The Fall of Intel

dada_dave said:
Click to expand...


I'm really not sure what the game plan is here.

If they'd provided funds: i can see it as being a matter of national security to save intel.

But to not provide any funding? Though i do hear this was a demand to continue to get the CHIPS act funds.
 
Cmaier said:
arstechnica.com

Intel details everything that could go wrong with US taking a 10% stake

Intel warns investors to brace for losses and uncertainties.
arstechnica.com arstechnica.com

… in which it is revealed that the real reason Trump wants the stake in Intel is because the gov’t would then have inside information that he can trade stocks on in order to personally profit….
Click to expand...
abovethelaw.com

Intel Drops Funding HBCU Legal Pipeline - Above the Law

Surprise surprise.
abovethelaw.com abovethelaw.com

I’m not sure Intel needed to be told that they had to drop any diversity initiatives as part of the deal, but I’d bet the WH did so anyway.
 
www.tomshardware.com

Intel's 14nm+++ desktop CPUs are making a comeback — chipmaker inexplicably resurrects Comet Lake from five years ago with 'new' Core i5-110

So, should we call this 10th Gen or Core Series 1?
www.tomshardware.com www.tomshardware.com

Intel taking recycling seriously. Rereleasing a Comet Lake chip.

The Core i5-110 is a desktop processor, meaning it is compatible with an LGA1200 socket and either an Intel 400-series or 500-series motherboard. However, Intel has introduced two new sockets since LGA1200, so it's a mystery just how many consumers still have a LGA1200 motherboard that can accommodate the Core i5-110.
Click to expand...

Despite the Core i5-110 being a blatant rebrand, Intel is still charging the same price for the chip as it was when it launched five years ago. The RCP (Recommended Customer Price) for the Core i5-110 is $200, which falls within the same range as the Core i5-10400's $200 to $210. It's an insane price considering that 14nm+++ chips should be dirt cheap to produce by now.
Click to expand...

At $200, the Core i5-110 is supposed to be a value processor, but it's hard to see the value in it.
Click to expand...
 
www.tomshardware.com

Intel takes the wraps off Panther Lake — first 18A client processor brings the best of Lunar Lake and Arrow Lake together in one package

Panther Lake brings the power efficiency and graphics performance of Lunar Lake and the platform flexibility of Arrow Lake all together on one SoC
www.tomshardware.com www.tomshardware.com

Intel offered a high-level preview of single-threaded performance with Cougar Cove, suggesting that the new P-core can deliver 10% higher performance at similar power to Lunar and Arrow Lake, or 40% lower power at similar performance in less demanding workloads.
Click to expand...

The multi-threaded performance story for Panther Lake is a bit muddier. Intel claims that Panther Lake can deliver 50% more performance at similar power to Lunar Lake, or 30% lower power for multi-threaded performance similar to Arrow Lake-H.
Click to expand...

Of course, if we consider these charts in their entirety, Panther Lake can also deliver higher absolute performance at similar power to Arrow Lake-H, emphasizing the greater performance scalability of what we assume to be the 4P+8E+4LPE package.
Click to expand...

So a tweaked core design for both E and P cores and new node, but I’m not sure that even Intel’s own relative performance prognosticators are all that impressive?

Starting with multicore:


Unless I’m misunderstanding Intel is comparing a 4+8+4 Panther Lake chip to a 4+4 Lunar Lake chip to get the 50% performance improvement at the same power, so that number doesn’t really seem useful? The second number comparing it Arrow Lake are like more germaine.

The ST is also odd because according to notebookcheck data on actual laptops Arrow Lake and Lunar Lake had vastly different efficiencies in the wild and yet here they are comparable?


The ARL-LNL curves do separate (and PTL gets better too relative to both) at lower power but overall this must be an Int vs FP issue since the Intel curve in the chart is on SpecInt while the notebookcheck data is using CB R24, mostly a floating point test.

It’s also odd that they chose such a high power draw for their main stat given that again it seems that PTL actually gets better at lower power draw. Advertising 10% performance improvement for the same power draw doesn’t seem that great? The 40% power efficiency is nice, but overall it seems this is mostly down to the node improvement. And even then it’s not clear how big an improvement that node is, given that Intel is jumping from N3B. So if this is the result how will Intel’s 18A compare to TSMC’s upcoming 2nm? -though of course 18A is launching first and will be competing again N3P. I’m just not sure TSMC should be worried by that even if Intel had its act together in terms of being a foundry for 3rd parties?

I do like the tile system (unsurprisingly I know) but basically I feel like I read a list of extremely promising sounding improvements - first to ribbon-FET, first to BSPD, new cores (okay even Intel admits they’re optimized designs) - to get just a decent but not massive upgrade in the end. Like I felt it was building to a much bigger reveal that what it was and maybe than it needed to be. But we’ll see.
 
www.theverge.com

Intel’s tick-tock isn’t coming back, and everything else I just learned

I listened to Intel’s earnings call.
www.theverge.com www.theverge.com

Tan revealed today that Intel will also release new AI GPUs each and every year, following Nvidia and AMDin shaking up their traditional cadence to address the huge demand for AI servers. It’s not clear what that might mean for those hoping for more Intel gaming GPUs.
Click to expand...

the company reiterated it’s only launching one SKU of Panther Lake this year and slowly rolling out others in 2026. Here’s another possible reason why: Zinsner hinted today that Panther Lake will be a “pretty expensive” product to start with, and Intel’s going to have to push its existing Lunar Lake chips instead “in at least the first half of the year.”
Click to expand...

