News Intel 2Q25 Earnings

[poke01]

Intel doesn't need it, but saying it isn't a value add is an overstatement.
Google "site:reddit.com mediatek wireless".
Or "site:reddit.com apple broadcom wireless bootcamp"
Or "site:reddit.com atheros linux"

Intel wireless chipsets are a pretty safe recommendation. That Intel laptops almost always come with a good wireless chip is some value.

oh good wifi chips are super important. I had a lenovo that a broadcom wifi card and its was BAD. When I was still in high school, I had to go to IT a lot to get the wifi card replaced.

Even with a new card the problems did not stop cause broadcom can't make good drivers.

-----------------

Intel's vertical integration of CPU, GPU, Thunderbolt, Wi-Fi and laptop design support make them unstoppable. I 100% believe Panther Lake will be a success due to this, AMD nor Qualcomm even come close. Its also the scale they operate in, the OEMs give Intel 1st dibs
Sucks Intel has to deal with Microsoft though thats a bottleneck they will never get rid off

 

[itsmydamnation]

Probably the right decision at the time.

But a lot has changed in very short time, and Broadcom went from partner to competitor - to some extent eating AMD lunch in AI deployments.



We will see. The one piece of puzzle (*) that AMD is missing badly (UA Link switch) is the the segment that Broadcom exited (stabbed AMD in the back), so buying one of the other vendors with promising solution would not even be "angering Hock Tan"

(*) AMD would lo like to have a full rack AI solution with all of the crucial parts being AMD.

Ultra Ethernet Switches - a lot of companies are doing, some even in collaboration with AMD Pensando, but, whoever has a good UA Link switch, AMD will have to buy them eventually anyway, so why not do it sooner (for less money) than later (for a lot more money).

So IA fabrics are the dumbest /simplest IP networks, it really is all speeds and feeds for the SOC ( switch on Chip) switches which is by far and away the easiest things to do in network forwarding hardware. Out of the entire AI stack switches are the least important / most commodity of the entire stack ( i say this as a ~20 year network eng/consultant/architect ) and also there are many vendor ASIC's that AMD can buy as whiteboxes if need be ( can thank the hyperscales for that).

AMD has the important bits , the GPU / the DPU/smart nic and the pci fabric/cpu. the ethernet fabric is just an underlay for RDMA and so long as it has the speeds and feeds + whatever basic network stack they run ( TRILL / EVPN / whatever DCB ) can be replaced easily. Let me tell you mellenox IP switches (because they were InfiniBand switches running IP code ) are the biggest hunk of **** i've never had more frustration and displeasure and i've admin'd switches running catOS.

To bring this back on vendor, the stupid thing is for like 15-20 years Intel made the best Nic adapters, its insane how they had all the pieces of the puzzle like a decade before anyone else and couldn't put it all together.

 

[igor_kavinski]

To bring this back on vendor, the stupid thing is for like 15-20 years Intel made the best Nic adapters, its insane how they had all the pieces of the puzzle like a decade before anyone else and couldn't put it all together.

Similar to their other failures. Complacency and resting on their laurels and not even caring seriously about what the world was doing because their monopoly machine aided by their legal and marketing departments ran too well for too long. Long enough to bring about their ruin.

 

[coercitiv]

Intel wireless chipsets are a pretty safe recommendation. That Intel laptops almost always come with a good wireless chip is some value.

Only one thing was bad in the last laptop I got from Lenovo - the Atheros wireless card. The fix was not more driver or more firmware, it was Intel wireless card.

I think it will be a loss for them, not financial but rather in terms of perception, it makes their laptops look more stable and reliable.

 

[jpiniero]

Despite the name, I don't think the wireless and ethernet chips for PCs are a part of NEX.

They may still get out of that anyway.

 

[igor_kavinski]

Killer NICs, no!

 

[511]

btw the wireless chips are already absorbed into clients iirc need to check cause they have cut the important part of NEX into CCG and DCG

 

[511]

Only one thing was bad in the last laptop I got from Lenovo - the Atheros wireless card. The fix was not more driver or more firmware, it was Intel wireless card.

