Recent breathless reports have come out about Altera dumping Intel as a foundry and going back to TSMC. If any of the initial analysts had even tried to analyze the reams of publicly available information on this topic, this debacle and the resultant echo chamber would have never happened.
Lets start this out by saying that the majority of the tech-oriented financial analyst community seems to have trouble finding two live brain cells to keep each other company. There are some that have a clue, and a few that are very sharp, but the majority are either paid off completely or rank idiots. Apologies to the actual no-analyst idiots out there if you feel affronted but it needed to be said.
Worse yet you get an echo chamber of sites like Seeking Alpha, Forbes, and Motley Fool who will publish things from any idiot out there with a grammar checker. These bottom dwellers of the financial web sites are funded simply on hits, they will put up anything from nearly anyone, and pay by the view. The more inflammatory the headline, the more hits, and the more money both sides make. In short they don’t seem to give a rats ass about accuracy, content, or simple logic, it is all about hits.
Worse yet there are a lot of these echoes, and each one tries to hire the most prolific idiots they can. News feeds like Yahoo and Google finance love them too for the same revenue per view reasons leading to an exponential growth of low IQ pablum. This is the sad state of modern tech-oriented financial ‘news’, idiots feeding louder idiots, feeding more idiots. And no one calls them on it, probably because no one in the chain has the brainpower to actually do so, much less the motivation.
Last month it was Intel’s so called delay of their 14nm process and the Broadwell chip made on it. We pointed out that the CEO of Intel had directly said it was delayed by a quarter in an analyst call a month prior, something that most of these echo chambers of rank stupidity reported themselves. For some reason, prior to running with the headline about the 14nm sky falling, they didn’t actually bother to do the most basic research. By that we mean looking at their own site much less calling any company involved. Neither the authors, editors, or anyone else did the barest minimum research, if they did it would not have been such a prolific public display of ignorance.
Worse yet, none of them seem to have the barest technical knowledge about the things they are pontificating about or they would have done the same 30 second back of the envelope calculations SemiAccurate did and realized yet another way that there is no story. In short this isn’t journalism, it s something between rank sensationalism and lying for profit. OK, maybe lying is too strong a phrase, it presumes that these nitwits have enough logical capability to construct a fabrication, something that may be too high a bar.
Step forward to this week, and we have a story in Digitimes that says JJ Park of JP Morgan Securities is claiming Altera is annoyed at Intel’s decision to delay their 14nm process until the 4th quarter of 2014. As a result it is dumping Intel’s 14nm process and moving to TSMC’s 20nm process. If this is true, it should be really bad news for Intel but really good news for TSMC. Only one of the echo chamber residents bothered to so much as check with Intel, and they denied it completely. From what SemiAccurate has seen, no one bothered to ask Altera about it at all.
Why is this such rank idiocy and ignorant misinformation? If any of these people had bothered to do the basic research, they would have realized that there is no story here, it is all a bunch of half-truths and badly connected dots. All of these dots are public and found in larger presentations that connect them directly by explaining Altera’s moves in great detail. This includes their use of Intel, TSMC, and other related items.
Lets step back a bit to ARM’s Techcon 2013 that took place last October. SemiAccurate wrote up the news of their Arria 10 and Stratix 10 line a few days later, complete with a very telling picture reprinted below. This picture was one of a large deck with much more information about the company’s foundry and technology plans, and was on the Altera web site and presented at Techcon to at least the press and analysts if not all attendees.
Note the foundries listed in columns 2 and 3
What did it say? The high-end Stratix 10 is being fabbed on Intel’s 14nm process, that was the bit that grabbed headlines all over. The mid-range Arria 10 is being fabbed on, wait for it, TSMC’s 20nm process. See an eerie parallel to the ‘news’ of today that only a rank idiot could misinterpret? If not we will point it out later, go play with something sharp until we do.
So Altera is fabbing their big, high performance, high cost parts at Intel and the less performance sensitive but far more cost sensitive mid-range devices at TSMC. If you think that this may be related to the amount each foundry is charging for wafers, that would likely be right. In short for the cost no object Stratix 10, Altera is using the best process available, Intel’s 14nm. Where cost is an issue, they are going to TSMC.
