Intel 14nm CPUs in short supply, claims report

[LemmingOverlrd]

Why it matters: When it rains, it pours. Intel is fending off investors and partners on 10nm because of poor yields, and now, it seems, it has irked Taiwanese partners about the share of 14nm CPUs coming their way. Partners are publicly complaining about it, trying to drum up some more inventory from Intel.

Taiwan's Digitimes has said that Intel is coming up short on the chip supply front, but this time on 14nm, and that the recent release of Whiskey Lake and Amber Lake CPUs has not helped.

As one of the primary go-to spots for ODM and OEM manufacturing in the tech industry, Taiwan is a good place to take the pulse of the CPU industry. With Intel's upcoming launch of "9th generation" CPUs, some partners in Taiwan feel short-chagned and have become quite public with warnings of doom for Intel.

Quoting sources from Acer and Compal Electronics, the latter one of the biggest ODMs who does a lot of manufacturing of notebooks for HP, Dell and Toshiba, the article states that Intel is unable to supply enough CPUs to go around. While the article is not specific about which CPUs are supply-constrained, it is clear about these being 14nm SKUs, and hints that this may be related to newly-released Amber Lake and Whiskey Lake CPUs used in their own soon-to-be released devices.

Just days ago, Intel introduced its latest low-power and ultra-low power CPU designs, Whiskey Lake and Amber Lake, only to have its Taiwan partners complain about supply

Acer CEO Jason Chen weighed in on the matter, stating that while “the global PC market is expected to pick up in the second half of 2018, the tight supply of Intel's 14 nm processors will pose a significant challenge to the supply chain management capability of brand vendors.”

Compal Electronics, by way of its President, CP Wong, added that Intel’s (in)ability to supply CPUs would eclipse even the US-China trade dispute.

The report also seems to confirm what we’ve been hearing through the grapevine: that Apple’s upcoming lower-cost notebook will be powered by one of Intel's brand new processors. Despite these low- and ultra-low power CPUs, not having even hit the market, they are already bringing problems to Intel’s doorstep. This ties in to the fact that Compal's direct competitor, Quanta Computers, is in charge of manufacturing the MacBooks for Apple, and is expected to see a revenue surge next quarter which will eclipse Compal.

Despite this being a rude wake-up call from Intel partners in the Far East - which should be taken with a grain of salt - what is obvious is that Intel seems to prolong a multi-pronged strategy when it comes to market segmentation. With this it currently maintains seven CPU families based on a similar number of microarchitectures, despite consumers still having access to previous generation stock at major etailers. AMD, on the other hand, have designed a successful reuse strategy with its Zen architecture, using a single design to power all the possible SKU variations of their CPUs.

Permalink to story.

https://www.techspot.com/news/76241-intel-14nm-cpus-short-supply-claims-report.html

 

[Outlander04]

Good sign for the industry that demand is there

 

[Evernessince]

Good sign for the industry that demand is there

I'd recommend you not read one step past what is explicitly stated. Supply issues could be due to demand, yield issues, or Intel biting off more then it can chew at once.

In any case the only thing we do know is that Taiwanese partners are not getting enough, that's it.

 

[koblongata]

Looking at the specs those 14nm parts are insanely clocked, 5W TDP for 4.2Ghz kind of stuff, so fast all of them are like overclockers' dreams. Intel really pushed the limit of 14nm with this generation.

 

[Evernessince]

Looking at the specs those 14nm parts are insanely clocked, 5W TDP for 4.2Ghz kind of stuff, so fast all of them are like overclockers' dreams. Intel really pushed the limit of 14nm with this generation.

Intel's TDP numbers are at base clock only. The model you are referring to has a base clock of 1.1 GHz and boosts to 3.4. Given that power consumption and TDP scale exponentially, expect a much higher power consumption at the boost clock and/or a very short boost clock duration.

The wide range in base/boost clock should give you the immediate indication that these are intended for mobile devices like ultrabooks and tablets. These devices have to meet a wide range of performance and TDP targets. Of course, Intel themselves are touting these for portables so that should have been clear for the get go.

These are more the antithesis of an overclocker's dream. Power throttling, short boost duration, and they are locked.

 

[cliffordcooley]

Intel's TDP numbers are at base clock only. The model you are referring to has a base clock of 1.1 GHz and boosts to 3.4.

I've never really thought about it. That does make sense. Do they though have a hard wattage limit to throttle them?

