News Bill Gates says Intel has lost its way, hints that 'brave' Pat Gelsinger exited too soon

[bit_user]

I'd posit that it's specifically setbacks in process tech advances that is the current cause of their woes.

Probably. However, they tried to stick with x86 for too long. As long as they had an unassailable process lead, that could work in the server market, but even that wasn't enough to make Xeon Phi viable against Nvidia. Had they moved beyond x86 in the server realm, they might also be more competitive on their Intel 3 node.

Those CEOs allowed Intel to fall behind, and now they're seen as trailing TSMC,

It's not just appearances. By Intel's own admission, its Intel 3 node isn't on par with TSMC's latest.

I do think Intel can come back, and their new stuff shows promise.

Depends. Intel 18A needs to ramp before TSMC has already surpassed it. Let's just say the past several years don't instill me with confidence that'll happen. Intel 14A will be the next chance for that to happen, but I wonder whether it could be suffering from their current funding woes.

 

[Eximo]

Or, maybe they get too profitable, then Wall St. gets too greedy and starts siphoning off too much (in the form of dividends and stock buybacks), and the company is slowly starved of the means to continually reinvent itself, in order to stay relevant.

Maybe Xerox, that one seems to be a series of mismanaged attempts to merge with other hard copy and printing companies. Certainly seems like they are playing to the stock price.
IBM has shifted to a services company and sold off their PC division creating Lenovo. So from a management area they did well, but IBM Compatible PC outlasted them which is why I was mentioning them.
Kodak just didn't bother getting into digital cameras because their main business was film, it was too late by the time they realized people were making the switch.

 

[bit_user]

Maybe Xerox, that one seems to be a series of mismanaged attempts to merge with other hard copy and printing companies. Certainly seems like they are playing to the stock price.

IBM has shifted to a services company and sold off their PC division creating Lenovo. So from a management area they did well, but IBM Compatible PC outlasted them which is why I was mentioning them.

Kodak just didn't bother getting into digital cameras because their main business was film, it was too late by the time they realized people were making the switch.

Xerox and Kodak both predicted their own demise and tried to adapt, but it just proved too big of a leap for them to ramp revenue in the new industries quickly enough to offset their losses in the old ones. Xerox famously had PARC, where most of the innovations that later informed the Apple Macintosh originated. Kodak actually did do R&D into digital cameras, but I can't say why those didn't come to fruition early enough and well enough for it to take a leading role in that market.

Don't forget that IBM also had a substantial microelectronics division. I think they sold their fab to GloFo? IIRC, its 7 nm process was a continuation of work originally done by IBM. They still design mainframe and POWER CPUs, to this day. They also own Redhat, which has been a profitable acquisition for them.

 

[Eximo]

Xerox and Kodak both predicted their own demise and tried to adapt, but it just proved too big of a leap for them to ramp revenue in the new industries quickly enough to offset their losses in the old ones. Xerox famously had PARC, where most of the innovations that later informed the Apple Macintosh originated. Kodak actually did do R&D into digital cameras, but I can't say why those didn't come to fruition early enough and well enough for it to take a leading role in that market.

Don't forget that IBM also had a substantial microelectronics division. I think they sold their fab to GloFo? IIRC, its 7 nm process was a continuation of work originally done by IBM. They still design mainframe and POWER CPUs, to this day. They also own Redhat, which has been a profitable acquisition for them.

Kodak invented the digital camera, and sat on it for the most part. While their competitors started shrinking them down into handhelds and displacing the physical film market, where they made their money.

Yes, these are companies that used to innovate and then failed to compete, and ended up moving into other businesses through acquisitions, etc. The whole point was mentioning successful companies that innovated, and then fell behind.

Last time I saw a Xerox product was actually an amazing piece of technology, but low volume. Plotter sized color photo copier. You could literally print out a roll of color blueprints, and feed it into this thing and it would roll out another copy for you. Crazy expensive, but cool. Allowed someone to mark up a blueprint by hand and make copies.

IBM shifted to services and data science. I already mentioned they are doing fine. They used to make typewriters too.

My point was that Intel doesn't seem to be going down these paths. Everyone is saying doom and gloom, but they still have a lot of assets and innovation happening. They just aren't #1 any longer (technology wise), but they haven't faded away in their main industry. Not like we are going to see them stop making CPUs anytime soon. And even if x86 fades away, there is still a lot that Intel could do in the computer components business besides CPUs.

