Intel’s analog shrink is for real
MOAR smaller
Nov 22, 2011 in Chips, Microprocessors, Mobile
We have previously reported that Intel (NASDAQ:INTC) was planning an aggressive shrink of their analog designs and now the first project seems to be in the pipeline.
According to the advance program from ISSCC Intel will present news about a dual core Atom processor with integrated Wi-Fi. This is, as far as we know, the first time that analog has been shrunk below 65nm as the Atom processor will actually be fabbed in 32nm.
Shrinking the analog parts has high priority at Intel as it will allow, in the future, for complete chipset integration in their processors thereby saving both design time, space and power.
As it stands right now analog designs are several generations behind digital designs, but Intel thinks that they can continue the aggressive analog shrink for several more generations. This aggressive scaling will for sure give both AMD and ARM a run for their money.S|A
Gentlemen,
I am not a cmos/engineer or any type of chip designer, however if my memory serves me correctly,
The original netbook designer was AMD whom offered its low power “broadcom SOC” chipset for the “One laptop for child project” that intel jumped on afterwards with ASUS. It was when Intel saw the potential of selling a slightly updated PIII, shrunk down a bit to peddle off as some new design, and wonderful offer for folks that wanted a laptop, but couldnt afford them. All while selling a limited computer, low memory, barebones operating system, low res screen, tolerable battery power. Actually, if AMD’s new head is so enthralled by the tablet, smartphone, netbook market, he should update the hell out of bobcat and push that out as SOC. Its much more efficient, and advanced than the atom.
Many words, oh. Stupid boboo not want many words read. boboo would explain to dave but boboo still learn how read some letters.
This was a reply to
http://semiaccurate.com/2011/11/22/intel’s-analog-shrink-is-for-real/comment-page-1/#comment-13344
Well, someone explained earlier how moronic it is to suggest ‘digital’ chips lack any analogue circuits, but this reminds us all that this site is proud to comment on things with no actual technical understanding of the subject whatsoever.
What shames commentators like these is that the age of the Internet is the age of the world’s best and most accessible library, allowing anyone prepared to make an effort to improve their understanding on any subject. Why is it, therefore, that too many ‘blog’-like sites are run by people who would clearly rather do anything other than spend a few hours in honest research.
Ignorance as a result of pure laziness can never be forgiven. And how many articles does this site publish a week? And by how many different people? Obviously some people think a work-day is excessive if longer than one hour.
Intel’s Atom is horrible, and worse, has played a massive role in getting the wider industry behind ARM in higher (computing) power devices. Save for going dual-core, the Atom has seen no significant architectural improvements whatsoever. The chipset support for the Atom is vastly worse, from a technical POV, and has been the Atom’s greatest shortcoming.
Clearly the market wants (or did want once upon a time) not Atoms, but ULV x86 chips from Intel with low prices, and at least a core2 architecture. All new Atom designs can to at this time is help sell even more ARM chips. The new markets do not even want good x86 chips, let alone the worst x86 chip ever made, the Atom.
Even for the slow witted, who believe every pro-Intel shill article blogged for cash, it has to be obvious that Microsoft’s adoption of ARM for Windows8 meant something massive.
Intel is notorious for keeping small engineering teams active in connection with long cancelled or dead-ended projects, like Netburst, Itanium, Larrabee or Atom. This is a business strategy designed to disguise the project failure rate at Intel. It also allows Intel to continue getting some sales for dead products (just consider Itanium).
Take Larrabee. The chip actually failed almost immediately at Intel, but for several years afterwards, sites like this were telling you that Intel would destroy AMD and Nvidia in the GPU (real, as against broken integrated) marketplace. And on what basis? That Intel had loads more money to spend! And that ‘ray-tracing’ (which actually Larrabee is horrible at as well) was the future.
30 mins of research would have demonstrated that being very wealthy is no indication that a company can successfully move into new product areas, and that ray-tracing is the worst approach to general real-time rendering output imaginable.
Sites like this assume their readers are very stupid and uninformed- too lazy to do their own research, and too incapable of understanding meaning if they did try.