 

[EEatGDL]

Intel's TDP numbers are at base clock only. The model you are referring to has a base clock of 1.1 GHz and boosts to 3.4.

I've never really thought about it. That does make sense. Do they though have a hard wattage limit to throttle them?

I think it was 1.5 x TDP for long periods of time, but I can't quote that info, once read it in 01.org. But for small bursts it could go >= 2 x TDP (don't remember the exact limit for burst).

 

[Evernessince]

I've never really thought about it. That does make sense. Do they though have a hard wattage limit to throttle them?

Depending on the SKU purchased, the processors will limit frequency / power consumption based on internal sensor readings like temperature. Many of the mobile processors will only boost to max clock for 5 seconds at most. I'm not sure about a hard wattage limit built in but what I do know for sure is that OEMs always build their systems with the TDP and wattage Intel provides. This means that even without a hard limit set by Intel, the motherboard will throttle the processor to stay within Intel's specifications.

I think it was 1.5 x TDP for long periods of time, but I can't quote that info, once read it in 01.org. But for small bursts it could go >= 2 x TDP (don't remember the exact limit for burst).

It varies processor to processor. Each mobile Intel processor has a built in Turbo table that tells it what it can boost to, for how long, ect. The amount is going to vary on the exact config and if the mobile device can actually provide that much power.

 

[Sausagemeat]

Intel chips are the best overclocking chips around at the moment. The unlocked parts can reap some real rewards if you put want to put the time and effort into building the system sufficiently and squeezing the most out of your CPU. They may use a bit of power to do so but they can do it quite well, some chips even get to 5.3ghz!

 

[hahahanoobs]

Too many 14nm SKU's?

 

[misor]

"short supply' ...right...just like what apple practices....despite the massive sales.. (on both apple and intel)

 

[CaptainTom]

This should surprise no one. It's abundantly clear that CoffeeLake/KabyLake-R are just Skylake CPU's with incredible binning. That's literally the only way intel can release 15w quad-cores to keep up with Ryzen in laptops.

I just bought a laptop with an i7-8550U, and this thing is amazing after you undervolt it. I cannot believe I have a 2GHz 8-thread chip that only uses 7w while gaming!!! But I am aware they can't make a lot of these...

 

[Outlander04]

I'd recommend you not read one step past what is explicitly stated. Supply issues could be due to demand, yield issues, or Intel biting off more then it can chew at once.

In any case the only thing we do know is that Taiwanese partners are not getting enough, that's it.

Demand is demand . If you were running a business you'd want people to be looking forward to owning your product too .

 

[LemmingOverlrd]

Intel's TDP numbers are at base clock only. The model you are referring to has a base clock of 1.1 GHz and boosts to 3.4.

I've never really thought about it. That does make sense. Do they though have a hard wattage limit to throttle them?

I think it was 1.5 x TDP for long periods of time, but I can't quote that info, once read it in 01.org. But for small bursts it could go >= 2 x TDP (don't remember the exact limit for burst).

Here's a video https://goo.gl/3v2jGP of how the throttling happens. In this specific case it's an Apollo Lake SoC (also a 14nm chip with a low SDP). It throttles really quick.

Using a little readwriteeverything magic, the author removed the cap, but then had to account for the extra heat. It's a good video that clearly shows what's happening with the CPU.

 

[fktech]

So are prices going up?

 

[oldtink]

This should surprise no one. It's abundantly clear that CoffeeLake/KabyLake-R are just Skylake CPU's with incredible binning. That's literally the only way intel can release 15w quad-cores to keep up with Ryzen in laptops.

I just bought a laptop with an i7-8550U, and this thing is amazing after you undervolt it. I cannot believe I have a 2GHz 8-thread chip that only uses 7w while gaming!!! But I am aware they can't make a lot of these...

Thermal throttling happens very quickly with U-series chips (especially in thin & lights). Run Prime95 then run it a second time immediately after the first. Post the two "amazing" stress test scores. I am curious to see just how much different they are.

 

[CaptainTom]

Thermal throttling happens very quickly with U-series chips (especially in thin & lights). Run Prime95 then run it a second time immediately after the first. Post the two "amazing" stress test scores. I am curious to see just how much different they are.

Well first you should learn to read with your "amazing" eyes. My chip is undervolted, and it never throttles. It holds 2GHz no problem at 70c in a 13" netbook smaller than a MacBook Air (With IBT). I can even run it at 3GHz+ for extended periods.

