Interview with AMD CEO Lisa Su | Hacker News Reader

Interview with AMD CEO Lisa Su

anandtech.com

172 pointsvivekchandsrc43 comments

It's pretty impressive to see a CEO being able to answer a question about CPU vs GPU power envelope targets just of the cuff.

The interviewer asks why there are no CPUs targeting higher wattages, like how CPUs are less than 150W but there are 300W GPU. She says she'll need to think about it but she's pretty sure that the high parallelisation of GPUs makes them easier to scale up in terms of wattage.

Granted, that answer could be wrong (I know nothing about CPU engineering), but either way it's a really well considered one to be thought up when "put on the spot" in an interview with a question you haven't considered ahead of time. I'm sure that's one of the things which separate good from great in CEOs: an understanding not only of the broad market decisions made by their company but even the details of low level architecture decisions.

> Ph.D. in Electrical Engineering

i mean, you could expect someone with that degree from MIT have at least a general idea about the technicalities of her flagship products.

Yes, you would expect it, but it's still rather atypical; most people become CEOs because of their business and political acumen, whether or not they understand the technical bits is maybe 10th priority.

They say about companies that you can tell their culture from the outside by who they value and who they promote. Some engineering companies promote MBAs, and some promote engineers. Promotions and who holds high rank is an external cultural signal to the world that is really hard to fake.

This cuts both ways though: if you see a company that hires and promotes engineers, where the CEO is this technically deep, you can probably conclude that (a) they value engineering but also that (b) they're probably not that great in terms of financial management & project management.

They're radically different skill sets, and opportunity cost abounds. Almost no one is awesome at everything, because there are only 24 hours in the day, so being awesome at one thing often does mean you have some trading area where you're weak.

Engineering companies run by MBAs are not always so bad as they sound. It sucks as an engineer if your considerations are not always front and center, but you get other advantages too; better strategy/insight on fundraising, better overall management, and sometimes vision too, many engineers are too narrowly focused on incremental improvements, and are missing the "dreamer" component, aka the "Jobs" not the "Wozniak".

And how many MBAs are "dreamers"? Then again, i don't think Jobs was much of a dreamer either, he was just damn good at talking the talk.

If anyone was a dreamer it was Woz back then. Dreaming of having his own computer, dreaming about computers improving life for everyone, etc etc etc.

Jobs are all about appearance. And not just visual. He had a near pathological loathing towards fan noise for example. To the point that one AppleII variant had problems with excessive heat buckling the logicboard and unseating components because he refused the engineers to install even the most unobtrusive of fans.

I suspect we can see this in how he was quite hung up on the GUI but completely missed networked computers while touring Xerox PARC.

An MBA doesn't teach you to be a dreamer. Neither does an engineering degree.

But consider what personality types are drawn to different businesses...and I think you'll find a larger (but still small) percentage of the dreamer type within the business community than within engineering.

This is not to say that there aren't also engineers who think this way, of course there are. It's more about demographics, and which personality type is attracted by which type of work.

> I think you'll find a larger (but still small) percentage of the dreamer type within the business community than within engineering.

Can you expound on that a little? It's completely contrary to my life experience, where people with business degrees are mostly system followers and ladder climbers.

I can't speak for moxious, but I agreed with their post because in my understanding of business and engineering and from being around people of either backgrounds, engineering seems to firstly prize hard, analytical skills and only later on considers creative solutions, while business gets around to analyzing things after considering the things within a/the bigger picture.

In terms of seeing the forest vs. the trees: My thought is that engineering might be so concerned with seeing and analyzing the trees that they do not get around to considering the forest as a whole and recognizing where no tree presently (but could eventually) resides within that forest. (I think that "recognizing where no tree presently resides" is what a dreamer type might do.)

A big issue that I see with framing business people in this light is that few of them do seem to think entrepreneurially, but that may explain why you haven't run into more dreamer types.

> Jobs are all about appearance.