 

[JamesJones44]

I think what they did with Hololens and VR was very impressive! Hololens was truly ground-breaking. Their VR platform was the first to rely entirely on inside-out tracking. Since then, they've been eclipsed by Apple, but MS got there way sooner.

That was really a case of mispredicting the market. VR and AR just never took off like they thought they would. I wish MS didn't shutdown those programs, but I understand they just couldn't keep shoveling more money into that pit, without any foreseeable way to turn a profit.

Very true! Hololens was definitely ahead of its time.

 

[bit_user]

My point was that Intel doesn't seem to be going down these paths. Everyone is saying doom and gloom, but they still have a lot of assets and innovation happening. They just aren't #1 any longer (technology wise), but they haven't faded away in their main industry. Not like we are going to see them stop making CPUs anytime soon. And even if x86 fades away, there is still a lot that Intel could do in the computer components business besides CPUs.

Their CPU core design team is strong. Lunar Lake is good. Arrow Lake... less so, but I believe the problem is neither Lion Cove nor Skymont. I'm eager to see how Nova Lake turns out.

 

[JamesJones44]

Kodak invented the digital camera, and sat on it for the most part.

The work Kodak did on OLED alone tells the story that they didn't just sit around after creating filmed cameras. LG's OLED business is what it is today thanks in large part to Kodak. The digital camera was really their down fall, the shift happened too fast with too many players for them to overcome it. They didn't have enough business diversity and OLED wasn't big enough at the time to keep them from hemorrhaging money.

Don't get me wrong Wall St. played a part too, but it wasn't due to just simply sitting back and collecting and redistributing cash.

 

[phead128]

I do think Intel can come back,

I doubt it. They replaced an engineer's engineer CEO with two interim bean counters CEO (the same types that got Intel into this trouble in first place). These bean counters are salivating at the prospect of spinning off IFS foundry as separate entity, and focus on lucrative Intel products.

 

[Eximo]

Their CPU core design team is strong. Lunar Lake is good. Arrow Lake... less so, but I believe the problem is neither Lion Cove nor Skymont. I'm eager to see how Nova Lake turns out.

I'm still optimistic about Intel.

Arrow Lake is like a lot of their other first run major architecture changes. A drop or equality with previous gen hardware. It shows potential. If they continue it will be interesting to see if they can claw back the clock speed advantage.

Alder Lake was good in the end, but in those early days with all the software issues it wasn't a great pick.

Raptor Lake, technically fine, they just made mistakes on the firmware side of things. Raptor Lake refresh was unexpectedly nice for LGA1700 board owners, but wasn't the original plan, so not sure that is a positive.

Nova Lake looks good on paper. But if they can bring i7 like performance to the i5 class of chips, I think they will have a winning solution.

 

[DS426]

Maybe Bill Gates should also comment on Microsoft losing it's way.

 

[flowingbass]

When IFS wanted to follow suit with Global Foundries short of idolizing them, that's when Intel started to decline rapidly.

I am still baffled and amused that they suddenly decided to be lazy and hang back from being on the cutting edge like they been good at for so long.

Just letting themselves get overwhelmingly overtaken by competing foundries.
They had the money and talent so seriously what gives. A real head scratcher

 

[thestryker]

Intel and Boeing, what happened man.

The exact same thing is at the core of both company's rot: emphasis on profit above all else. While with Boeing you can look at the McDonnell Douglas merger as the catalyst there isn't something of that sort with Intel. For Intel it was as they became a dominant player in data center (they were already in client) money people became more dominant on the board. This caused a shift of overall company focus and led to financial oriented CEOs being appointed.

 

[emike09]

Also, in the same way, camera manufacturers are stagnant compared to cell phones. Canon, Nikon, and even Sony are lagging behind in photo and video production. I get that you can do a lot with what you have, but it requires professional level understanding to maximize the results. Besides Depth of Field, I can produce something incredible from the lenses on my phone in an instant. With my Sony A7c camera, it takes me hours to recreate a shoot that still doesn't look as good.

 

[rluker5]

I'm still optimistic about Intel.

Arrow Lake is like a lot of their other first run major architecture changes. A drop or equality with previous gen hardware. It shows potential. If they continue it will be interesting to see if they can claw back the clock speed advantage.

Alder Lake was good in the end, but in those early days with all the software issues it wasn't a great pick.

