A visit with Apple’s chief chipmaker
bloomberg.com
bloomberg.com
Choice quotes
"Apple had a problem". The blame always seems to be shared. How about - The management fucked up the release schedule.
"That gave most of Apple’s engineers more time."
"It gave a little-known executive named Johny Srouji much less."
So this makes it seem like it was harder on the executive than the people doing the actual work.
Unfortunately this is how it almost always works in business. I'll never forget the times I've worked harder to make a newly created deadline only for the management to gain a reward and nothing for me. Who cares I worked until 2am every day for weeks while they left at 5pm because they "weren't coders".
I'm sure most in the tech industry goes through this and it's not healthy. Yeah yeah I know you're salaried so you shouldn't expect more but when management gains more because they imposed new deadlines for you to hit, it creates resentment among the developers and eventually an exodus.
I've seen it happen 3 times so far in my career. You'd think I'd learn my lesson but it's hard finding a place that actually rewards both employees and management (or rewards none of them which would be more fair in someways and worse in others considering only the top executives benefit at that point).
Sorry, ranting a bit.
> Sorry, ranting a bit.
Rant away! I don't want to hijack this thread, but there's a story I like to share with people when they become new managers. It's from Herb Ryman, one of the first Disney Imagineers.
"The year was 1953. Walt [Disney] said, ‘Hi Herbie. I’m over here at the studio. I wonder if you could come over here. Just come the way you are, I’ll be out front waiting for you.’ I was curious, and flattered that he picked up the phone and called me. I had no idea what he wanted.
He met me out front and shook my hand saying, ‘Hi Herbie, we’re in the Zorro building.’ So we went in, and I asked what this was all about. He said, ‘My brother Roy has to go to New York on Monday morning. He’s got to talk to some bankers there. You know bankers don’t have any imagination, none at all. You have to show them what you are going to do. Roy has to show them what this place is going to look like.’ And I said, ‘Well, I’d like to see what this place is going to look like, too. Where have you got all this stuff?’ I thought maybe it was in the other room. Walt said, ‘You’re going to do it.’ I said, ‘No I’m not.’ There was a brief pause. Walt paced back and forth- we were alone in the room.
He went over to the corner and he turned his head around with his back to me and said, ‘Will you do it if I stay here with you?’
‘Yes, I’ll do it if you stay here.’"
One is a google book with the exact passage, and the last is a blog post quoting said book.
Just something interesting I noticed.
I don't know if the Yesterland blog post is the one you came across, but they do a good job cataloging the different retellings of the story and even include some of the artwork Herb put together that weekend. Definitely worth a read.
When writing software, I almost always reach first for Erlang these days—not because it's the most performant platform, but because it allows for an architecture where individual Erlang processes (tasklets/green threads) can crash and be restarted producing only events in a log that I can debug at my leisure, rather than waking me up at 3AM to fight a fire.
If I were a CEO (and I have been, but only as a single founder, so I've never felt that "others depend on me to not go hungry" feeling), I'd be looking for similar wins in that dimension: fault-tolerant strategies allowing me to know that nobody's going to go hungry because of one mistaken assumption, so that I can sleep at night.
If I was doing consulting, I'd try to productize it; if I had a product, I'd try to sell it to enterprises; if I had enterprise customers, I'd try to get them on a yearly pre-payment schedule; if I was in one vertical, I'd try to "clone" my product for other verticals with their own brands and marketing; etc.
Are you looking for these? Or are you prioritizing growth: trying to, effectively, sprint for three years, rather than figuring out how to run a marathon?
We were near the end, made our schedule, when I was called into a meeting. My schedule had been cut by 2 weeks. I ran out of the meeting and called 4 software development companies, the 2 hardware teams. We pulled together and finished with the new deadline.
The day I signed the paperwork for release, I was told the reason my schedule was cut was because a senior VP was going on vacation and he wanted to take the product with him to show his grandkids. He didn't want to show them something that might have bugs...