Job was all about design, and design for him went from the very simple, raw components to the final packaging and marketing. For him there was no boundary between engineering and design, between software and hardware, they were all part of the same product.

It's easy to dismiss his views as superficial, but when he wanted to have a board look a certain way, or for a case to be a certain size, he'd press for it. When there were technical problems that prevented that from happening he wanted to understand why. He wanted explanations, and he'd listen to them, then make his own judgement based on that information.

> ...but completely missed networked computers while touring Xerox PARC.

The Macintosh famously shipped with very high speed serial ports for that time, up to 230Kbaud, which was vastly faster than any modem or other serial device around in the 1980s. Why?

LocalTalk: https://en.wikipedia.org/wiki/LocalTalk

Apple may not have had Ethernet in their early computers, but they were absolutely aware about the importance of networking. Sending files from one Mac to another was as simple as plugging them together. While the Macintosh didn't have a lot of games, it did have multi-player ones over a local network long, long before Windows did.

>Job was all about design, and design for him went from the very simple, raw components to the final packaging and marketing. For him there was no boundary between engineering and design, between software and hardware, they were all part of the same product.

+1 Somebody who actually understands Apple and Steve.

I don't know about the time or Parc, but he was a proponent of thin terminals and having all of your data being available to you everywhere.

>that (b) they're probably not that great in terms of financial management & project management.

That's hardly a conclusion you can make at all. Only on the Internet could someone claim with a straight face that engineers as a rule are not also competent at financial and project management. There are millions of counterexamples. Further, engineers do on occasion go back to school to get MBA's and take a management track. Does that make them an engineer or a business person for purposes of this discussion?

Yes absolutely, I think that's the key. It's cool to see very tech-y CEOs like this. I feel like many CEOs would defer a hypothetical question about how a product could have been made.

I mean it's like asking Tim Cook about a different strength cellular antenna in the iPhone. Probably he could give a reasonable answer about what the iPhone antenna is actually like and why it's like that, and how many tests they ran to confirm it's the best design or whatever. But could he talk about an alternative antenna that could have been, on the spot?

No, but I bet he could talk about the messy complexities of the Apple supply chain. Can technical CEOs do that?

My point isn't to get into a skills based dick measuring contest, but rather to point out that even at huge technology companies*, is being technically adept really a requirement for the job?

Another one for discussion: is Apple a "technology" company? I put forth that a company is defined by the most difficult problem they need to solve / continually solve in order to stay in business. That tells me Apple is more of a manufacturing/ marketing company than technology.

I wasn't saying Tim Cook is lacking in skills. As the CEO of one of the world's most successful companies he certainly has his own unique and impressive skills.

Lisa Su demonstrated an impressive skill not every CEO could pull off and I wanted to highlight that. That's all.

Whether being technically adept is a requirement or not, I'd say no in general for tech companies (and I'd count Apple as one) but for a certain subset of the tech companies (AMD, Tesla) it certainly seems to help.

She's essentially correct about it. Scaling a gpu Basically amounts to more compute units.

I happened to watch the Ryzen keynote video last week [1], and she's also an impressive presenter.

[1] https://www.youtube.com/watch?v=1v44wWAOHn8

Aren't the CEOs of most successful technical companies engineers?

She has also increased AMD's market cap by 400% since she was made CEO 3 years ago. Pretty impressive person all around.

As nothing material about their business has changed in 3 years, I think it's safer to say Nvidia has raised AMD's stock price 400%.

See http://imgur.com/a/ul9LE (nvidia green, amd red)

Nvidia clearly started ramping first, after which investors looked at AMD as a "cheaper alternative". Apparently a bunch of people think they're the same company, but Nvidia's recent rise is a result of their work in AI while AMD has totally missed that boat with both developers and industry.

Surely none of that rise has come from their best in class APUs, open source support, or a headline stealing new architecture that is stomping on Intel benchmarks. Surely their rise is because investors in huge masses are buying them up mistakenly thinking they're nvidia.

