Intel Sunsets Network Switch Biz, Kills RISC-V Pathfinder Program (tomshardware.com) 33
Intel's disastrous Q4 2022 earnings found the company losing $661 million and its margins crashing to the lowest point in decades, so it isn't surprising that the company announced new cost-cutting measures. From a report: That includes news that it would no longer invest in new products for its networking switch business, effectively sunsetting the unit much like it recently decided to end its Optane Memory business. Surprisingly, Intel also pulled the rug from under its respected RISC-V Pathfinder program without a formal announcement, raising questions about its commitment to its other broad investments in the RISC-V ecosystem.
"NEX continues to do well and is a core part of our strategic transformation, but we will end future investments in our network switching product line, while still fully supporting existing products and customers," said Intel CEO Pat Gelsinger. "Since my return, we have exited seven businesses, providing in excess of $1.5 billion in savings," he added. However, Gelsinger also noted that he is still doing a thorough analysis across Intel's portfolio to look for other cost-saving measures in areas that don't generate strong returns. Intel's networking switch business stems from acquiring Barefoot networks in 2019 for an undisclosed sum (the company had raised $144 million over several investment rounds). The Tofino series of network switches gave Intel yet another tool in its arsenal of data center 'adjacencies' that it could leverage to expand its data center revenue. However, this unit faces stiff competition from entrenched players like Broadcom, Cisco, and Nvidia's Mellanox, making it an easy cost-cutting target.
"NEX continues to do well and is a core part of our strategic transformation, but we will end future investments in our network switching product line, while still fully supporting existing products and customers," said Intel CEO Pat Gelsinger. "Since my return, we have exited seven businesses, providing in excess of $1.5 billion in savings," he added. However, Gelsinger also noted that he is still doing a thorough analysis across Intel's portfolio to look for other cost-saving measures in areas that don't generate strong returns. Intel's networking switch business stems from acquiring Barefoot networks in 2019 for an undisclosed sum (the company had raised $144 million over several investment rounds). The Tofino series of network switches gave Intel yet another tool in its arsenal of data center 'adjacencies' that it could leverage to expand its data center revenue. However, this unit faces stiff competition from entrenched players like Broadcom, Cisco, and Nvidia's Mellanox, making it an easy cost-cutting target.
I used to think highly of Intel (Score:2)
Re: (Score:2)
What would it take to make you care even less?
Re: (Score:2)
Re:I used to think highly of Intel (Score:4, Interesting)
Same. Lying and cheating on benchmarks, intentionally crippling software so that it performed poorly on AMD by not using CPU instructions even though AMD CPUs supported them, holding the entire PC industry back with quad-core for almost a decade, constantly changing sockets forcing artificial motherboard sales, and incremental upgrades, is why I don't support Sintel.
A few years back I upgraded from an i7 4770K to a few Threadrippers and haven't looked back. For the family I built multiple AMD X570 gaming systems. AMD's socket AM4 is how you "win" customers -- by respecting their purchase so they can upgrade and giving them performance uplifts. Ryzen's performance uplift from 2xxx, 3xxx, 5xxx, and 7xxx has been awesome even if AMD has been jacking up prices lately. For gaming 5800X3D is a fantastic bang/buck.
I like to joke that AMD puts the Amd in Amdahl's Law. :-)
---
Someday Never A Straight Answer will be known as Notable Aliens Supplement Answers. As long as Scientists are ignorant about the extraterrestrials in our solar system Cosmology will remain an utter joke SWAG.
Re: (Score:1)
This is how you win (Score:3)
Not surprising... (Score:4, Interesting)
Torfino wasn't getting anywhere in the market, and didn't really seem to have an 'in', at least not one consistent with Intel business goals (between Mellanox, Broadcom, and Marvell, the gamut of switch silicon is pretty well covered.
RISC-V represents nothing but a threat to their cash cow. Again, the path for RISCV to succeed is as a platform less subject to licensing terms/costs than ARM, and Intel prefers even more locked down scenario than ARM.
In general, Intel has been like Google over the past few years. If it's a project outside one of a couple of their 'big' businesses, don't plan to count on it because it'll probably get killed. Optane, Omnipath, Nervana, Torfino, Knights series. The clock is ticking on their discrete GPUs...
Re:Not surprising... (Score:4, Interesting)
Intel and ARM are both underestimating RISCV. Sure, 5 years ago it was a big maybe, but I just watched an EEVblog this week featuring a small RISCV microcontroller going for $0.10. Not an announcement, actual silicon in hand.
Not too surprising on Intel's part, they have a history of fumbling in the embedded market. ARM may be resting on their laurels. I suspect that their price hike in 2020 was a big boost for RISCV, not because of it's size, but because it reminded a bunch of people that without a real competitor they would be at ARM's mercy.