Part of me thinks I should start invoicing employers for time spent at work over the 40hrs, charging an overtime rate, or something to increase compensation in exchange for the lost time, or modifying the contract accordingly if there's not an alternate compensation mechanism like comp time offered. The one employer I've worked at who was good about this let you disappear from the office if you hit your 80hrs for a 2wk pay period.
That's what unions are for.
Obviously there's a limit, and this is highly dependent on where you work, but I love being salary vs being hourly.
That's only because if you refuse someone else will step in to take your place under those terms. Collective bargaining can be a powerful aid with things like that.
Don't work if you don't get paid.
I'm lead developer of a project that was just submitter to Apple review an hour ago. This week haven't been easy, but both my direct manager and company CEO were in the office with us, even if they really didn't have anything to do. And needless to say, they ordered food in around 20-21 for everyone. And this has only been one week.
To some degree. The problem is every place you talk to there is a "well there could be crunch times when we're busy and you'll have to put in extra hours but those won't be frequent" conversation which seems okay at first. Then you join and you find out they're fighting fires almost every week and depending on the fire management may get compensated whereas developers do not.
It's not always an easy thing to avoid. I've seen it at big companies and start-ups. Start-ups typically seem the worst with this type of thing in fact, at least in my experience.
I hear they have been designing their own GPU for many years now, and I'm quite (half)surprised they haven't released anything yet: one does not simply acquire the 30-year expertise of Imagination in such a short timespan.
NVIDIA tried going in the mobile SoC business, and after 5 generations and billions burned they finally gave up. What Apple did (and keeps doing, by putting Samsung, Qualcomm to shame[1]), is mind boggling. They are even approaching Intel's perf-per-watt[2] ratio.
[1] http://anandtech.com/show/9686/the-apple-iphone-6s-and-iphon...
[2]: http://anandtech.com/show/9766/the-apple-ipad-pro-review/4
>> be the first to bet on powerful GPUs
Sure they are the first to bet on expensive stuff - they have the most margin and sales, a more profitable ecosystem - so it's easier to justify the risk of more complex coding(by an app developer), etc.
>> I hear they have been designing their own GPU for many years now
From what i hear, they use the best of Imagination , while locking it out of the industry. But i would be happy to learn otherwise.
And i'm not saying Apple didn't create an impressive chip, but creating impressive chips is much easier than sustaining a good business for them.
Also ,the real power of the chip is for marketing - one critical tool in building luxury brands is a story about unique creation ,creating stuff that nobody else can do ,etc. So that's a very valuable story.
If they ever chose to release their own GPU, they'll be sure to sell their stock on the same day the device using it is released.
I don't see why it would be more complex coding, I'm pretty sure the move to ARMv8 was transparent to most developers (as it should be).
They use the best of Imagination because of the ambitious goals, and because no one is even trying to push the boundaries; I'm not sure they have exclusivity on the designs, they just have a better SoC integration team, while TI choked on OMAP5 after the years of delay.
Another bet for which they were the first: UFS PCIe storage, on mobile. The storage performance crushes all competition, it's sad.
Honestly, most people don't know what a SoC is, so no, it's not a marketing ploy, even if Apple is marketing company. It's a (huge) technical bet.
... Eh? Which bet are you referring to?
At the time almost all SoCs had anemic GPUs, and Android was betting on a software rendering pipeline.
Wonder what his team got that pulled of the needed engineering 6 months early? I'm sure they missed a lot of family life.
What a wonderful statement. I like this spirit, especially that he only mentions physics, not money.
I'm optimistic that many things people say today that are impossible will eventually become possible. Except time travel to the past; that shit just hurts my head :)
Unfortunately, physics has not changed much. There is a great speech from 1959 about how small you can make things: http://www.zyvex.com/nanotech/feynman.html
People don't say that anymore, because we are already building really tiny things. Sure when your using ~10^20th atoms in a transistor there is a lot of room to grow. In 2008, researchers made transistors one atom thick and ten atoms wide. Where do you go from there?
People want to say 3d, but we already make chips in several layers and heat is an issue.
