Intel Provides Update On 7nm, New US Fabs, "Intel On" Event

[piotrj3]

For me, the difference between the two is stark, and the fact that I can buy a 65w CPU and by default it will try sucking down >200w (per Anandtech testing) is frightening. I was actually toying with getting a 65w Intel CPU for a tiny rig (which I would power with a 120w pico-PSU-alike) until I saw that. It doesn't matter whether Intel or the motherboard manufacturers are the root cause.

There is a world of difference between a 105w CPU drawing 136w when boosted, and a 65w CPU sucking down triple. Another interesting investigation is this one.

And thanks for the video, but I can read faster than he can talk. Video reviews are torturous.

That video exactly explains why anandtech review is simply wrong. Anyway 6:30 graph should be enough.

[stiiixy]

Agreed, video 'guides' are extremely boring, torturous and generally uninformative compared to simple text. I don't want your (the content creator) emphasising your opinions over your "data's", either, as so often happens.

I can make up my own mind.

[coder]

Raspberry Pi 4 is built on a 28nm process and it doesn't run like a furnance.

It's actually the hottest Pi, yet. Also, it had serious issues with thermally-induced clock throttling when released. Still, the only way to completely avoid clock throttling is with an aftermarket cooler.

But that's only because (like Intel), they decided to push clock speeds close to what the manufacturing technology and circuit design could support. If you run it a bit slower, no such problems.

Unfortunately, we can look forward to a couple more Pi generations at 28 nm. They've said they're going for the cheapest (per transistor) manufacturing technology. At the release of v4, it looked like that would remain 28 nm for years to come. All previous generations were 40 nm.

[coder]

you need to watch video. Thing is in case of Intel, Intel makes a guidance that motherboard manufactures don't have to follow regarding turbo etc.

However, even strictly following Intel's guidance means a 65 W Comet Lake i7 can hit 224 W, for up to 28 seconds!

https://www.tomshardware.com/news/in...ake-processors

And, with AVX-512, Rocket Lake looks like it could be even worse. Even with the updated BIOS, Anandtech saw spikes of up to 276 W! A few years ago, only GPUs were pulling those kinds of numbers at stock settings!

Don't get me wrong, AMD is more energy efficient on 7nm with according to guidance vs according to guidance, but diffrence is way smaller then it is rumored to be.

This is nonsense. AMD does boost their CPUs, but Intel is in a league of its own!

Source: https://www.anandtech.com/show/16549...-core-i711700k

Just to be clear, that's showing an AMD CPU boosting by about 35%, while the Intel CPU boosts by 133% (and more than 2x of the absolute power as the AMD CPU)! Both are high-end 8-core models, Intel's a "K" and AMD's an "X".

[piotrj3]

However, even strictly following Intel's guidance means a 65 W Comet Lake i7 can hit 224 W, for up to 28 seconds!

https://www.tomshardware.com/news/in...ake-processors

And, with AVX-512, Rocket Lake looks like it could be even worse. Even with the updated BIOS, Anandtech saw spikes of up to 276 W! A few years ago, only GPUs were pulling those kinds of numbers at stock settings!


This is nonsense. AMD does boost their CPUs, but Intel is in a league of its own!

Source: https://www.anandtech.com/show/16549...-core-i711700k

Just to be clear, that's showing an AMD CPU boosting by about 35%, while the Intel CPU boosts by 133% (and more than 2x of the absolute power as the AMD CPU)! Both are high-end 8-core models, Intel's a "K" and AMD's an "X".

Measuring peak power is stupid, and this type of metric is only useful for knowing if your power supply has bad overcurrent protection. Intel since Haswell (rightfully) extremly dynamicly switches frequency and every PSU since around Haswell times will be fine with that since every power supply since then uses DC voltage conventer.

The worst part is that Anandtech doesn't even specify about what period of time is measured in each peak power interval as well doesn't mention even motherboard used for testing or if they are compliant with Intel's boosting guidelines.

However, even strictly following Intel's guidance means a 65 W Comet Lake i7 can hit 224 W, for up to 28 seconds!

And it is absolutly ok, as for normal everyday use you have often small very short spikes of usage when CPU goes. Think about it from benchmarking perspective, it is very easy to pull average result of cinebench or 7zip compression or some big data set. But user most of time cares does his webrowser opens fast, does his page load fast, does archive with less then 100MB of documents decompress fast, can he compress something fast, and it is absolutly ok that for this very short time CPU works harder

[CochainComplex]

Is there any reason why does Intel take forever with 7nm and why are the other brands ahead of Intel?

Are they trying to achieve true 7nm (as in very tiny spacing) or something?

too much bean counters I guess

[coder]

And it is absolutly ok, as for normal everyday use you have often small very short spikes of usage when CPU goes.

Maybe you don't mind your CPU cooler sounding like a jet engine reving up for takeoff, but I do. Or maybe you have water cooling, but most don't.