Raptor Lake, technically fine, they just made mistakes on the firmware side of things. Raptor Lake refresh was unexpectedly nice for LGA1700 board owners, but wasn't the original plan, so not sure that is a positive.

Nova Lake looks good on paper. But if they can bring i7 like performance to the i5 class of chips, I think they will have a winning solution.

I too am still optimistic about Intel. I just saw KitGuru's rereview of Arrow Lake and the small gain in efficiency over Raptor Lake is pretty striking considering 7nm DUV vs N3B EUV. 18A could be a leap in efficiency that will be somewhat masked by Intel's insistence on the highest possible clocks no matter how far outside the efficiency range they go.
And the firmware on Raptor Lake and associated motherboards is really bad. The chips are fine, but the whole firmware and default LLC settings are a complete mess. Once you fix those you can run faster than stock with easily safe volts. As the years go by and RPL chips are still fine and fast, I hope the associated power management debacle gets the credit it is due for being as bad as it is.

But this keeping the memory controller off of the compute die like in Arrow Lake will have to be dealt with. Either kill the added latency with faster communication or bury it with cache. In my fantasyland they would just link the memory controller to the compute tiles by having it be in the base die, but who knows what will actually happen with those latency issues.

Battlemage is also really efficient in games. Close to the 40 series. Arc is catching up fast. And as far as AI upscaling, Intel has 150 games and AMD has 40 that are currently eligible and 11 more upcoming if all FSR 3.1 games can be converted to FSR 4 when that comes out. Not going to be a good revenue source at the low end unless Intel starts making the chips for one of the two consoles though. Which doesn't seem that far fetched with my B580 usually pulling in a little over 100w in games. Just add 16 e-cores for another 50w. Celestial should be even better.

 

[bit_user]

Maybe Bill Gates should also comment on Microsoft losing it's way.

He's not going to throw MS under the bus, however he has no such entanglements with Intel. That's why his take on Intel is worth noting - because he's probably not holding back.

 

[bit_user]

When IFS wanted to follow suit with Global Foundries short of idolizing them, that's when Intel started to decline rapidly.

I am still baffled and amused that they suddenly decided to be lazy and hang back from being on the cutting edge like they been good at for so long.

That timing is coincidental. It came just as the pandemic-era surge in cloud computing demand was waning. That's also when Gelsinger got his legs under him and had really come to grips with the scale of the structural problem he faced.

To put it another way, Intel dropped not because of their IFS strategy, but their IFS strategy was a response to the ballooning fab costs that Intel was increasingly struggling to support on the backs of its own products.

Just letting themselves get overwhelmingly overtaken by competing foundries.
They had the money and talent so seriously what gives. A real head scratcher

A decade of bad management (i.e. the ones preceding Gelsinger) squandered it, back when they most needed to be investing in their future. Look at the 10 nm debacle.

 

[bit_user]

Gates just could not bring himself to praise or even mention AMD in any way.

It wasn't the focus of the interview, but rather a side-question some one asked him during an interview about his memoir. If he really did an interview on the modern computing landscape, I'm certain he'd mention AMD.

Still Macroshite ignore AMD and cosy up to a rapidly fading giant.

They used AMD for their consoles over the past dozen years, and ATI before that. They don't ignore AMD, but Intel has been their focus due to its market dominance. Software developers need to prioritize support for the hardware that the substantial majority of their users have, and that's been Intel.

The weirdest part of your remark is that it completely ignores Microsoft's pivot towards ARM and their close partnership with Qualcomm in porting & optimizing Windows on ARM. Did you know Microsoft even has its own ARM-based cloud CPUs?

 

[bit_user]

keeping the memory controller off of the compute die like in Arrow Lake will have to be dealt with. Either kill the added latency with faster communication or bury it with cache.

Arrow Lake's problems start at L3.

"L3 load-to-use latency on Arrow Lake is north of 80 cycles from a P-Core, compared to about 52 cycles on Lunar Lake. The cycle count penalty is high enough that Arrow Lake’s actual L3 latency is higher than Lunar Lake’s, even though Arrow Lake runs at higher clocks."

Source: https://chipsandcheese.com/p/examining-intels-arrow-lake-at-the
​

In this case, I take the stronger result for Lunar Lake to indicate the ring bus is having scaling problems.

