Is Intel's "Process Lead" Somewhat a Sham?

[Khato]

TSMC already have taped out test chips on 16FF

Indeed, just like they taped out a 20nm A15 back in October of 2011 - http://arm.com/about/newsroom/arm-a...t-20nm-arm-cortex-a15-multicore-processor.php

So if they're on track for 3 years from tape out of a test ARM core to production for 20nm why exactly should we expect anything different for their '16nm finfet'? Which would put them somewhere in the first half of 2016...

 

[Arachnotronic]

Thanks a lot :thumbsup:

Anytime.

 

[Arachnotronic]

Which would put them somewhere in the first half of 2016...

Right, that's when I expect the first 16 FinFET products in market.

The problem is, Intel is going to screw up Broxton as it has always screwed up every. single. mobile. launch.

This means that the 14nm Broxton will "launch" in 2H 2015, but won't be available in devices until late 2015/early 2016, which is roughly the timeframe that a 16 FinFET Snapdragon is likely to hit the market. And then what? Intel has a slight density advantage, no transistor performance advantage, and a proven track record of SoC architecture slip ups?

You can see why from an investor's standpoint, this really doesn't look all that great. I don't know why Intel didn't do a smartphone version of Cherry Trail for late 2014 launch and early 2015 device availability. This would have given them a real weapon to fight with to start taking share in the market that's driving TSMC's leading edge volumes: smartphones.

 

[NTMBK]

You can see why from an investor's standpoint, this really doesn't look all that great. I don't know why Intel didn't do a smartphone version of Cherry Trail for late 2014 launch and early 2015 device availability. This would have given them a real weapon to fight with to start taking share in the market that's driving TSMC's leading edge volumes: smartphones.

And the irony is that Intel's smartphone SoC which is targeted at high volume markets... is going to be fabbed at TSMC, and drive even more of their volume. *facepalm*

 

[Arachnotronic]

And the irony is that Intel's smartphone SoC which is targeted at high volume markets... is going to be fabbed at TSMC, and drive even more of their volume. *facepalm*

That's correct. However, that will be temporary as Intel does plan to move this to 14nm by the end of 2015/early 2016. While I understand that this was largely due to the XMM 7260 IP being built on TSMC to begin with, I'm surprised that porting it to 22nm was not priority #1 when they acquired Infineon Wireless.

 

[Nothingness]

Indeed, just like they taped out a 20nm A15 back in October of 2011 - http://arm.com/about/newsroom/arm-a...t-20nm-arm-cortex-a15-multicore-processor.php

So if they're on track for 3 years from tape out of a test ARM core to production for 20nm why exactly should we expect anything different for their '16nm finfet'? Which would put them somewhere in the first half of 2016...

Oh I do agree, it was just to say that if they wanted to claim production they could according some say wafer out is enough

 

[NTMBK]

That's correct. However, that will be temporary as Intel does plan to move this to 14nm by the end of 2015/early 2016. While I understand that this was largely due to the XMM 7260 IP being built on TSMC to begin with, I'm surprised that porting it to 22nm was not priority #1 when they acquired Infineon Wireless.

It's interesting that porting a CPU to TSMC was deemed cheaper than porting a modem back to Intel, even with TSMC taking a cut of the profit margins. I wonder whether there is some deeper issue- that Intel can't get its wafer costs low enough to compete in the rock-bottom smartphone market where SoFIA is targeted.

But yeah, it should just be a temporary blip. Still, one of those moments that makes you wonder what on earth is going on with Intel's modems!