So, what it comes down to is dishonesty by Intel. In order to sustain these peaks that they use to gin up their benchmarks, you need a high-end cooler. Otherwise, it ends up being thermally-limited before it can hit Tau. And it still has plenty of potential to annoy, in the process. And the worst part is that it falls short of the advertised performance!

But user most of time cares does his webrowser opens fast, does his page load fast, does archive with less then 100MB of documents decompress fast, can he compress something fast, and it is absolutly ok that for this very short time CPU works harder

I'm not objecting to any sort of turbo -- just the extreme degree to which Intel has taken it, in the past few generations! Your argument lacks any sense of scale or degree (not to mention the unheard of 104 degree temp Anandtech's CPU hit, in their test!).

Also, speaking of temperatures, I like reliability in my tech. Most computers I end up keeping around for the better part of a decade. I don't want a CPU that's just going to burn out shortly after the end of its warranty period. And it should be noted that we don't know the long-term implications of this kind of boosting behavior. It's not going to be kind on your CPU fan, I can assure you of that!

[Paradigm Shifter]

And it is absolutly ok, as for normal everyday use you have often small very short spikes of usage when CPU goes. Think about it from benchmarking perspective, it is very easy to pull average result of cinebench or 7zip compression or some big data set. But user most of time cares does his webrowser opens fast, does his page load fast, does archive with less then 100MB of documents decompress fast, can he compress something fast, and it is absolutly ok that for this very short time CPU works harder

I really hope you are joking. I have already mentioned a scenario where what you say is "absolutely OK" is anything but. Running a "65W" CPU should be completely possible at stock on a 120W PSU, but if it boosts up to 224W, that is likely a very unhappy 120W PSU, at best you'll drastically shorten its life, what I would expect would be the overcurrent protection to kick in, and worst case would be something goes pop somewhere.

I'm getting flashbacks to days of arguing about PMPO vs. RMS for amps/speakers...

[piotrj3]

Maybe you don't mind your CPU cooler sounding like a jet engine reving up for takeoff, but I do. Or maybe you have water cooling, but most don't.

So, what it comes down to is dishonesty by Intel. In order to sustain these peaks that they use to gin up their benchmarks, you need a high-end cooler. Otherwise, it ends up being thermally-limited before it can hit Tau. And it still has plenty of potential to annoy, in the process. And the worst part is that it falls short of the advertised performance!


I'm not objecting to any sort of turbo -- just the extreme degree to which Intel has taken it, in the past few generations! Your argument lacks any sense of scale or degree (not to mention the unheard of 104 degree temp Anandtech's CPU hit, in their test!).

Also, speaking of temperatures, I like reliability in my tech. Most computers I end up keeping around for the better part of a decade. I don't want a CPU that's just going to burn out shortly after the end of its warranty period. And it should be noted that we don't know the long-term implications of this kind of boosting behavior. It's not going to be kind on your CPU fan, I can assure you of that!

To be honest, Anandtech worked out of spec, at this point you were not seeing review of 11700k, but review of 11700K OC-ed by mobo manufacturer which was not provided by name.
They went 270W+ instead of 225W that is max peak power spike Intel can have, and thermal throttling normally for intel happens agressivly at 90C, so 95C i could believe was happening but 105C on stock is kinda hard for me to believe unless they used some really garbage cooler.

[piotrj3]

I really hope you are joking. I have already mentioned a scenario where what you say is "absolutely OK" is anything but. Running a "65W" CPU should be completely possible at stock on a 120W PSU, but if it boosts up to 224W, that is likely a very unhappy 120W PSU, at best you'll drastically shorten its life, what I would expect would be the overcurrent protection to kick in, and worst case would be something goes pop somewhere.

I'm getting flashbacks to days of arguing about PMPO vs. RMS for amps/speakers...

First. Check Intel's definion of TDP : https://ark.intel.com/content/www/us...-5-30-ghz.html

"Thermal Design Power (TDP) represents the average power, in watts, the processor dissipates when operating at Base Frequency with all cores active under an Intel-defined, high-complexity workload. Refer to Datasheet for thermal solution requirements."

Intel is NOT saying 65W CPU is consuming only 65W even during turbo. It is saying it is consuming up to 65W on base frequency, for example

https://ark.intel.com/content/www/us/en/ark/products/212277/intel-core-i5-11500-processor-12m-cache-up-to-4-60-ghz.html

means consumes 65W maximum on all cores used maximum frequency of just 2.7GHz.

Intel does define spec, but doesn't force motherboard manufactures to hold up to it, weak mobos with weak VRMs etc. can turn down turbo limits below intel spec (actually happens often in laptops when manufactures set limits lower to be fine cooling/noise/power wise), and strong VRMs motherboards can go beyond the spec. Reviews almost always take overpowered motherboards which often push limits (and rightfully should as who buys extremly expensive motherboard but cheaps out on PSU/cooling.

Another thing is, we talk about K processor, that intel only allows overclocking on Z motherboards which are supposed to be top of the line.

And btw 11700k has TDP not of 65W but 125W, and during boosting after first 28 seconds, it does hold to it under spec.