One last thing. All digital circuits in electronics are ‘analogue’ (it is the function that is ‘digital’, not the components used). And all analogue components, in their turn, are ‘digital’ (because of the quantum nature of reality that means that only certain discrete states exist).
Respect the truth, not the stunt wording used in cheap PR.
tl;dr – people come here for the opinions, not accurate facts, hence, Semi Accurate.
whoa yay, I thought you would never get to the point, but you made it. And made it well…
Dood, that is wa-ay to much wordage.
There’s a difference between analog circuits and digital circuits… that’s why we differentiate the two. Analog circuits operate under continuously variable voltages and currents, digital circuits operate under discrete values. There’s a physical difference in the hardware that makes up the circuit.
Nobody is claiming that digital and analog circuits aren’t similar; however they’re functionally and physically two entirely different classes of circuits. I think you’ve completely missed the point of differentiating them. I don’t think anyone would argue that vacuum tubes and transistors are the same thing.
So you do understand analog a little bit. But, to be honest, your analog commentary is entry level engineering.
Maybe you were trying to dumb it down so everyone could understand, but I’m not yet fully convinced you have a solid understanding of analog.
Entry level? So beyond your grasp, obviously.
For instance, the fact that you fail to specify circuits when talking about “analog” demonstrates to me that you don’t even understand the difference between “analog” and “analog circuits”.
I do CMOS analog circuit design for a living. How about you?
I’m a rock star.
Which band? Or do you have a solo career?
Right over your head, huh?
Just giving you the benefit of the doubt.. There was a 95% chance that you were being a sarcastic a$$hole, but because of that 5% I decided not to jump to a conclusion, and instead asked you a honest question.
It’s disappointing that you did turn out to be a sarcastic a$$hole. For what it’s worth, I actually am a CMOS analog circuit designer – that was NOT a lie or a joke.
Everything I’ve ever said on the internet is absolute fact. I don’t know what you’re talking about! I’m a CMOS analog circuit designer, too!
Ok. Let’s test that a bit, just so we know for sure.
How to determine the IIP3 of an amplifier?
Man, I can’t believe it. Every time I think “Jesus Christ, he’s posted something even DUMBER than before”, you go and do it again. You just keep setting the bar lower and lower.
This is getting boring. Obviously I don’t need to contribute here. You should just keep replying to yourself with more stupid comments.
I don’t even know what the hell you are after. In your first post you babble something basic about analog and digital, and after people point out you are incorrect you start the personal attacks.
You know very little about analog design, so it would serve you well to not say things like “A transistor is a transistor” and “It’s not like they’re using larger transistors because they have to or anything”
And my favorite: “I think what people are failing to understand here is the fact that digital circuits are easier to make because they operate under discrete voltages and currents.”
You try to make yourself sound like an expert when you don’t really understand this yourself. Discrete currents? You don’t see those in digital circuits unless you consider DACs to be digital circuits.
+1 and topped ROFL :)
Yeah, except I was being truthful.
I’m placing this comment here because I’m tired of the indentation.
I called you dumb because you take absurd internet claims seriously. I’m sorry for assuming you are a moron simply because you responded with a moronic reply. I realize that sarcasm is hard to read through text, but I would have thought the sheer absurdity would have made it extremely obvious.
If you’re actually interested in discussing things, you should be more explicit and less ambiguous. For instance, “your analog commentary is entry level engineering” doesn’t add anything at all to the conversation (other than general abrasiveness).
Judging from prior posts, however, I’m inclined to believe you just want to impose your aspirations on superiority onto others. If that’s the case, let me know and I’ll refrain from ever responding to you again.
Ok, yeah, fair comment. Sorry about being a jerk.
Let’s try a clean debate some time in the future.
Other than for the fact that Atom is pretty much single-handedly responsible for the netbook phenomenon and was the lowest TDP chip for its time that could actually offer passable performance (which is not something you would have said about the 600MHz single core ARM that was available back then in 2008), you’re right :)
Negatory!
