You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
- There have been endless attempts to replace x86 in Windows. Intel themselves couldn’t even do it at the height of their powers.
- Nvidia’s been largely unsuccessful in launching their ARM chips into markets in the past (Tegra)
- Nvidia launching with those specs sounds premium/expensive - so limited market for a while
- Right now the AI market is ‘all new’, so there’s no market for Intel to lose (yet).
I think ARM on Windows might be getting mature enough to get a slow but steady market share — like ChromeOS, but I don’t see it really breaking Intel’s back. If Intel executes their 20A/18A as well as they expect (not to mention N3 products), they’re likely to take back more market share from AMD than they’ll lose to Windows on ARM over the next 2-3 years.
If you look at it from a fab standpoint, Intel is already going to be competing against a whole host of N3 products over the next couple of years.
This doesn't make a lot of sense to me. For example, my understanding is that the new Cortex X5, which itself is an unknown quantity, is aimed at Android applications, not an OS like Windows with its very complex memory management. However, Arm's web site says nothing about the -X5, it's all leaks and speculation, so no facts are available.
Qualcomm and Apple use custom-designed Arm microarchitectures with an Arm architecture license for their laptop SoCs. So, Nvidia beats Qualcomm and Apple with just an off the shelf Arm IP CPU? Then how is Nvidia going to differentiate against Qualcomm, no less Intel and AMD. Especially AMD. There is the Arm Cortex-X4 program, which is customizable:
The Arm Cortex-X Custom (CXC) program allows partners to shape the target performance point of Arm Cortex products.
www.arm.com
But still mobile/Android targeted.
Then there's the memory in package thing. I'm a huge fan of that for lower latency and lower power, but Apple makes it work with its sophisticated Unified Memory Architecture, which takes specific memory controller support and OS support. So Microsoft is going to put something like that in Windows? If not, Nvidia will have the same memory disadvantage Intel and AMD have, meaning they need more DRAM than Apple does for similar performance.
I'm just not convinced of anything yet, including the new Cortex-X5 popping out and being a Qualcomm competitor to anyone with license dollars.
By now, everyone’s heard that Nvidia is coming next year in the Windows ARM arena with new SoC’s to compete with Qualcomm and others.
They’ll be using, TSMC N3P, Cortex X5 BlackHawk CPU cluster, Blackwell GPU, and LPDDR6 on package.
Microsoft is porting all Windows 11 AI stuff to Nvidia hardware.
Is this the straw that breaks Intel’s back, or a snowball that evolves into an avalanche?
Certainly not a nail in a coffin but x86 is under attack from all sides and it is not just Arm. RISC-V is a tsunami and I don't see how x86 or Arm is going to stop it. On the other side Google, Meta, Amazon, Microsoft and maybe even Apple are making silicon for data centers. Not to mention the China cloud companies and Huawei. With AMD getting more competitive, Intel is in a tough spot for sure but sometimes the best innovation comes when your back is to the wall, been there done that.
Fair though an install base of billions is a few orders of magnitude different than what the mainframe had at its peak. Only now are we even talking about fully removing 16-bit capability on x86 chips.. X86S.
Certainly not a nail in a coffin but x86 is under attack from all sides and it is not just Arm. RISC-V is a tsunami and I don't see how x86 or Arm is going to stop it. On the other side Google, Meta, Amazon, Microsoft and maybe even Apple are making silicon for data centers. Not to mention the China cloud companies and Huawei. With AMD getting more competitive, Intel is in a tough spot for sure but sometimes the best innovation comes when your back is to the wall, been there done that.
Are you thinking the RISC-V tsunami hits x86 mainly in dense data center, or more than that? (It seems like RISC-V has the capability to hit ARM everywhere)
Certainly not a nail in a coffin but x86 is under attack from all sides and it is not just Arm. RISC-V is a tsunami and I don't see how x86 or Arm is going to stop it. On the other side Google, Meta, Amazon, Microsoft and maybe even Apple are making silicon for data centers. Not to mention the China cloud companies and Huawei. With AMD getting more competitive, Intel is in a tough spot for sure but sometimes the best innovation comes when your back is to the wall, been there done that.
Given this rumor is 3 weeks old, and INTC has been trading at same dismal $30 (down 55% since Pat) for the last 4 weeks, it’s clear that ~nobody believes Intel will be receiving an order from Nvidia, anytime soon, if ever.
TSMC makes +95% of Nvidia wafers, Intel makes zero
TSMC makes +95% of MediaTek wafers, Intel makes zero
TSMC makes +95% of 3nm wafers, Intel makes zero
Given this rumor is 3 weeks old, and INTC has been trading at same dismal $30 (down 55% since Pat) for the last 4 weeks, it’s clear that ~nobody believes Intel will be receiving an order from Nvidia, anytime soon, if ever.
TSMC makes +95% of Nvidia wafers, Intel makes zero
TSMC makes +95% of MediaTek wafers, Intel makes zero
TSMC makes +95% of 3nm wafers, Intel makes zeroearly
Given this rumor is 3 weeks old, and INTC has been trading at same dismal $30 (down 55% since Pat) for the last 4 weeks, it’s clear that ~nobody believes Intel will be receiving an order from Nvidia, anytime soon, if ever.