“But wait!” you may think, the article mentioned 16nm. That is true but for all of the non-Intel foundries that SemiAccurate is aware of, 16nm is just 20nm with a new FinFET transistor. The rest of the process, also known as the back-end of line, is the same as 20nm. Shrinks are minimal and so are most performance gains. Even the most aggressive foundry out there, Samsung, is only claiming a 7-15% shrink, and that only if the customer really wants to push it. For most it will be a low single digit shrink.
Why bother if you get so little out of it? The transistor replacement means lower leakage and can have active power savings too, but the big reason is that it is really easy. Since the back-end of line, everything but the transistor that is, is unchanged, moving to a “new node” is relatively simple. Cost should only go up minimally, and price/performance should be a pretty clear win. Basically it is an easy way to squeeze a bit more out of a mid-life device refresh.
The first point we will make, and sadly not in a way that will protect the gene pool from the analysts above, is that a mid-life kicker would take this path. If Altera wants to do an update to the Arria 10, it is a no brainer to do it on TSMC, and borderline insane to do otherwise. If they want to do a mid-life kicker on the Stratix 10, well lets hope Intel has something for them otherwise it will just have to stay ‘only’ 2-3 effective nodes ahead of the TSMC product. In short TSMC 20 -> TSMC 16 is a given, anyone not clear on why should be in another line of work.
Then there is the point about the “delay” in Intel’s 14nm process causing this waterfall of process changes. First is the simple fact that the process is only a tiny bit late, about a quarter, 5 months tops. This is not a game changer, it is a minor hiccup, and nothing to rip and replace foundries over. Worse yet is that this late in the game, such a change would be problematic enough to delay the Stratix 10 far more than the minor delay from Intel. In short, if you have a clue about the tech, you would not write such babble, you would realize that it would make you look like an idiot.
The kicker however is one that Altera clearly stated, SemiAccurate published, and that none of these below average chimps seemed to ask. That is the release date of the Stratix 10 on Intel’s 14nm process, something that Altera would not comment on at Techcon. If Altera would not comment on release dates, why does this make the “analysts” look bad? Easy, because Altera did say that the Stratix 10 line would not tape out until Q4/2014, the same quarter Intel will have volume products on 14nm out in massive volumes.
If you read our earlier piece about the 14nm non-delay, you would realize that those wafers are going in to the oven in a few weeks, it takes time to ramp a process up. Intel’s 14nm process will be ready for prime time about six months before Altera even tapes out, and that is inclusive of a worst case five month delay. Only a moron would pull a design from a process this late in the game because a process is ready six months before it is needed instead of nine months prior. On top of that, the time from tape out to salable devices is 6-12 months, so we are looking at late 2015 before Altera needs 14nm wafers in volume. And yet these ‘analysts’ and ‘journalists’ are once again claiming that the sky is falling. Someone please put them out of my misery.
So in the end this all came from a single idiot that obviously doesn’t have the barest minimum clue about the tech he and his company are pontificating about. It is embarrassing for anyone in the tech field, yet two more layers of idiots so far, are repeating it breathlessly. Their vapid screeds are actually more understandable, they are paid by the hit instead of by accuracy, integrity, or common sense like the analysts are purported to be.
For the second time in a month these dim but provocative bottom-dwellers are hyping Intel’s purported failures that don’t exist. They didn’t do even the basic, bare minimum research, ask anyone with basic domain knowledge about the tech, or anything else. Neither did their editors. It makes us all look bad, it hurts Intel and Altera, and makes you wonder if they don’t have undisclosed ulterior motives.S|A
Have you signed up for our newsletter yet?
Did you know that you can access all our past subscription-only articles with a simple Student Membership for 100 USD per year? If you want in-depth analysis and exclusive exclusives, we don’t make the news, we just report it so there is no guarantee when exclusives are added to the Professional level but that’s where you’ll find the deep dive analysis.
Latest posts by Charlie Demerjian (see all)
- You can now buy bare Snapdragon SoCs with 410E and 600E - Sep 28, 2016
- Spin Transfer Technologies talks about their ST-MRAMs - Sep 27, 2016
- ARM adds CMN-600 interconnect and DMC-620 memory controllers - Sep 27, 2016
- Globalfoundries 7nm process isn’t even close to the name - Sep 26, 2016
- ARM upgrades realtime offerings to v8-R and adds Cortex-R52 - Sep 21, 2016