ASUS started the netbook phenomenon using a Celeron CPU.
Somewhat after that (many months) Intel introduced the Atom and, ahem, design guidelines for netbooks.
Intel’s sole purpose in doing so was to castrate the new form factor & price point in order to protect the higher margin thin & light market.
With all this talk of integrating parts to save money and space, why doesn’t someone integrate a CPU, GPU, memory and a hard drive into a single chip. Because it will be so small it will save on power too, and then you will only need whip it out of your pocket to connect to a monitor. Where can I get a patent for this?
Von Drashek, who never speaks garbage.
Because the chip die would be so huge that it would cost you an eye if it fails to work for some reasons. And having so much transistors on it it has many many reasons to stop working.
If you just mean putting different dies on the same chip (with different buses), it would require some more hundreds pins below.
All-in-one is not always good.
BTW: in the ’70-ies IBM made a revolution by introducing a computer that could be upgraded in pieces with standard parts.
Now, we are returning to the point were everything is inside a single component which is not upgradeable? I can accept that only for embedded devices, not for general purpose ones, like PCs are.
If you can do all that, wouldn’t you be able to integrate a wireless display feature as well, so you don’t have to use a cable to connect to a monitor?
drashek I Vs. von drashek 0.
ivy out much sooner than expected, already on test bench
22nm Quad Core B3 stepping processor running clock speeds of 2.0GHz at 1.056V
notice low power.
drashek Iologist….
If you want it all on one die it’s not going to be cost effective for the foreseeable future. If you accept multiple connected dies (possibly stacked using TSV) it shouldn’t be that hard to do, but storage read/write will be relatively slow. When non-volatile RAM (memristors?) become an affordable competitive option in the main stream this would seem the natural outcome. You’ll probably still want additional mass storage, but a few gig (say between 4 and 128) for main memory and storage would still allow a great deal of functionality without extra storage. For real world use I’m not sure how different that would be from something like Atrix. It will still need a case and battery, and it would benefit from some form of wireless and at least one physical port; add a screen, mic, and speaker (or just Bluetooth) and it’s a phone.
Well why don’t they also integrate the Intel GbE of the sandy bridge architecture into the 22nm Ivy bridge? Wouldn’t that also lower costs and power consumption? Besides, I believe that the Sandy bridge-E could have rather been released directly on 22nm, as Intel has had working samples of Ivy already for a (relatively) long time. The process was mature, and the Sandy bridge design on 32nm seemed to be working well.
No offence about Sandy bridge, which I have myself, but they could have done these things to benefit consumers and not just end of next year release 22nm Sandy bridge-E.
Possibly, but analog will be challenged by increasing quantum behaviour as currents and voltages become very small
Yeah, because there is no analog at all in all those PLL’s, PCI-Express PHY’s, Memory I/O’s, etc. that’s been shipping out on CPU’s at 45nm, 32nm, and soon to be 22n.
Check your facts on what constitutes analog circuitry before you report something as news.
Be nice – not everyone can understand everything about engineering.
Although, Mads has reported on these “analog breakthroughs” before, and was thoroughly blasted in the comments for not knowing what analog is. I’m a little surprised he didn’t fact check back then..
Like you, for instance.
Well thanks, I’m flattered – I admit I do understand a lot about engineering, but saying I understand “everything” would be exaggerating a tiny bit..
But thanks for the vote of confidence!
I don’t understand. Obviously I was insinuating that you DON’T know everything. Like “what analog is”.
Since you say that Mads doesn’t know, then I’d love to hear you explain why you think that is the case.
Why? Because in the comments on his last “Intel analog breakthrough” the same thing was mentioned – those CPUs have had a ton of analog circuits in them already (PLLs, I/O PHYs etc.)
From people like you, who don’t know what they’re talking about. Who cares? You still haven’t explained anything.
“What I said is right because I said it previously.” You just keep getting dumber and dumber!
I didn’t say it last time, it was someone with a screen name AnalogKid. Try to focus.
Not once you have demonstrated that you know what sort of issues analog design might have, particularly at deep-submicron process nodes.