I think it will be a loss for them, not financial but rather in terms of perception, it makes their laptops look more stable and reliable.

i have a Intel wifi card as well and it's more advanced than my router by 2 Gen

 

[NTMBK]

Intel needs to make a big deal to fill their fabs, and they need to make savings by getting rid of products that distract from their core focus. I think that they should make a big deal with Nvidia to do both in one go.

• License Nvidia graphics for their integrated graphics, and get out of graphics development entirely.
• Help Nvidia port their consumer GPUs to Intel 14A, and produce their next mainstream GPUs

It gives Nvidia capacity outside of TSMC for their "less important" non-AI GPUs, it secures a big customer for Intel, and it helps make their laptop chips more competitive against AMD graphics.

 

[jpiniero]

Intel needs to make a big deal to fill their fabs, and they need to make savings by getting rid of products that distract from their core focus. I think that they should make a big deal with Nvidia to do both in one go.

• License Nvidia graphics for their integrated graphics, and get out of graphics development entirely.
• Help Nvidia port their consumer GPUs to Intel 14A, and produce their next mainstream GPUs

It gives Nvidia capacity outside of TSMC for their "less important" non-AI GPUs, it secures a big customer for Intel, and it helps make their laptop chips more competitive against AMD graphics.

The IGPs as is are very competitive in everything but gaming.

 

[511]

They are not canni their Xe IP Unless Nvidia give them a big foundry deal.

 

[Win2012R2]

License Nvidia graphics for their integrated graphics

Anti-trust won't let this happen

 

[igor_kavinski]

Intel needs to make a big deal to fill their fabs, and they need to make savings by getting rid of products that distract from their core focus. I think that they should make a big deal with Nvidia to do both in one go.

• License Nvidia graphics for their integrated graphics, and get out of graphics development entirely.
• Help Nvidia port their consumer GPUs to Intel 14A, and produce their next mainstream GPUs

It gives Nvidia capacity outside of TSMC for their "less important" non-AI GPUs, it secures a big customer for Intel, and it helps make their laptop chips more competitive against AMD graphics.

Nvidia has shown with its past dickish behavior (original Xbox, Apple laptop solder fiasco etc.) that it cannot be trusted as a reliable partner. Plus, they have bad blood with Intel. They would either charge wayyy too high royalties or ask for something Intel wouldn't want to give them, like asking to be the No.3 x86 chipmaker with same level of patent portfolio access that AMD enjoys and AMD probably would object to that too.

 

[511]

x86 license is off they table what you can get is Intel CPU Chiplets

 

[jpiniero]

I was saying a couple years ago that Intel needed to dump Altera

Altera (until they ran it into the ground lol) was a profitable fab filler... and the products are very yield friendly. It's exactly the kind of business you would want to help keep the fabs full.

 

[511]

Man Intel's braindead take of using custom pdk and design rules screwed everything.

 

[NTMBK]

The IGPs as is are very competitive in everything but gaming.

Nobody cares about IGP performance outside of gaming. Spreadsheets run just fine on an Ivy Bridge IGP, and video decoding is handled by hardware decoder blocks.

They are not canni their Xe IP Unless Nvidia give them a big foundry deal.

I mean that's what I was proposing

Anti-trust won't let this happen

Why? There are plenty of companies making their own graphics architectures - AMD, Nvidia, ARM, Imagination, Qualcomm, Apple. And anti-trust were fine with Samsung licensing AMD IP for their smartphone chips. It's a deal that helps two US companies and helps keep US chips manufacturing afloat, the US government will wave it through.

 

[igor_kavinski]

Man Intel's braindead take of using custom pdk and design rules screwed everything.

Their custom PDK is far superior to the rest of the world. That's how they got to 6.2 GHz on Intel 7 and baked their CPUs at the same time.

You can't get that with any other PDK. Pay respects where due

 

[511]

Their custom PDK is far superior to the rest of the world. That's how they got to 6.2 GHz on Intel 7 and baked their CPUs at the same time.