The difference in memory latency between Intel and AMD isn't as big, as you can more easily see here (also, note the effect of Lunar Lake using LPDDR5X):

Source: https://chipsandcheese.com/p/analyzing-lion-coves-memory-subsystem
​

Arrow Lake did add an entirely new layer to their cache hierarchy, enabling them expand the P-cores' L2 cache to 3 MB/core. That really wasn't enough to offset its relatively weak L3 bandwidth and high L3 latency. IMO, that's what they need to fix.

Raptor Lake had these same issues. This tells me the problem is more complex than the memory controller being on a separate die from their CPU cores.

 

[OldAnalogWorld]

In fact, the sly Bill Gates makes such injections only for one purpose - to confuse the masses, while he quietly buys up Intel shares at the bottom of the market, realizing that they will soon start to grow as soon as the lag is over. And he, unlike ordinary people, has criminal insider information about the state of affairs in Intel and AMD. AMD shares have been falling continuously for 3 years, this does not look like a market leader and its market conditions are improving. Perhaps he buys up AMD, because no one is going to stop printing money in the USA. And that means the market has only one way - up in nominal prices...

 

[jp7189]

I think what they did with Hololens and VR was very impressive! Hololens was truly ground-breaking. Their VR platform was the first to rely entirely on inside-out tracking. Since then, they've been eclipsed by Apple, but MS got there way sooner.

That was really a case of mispredicting the market. VR and AR just never took off like they thought they would. I wish MS didn't shutdown those programs, but I understand they just couldn't keep shoveling more money into that pit, without any foreseeable way to turn a profit.

I had a chance to try hololens, and it was really good, but limited in way that made it hard to envision a lot of use cases (it had some). I'm still an avid vr guy, and am insanely furious about MS decison to remove WMR from the latest Win11 build. It's completely bricks VR headsets that rely on it, including the still excellent Reverb G2.

 

[bit_user]

I had a chance to try hololens, and it was really good, but limited in way that made it hard to envision a lot of use cases (it had some).

Which version? Remember that v1 was just a development platform and not a real product. V2 was offered mainly for business customers. I think if MS had kept working, with the same intensity, they could be at a point of offering something almost comparable to Apple's.

I'm still an avid vr guy, and am insanely furious about MS decison to remove WMR from the latest Win11 build. It's completely bricks VR headsets that rely on it, including the still excellent Reverb G2.

Hmmm... it seems that the Monado OpenXR framework has experimental support for WMR HMDs.

https://monado.freedesktop.org/#supported-hardware
​

I'm not sure whether you can use it for SteamVR on Linux, but I believe there's some path for getting that working.

 

[castl3bravo]

"Boeing, and a lot of other heavy engineering industries are suffering a bit from brain drain. Last 5-10 years has seen retirement of a lot of experienced engineers, and in a lot of cases, wholesale outsourcing of engineering overseas."

There are unintended consequences in an MBA's biz plan or a VC's "private equity" scam.

 

[jp7189]

Which version? Remember that v1 was just a development platform and not a real product. V2 was offered mainly for business customers. I think if MS had kept working, with the same intensity, they could be at a point of offering something almost comparable to Apple's.


Hmmm... it seems that the Monado OpenXR framework has experimental support for WMR HMDs.

​

https://monado.freedesktop.org/#supported-hardware​

​

I'm not sure whether you can use it for SteamVR on Linux, but I believe there's some path for getting that working.

V1. In my opinion, the biggest strength is the biggest problem too... not being able to block out the real world. Really cool for some applications, but also really limited. I think hi res pass-through is the way to go rather than a micro lens array which I'm not sure could ever the fov and immersion that's needed for some things.

I'll have to check out monado. Thanks for the heads up. I hadn't seen that yet.

 

[bit_user]

V1. In my opinion, the biggest strength is the biggest problem too... not being able to block out the real world.

Doesn't Apple's HMD have a way to do that?

I think hi res pass-through is the way to go rather than a micro lens array which I'm not sure could ever the fov and immersion that's needed for some things.

I prefer Magic Leap's approach.

I'll have to check out monado. Thanks for the heads up. I hadn't seen that yet.

If you like getting your hands a bit dirty, then it's probably worth a look. If you want something that just works out-of-the-box, you might be disappointed.

 

[DS426]

He's not going to throw MS under the bus, however he has no such entanglements with Intel. That's why his take on Intel is worth noting - because he's probably not holding back.

No, of course he won't. 😛