[shiroiushi]

>Also, I’m not sure good engineers always make good managers.

I worked at Intel when Craig Barrett was CEO. He was also a former engineer, and he was a terrible CEO. Intel got involved in a bunch of stupid stuff during his tenure, like the P4 Netburst architecture, the RAMBUS fiasco, and lots of side businesses that didn't pan out. Otellini took over and fixed things by going to the Core architecture, IIRC.

I'd say that good engineers becoming good managers (esp. upper managers) is the exception, not the rule.

[tgma]

> I'd say that good engineers becoming good managers (esp. upper managers) is the exception, not the rule.

Good managers are exceptions, period. Still, good engineers are leaps and bounds more likely to be effective as managers, even if they make managerial mistakes. This is probably even more true for high level executives than line managers.

There's a common bigco school of thought that does not say this explicitly, but believes "nerds are bad managers," so if you are good at engineering you're less likely to be good at management, so let's go hire pro managers (that end up being schmoozers and less technical). In my experience, the less technical managers have been bozos AND bad at management.

[lnsru]

Technical founder sold a company (where I currently work) to a big corp. Big corp sends here regularly directors with MBA to fix things, but things are worse and worse every year. People ask me why I joined a company in free fall every week. The thing is that technical understanding of MBAs is close to zero. Result will not improve throwing away 15% of the employees (the guy before current director did this). It’s obvious to me as an engineer with experience in operations - the products are not designed for easy production and manual testing takes too long and has bad error coverage. I have experience to fix obvious technical problems. No MBA will see them in first place. No MBA with their arrogance will listen to an engineer. Let’s wait for recession to end and look for new jobs.

[oneshtein]

Talk to your manager manager. Talk in manager's language. Be rich. :-/

[silvestrov]

It's a very common business school line of thought that managers are always smarter than the people they manage. Employees are seen as "resources", not as talents.

So by definition nerds (and other non-business school people) cannot be as good.

[packetlost]

I had a conversation today about that topic sorta. I find the abstract tradeoffs of writing software (ie. efficiency vs flexibility) to be inherent to company social structures as well. I suspect the intuition that engineers build up of knowing when to favor either side is applicable, just with different error signals.

[YZF]

There are a lot of similarities between software and organizations. But being a CEO also requires people skills. Many of us software folks have good intuition about how to make technical projects successful but the way to get there often involves getting other people to cooperate. AKA leadership. That's the tough one.

That said I agree that a CEO of a company like Intel has to be technical. I mean we saw what happened to Apple when the CEO came from Pepsi. But you need someone technical, in the domain, who is also a strong leader and a visionary and those people are not that common.

[PittleyDunkin]

Reminds me a lot of Conway's Law: https://en.wikipedia.org/wiki/Conway%27s_law

[bjarneh]

> the less technical managers have been bozos AND bad at management

The golden combo.

[ethbr1]

The difference between immediate pre-P4 and now Intel problems is stark though.

Immediately pre-P4, Intel need a better architecture than K6-* or AMD-next (revealed as Athlon). That's not a process problem.

Now, Intel needs their leading node to work. That's a process problem.

Arguably, none of Intel's business model works if their process is substantially behind TSMC.

Consequently, the CEO and board should be 100% focused on returning to process parity.

Gelsinger at least had an economically viable plan for that. Without him? No strategic plan: the board sure as shit isn't going to come up with one.

[YZF]

Intel historically was the best at semiconductor manufacturing processes. That's what got it where it got to.

I bought some Intel shares when it crashed last time but the latest events are changing my mind on the prospects of this company. Intel getting to be the best at process happened over years or decades of people and culture. If that's gone there's no way to reproduce it. Looking from the outside it looks like through offshoring, layoffs and culture issues they've basically lost what they had.

I think AMD managed to make it through tough times (and Intel way back then as well) because even when the company was struggling financially it managed to maintain people and culture that enabled it to overcome the challenges. But now the times have changed and executives don't see people or culture as leverage, and are happy to get rid of those. You better not need any amazing execution from that point on.

[tgma]

> Intel historically was the best at semiconductor manufacturing processes. That's what got it where it got to.

Exactly. AMD has been executing perfectly on the CPU side, but they also happened to be a key beneficiary in an unprecedented event in the past few decades: TSMC overtaking Intel in their process which would likely not have happened without Apple betting on them.