What time frame are we talking here? We're constantly learning about how more stuff works. Physics has had plenty of changes[1] but perhaps you're only speaking to very specific theories that have not changed such as relativity? In which case yeah it hasn't changed much but there is also a disconnect between relativity and quantum mechanics and it seems a pretty large majority of scientists think there is a unified way to make them one theory so obviously our understand of both is not entirely accurate (though our observations with relativity are pretty spot on).
> Where do you go from there?
Hard to say since I'm not a physicist or hardware engineer but I've seen lots of interesting talks about scaling down more and more when we didn't think we could before. Ultimately there may be a size in which we can't get any smaller with conventional processor architecture but that doesn't mean we can't build new ones. Optical based processors, built in a 3D way, allows for far more overlap than transistors since light passing through each other doesn't interfere with one another[2] so that looks really interesting.
Quantum computing also promises great speed increases which perhaps could eventually be scaled down to closer to quantum sizes than traditional computing.
I think everything is getting much harder than the breakthrough before it but we seem to keep going and I would be surprised if we end up completely stopped due to hitting some physics block. We'll find a workaround!
[1] http://quantumfrontiers.com/2013/10/23/the-10-biggest-breakt...
Quantum Computers are generally far less useful than you might think and go back more than 25 years.
Finding planets is hardly a physics breakthrough.
Bose Einstein Condensates reference Einstein for a reason being predicted in 1924-5.
Anyway, I don't mean physics is static, rather we have been very good at a range of timescales, speeds, and sizes for quite a while. Things you notice with a particle accelerator at 99.999x+ of the speed of light are only so meaningful.
Clarke's first law: When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.
Clarke's second law: The only way of discovering the limits of the possible is to venture a little way past them into the impossible.
In the past, this type of work might have been limited to a smaller set of government-funded researchers, for example defense contractors.
“Hard is good. Easy is a waste of time.” -- If I had the motivation and guts, I'd put this on my coffee cup. But I'm sure it would make me a magnet for abuse.
Crush them.
The result was a lower average hourly rate to them, but it helped them stay up to date--and in some cases on the cutting edge. Why? Because it's easier to learn a new technology when there is a real project at stake. The deadline and accountability provides strong motivation, and there are real problems that need to be solved along the way--not just self-directed exercises.
I'm not a freelancer but I can see the long-term advantage of that approach. If you're one of just a few people who can do the cool new thing, then you can charge higher margins on easier projects, because you have less competition for the jobs.
Personally, I have never learned as much as when I am working a real project (i.e. paid, deliverable) that is a bit outside my current comfort zone.
A very perceptive mindset, in my opinion.
Tagging pointers on 64bit ARM does allow some new techniques for compactness but it'd be hard to claim this overcomes the doubled pointer size. Does anyone have data on how memory use was impacted between architectures for the same code base? I recall the Objective-C runtime taking advantage of this so perhaps it's only a slight tax if you rely Apple's frameworks.
Why is that so? For ages we have relied on Intel and AMD, and that worked pretty well.
Given their volume, Apple could probably have outsourced design and still gotten reasonable results, but look at what happened when they outsourced their chip production to Samsung: they helped create their closest competitor.
And it helps their overall business model to keep competitive/strategic decisions under wraps as long as possible, which would be harder if someone else was apprised of their chip direction.
This words are very encouraging for entrepreneurs. No matter you like Apple or not, this philosophy made Apple succeed.
Sounds like an non-sequitur to me: what's the link between Apple and entrepreneurs, and why should they be encouraged? Shouldn't they be encouraged by commoditization of silicon instead? >90% of kickstarted/'maker' projects would vanish if step 1 were "have your own silicon"
Although Israel grapples with Jewish-Arab tensions all the time, none of it mattered in Srouji’s world. Cohn, who remains friends with him, says their different backgrounds never came up. “Technical people treat technical people based on personality and technical ability,” he says. “You don’t think about it. You just work together. The rest goes away.”
And when Israel finally decided that it was time for the "mighty" IDF ground forces to enter the stage... nothing really happened. The rockets just continued to be launched from Gaza for a whole month. It was a really amusing "war" to watch in my opinion :D
What I do know is that Israel was forced to agree on a ceasefire with a a bunch of "terrorists". Quite hilarious.