 

[bullzz]

@Intel17
"This means that the 14nm Broxton will "launch" in 2H 2015, but won't be available in devices until late 2015/early 2016, which is roughly the timeframe that a 16 FinFET Snapdragon is likely to hit the market"
intel has a execution problem. not performance related. moorefield would be competitive when it comes out but it is late. as long as intel sticks to a schedule, it should catch up on performance. by 2016/17, 10nm should be available. assuming they deliver broxton on time, even it is not competetive, it will have an advantage at 10nm

"I don't know why Intel didn't do a smartphone version of Cherry Trail for late 2014 launch"
considering tablets are their priority, i think they wanted to give tablets and SFF desktops priority. didnt they mention cherry trail is EOY? i would be surprised if they deliver cherry trail on time for tablets

"I'm surprised that porting it to 22nm was not priority #1 when they acquired Infineon Wireless"
again, execution. as long as they are able to put out products on time, they can always figure out porting into their fabs. that is better than pushing out schedule

 

[Idontcare]

IDC,

Thanks for your great posts. I guess the question is, how does TSMC get away with making statements such as the following on its earnings call:

Then I will cover the updates on 16-FinFET plus and I’ve two general offers for customers, 16-FinFET and 16-FinFET plus. 16-FinFET plus offers 15% speed improvement the same total power compared to 16-FinFET . More importantly 16-FinFET plus offers 30% total power reduction at the same speed compared to 16-FinFET. Our 16-FinFET plus matches the highest performance of both 60 nanometer and 40 nanometer technologies in the market today. Compared to our own 20 SoC, 16 16-FinFET plus offers 40% speed improvement.

More importantly, if this is true, how would they have pulled it off and left Intel - which had FinFETs at 22nm - without a process node advantage?

Read the TSMC quote very carefully...they are saying their 16-FF+ matches the highest performance of the best "60nm and 40nm technologies".

Does Intel have a 60nm or 40nm process node? (answer: no they don't, so TSMC is avoiding all comparisons to Intel's xtor performance at 65nm and 45nm)

Is it really an achievement to have your 16nm node best someone else's antiquated 60nm and 40nm nodes? (these would be the other foundries such as UMC and SMIC)

Why don't they boast that it is better than Intel's 22nm process, or for an actual apple-to-apple comparison make claims regarding 16-FF+ versus Samsung's 14nm or Intel's 14nm?

What does 16-FF+ best? In TSMC's own words their 16FF+ only bests TSMC's own 20nm node...well that doesn't sound like much of a ringing endorsement for their own 20nm node then.

But at no point are they drawing any comparisons to Intel's transistors, past present or future...or perhaps more tellingly is that they are intentionally avoiding all such comparisons.

 

[witeken]

@Idontcare:

Why would they compare their 16nm FinFET(+) to old 40 and 60nm transistors? I was reading Intel17's text before reading your comment, and I was wondering about it, and I think it's just a wrong transcription.

 

[lefty2]

@Idontcare:

Why would they compare their 16nm FinFET(+) to old 40 and 60nm transistors? I was reading Intel17's text before reading your comment, and I was wondering about it, and I think it's just a wrong transcription.

That's because the speakers native language isn't English. There's tonnes of mistakes like that in the transcript

 

[NTMBK]

Yeah, comparing to 40nm and 60nm technologies is just crazy. It only makes sense if they actually said "16 nanometer and 14 nanometer", and it was transcribed wrong. Does anyone know where the audio from the concall is available?

 

[AtenRa]

Just heard the audio and i believe he said 16 and 14 nanometers.

Link bellow

http://www.tsmc.com/tsmcdotcom/ExtI...rAndQuarter&year=2014&theQuarter=1&language=E

It is around 16:35

 

[Idontcare]

Just heard the audio and i believe he said 16 and 14 nanometers.

Link bellow

http://www.tsmc.com/tsmcdotcom/ExtI...rAndQuarter&year=2014&theQuarter=1&language=E

It is around 16:35

The audio won't work for me for some reason but it would certainly make more sense if they were referencing 16nm and 14nm...except the claims are still unsubstantiated because TSMC has precisely zero 16nm or 14nm competitor samples to analyze and compare their 16FF+ transistor against.