I did chuckle reading that, because yes it sounds ridiculous at first. Buuuut...

> best in class APUs

Best in class because no one else wants the market, there's no profit margin there (as AMD has proven).

> open source support

Just because you support it don't make it good. While I think it's lame Nvidia's stuff is closed source, they have such a superior eco-system that I'm willing to overlook it (along with most of industry). If AMD could somehow get around this, by having an open source ecosystem at parity or better than nvidia, they'd be golden. But as its stands, it's not enough to make a dent.

> huge masses are buying them up mistakenly thinking they're nvidia

Walk that back to "investors think the semiconductor industry is slightly more commoditized than it is and they're misjudging the current landscape" and yes.

> a headline stealing new architecture that is stomping on Intel benchmarks

Stealing headlines and stomping on single benchmarks does not a good product make. Jury's out on if they can come up with a winning flagship product, but I'm not holding my breath.

Also, they need to focus on stomping Nvidia benchmarks; Intel is not their competitor. By relative market cap, it's like saying "yahoo is google's competitor". Nvidia is AMD's "big" competitor.

Current CEO is an engineer (great!) but lacks the ability to see where things are heading; Nvidia's CEO founded the company. They need Jerry Sanders' younger clone back in there or something.

https://en.m.wikipedia.org/wiki/False_consensus_effect

Just because you care about AI doesn't mean anything about market size or even potential market size. GPUs are still niche compared to CPUs, GPGPUs even more so, and AI specialization is a fraction of that. I mean you're talking about a market measured in hundreds of millions today, a pittance compared to x86...and even assuming massive growth for a decade or more still couldn't catch up to the current market size for x86.

And as the only company in the world that is even legally allowed to compete with Intel on x86, they would be stupid to not try. The fact that they're catching headlines suggests that they actually have a chance of recapturing a portion of that $60B market...something unthinkable 3 years ago, especially for a company with yearly revenues lower than the monthly R&D spend of its competitor. To think that AMDs growth is entirely due to the market misjudging the level of commoditization of GPUs is absurd. If you're that confident of the market being so absurdly off base, some simple bets should put you in the Forbes 400 within a year or two. Get on it.

> GPUs are still niche compared to CPUs, GPGPUs even more so, and AI specialization is a fraction of that.

Yes, that's definitely AMD's thought process. The apocryphal IBM quote "There's a world market for perhaps only five computers" is why they're wrong.

My main point is this: AMD needs a new CEO; they had an opportunity that they've completely fumbled so far; this piece is so much fluff. You're free to disagree of course, but I don't think my opinion on this is going to change based on your arguments.

> If you're that confident of the market being so absurdly off base, some simple bets should put you in the Forbes 400 within a year or two. Get on it.

I'd be buying put options if the market couldn't stay irrational longer than I could stay solvent. Unfortunately the quality of their decisions as a company has nothing to do with their stock price right now. They were priced right a year and a half ago.

Since I stopped gaming, I stopped following the high-end market. However, it is nice to see that AMD is trying to get in there again, some competition is good for the market in general.

If AMD can deliver, then Intel might have to respond with lowering prices on their products, which I believe might be a bit overpriced now due to there no being a real vialbe alternative.

Whatever happens, some competition is bound to be a good thing imo.

They actually already have: http://wccftech.com/intel-amd-price-war-ryzen-processors/ and that was before the reviews even went live. It should only get better since impressions seem favorable.

Oh thanks for pointing that out, I missed that :-)

You should come back, even just for a few games a year. There are some incredible experiences available these days.

It is something I sometimes want to do, but find hard to justify. First of all, when I played, it was competitively and I found it hard to get fun out of just playing casually. In fact, I have quite addictive behaviour so I usually ended up addicted to games I played competitively. (CS:S, L4D(2) and COD2/4).

I know myself well enough to know that I can not play "in moderation" and would tend towards the competitive aspect. (Single player games never really entertained me that much)

Overwatch would be right up your alley :)

Hehe, I have heard that before, but I have actually not seen any gameplay of that :-)

I'm curious if you could elaborate on this.