Re: (Score:2)
Intel's challenge is right in the summary, they demand huge gross margin. There's no room in embedded for that sort of margin. In short, their business expectations are fundamentally incompatible with trying to compete with the standing embedded players.
Which long term is bad news for Intel, as the gap between embedded and 'performance' segments evaporates...
Re: (Score:2)
Very true. There's still a pretty big gap but the larger microcontrollers have been well able to handle light computing for a while. For example Raspberry Pi.
Today's cellphones would have been reasonable servers in 2000.
Doesn't matter (Score:3)
The suits spent billions on stock buybacks and they all get rich while gutting the company. https://www.reuters.com/busine... [reuters.com]
They could have spent the money on r&d but that never makes a quick buck.
Re: (Score:3)
And unfortunately it is hardly an Intel specific story, you can apply that statement to dozens of firms over the past few years and decades.
It's really an institutional rot in our corporate culture, I am not entirely sure the best method to turn the dial down on the shareholder-value-above-all operating principle but it's looking more and more like it has to be legislative in nature.
Re: (Score:2)
That isn't exactly "shareholder value". That's more "management value". And it is a for of rot designed into the corporate system. What to do about it is less clear. The incentives are misaligned, but how should they be aligned?
The only thing that occurs to me is that "performance bonuses" should be taxed at a rate dependent on the period covered. A 10 year coverage should be tax free, and a one quarter bonus should be taxed at over 50%. I'm not quite sure what shape that curve should take for interve
Re: (Score:2)
Over $10 billion in 2011 alone https://ycharts.com/companies/... [ycharts.com]
Re: (Score:1)
Not sure where you get the information on overcapacity from, but it definitely does not apply to reasonably current-process fabs (7nm/10nm). And one cannot use older-process lines to build chips that would meet today's requirements w.r.t. speed or power consumption.
Re: Let the death-spiral begin! (Score:1)
Now with domestic plants possible stop straining to meet gaming machine requirements it's just silly. There are more important benchmarks out there.
Short-sighted. (Score:4, Interesting)
Intel is acting like they are still top dog, relying on server processor sales. However, they seem to have a glut of their inferior and overpriced CPUs which tells me that AMD has done considerable damage to their core business. What they are doing now is effectively gambling on the hope that they will return to the top position while retaining their reputation (which they bought and paid for through anti-competitive business practices). The question is if they can really build a new microarchitecture without taking shortcuts that will end up as CVEs.
Re: (Score:3)
However, they seem to have a glut of their inferior and overpriced CPUs which tells me that AMD has done considerable damage to their core business.
My interpretation of the tea leaves is a bit different.
Intel's single biggest competition is...last season's Intel chips. A 4th-gen i5 is starting to show its age now...despite it being a decade old. While enthusiasts, video editors, and CAD pros might be able to tell a difference between a 12th-gen i7 and a 7th-gen, the market as a whole would be hard pressed to sense a performance difference.
Intel had a fantastic 2020 and early 2021, as the aging XP/Vista era P4 in the corner suddenly needed to be replace
Re: (Score:3)
Back in December we saw that one Genoa can damn nearly keep pace with TWO top-end 83xx Xeon
Re: (Score:2)
Your argument doesn't make sense because it's based on the assumption that the desktop/laptop/tablet market matters. However, 95%+ of revenue for Intel [theregister.com] comes from CPUs for the server market. AMD has started eating the server market [theregister.com] with it's EPYC processor line. Unless they can turn things around fast then they are in big trouble.
Histroy repeats? (Score:5, Insightful)
Re: (Score:2)
Intel sold ARM based CPUs for a while then, just as ARM usage started to take off, Intel sold off that division.
StrongARM didn't scale down as well as other ARM implementations for power efficiency, and they were going to have problems scaling it up too, so they dropped it.
Re: (Score:2)
Bonuses are tied to quarterly profits so that's what they'll maximize.
Within the decade Intel stock will be a hot potato.
Jump on RISCV now if you're retraining.
Intel network switches? (Score:2)
I'm quite familiar with Intel NICs. Never heard of Intel network switches before this.
Re: (Score:2)
In 2019, Intel bought an ethernet switch silicon startup.
They saw it as a way to jump into the ethernet switch ASIC market served by Marvell, Broadcom, and now nVidia and Cisco.
Problem was between those 4 players, there's hardly room for another player to be worth it. Particularly Intel isn't a player people have a lot of excitement or confidence in.
Re: (Score:2)
Re: (Score:2)
Re: (Score:1)
Intel was making chips with a relatively powerful CPU with switching fabric and things like NICs and wifi. I had a SoHo router with one of these chips and was looking forward to seeing something like this powering something like a banana pi or at least something that could take open firmware. Would have been a nice platform for running a few home services and a NAS but looks like it wont ever happen.
Shame to see them get rid of optane. I purchased several of these and they’re freaking fast if you d