 

[Arachnotronic]

The audio won't work for me for some reason but it would certainly make more sense if they were referencing 16nm and 14nm...except the claims are still unsubstantiated because TSMC has precisely zero 16nm or 14nm competitor samples to analyze and compare their 16FF+ transistor against.

Intel is a "foundry" now, so I'm sure one of the customers that evaluated Intel's tech reported back to TSMC.

 

[Khato]

Intel is a "foundry" now, so I'm sure one of the customers that evaluated Intel's tech reported back to TSMC.

I wouldn't be so sure about that.

 

[Arachnotronic]

I wouldn't be so sure about that.

How else would TSMC find out?

 

[Ajay]

How else would TSMC find out?

Vendor salesmen will often drop a hint or two about the competition when trying to close a sale in tech industries (not sure if this is verboten in microelectronics or not). They are very clever about it. It was a big shocker to me early in my career.

The other thing is that customers sometimes have early samples that, uhm.., get lost somehow.

 

[Arachnotronic]

Vendor salesmen will often drop a hint or two about the competition when trying to close a sale in tech industries (not sure if this is verboten in microelectronics or not). They are very clever about it. It was a big shocker to me early in my career.

The other thing is that customers sometimes have early samples that, uhm.., get lost somehow.

I suppose that's likely! But the point is that TSMC seemed to have some degree of confidence in their statement(s).

If TSMC really has caught up with a much lower R&D budget and coming from behind, then Intel really screwed the pooch.

 

[Idontcare]

I suppose that's likely! But the point is that TSMC seemed to have some degree of confidence in their statement(s).

If TSMC really has caught up with a much lower R&D budget and coming from behind, then Intel really screwed the pooch.

AMD did it in the late 90's, so I suppose it could happen again.

It would just be so un-TSMC like though, they aren't focused on high performance drive currents the way Intel is.

Never say never I suppose.

 

[kagui]

didnt intel bet the farm on UV tech and it just didnt work?.
also intel lacks the soc on theirs socs no gpu and no modem but companies like qualcim offer

 

[Abwx]

It would just be so un-TSMC like though, they aren't focused on high performance drive currents the way Intel is.

High current drive means high transconductance and low Vth , for thoses unfamiliar with thoses enginering terms transconductance is the current through the device, a FET, for a given command voltage , while Vth is the command voltage at wich the device start to conduct according to a square law, that is , the current increase as a square of the command voltage, below this level the current through the device will decrease almost exponentialy as the command voltage is reduced but conduction will not be zero when the command voltage reach zero volt, a very low current will still exist through the device , this is called leakage and the lower the Vth the higher the leakage.

Now Vth is inherently a compromise, the lower the Vth the higher the possible frequency but also the leakage, the higher, to some extent, the Vth and the lower the leakage and the frequency capability so this second case is the one wich is relevant for low power devices with lower frequencies, that is what we are used in most conventional PCs CPUs but it happen that 2 cores at frequency 1.25X and voltage Y with a 80% software scaling will outperform a single core at frequency 2X and 1.4Y voltage power efficency wise so the pursuit of the highest frequencies/I drive is somewhat at odd with the current (!) industry trends..

 

[Erenhardt]

High current drive means high transconductance and low Vth , for thoses unfamiliar with thoses enginering terms transconductance is the current through the device, a FET, for a given command voltage , while Vth is the command voltage at wich the device start to conduct according to a square law, that is , the current increase as a square of the command voltage, below this level the current through the device will decrease almost exponentialy as the command voltage is reduced but conduction will not be zero when the command voltage reach zero volt, a very low current will still exist through the device , this is called leakage and the lower the Vth the higher the leakage.

Now Vth is inherently a compromise, the lower the Vth the higher the possible frequency but also the leakage, the higher, to some extent, the Vth and the lower the leakage and the frequency capability so this second case is the one wich is relevant for low power devices with lower frequencies, that is what we are used in most conventional PCs CPUs but it happen that 2 cores at frequency 1.25X and voltage Y with a 80% software scaling will outperform a single core at frequency 2X and 1.4Y voltage power efficency wise so the pursuit of the highest frequencies/I drive is somewhat at odd with the current (!) industry trends..