Learn to overclock, it pays dividends

 

[oldtink]

Assuming that your two PRIME95 "Stress Test" scores would be the same then?

Looking at this review:
https://www.techspot.com/review/1500-intel-8th-gen-core-quad-core-ultrabooks/page5.html
The i7-8550U throttles and benchmarks a sustained clock of about 2.3 - 2.7ghz, providing around 30% less performance than a i7-7700HQ. You claim to get somewhere near 3.0ghz. What are the numbers?

Better yet, what scores do you get for x264 encoding test?

 

[CaptainTom]

Assuming that your two PRIME95 "Stress Test" scores would be the same then?

Looking at this review:
https://www.techspot.com/review/1500-intel-8th-gen-core-quad-core-ultrabooks/page5.html
The i7-8550U throttles and benchmarks a sustained clock of about 2.3 - 2.7ghz, providing around 30% less performance than a i7-7700HQ. You claim to get somewhere near 3.0ghz. What are the numbers?

Better yet, what scores do you get for x264 encoding test?

Hmmm. Do you not understand what undervolting is?

If a stock one gets ~2.3-2.7, then one running -90mV can very easily hold 3GHz. However I prefer to just lock it at 2GHz to allow more thermal room for my MX150, and also I find the framerates feel smoother when clocks are locked.

 

[oldtink]

Yes, "over-clocking and under-volting" is really just balancing voltage/heat dissipation to maximize some performance metric. Intel "warps" SKU's to provide a range of TDP's for different use cases. Thin and lights are particularly challenging given that they still are subject to the same physics as any computer but with more limiting constraints on current draw and thermal dissipation solutions. In order to get to a 15 watt TDP, Intel defines a set of parameters (base clock, base/maximum voltage, single and multi core turbo clocks) and the OEM manufactures try to balance "battery life" and "performance" and "thermal heat dissipation" to create a viable product. My first over-clocking experience was with an Intel PIII Pentium in the early 2000's, my first under-clock was with a AMD Turion x64 a few years later. Things have change quite a bit since then. All of this neither here nor there.

I have used an i5-xxxxU (don't remember the exact model) thin and light (it was provided, I had no "control" over it) and found it to be mostly a "dog" (though most dogs would probably be insulted). Under any kind of load, particularly multi-tasking, it would thermal throttle so bad. I see this review (https://www.amazon.com/Dell-Inspiro...r&reviewerType=all_reviews#reviews-filter-bar) of an i7-8550U Dell laptop where several negative comments say it is a "dog", presumably because it thermal throttles under load.

My main question is still, what does "amazing" mean? What kind of performance "bump" (percentage) have you gotten by "under-volting" in order to boost the length of time the processor is able to increase the clock while living under the voltage/heat dissipation constraints? A significant "bump" would suggest to me that Intel has "tightened" up its "performance governor" (rather than just provide more cores) probably due to recent competition from AMD. More importantly, does a performance metric improve (x264 for instance)? Does your particular laptop model? allow for a sizable increase in performance over stock settings without "burning my lap"? Could you provide a number rather than a subjective review?

I took this particular opportunity because I though you might be knowledgeable and be able to answer my specific question. I am primarily interested in more conventional workloads. I agree, gaming performance is a bit of a different beast with more things to consider.

 

[cliffordcooley]

My main question is still, what does "amazing" mean? What kind of performance "bump" (percentage) have you gotten by "under-volting" in order to boost the length of time the processor is able to increase the clock while living under the voltage/heat dissipation constraints?

I think you are asking for a response that is just not available. I'm fairly certain CaptainTom used the word "Amazing" in reference to how much better the machine runs that it would have with default voltages.

I think you are asking. What is so "Amazing" about this under-volt setup, than a comparable under-volt setup. The answer to that question is likely nothing.

 

[oldtink]

After searching for numbers, I found this:
https://www.notebookcheck.net/More-...d-X1-Carbon-2018-ThinkPad-T480s.296752.0.html

They "tuned" an i7-8550U and got either a 8% increase in performance (Cinebench) or a 12 degrees C reduction in CPU temperature over an extended test period. So "Amazing" equals about 8% on a ThinkPad X1 Carbon.

or this using a i7-8650U:
https://medium.com/@n4ru/45w-performance-from-15w-kaby-lake-r-d8d5e7ea4fad.