The only thing that is required to be smuggled are rockets.
Conclusion: Gaza is completely isolated from the outside world. 1 million+ citizens in a confined space. Long live Israel.
[1]: http://www.jpost.com/Breaking-News/Rafah-crossing-between-Eg...
[2]: https://en.wikipedia.org/wiki/Israel%E2%80%93Gaza_barrier
"Completely isolated from the outside world"? Not even close. 3000 truckloads a week. And growing.
> Now, if Egypt's crossing is closed, do you think Israel would keep theirs open? Hmm.
You seem to imply something here; as if Egypt-Gaza relations should be expected to better than Israel-Gaza. Why?
> 1 million+ citizens in a confined space.
Yeah, this is called a "country". Exactly what Hamas is, if you believe it, wants. And this "barrier" would be called a "border" then. You know, like countries have.
I apologize. You are right, Israel seems to at least let some aid through.
A country border? What kind of country lets another state build a huge wall around it and doesn't have control of the crossing points? I'm sorry but you're spewing nonsense.
The "rockets" that Hamas fires from gaza are little more than homebrew model rockets. They are scarcely larger or more powerful than the kind ameature models that rocketry enthusiasts fly. For some time now the Hamas rockets rarely even carry explosive payloads, and are only meant as decoys to trigger the Isreali Iron Dome defense system in a battle of attrition-- Iron Dome tracks and intercepts the rockets with more modern projectiles of it's own-- I've seen estimates that each Iron Dome response costs upwards of $20,000. I doubt Hamas has to spend more than $300 for each decoy.
This is not acceptable provocation for the kind of warfare that IDF routinely makes against Gazans.
Quite the contrary. A lot of them are not home-made, they are imported from countries like Syria and Iran. Sometimes, they end up even with Russian made rockets. And they're getting bigger, and have a wider range, each year. Here's the infographic: http://www.thetimes.co.uk/tto/multimedia/archive/00731/inlin...
I think you made an honest mistake because a lot of rockets are called "Qassam" by the media, while the name itself denotes the most basic, home-made rocket variety.
> This is not acceptable provocation for the kind of warfare that IDF routinely makes against Gazans.
No, IDF response is an adequate minimum. You're talking about the "battle of attrition", and you're right; however, you're thinking about money, while the deal is about overall capacity. Iron Dome does it job well, but it has a limited output. When it's out, it means that those rockets will start killing people, and that's exactly the point when IDF starts destroying rocket stockpiles and infrastructure. It's tragic, because despite all IDF's efforts, it means that innocent people will die — but at that point, lack of action would lead to a larger amount of civilian deaths.
> against Gazans
And, finally, this depiction of sides of the conflict is a great demonstration of your bias. The war is started by Hamas, who wasn't elected democratically and is holding power by extreme totalitarian violence. IDF, on the other hand, is a military force of a democratic country. Unlike US or other first-world country, all israelis go to the army, so IDF is even more representative of society as a whole. So, "warfare that israelis make against al-Qassam Brigades" would be a much more fair description.
That same article describes the rockets you reference, but with the added context of relative quantities:
"According to the IDF, around six rockets are being fired at Israelis every hour. However, many of these rockets are not sophisticated, and they either fail to land in populated areas or lack the firepower to cause casualties when they do: Sometimes, the payloads are removed from missiles in a bid to increase their range."
[1] https://en.wikipedia.org/wiki/Palestinian_rocket_attacks_on_... [2] https://www.washingtonpost.com/news/worldviews/wp/2014/07/11...
I worry that Bloomberg isn't sure if it wants its website to be Buzzfeed or the Atlantic.
Also Buzzfeed has grown into a serious news operation, investigating many stories (Tennis Fixing), and breaking others.
And according to RealWorldTech, ARMv8 is simply a work of art.
Notice the annotated die shot chip works provides. Even if it were true that "Apple's custom SoC's are still mostly a collection of licenced ARM IP blocks " that would constitute only a small fraction of what is in the SOC overall.
Submitters: the HN guidelines ask you not to use the article title when it is misleading or linkbait: https://news.ycombinator.com/newsguidelines.html