Thank you. I am reading this thread like it was a black magic. Posts like this clear things a little

 

[witeken]

I suppose that's likely! But the point is that TSMC seemed to have some degree of confidence in their statement(s).

If TSMC really has caught up with a much lower R&D budget and coming from behind, then Intel really screwed the pooch.

Extraordinary claims require extraordinary evidence. Not marketing talk or vague slides.

 

[raghu78]

You're a 'victim' of TSMC's/Samsung's/GlobalFoundries' marketing lies. 14nm FinFET is nowhere near Intel's 14nm process. For starters, it's made on the same 20nm process as the foundries' 20nm process, which is 1.5x less dense and much costlier. Even if we take their FinFET+, but we don't know when that will be released, it might be immediately or 1 year after FinFET, Intel's process is still 1.3x denser. Secondly, Intel's 14nm will be their 2nd generation Tri-Gate, compared to TSMC's 1st gen.

Intel's node naming is extremely questionable.

http://www.electronicsweekly.com/mannerisms/general/the-intel-nanometre-2013-02/

Intel 22nm FINFET transistor density was comparable to TSMC 28nm planar. Both used single patterning immersion lithography. btw Intel cannot have a major lead at 14nm as it uses dual pattern immersion litho just like TSMC/Samsung. All three have a M1 metal pitch of 64 nm. Intel's lead if any is marginal.

Yeah, and it's also extremely power hungry. Cyclone is 1.4x faster than 2.4GHz Silvermont, but it consumes ~3x as much energy per core, which means the performance/watt is awful.

Where did you get the 3x power per core number ? do you have proof to back your statement

And you forget that Intel also has Core, which in its turn demolishes Cyclone. So Apple could also ask Intel for a dualcore Broadwell or Skylake.

Cyclone is already on par or better than Ivy bridge on a clock for clock basis

http://browser.primatelabs.com/processor-benchmarks
http://browser.primatelabs.com/ios-benchmarks

Single thread integer

core i3 3217u (1800 mhz) - 1608 (64 bit score)
A7 (1400 Mhz) - 1392
Baytrail z3770 (1460 mhz, 2400 mhz turbo) - 935 (32 bit score)

Multithread performance

core i3 3217u (1800 mhz) - 3370 (64 bit score)
A7 (1400 Mhz) - 2519
Baytrail z3770 (1460 mhz, 2400 mhz turbo) - 2967 (32 bit score)

1.8 Ghz core i3 3217u (ivybridge) has a 15% higher single thread performance than 1.4 Ghz A7 (Cyclone) while running at a 28% higher clock speed . Multithread performance is 33% higher on 1.8 Ghz ivybridge primarily because ivybridge has Hyperthreading and can support 4 threads while Cyclone can only support 2 threads. Also the A7 is definitely a lower TDP part than core i3 3217u which is a 17w SKU

http://ark.intel.com/products/65697/Intel-Core-i3-3217U-Processor-3M-Cache-1_80-GHz

Your 20$ estimate is way too optimistic. And comparing the price of a high-end chip too a slow ARM one doesn't make much sense either.

there is a teardown by IHS. read it. Comparing the A7 with core i3 makes sense when they are performing similar. A8 and A9 are going to be even more impressive.

http://press.ihs.com/press-release/...sts-less-make-third-generation-ipad-model-ihs

http://press.ihs.com/sites/ihs.news.../file/Preliminary_iPad_Air_Cost_Estimates.PNG

I don't think so. Intel's at a comfortable 2 year lead, which will increase to 3 year at 14nm (TSMC's 10nm is the equivalent of Intel's 14nm) and further expand at 10nm.

Keep believing what you want to. Your illusion will be shattered in late 2015 when Broxton goes up against the Apple A9 built at either TSMC 16FF+ or Samsung 14 FINFET.