News Intel Disconnects 5G Modem Business, Sells to MediaTek

[Admin]

Intel set to exit PC modem business as part of IDM 2.0 strategy.

Intel Disconnects 5G Modem Business, Sells to MediaTek : Read more

 

[sepuko]

Wireless wire...

 

[Alvar "Miles" Udell]

Think if I were Intel investors I'd sue to prevent this and instead sell it to Apple, who is in dire need of 4G and 5G tech to develop their own modem.

 

[InvalidError]

Think if I were Intel investors I'd sue to prevent this and instead sell it to Apple, who is in dire need of 4G and 5G tech to develop their own modem.

MediaTek supplies chips for anyone who wants them, Apple keeps everything to itself to turn its products into impenetrable and unserviceable black boxes. For consumers as a whole, I think MediaTek was among the best possible outcomes.

 

[suryasans]

This is an inevitable outcome after Intel gave up its Android based Smartphone business in 2016.

 

[digitalgriffin]

It was way behind the 8 ball delivering the 5G solution. Apple was not pleased and partners were backing out.

Seems to be a trend for Intel. While Pat is turning thing around and investing heavy, it may be too little too late. Them still pushing xe graphics was pretty vital. But the timing, environment, and Kadori's poor leadership has permanently closed the door. It is one that I seriously doubt will ever be opened again unless Intel gains huge margins on their CPUs again. And that isn't going to happen.

With the confirmed loss of Intel sgx/xe graphics this week (and I KNEW Intel would abandon it). It looks like we are stuck to a duopoly.

 

[InvalidError]

It is one that I seriously doubt will ever be opened again unless Intel gains huge margins on their CPUs again. And that isn't going to happen.

With the confirmed loss of Intel sgx/xe graphics this week (and I KNEW Intel would abandon it). It looks like we are stuck to a duopoly.

Pretty sure Intel still makes more than decent margins on most of its CPUs.

Last I heard, Xe is still planned at least until 2025, albeit with only single smaller designs for BMG and Celestial, which makes sense if Intel realized it cannot charge much over $250 for GPUs with half-baked drivers and basically no brand recognition in that product category. I'd argue that is what Intel should have aimed for all along: the A380 was far too weak to be of much interest and the A750+ are too expensive for most people to give the beta-test with unknown support life a shot. Arc would likely have fared much better had Intel aimed for $200-250 from the start and launched the A580 instead of the A380 using a sub-300sqmm die. No point in going after the medium-high-end that won't tolerate game support issues half as much until it is confident all of its ducks can hold a row.

AFAIK, Druid still has three chip designs planned, which means the final call hasn't been made yet. We'll see whether that remains as-is, gets whittled down to one chip or cancelled altogether.

By far the single strongest argument for Xe sticking around for the foreseeable future is that Arrow Lake is slated to use both BMG and Celestial GPU tiles, with little doubt that Intel will continue developing Xe for its future IGP needs. With IGPs now existing as GPU tiles, it would be pretty easy for Intel to slap an IO tile (possibly the exact same one CPUs would use, just add GDDR6/7 support to the memory controller if the GPU tile doesn't have its own) next to one and spit that out as dGPUs for those who want them for minimal additional R&D cost. It won't be breaking world records at ~2.5x the A380's performance for the biggest Celestial IGP but should be enough to keep the ~$200 crowd happy.

If you think putting GDDR7 capability on a CPU sounds silly, a Celestial-192 IGP would require ~50% more bandwidth than an A380 and DDR5 won't get you even half-way there on 128bits bus width. It would make sense for a non-upgradable budget gaming laptop to have 16-24GB of GDDR7 as system memory.

 

[kjfatl]

If Intel is successful with IDM-2, it is quite possible that they will buy MediaTek. and make their technology available to other Intel IDM-2 customers. For the moment the goal seems to be catching up and surpassing TSMC and Samsung by at least a year. Will they succeed? It all comes down to cash on hand to invest in bleeding edge manufacturing equipment.

 

[TechieTwo]

Pump and dump is the Intel way.

 

[bit_user]

As a part of Intel's strategy of devouring businesses that do not exactly fit its IDM 2.0 strategy

LOL, I think the author means to say divesting.

"Devouring" is more how I'd describe a company that's on an acquisition binge.

 

[bit_user]

With the confirmed loss of Intel sgx/xe graphics this week (and I KNEW Intel would abandon it).

Source? (other than youtube, if you can)

 

[bit_user]

If you think putting GDDR7 capability on a CPU sounds silly, a Celestial-192 IGP would require ~50% more bandwidth than an A380 and DDR5 won't get you even half-way there on 128bits bus width. It would make sense for a non-upgradable budget gaming laptop to have 16-24GB of GDDR7 as system memory.

Playstation and XBox have used GDDR for system memory for a while, now. There was even a Chinese console with a custom AMD APU that had the ability to boot desktop Windows, and I was sure hoping to see benchmarks to get a sense of how well it worked.

Pointedly, Microsoft switched from their strategy of using eDRAM, like in the XBox 360 and XBox One, over to Sony's model of unified GDDR. However, XBox One came along before HBM was really a thing, so we can't really read into it as an indictment against that approach.

 

[bit_user]

If Intel is successful with IDM-2, it is quite possible that they will buy MediaTek. and make their technology available to other Intel IDM-2 customers.

This would probably raise anti-trust flags, not to mention I highly doubt the Taiwanese government would allow the sale to go through. That's even before we get to the topic of whether it makes business sense, which I somewhat doubt - MediaTek is a SoC company that I think doesn't traditionally design much of the IP in its chips.

It all comes down to cash on hand to invest in bleeding edge manufacturing equipment.

No, it doesn't. To have a competitive fab business, you definitely need the latest equipment. However, that's only one ingredient.

Fabs aren't much different than other manufacturing businesses, in that the way you setup and use the equipment is key to what comes out at the end. The equipment is just a tool - it doesn't tell you how to use it. The equipment doesn't stipulate what your transistors look like, what they're made of, or how you build them. Among many other things.

 

[-Fran-]

So they have made it official now. Good riddance.

Regards.

 

[dalek1234]

...
"Devouring" ...

Freudian Slip, perhaps? Intel is kind of devouring itself slowly, but steadily.

 

[digitalgriffin]

Source? (other than youtube, if you can)

I do know someone. But I can't quote this person. But they were hearing hints for a while. Looks like Intel is aiming solely at the middle to lower ground only. Despite the "leak" that occured yesterday about 4070ti levels of performance coming, Intel is doing the opposite and going entry level only.

Anything else more than that is still up in the air. But it's certainly not good looking unless there is some miracle where everyone starts buying Intel xe chips.

The setup of xe being a separate division is proof Intel is looking to divest themselves of the unit financially. They might be looking for people to licence of their ip which is quite large.

 

[InvalidError]

The setup of xe being a separate division is proof Intel is looking to digest themselves of the unit financially. They might be looking for people to licence of their ip which is quite large.

I can't imagine Intel ditching its ability to grow its own in-house IGPs, especially as they plan to scale up their IGPs as separate tiles in the future. Though Intel retrenching Xe development to its IGP needs does mean we may not be seeing anything much beyond 50(0)-tier equivalent any time soon. Which is just as well since there is little chance of Intel catching up with AMD and Nvidia on driver maturity to a truly satisfactory level within the next 2-3 years anyway.