Intel did not benefit from the stagnation they had which was enabled by their success and monopoly. Internet Explorer 6-level stagnation: from Haswell to ~2018 they successfully shipped the same product over and over again and when AMD pressure started to mount, they basically trickled down higher core counts into cheaper SKUs. It would have worked, except at the highest end, they needed the process to be competitive and it wasn't, which messed up everything.

[hedora]

It’s not just a process problem. Unless you are on windows or playing steam games on linux, there is zero reason to prefer x86.

Those markets add to way less than 50% of consumer market share. X86 is rapidly declining in cloud too.

It reminds me of sun, dec and hp fighting to have the best commercial unix OS in the late 1990’s. Even if you win, you loose.

[adrian_b]

Besides Windows or gaming, there is another market for x86.

If you are interested in technical and scientific computing, but you cannot afford to spend amounts of $ written with 6 digits or much more, then x86 is the only solution.

The only Arm-based CPUs with decent performance are those made by Fujitsu, which are not available at retail.

The "datacenter" GPUs from AMD and NVIDIA are much too expensive for small businesses or individuals. Even for a bigger business their price is justifiable only when they are busy close to 24/7.

So for most people only the x86 CPUs provide an acceptable performance for computations that use either FP64 numbers or large integers.

[mkesper]

Things are changing on that front, too: https://www.servethehome.com/supermicro-megadc-ars-211m-nr-r...

[adrian_b]

The Ampere CPUs may be great for something like Web servers, but they are pathetic at number crunching. The same is true for the CPU cores designed by Arm, which are used in the server CPUs of other cloud vendors.

[kimixa]

The only way you can get close to x86 being "less than 50% of the consumer market" is by including phones, but I'd argue that's a new market on top of the "consumer computing" market. And even then, in "unit count" rather than revenue.

[shiroiushi]

>Unless you are on windows or playing steam games on linux, there is zero reason to prefer x86.

This is completely false, unless you're just speaking theoretically. In the real world of 2024, there's no real alternative to x86 for general PC computing, even if you're running Linux. Where are the laptops with ARM CPUs? (No, Macs don't count; can't easily run Linux on those plus they have a huge price penalty.) Where are the consumer-grade ATX motherboard + CPU combos for building desktop PCs? None of this stuff exists, at this time.

Sure, Apple's made some really nice ARM CPUs, but that's pretty useless if you don't want to run all-Apple hardware and use MacOS. And Windows keeps using x86 for whatever reason, even though technically they could target other architectures if they really wanted to (and they have, in the past). Linux basically just uses whatever Windows uses, because unlike Apple it's just not nearly big enough to command marketshare all by itself outside of dedicated servers. (And even here, for some reason, ARM hasn't made much headway, though I think it is pushing into this market slowly.)

[wbl]

Hindsight is 20/20. If higher clocks won out because extractable IPC was lower Netburst was the right design.

[Earw0rm]

The problem with Netburst was the clock speed wall. They designed it with a view to continued doubling which has been a trend up until then. It turns out once you go past 3GHz improvements are much, much slower.

We hit 3GHz in 2003. If you go back ten years from that, the top speed available was 66MHz, ten years before that and it was 8MHz. Fast forward 20 years and we're just barely scraping 6GHz.

[Earw0rm]

Doing the math - even a conservative estimate for 1983 to 2003 is eight doublings.

That would give us a hypothetical Pentium 4 at 768GHz today. Which, Netburst or no, would wipe the floor with the fastest available multi-core chip.

While also putting out enough heat energy to keep a small town in Northern Canada warm all winter long.

[giantrobot]

Foresight (at the time) told you NetBurst was a dead end though. AMD made a lot of sales from people avoiding NetBurst, ignoring clock speeds and looking at actual performance numbers.

NetBurst looked good on benchmarks; code with few/no branches that fit in the L2 cache if not the L1 cache. On real world code consisting almost entirely of branches and pointer chasing the NetBurst chips performed horribly compared to lower clocked Athlons.

Intel's promises of 5+ GHz chips got more laughable as the line went on because they were already straining cooling and power supply options at less than 3GHz.

[wbl]

Netburst's branch predictor choices survived. It was the penalties that were a bad idea.