You can't get that with any other PDK. Pay respects where due

Oh yes but that doesn't mean they need to continue doing it a big reason in the difference between PPA for Internal vs External for 18A even though the PDK is industry standard still has some custom Intel stuff for Intel only use.

 

[Joe NYC]

So IA fabrics are the dumbest /simplest IP networks, it really is all speeds and feeds for the SOC ( switch on Chip) switches which is by far and away the easiest things to do in network forwarding hardware. Out of the entire AI stack switches are the least important / most commodity of the entire stack ( i say this as a ~20 year network eng/consultant/architect ) and also there are many vendor ASIC's that AMD can buy as whiteboxes if need be ( can thank the hyperscales for that).

AMD has the important bits , the GPU / the DPU/smart nic and the pci fabric/cpu. the ethernet fabric is just an underlay for RDMA and so long as it has the speeds and feeds + whatever basic network stack they run ( TRILL / EVPN / whatever DCB ) can be replaced easily. Let me tell you mellenox IP switches (because they were InfiniBand switches running IP code ) are the biggest hunk of **** i've never had more frustration and displeasure and i've admin'd switches running catOS.

To bring this back on vendor, the stupid thing is for like 15-20 years Intel made the best Nic adapters, its insane how they had all the pieces of the puzzle like a decade before anyone else and couldn't put it all together.

Funny then how AMD can deliver 2 generations of datacenter GPUs and in the same time frame, no one can deliver the network switch for UA Link consortium.

BTW, I have also had good experience with Intel Nics, but don't know how well Intel kept up with the DPU trends, or if what Intel now has is more commoditized.

Also interesting that wile Intel is selling these (networking divisions), AMD, is buying networking companies. Pensando was the big one and more recently some fiber optic co-packaging related company...

 

[gdansk]

And anti-trust were fine with Samsung licensing AMD IP for their smartphone chips.

Why is that similar? This deal would involve Intel leaving the graphics business. Nvidia wouldn't touch it with a 10 foot pole. The potential cost of any regulator attention is not worth whatever Intel could offer them.

 

[Win2012R2]

until they ran it into the ground lol

Thanks to 10nm...

Who is going to prepay big money for newest 14A node, wait years praying it will actually work when they can do the same with TSMC and no exec will lose their bonus if suddenly TSMC misses deadlines?

External juicy premium customers won't touch it with a bargepole, and bargain basement won't be going for such new node.

 

[Win2012R2]

And anti-trust were fine with Samsung licensing AMD IP for their smartphone chips.

It's AMD, not Nvidia (dominant player in graphics - including client) - anti-trust won't have a problem for Intel to license AMD gpu, which already happened once, when Intel clearly used that to learn things and screw them over later, luckily for AMD they got Raja to do it.

Having said that these days lots of things are transactional and laws and common sense count for ****. Seems like very low priority market for Nvidia right now, why would they want to dilute their hefty margins and tank stock? Maybe they just should spin off consumer graphics division into separate firm.

It's exactly the kind of business you would want to help keep the fabs full.

What's the sell volume of FPGA like Altera or Xilinx even? These products so expensive they don't have price published, they are making low volumes of highly sophisticated stuff that isn't for everyone. I don't mean here old cheap high volume lines, those ain't being made on latest node.

The potential cost of any regulator attention is not worth whatever Intel could offer them.

How much would Intel even offer them? Nintendo totally cheapened out on Switch 2 (as they always do), this uses basically Ampere on 8nm process (5 years old and it was not the best at the time!), would Nvidia even get $50 per Intel CPU with their iGPU? Seems very doubtful, and fixed costs won't exactly be trivial, Jensen will simply not invest his engineers for this.

 

[adroc_thurston]

Altera (until they ran it into the ground lol) was a profitable fab filler... and the products are very yield friendly. It's exactly the kind of business you would want to help keep the fabs full.

That's why Intel bought them to begin with, lmao.

 

[Thibsie]

That's why Intel bought them to begin with, lmao.

Then Intel is even more clueless than I thought.