Intel is notorious for having bloated R&D and keeping those projects funded despite not going anywhere, while missing mobile completely, perf/watt and now missing AI/ML as well.
There is this interesting phenomenon where people outside an organization view R&D potential as elastic. That is to say, there is always a positive return on investment for the marginal R&D dollar.
It is clearly not in the case or every company would have exponential growth simply by reinvesting in R&D.
In reality, it is difficult to create growth from R&D at all, let alone the marginal dollar.
>There is this interesting phenomenon where people outside an organization view R&D potential as elastic. That is to say, there is always a positive return on investment for the marginal R&D dollar.
I don't think that's the view here; instead, I think the idea is that R&D personnel could be reassigned to work on something more profitable, rather than simply laying them off. Assigning R&D people to work on something unprofitable or a dead-end is the fault of management and their poor vision, not the people assigned to research that stuff.
That implies that there are job openings in those "more profitable" areas.
Indeed, wouldn't every company love to just reallocate all of its employees to the areas that are most profitable. But alas, adding eight more women to the project of birthing a baby does not reduce pregnancy time to one month.
>That implies that there are job openings in those "more profitable" areas.
if they are more profitable, they probably have more space and budget by defauly. Outside of these strange times, it's not like Tech has historically been oversaturated and companies shy away from hiring new talent.
> if they are more profitable, they probably have more space and budget by defauly
Not sure what you're basing this assertion on. I've never been at any tech company where the fact that a project is profitable somehow inherently means that we can just keep adding engineers to it and those additional engineers will magically increase the revenue the project generates.
In fact there's almost no industry where such an assumption would be logical. Even if your work has a very straightforward financial model, like manufacturing where you simply make Y dollars for every X parts you produce, that still does not imply you can just keep adding assembly line workers. At a certain point you have to consider whether there are even enough purchase orders to cover the volume of goods you're creating, or if they are just going to become wasted surplus sitting in a warehouse.
What you're suggesting is the equivalent of the eight other mothers not working on the same most profitable project, but instead on other projects of their own. So you're agreeing with my point.
And then you have 9 mouths to feed when you really wanted one very healthy baby.
The corporate analog would be sustaining dozens of projects, partnerships, and business areas that never really took off and had the results you hoped for.
Why does every organization always have something profitable they could be working on?
In the case of firing,
either way, companies are not staffed by omniscient leaders. The are staffed by humans, which don't know where to put R&D today to guarantee profitable inventions tomorrow.
Human uncertainty and imperfection underlies all business risk.
It seems that the visionary management at Intel has been consistently failing at focusing R&D efforts in a productive and profitable way, so why are they still employed there? They should fire the management and hire some new management that can make better decisions.
This is nothing new: even when I worked there decades ago, they were always making stupid decisions, such as the whole Itanic idea, the dumb idea of making their P4 processors dependent on expensive RAMBUS memory, refusing to make a serious 64-bit version of x86 until AMD did it first, etc. It seems they still haven't learned (or hired some better executives).
They say the fish rots from the head. This is a common problem with mature companies with clueless owners and nobody competent enough to right the ship.
If the money printing machine works to well, everybody forgets how to repair it by the time it breaks down.
It's more like R&D is the lifeblood of a company like Intel. Intel is basically purely an R&D company, their entire future hinges on R&D. This is not typical of other companies who can usually coast for years or even decades on existing product or may not even require R&D for new products. R&D should really be Intels major spend.
They have been putting record funding into R&D for years, but it seems that simply throwing money at the problem isn't helping, so it makes sense that they would try something different.
Their press statement essentially R&D is too bloated, dysfunctional, and disorganized in the current state to produce.
Throwing money at any problem is ineffective by itself. Still throwing money at a problem is a prerequisite to get stuff solved. It may very well be true that Intel R&D teams are too bloated, dysfunctional and disorganized. But that is not solved with throwing less money at it. If anything doing that and nothing else just leaves with less teams that are too bloated, dysfunctional and disorganized.
Their R&D spend is 4x that of TSMC and 2X that of Nvidia.
If they cut 20%, they will still have an obscene budget.
If you read the press release the idea is to focus on their core competencies, like making good processors. That means laying off people that can't help with that goal.