Please hold the ad hominem and focus on facts, knowledge and true debate, ok?
“Not once you have demonstrated that you know what sort of issues analog design might have…”
That’s nice. I’m not the one making the claim, you are. If you think that Mads is wrong, then you need to explain why. You’ve failed to do so. Nice debate skills, dude.
IbleedOrange already listed the reasons. I agreed with him.
I’ll summarize for you: analog circuits such as PLLs, serial I/Os, memory I/Os, bandgap references etc. have been designed on 45nm, 40nm and 32nm, and have been included in products such as CPUs, GPUs and APUs.
Mads said in this article that “This is, as far as we know, the first time that analog has been shrunk below 65nm as the Atom processor will actually be fabbed in 32nm.”
As far as he knows, this is the first time, but it’s actually not. He was called on this the last time he reported about Intel’s analog breakthroughs (by AnalogKid), but Mads didn’t follow up, and still doesn’t know that <65nm analog is commonplace.
Analog does not necessarily shrink. Actually, most transistors in analog circuitry are pretty large, well above the minimum size rules. Since there’s relatively few of them, it’s not a big deal.
The problem usually is porting a circuit design to work on a new tech node, with different electrical parameters – gains, threshold voltages, leakage currents, etc.
And I’m guessing some analog has already been fabbed under 65nm – if nothing else, some humble USB PHYs …
A transistor is a transistor. The size depends on packaging constraints and cost. It’s not like they’re using larger transistors because they have to or anything.
I think what people are failing to understand here is the fact that digital circuits are easier to make because they operate under discrete voltages and currents. With an analog circuit, those values become continuously variable, which makes them harder to design.
I agree on you bro =)
Larger transistors => less variation.
Smaller transistors => more variation.
More variation => poor transistor matching.
Poor transistor matching => analog performance -> toilet.
There is a reason why these guys use large transistors, and why it’s difficult to do analog with small transistors.
Random statement => useless assertion.
Meaningless post => waste of time.
Internet claims => no credibility.
Find Diamond => Store in Chest
Find Creeper => Run Away
Creeper ‘splodes Chest => No Diamond.
Well played, sir.
Your debate skills are just awful. Do you just always ignore everything somebody said and respond with a personal attack?
Everything I said there was true and factual. And if you actually understood analog circuit design, you would know this.
Crawl back under your bridge, troll.
I ignore stupid posts. You should have figured that out by now.
EVERY SINGLE TIME YOU POST, YOU MAKE YOURSELF SEEM DUMBER. The above is no exception. Keep it up.
“Ignores stupid posts” , makes unclever ad hominem attacks. Doublethink is strong with this one.
You’re an idiot. An ad hominem is a logical fallacy, not an insult.
“ad hominem” is BOTH a logical fallacy, AND an insult.
do you realise all that applies to your own post?
BTW, NeelyCam is right, in case you did not know that already. Try to make a differential amplifier with poorly matched transistors and you will see the result.
Regards,
Jose
My reply had nothing to do with what he said, but with how he said it. If he was actually interested in meaningful debate, he would take the time to construct complete sentences. But that doesn’t seem to be why he’s posting. It seems as though he just wants to argue with people and spout nonsense about how great Intel is.
Then let him have his moment.
I have owned AMD CPUs for the last 10 years, since the 5K86-P90, currently own a Phenom II 955, but Intel has now the upper hand.
Just wait until AMD releases something better, then you will have your moment. Insulting people will take you nowhere, although it feels good :)
Adam Nov 23, 2011 at 11:46 am #
“A transistor is a transistor. The size depends on packaging constraints and cost. It’s not like they’re using larger transistors because they have to or anything.”
Actually they do. The analog properties of a transistor are highly dependent on its size. Using minimum size digital transistors would lead to amplifiers with no gain, no power handling capability, high noise etc. etc.
A lot of analog IC design is in fact deciding what size of transistor to use and how to lay it out on the die.
This.
Niilo understands analog design. Adam does not.
Thanks for actually making a coherent, informative response. Other people could learn something from you.