He concludes that the i7-8650U is just a "HQ part" with aggressive throttling provided by the OEM Lenovo. This is also at the heart of my original question, are these i7-XXXXU parts just "crippled" HQ type parts (ie to some degree, just marketing games by Intel)? In Thin and Light designs, there is probably no hope of getting much more performance out of them by "tunning" though 8% is something (amazing?).

Just always a little skeptical because I know "performance isn't cheap" but sometimes it is "crippled by the OEM".

Answered my own question. Thanks.

 

[CaptainTom]

Yes, "over-clocking and under-volting" is really just balancing voltage/heat dissipation to maximize some performance metric. Intel "warps" SKU's to provide a range of TDP's for different use cases. Thin and lights are particularly challenging given that they still are subject to the same physics as any computer but with more limiting constraints on current draw and thermal dissipation solutions. In order to get to a 15 watt TDP, Intel defines a set of parameters (base clock, base/maximum voltage, single and multi core turbo clocks) and the OEM manufactures try to balance "battery life" and "performance" and "thermal heat dissipation" to create a viable product. My first over-clocking experience was with an Intel PIII Pentium in the early 2000's, my first under-clock was with a AMD Turion x64 a few years later. Things have change quite a bit since then. All of this neither here nor there.

I have used an i5-xxxxU (don't remember the exact model) thin and light (it was provided, I had no "control" over it) and found it to be mostly a "dog" (though most dogs would probably be insulted). Under any kind of load, particularly multi-tasking, it would thermal throttle so bad. I see this review (https://www.amazon.com/Dell-Inspiro...r&reviewerType=all_reviews#reviews-filter-bar) of an i7-8550U Dell laptop where several negative comments say it is a "dog", presumably because it thermal throttles under load.

My main question is still, what does "amazing" mean? What kind of performance "bump" (percentage) have you gotten by "under-volting" in order to boost the length of time the processor is able to increase the clock while living under the voltage/heat dissipation constraints? A significant "bump" would suggest to me that Intel has "tightened" up its "performance governor" (rather than just provide more cores) probably due to recent competition from AMD. More importantly, does a performance metric improve (x264 for instance)? Does your particular laptop model? allow for a sizable increase in performance over stock settings without "burning my lap"? Could you provide a number rather than a subjective review?

I took this particular opportunity because I though you might be knowledgeable and be able to answer my specific question. I am primarily interested in more conventional workloads. I agree, gaming performance is a bit of a different beast with more things to consider.

Then don't buy that DELL lol. I have an Envy 13t (My favorite laptop I have ever owned), and it actually keeps the CPU below 80c (Most laptops let them get to 95c!). I will say it does eventually thermal throttle under IBT/Prime95, but it takes a while. When I haven't locked the frequency, I lock the powerusage to 16w. At 16w it stays between 2.5GHz and 3.3GHz.

It can actually run at 4GHz all cores for ~30 seconds at 22w, but then throttles pretty quickly. Again keep in mind I use this to game while on business trips, and so I see no point in letting it go above 2GHz while gaming. At that speed my MX150 stays at 1200MHz and can play AAA games at medium-high settings in 900p no problem. Remarkable for a netbook only pulling ~25w total while gaming!

 

[oldtink]

Thanks for the info and testimonial, asking a tech sales guy at any brick and mortar is futile. I am thinking about upgrading and have been reluctant to buy a thin and light because of the negative experience in the past with U-series chips. I have an aging HP Probook that I have been pleased with, I mostly do productivity and development/IT stuff. It isn't huge but it is nice not to have to carry any extra weight when you do a lot of "mobile computing".

Under-volting and locking cores at reasonably maintainable speeds certainly would smooth out the performance particularly for gaming. Just looking at one of the graphs in the above link show how much fluctuation occurs after the thermal throttling start to interfere.

 

[CaptainTom]

Thanks for the info and testimonial, asking a tech sales guy at any brick and mortar is futile. I am thinking about upgrading and have been reluctant to buy a thin and light because of the negative experience in the past with U-series chips. I have an aging HP Probook that I have been pleased with, I mostly do productivity and development/IT stuff. It isn't huge but it is nice not to have to carry any extra weight when you do a lot of "mobile computing".

Under-volting and locking cores at reasonably maintainable speeds certainly would smooth out the performance particularly for gaming. Just looking at one of the graphs in the above link show how much fluctuation occurs after the thermal throttling start to interfere.

I have the 8w version of the MX150 (Gt1030) too. At stock settings it power throttled so bad intel graphics were almost better. But I did figure out a way to trick it into undervoling. Been meaning to upload a video that explains how...