That would be unfortunate. Without a competent GPU it will be increasingly difficult to compete against AMD and Nvidia who have all the pieces necessary to produce high end gaming PCs / workstations / datacenter GPU compute. Developing a competent GPU solution isn't really optional if one wants to compete in those spaces, despite the large up-front cost, which Intel seems to have mostly already paid.
Intel similarly cut their project Larrabee / Knights Landing / Knights Mill product line, which could have thrust x86 into the AI space in a way no other platform does offers today. They just quit too soon. Which suggests lack of vision or inability to properly communicate that vision to the management / board / investors.
Cutting all those projects seems more like an inability to communicate those products to developers. I'm hardly pro - NVIDIA but somehow they've really excelled at convincing software developers that their product is the platform to build on
Everything AI already runs on x86. Knight's Mill had 72 cores with 244 threads with AVX512, connected to 16gb of 400gb/s local DRAM and up to 384gb of 115gb/s DRAM. Very similar bandwidth to GPUs of the era but with the addition of more DRAM on a second performance tier a little further away from the CPU.
AMD's having a ton of success with Epyc and Threadripper putting up very similar numbers.
The development story for Knight's Mill was a lot simpler than for any GPU. Consequently, I don't think developers were the ones who needed convincing.
32 comments
[ 3.4 ms ] story [ 21.7 ms ] threadIsn't squandering incredible engineering resources exactly how Intel got into this situation in the first place?
There is this interesting phenomenon where people outside an organization view R&D potential as elastic. That is to say, there is always a positive return on investment for the marginal R&D dollar.
It is clearly not in the case or every company would have exponential growth simply by reinvesting in R&D.
In reality, it is difficult to create growth from R&D at all, let alone the marginal dollar.
I don't think that's the view here; instead, I think the idea is that R&D personnel could be reassigned to work on something more profitable, rather than simply laying them off. Assigning R&D people to work on something unprofitable or a dead-end is the fault of management and their poor vision, not the people assigned to research that stuff.
Indeed, wouldn't every company love to just reallocate all of its employees to the areas that are most profitable. But alas, adding eight more women to the project of birthing a baby does not reduce pregnancy time to one month.
if they are more profitable, they probably have more space and budget by defauly. Outside of these strange times, it's not like Tech has historically been oversaturated and companies shy away from hiring new talent.
Not sure what you're basing this assertion on. I've never been at any tech company where the fact that a project is profitable somehow inherently means that we can just keep adding engineers to it and those additional engineers will magically increase the revenue the project generates.
In fact there's almost no industry where such an assumption would be logical. Even if your work has a very straightforward financial model, like manufacturing where you simply make Y dollars for every X parts you produce, that still does not imply you can just keep adding assembly line workers. At a certain point you have to consider whether there are even enough purchase orders to cover the volume of goods you're creating, or if they are just going to become wasted surplus sitting in a warehouse.
The corporate analog would be sustaining dozens of projects, partnerships, and business areas that never really took off and had the results you hoped for.
In the case of firing,
either way, companies are not staffed by omniscient leaders. The are staffed by humans, which don't know where to put R&D today to guarantee profitable inventions tomorrow.
Human uncertainty and imperfection underlies all business risk.
This is nothing new: even when I worked there decades ago, they were always making stupid decisions, such as the whole Itanic idea, the dumb idea of making their P4 processors dependent on expensive RAMBUS memory, refusing to make a serious 64-bit version of x86 until AMD did it first, etc. It seems they still haven't learned (or hired some better executives).
If the money printing machine works to well, everybody forgets how to repair it by the time it breaks down.
Their press statement essentially R&D is too bloated, dysfunctional, and disorganized in the current state to produce.
If they cut 20%, they will still have an obscene budget.
If you read the press release the idea is to focus on their core competencies, like making good processors. That means laying off people that can't help with that goal.
sure, the standard press response to layoffs. They'll be back to chasing AI in a year, despite more or less having already lost the race.
That said, if they keep failing to turn a profit, it wont matter.
Intel similarly cut their project Larrabee / Knights Landing / Knights Mill product line, which could have thrust x86 into the AI space in a way no other platform does offers today. They just quit too soon. Which suggests lack of vision or inability to properly communicate that vision to the management / board / investors.
AMD's having a ton of success with Epyc and Threadripper putting up very similar numbers.
The development story for Knight's Mill was a lot simpler than for any GPU. Consequently, I don't think developers were the ones who needed convincing.
More discussion on official release: https://news.ycombinator.com/item?id=41133084