While Intel has repeatedly pushed back against the idea that its 18A process had poor yields, the company admitted to investors and analysts today that it’s not ready to be a huge financial success either: yields are “adequate to address the supply but not where we need them to be to drive the appropriate level of margins,” says Zinsner, suggesting that it might be 2026, or even 2027 for an “acceptable level of yields” in that regard.
Click to expand...

But don’t worry they say later that 14A is working out great!
 
www.macrumors.com

Apple and Intel Rumored to Partner on Mac Chips Again in a New Way

While all Macs are now powered by Apple's custom-designed chips, a new rumor claims that Apple may rekindle its partnership with Intel, albeit in a new and limited way. Apple supply chain analyst Ming-Chi Kuo today said Intel is expected to begin shipping Apple's lowest-end M-series chip as...
www.macrumors.com www.macrumors.com

Kuo now saying Apple might fabricate future base Mx chips with Intel in order to appease Trump (oh and diversify their manufacturing).
 
dada_dave said:
www.macrumors.com

Apple and Intel Rumored to Partner on Mac Chips Again in a New Way

While all Macs are now powered by Apple's custom-designed chips, a new rumor claims that Apple may rekindle its partnership with Intel, albeit in a new and limited way. Apple supply chain analyst Ming-Chi Kuo today said Intel is expected to begin shipping Apple's lowest-end M-series chip as...
www.macrumors.com www.macrumors.com

Kuo now saying Apple might fabricate future base Mx chips with Intel in order to appease Trump (oh and diversify their manufacturing).
Click to expand...
would make much more sense to use TSMC’s arizona plant. It’s not easy to redesign the physical blocks to accommodate two different sets of design rules. Hopefully Intel has at least stopped using mils in its design rules :-)
 
Cmaier said:
would make much more sense to use TSMC’s arizona plant.
Click to expand...
I suppose that’s where diversification comes in but yeah

Cmaier said:
It’s not easy to redesign the physical blocks to accommodate two different sets of design rules. Hopefully Intel has at least stopped using mils in its design rules :-)
Click to expand...

Supposedly Intel now uses “industry standard tools” in order to attract customers but, agreed, I can’t imagine designing the same generation processor under two very different fab’s rules would be easy even so. I believe Apple did that exactly once when switching between Samsung and TSMC.
 
dada_dave said:
I suppose that’s where diversification comes in but yeah



Supposedly Intel now uses “industry standard tools” in order to attract customers but, agreed, I can’t imagine designing the same generation processor under two very different fab’s rules would be easy even so. I believe Apple did that exactly once when switching between Samsung and TSMC.
Click to expand...
having done something like this (two different processes, albeit our own fabs), what ends up happening, probably, is you create your own “least common denominator” design rules that work with both fabs. But you inevitable give something up in doing this. For example, fab A might say minimum polygon area is 1nm^2 and minimum M0 spacing is 0,5nm, and fab B might say minimum polygon area is 1.2nm^2 and minimum M0 spacing is 0.3nm. So you end up with 1.2nm^2 and 0.5nm. Worst of both worlds.

And that’s assuming both processes give you equivalent number of metal layers and that the layer thicknesses are the same, dielectrics are the same, etc. If the layer thicknesses and dielectrics vary, then you have additional problems. You end up over engineering to make cycle time and satisfy hold times on both processes, and the design process takes a lot longer.
 
If they really are diversifying like that I feel like it would make sense to fab low performance need chips like S chips and H chips at Intel and A and M at tsmc
 
I don't know much about fabs, other than ES2 (European Silicon Structures) from long ago. But I suspect sometime in the future Taiwan will not exist as country, and will become another Chinese province. With TSMC becoming a Chinese company and having a different name.

As a result, it seems China would have a ton of leverage over US companies and their tech products, in both commercial and defense/aerospace/military spaces, having to rely on Chinese fabs (including TSMC's Arizona fab). Thus I can see the motivation for getting Intel and perhaps other US fabs (Texas Instruments) up to snuff.

The above is just me spitballing. Thoughts?
 
Isn't the point of pushing for TSMC fabs outside of Taiwan to enable to company to survive outside of a Chinese Taiwan?

As an aside, doesn't the fact that western companies are unable to compete with the Chinese model of the state sponsoring strategic industries (combined with stealing key know-how), allowing them to undercut pricing whenever they choose and driving the competition out of business, suggest we may need to adopt some of their tactics, like propping up a flailing Intel, heavy investing in rare earth metal production, etc, in order to avoid catastrophic dependencies?
 

Similar threads

Eric
The fall of Tesla
4 5 6
Replies
108
Views
4K
Eric
Eric
Back
Top