TSMC makes +95% of Nvidia wafers, Intel makes zero
TSMC makes +95% of MediaTek wafers, Intel makes zero
TSMC makes +95% of 3nm wafers, Intel makes zero
When the world’s innovation and technology future as well as todays economy is but a short few kilometers away from a country we are trying to prevent them from getting any of that goodness and future in what is a war, not good for the west to be happy and gloat.
Everyone should be rooting for Intel, even TSMC wishes for their return relevance and competitive technology.
Business ALWAYS benefits from competition and x86 when it was the monopoly under Intel saw what happened.
@fansink When the smoke clears, I believe we will see that TSMC is helping Intel survive. Intel now has leading silicon options available at a reasonable cost. We may see the impact this year with the ramp of new, very successful products based on TSMC silicon. TSMC use has been a huge benefit to Intel .
@BruceA The percent of sub Intel 7 at TSMC will actually increase until mid 2025. Intel has shown detailed ramp plans. It actually doesn't go down until 2026 when Intel ramps 18A and Intel 3. This will become more clear when we see Lunar Lake, Arrow Lake details and when we see the ramp of Meteor Lake. Most of the sub 7nm Silicon Intel sells is from TSMC through 2025
Given this rumor is 3 weeks old, and INTC has been trading at same dismal $30 (down 55% since Pat) for the last 4 weeks, it’s clear that ~nobody believes Intel will be receiving an order from Nvidia, anytime soon, if ever.
TSMC makes +95% of Nvidia wafers, Intel makes zero
TSMC makes +95% of MediaTek wafers, Intel makes zero
TSMC makes +95% of 3nm wafers, Intel makes zero
INTC is largely viewed as a Dividend stock and not a growth stock. They’re going to need to actually put real numbers on the board to move the stock price — 10% AI market share, >$10B in foundry wins at “3nm” or below, etc.
They are in the middle of their turn around plan now - the crux being 18A in 2025, products that put Zen on the back foot, and High NA 14A in 2027. 2024 is still a year driven by decisions made by Pat’s predecessor (TSMC N3 prebuys, lack of foundry capacity).
Also don’t forget, in the future Intel can certainly take customers from Samsung..
INTC is largely viewed as a Dividend stock and not a growth stock. They’re going to need to actually put real numbers on the board to move the stock price — 10% AI market share, >$10B in foundry wins at “3nm” or below, etc.
They are in the middle of their turn around plan now - the crux being 18A in 2025, products that put Zen on the back foot, and High NA 14A in 2027. 2024 is still a year driven by decisions made by Pat’s predecessor (TSMC N3 prebuys, lack of foundry capacity).
Also don’t forget, in the future Intel can certainly take customers from Samsung..
Intel’s fall from leadership is well documented both from narrow focus on process as well as singular x86 and now has missed two huge pivots mobile and AI.
It took decades luck, vision and great disciple and lots of crazy hard work for TSMC and Nvidia to roar to the top. Seemed only yesterday it was Apple and TSMC and a short decade ago WIntel.
In hindsight the failures are so obvious but looking into the future nothing really is predictable 10 years or even 5 years out.
Anyone who looks at Intel as a dividend play is misguided, maybe a decade value play. This is sure, TSMC is woefully under valued but again Buffett couldn’t be convinced it was a good risk value play giving the geopolitics. Nvidia kudos to them and their vision, perseverance and execution but they should not be a 5T company, simply would be crazy.
Speaking of TSMC and Intel manufacturing - does anyone find it interesting that even Intel, notorious for delays on its product roadmaps, has pushed/delayed everything but not Lunar Lake? Lunar Lake was planned to come out in Q4 2024 or H1 2025, and they've already announced it is coming out in Q4. Lunar Lake is all TSMC silicon, and we haven't seen Intel products actually launch on time for a while.
Really speaks to the fact that TSMC's "trusted foundry" should be considered almost completely trustable. Even Intel, which struggle to even release their products on time according to roadmap and has numerous delays - they still have not had a delay on their all TSMC products*.
*Unless you count the ARC A series on N7, which arguably was the driver teams fault.
Given this rumor is 3 weeks old, and INTC has been trading at same dismal $30 (down 55% since Pat) for the last 4 weeks, it’s clear that ~nobody believes Intel will be receiving an order from Nvidia, anytime soon, if ever.
TSMC makes +95% of Nvidia wafers, Intel makes zero
TSMC makes +95% of MediaTek wafers, Intel makes zero
TSMC makes +95% of 3nm wafers, Intel makes zero
Considering the revenue, profit, and technology strength, Intel is a product company, not a service company. Intel needs a lot innovative and attractive products to justify its value. IMO, building Intel Foundry Service is putting Intel's precious time, money, and management attention on the wrong place. There are many companies who own no fab but doing very well with their unique product offerings. Most of them are competing directly against Intel.
Pat Gelsinger became Intel's CEO on February 15, 2021. Here I compare the market caps between February 16 or 17, 2021 and May 29, 2024 among several key semiconductor players.