For the population I represent (never, ever, want a first person visual rendering, under any circumstance) but love high quality game mechanics, I have a hard time finding what I consider good games. Everybody wants to polish their fundamentally undesirable three dimensional scene.

While there is a lot of 3D experiences out there, a lot of the recent great games that show off interesting mechanics can be found in the 2D realm. While not an exhaustive list by any means, and not knowing your preferences, here's a handful: - terraria - stardew valley - undertale - rimworld - prison architect - nuclear throne

In addition to mysterydip: Many RPGs and strategy games use 3D engines, but only for more "isometric" or "top down" views (downside is that they require more GPU performance than you'd think sometimes)

Also would like to mention Factorio explicitly.

If you're a fan of Baldur's Gate and the like. Isometric CRPG's are back in full swing now. Pillars of Eternity, Pillars of Eternity Deadfire(just completed a fig.co funding), Tyranny, Torment. They are huge, deep games. Tons of under the hood mechanics going on. Definitely not point, click, kill. Requires lots of setup, strategy, on how you address each situation. Great stories.

Hmm, these last years were one of the best years in:

- strategy gaming (from Deserts of Kharak, Civilization, Paradox interactive grand strategy games, Eugen Systems games, Total War series, XCOM and more and more)

- RPG gaming (2D isometric RPGs are making a huge comeback with good writing and gameplay, things like Pillars of Eternity, new Torment, Divinty and more. 3rd person Witcher 3 collected record amount of awards and with good reason.)

- platformers (Ori, Shovel Knight), roguelikes (Darkest Dungeon, Don't starve, Spelunky, and more)

- experimental narrative games (Papers Please, This War of Mine, This is The Police)

- modern adventure games (Until Dawn, Telltales Wolf Among Us / Walking Dead / Game of Thrones and many more)

And those are only on PC and none of them are first person. Just like movies go beyond just Marvel and Transformers, gaming goes way beyond just Call of Duty and Battlefield. The AA market of smaller but still established companies (Paradox Interactive, Relic Entertainment, Telltale and many more) build great experiences and I don't think gaming market was ever so live and broad as it is these years.

I'd recommend Kentucky Route Zero, too, which is finally almost all out (boo episodic release models). It's what you get when people with vision, talent, some real book learnin', a good understanding of video games in general, and artistic taste set out to tell a story in video game form. Brain bleach for the various trying-too-hard, one-note "art" games out there.

Why the aversion to first person? I feel that something like Deus Ex would be a lot less immersive in third person.

“Sadly, playing games where we live imaginary, exciting lives having basically destroyed our own is insane.”

http://www.scarygoround.com/bobbins/index.php?date=20140120

The high-end market is actually not for gaming — both Ryzen and Intel's "enthusiast platform" (X99) are worse for gaming than Intel's mainstream (quad core) CPUs.

Ryzen is great for tasks like, uh, recompiling your operating system :D Or editing photos or reencoding your movie archive into whatever is the next awesome codec (Daala anyone?).

Care to actually back up that statement with any sort of citation or fact? According to the leaked 3dmark benchmarks, the top 3 Ryzens are on par with some of the top i7 quads... so honestly I'm not sure where you are getting this idea from. It also completely ignores the price disparity which makes the AMD slightly slower but much cheaper CPU's a doable upgrade for many. (And VR capable)

http://cdn.wccftech.com/wp-content/uploads/2017/02/AMD-Ryzen...

3DMark's Physics section is a pure CPU parallel benchmark (also known as "buy 6950X to win"), not a game-like benchmark.

Here's results from actual games:

http://www.gamersnexus.net/hwreviews/2822-amd-ryzen-r7-1800x...

Even the older 4790K is sometimes faster. And the i7 quads are cheaper. So games prefer higher clocks to more cores, that's still true.

I am looking forward more to their Naples cpus for servers.

Reply on news.ycombinator.com