both derivatives of Skylake.
11th is Rocket Lake (Cypress Cove), which is a backported Ice Lake (Sunny Cove) with some improvements. This is a different microarchitecture than Skylake.
12th is Alder Lake, which is Golden Cove.
This is a comparison of all four generations on Intel's website:
Intel product specifications
Intel® product specifications, features and compatibility quick reference guide and code name decoder. Compare products including processors, desktop boards, server products and networking products.ark.intel.com
Why Intel orders them 12, 9, 10, 11, is a mystery, but you can see what has been added/removed over time. All of them have AES-NI.
The only difference between a -K and a non-K CPU on boards that apply no limits is that the non-K CPUs have significantly lower base clocks, and lower boost clocks. Since the boost clocks are what push the CPU far beyond the sweet spot of the voltage/frequency curve, the non-K parts are less likely to burn themselves out, even when unlimited.
They do degrade over time. Silicon degradation usually happens over a longer timeline than a CPU is useful if they design it well, but I think it is a likely cause (with why that happens at such a fast rate here being a good question).
The non-K chips run the clocks a few hundred MHz higher than non-K, and hitting those last couple hundred MHz pushes you way down the VID curve, meaning voltage will be significantly higher. We don't know that the problem is voltage-related, but it seems likely, and everything will be less reliable if you're headbutting the absolute limit of what the chips can do.
They re-tested at Intel's new "published" profile limits:
It's because the CPUs are going well past the voltage and current limits that are detailed in the datasheet.
I do not think they were lying, and the intent was to let you juice the thing until it won benchmarks if desired even to this generation of processors.
9th and 10th gen are both derivatives of Skylake.
11th is Rocket Lake (Cypress Cove), which is a backported Ice Lake (Sunny Cove) with some improvements. This is a different microarchitecture than Skylake.
12th is Alder Lake, which is Golden Cove.
This is a comparison of all four generations on Intel's website:
Intel product specifications
Intel® product specifications, features and compatibility quick reference guide and code name decoder. Compare products including processors, desktop boards, server products and networking products.
Why Intel orders them 12, 9, 10, 11, is a mystery, but you can see what has been added/removed over time. All of them have AES-NI.
AES instruction set - Wikipedia
en.wikipedia.org
That is a decent place to start if you want Intel or AMD. Basically it is very well supported and will not be going away (other than to be ever more standardized and integrated on everything). Basically you should be fine with anything you buy that was made in the last 10 years or so and was not an extreme low power or low budget design.
No competent engineer is going to say you can slam unlimited voltage into a device and permit unlimited current draw, as long as the clock frequency abides by some internal table, and still stay "within spec".
PL2 for the 11900K was 251W. PL1 was 125W. Tau was 56 seconds. Spec, the 11900K would use no more than 251W, for 56 seconds, and then throttle to 125W.
The sad reality is that the -K SKUs regularly go for less than their non-K counterparts.
I do not know about unlimited voltage, but the common view right up until this issue was that Intel CPUs were difficult to destroy and mostly self limiting. It could be unstable, or too much power for the cooling available or some other component, but usually you would not be able to kill the processor with ease. Maybe you could call that overconfidence in that it became an assumption.
The 11th gen datasheet:
Yes because the hardware defaults are changed during initialization. Again, these are defaults, not specifications.
Elsewhere in the datasheet, Intel makes it very clear that your assumptions are wrong:
You want systems to ship with turbo boost disabled?
That is a fine opinion for you to have, but that is not what Intel has decided and ultimately it is Intel that decides what is the specification, not you.
The only thing I am making is screenshots of datasheets.
