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!
(Reuters) -Intel said on Wednesday that Microsoft plans to use its services to manufacture a custom computing chip and that the company expects to beat an internal deadline of 2025 to overtake its biggest rival, Taiwan Semiconductor Manufacturing Co, in advanced chip manufacturing. The American...
(Reuters) -Intel said on Wednesday that Microsoft plans to use its services to manufacture a custom computing chip and that the company expects to beat an internal deadline of 2025 to overtake its biggest rival, Taiwan Semiconductor Manufacturing Co, in advanced chip manufacturing. The American...
For context, the article is referring to the chip fab tech, not volume of Foundry sales:
"the company expects to beat an internal deadline of 2025 to overtake its biggest rival, Taiwan Semiconductor Manufacturing Co, in advanced chip manufacturing."
Today Nvidia announced FY2024 earnings and it's a blowout. Recently I'm thinking if IFS and 5 nodes in 4 years is a wrong focus or wrong subject for Intel?
I believe Intel, AMD, and Nvidia's AI chips are all manufactured by TSMC. While Nvidia grow revenue crazily, lntel's revenue shrunk. With the same availability of TSMC manufacturing capabilities to Intel, AMD, and Nvidia, it's the product design, product vision, and execution made the huge difference.
Intel urgently needs killer products to grow or at least maintain its revenue. IFS' revenue is too small to solve the much bigger Intel product problems.
I believe Intel, AMD, and Nvidia's AI chips are all manufactured by TSMC. While Nvidia grow revenue crazily, lntel's revenue shrunk. With the same availability of TSMC manufacturing capabilities to Intel, AMD, and Nvidia, it's the product design, product vision, and execution made the huge difference.
Intel urgently needs killer products to grow or at least maintain its revenue. IFS' revenue is too small to solve the much bigger Intel product problems.
I agree Intel needs killer products soon, but we really need to see what the Intel products do on 3, 20A, and 18A.
The Intel 3 server products close the gap substantially with AMD, Meteor Lake on Intel 4 is pretty good competition for AMD on mobile despite being a first generation product in many ways. Lastly on the desktop - Intel is mostly holding ground now despite being on an older process. AMD won’t be able to launch Zen 5 until 2H 2024 and it will be a TSMC N4 product facing Arrow Lake on Intel 20A. Intel should flip the table on at least 1 of those 3 markets by year end, maybe 2.
I think (hope) the board sees IFS as a long term sustainability play after this..
P.S. Gamers Nexus did an updated deep dive on Intel ARC graphics. The short version was, there is still some variability in quality of drivers for certain games (including AAA Starfield), but he also said in a lot of ways Alchemist is very dangerous for AMD on the graphics front, offering much higher performance per dollar. Hopefully Intel can keep going in this space..
I agree Intel needs killer products soon, but we really need to see what the Intel products do on 3, 20A, and 18A.
The Intel 3 server products close the gap substantially with AMD, Meteor Lake on Intel 4 is pretty good competition for AMD on mobile despite being a first generation product in many ways. Lastly on the desktop - Intel is mostly holding ground now despite being on an older process. AMD won’t be able to launch Zen 5 until 2H 2024 and it will be a TSMC N4 product facing Arrow Lake on Intel 20A. Intel should flip the table on at least 1 of those 3 markets by year end, maybe 2.
I think (hope) the board sees IFS as a long term sustainability play after this..
P.S. Gamers Nexus did an updated deep dive on Intel ARC graphics. The short version was, there is still some variability in quality of drivers for certain games (including AAA Starfield), but he also said in a lot of ways Alchemist is very dangerous for AMD on the graphics front, offering much higher performance per dollar. Hopefully Intel can keep going in this space..
"I agree Intel needs killer products soon, but we really need to see what the Intel products do on 3, 20A, and 18A."
The problem is that the market and users won't wait. They will buy the best chips that are available in order to compete and to survive.
If Nvidia can have TSMC to make the $40,000 H100 and the world desperately want to get it as many as possible, why Intel AI chips (made by TSMC) can't get the same market reactions? I don't think this is a manufacturing technology issue and I struggle to understand why Intel 3, 20A and 18A will solve it.
For context, the article is referring to the chip fab tech, not volume of Foundry sales:
"the company expects to beat an internal deadline of 2025 to overtake its biggest rival, Taiwan Semiconductor Manufacturing Co, in advanced chip manufacturing."
(Reuters) -Intel said on Wednesday that Microsoft plans to use its services to manufacture a custom computing chip and that the company expects to beat an internal deadline of 2025 to overtake its biggest rival, Taiwan Semiconductor Manufacturing Co, in advanced chip manufacturing. The American...
I was at the conference. The CEO of Microsoft did say via video call that they would "try one design on Intel 18A". I feel other companies will do the same, try a design on Intel 18A to see how well it performs. Intel did say they they feel 18A with backside power delivery will be ahead of TSMC N2 without backside power delivery. To that I agree. The question I have is how long will it take foundry customers to adopt Intel BSPD? Intel can certainly brag about internal Intel designs using BSPD but I want to see a customer use it.
The $15B Pat says he has for foundry business is for the lifetime of the orders, not one year. Which I took to mean the projected lifetime of the design.
It was a full day conference with many things discussed. Just keep posting these misleading stories and I will tell you what actually was said. It was a VERY interesting conference.
Also Sam Altman denied saying he was going to raise $7T to make AI chips. He said do not believe everything you read, especially if it is AI generated.
Today Nvidia announced FY2024 earnings and it's a blowout. Recently I'm thinking if IFS and 5 nodes in 4 years is a wrong focus or wrong subject for Intel?
I believe Intel, AMD, and Nvidia's AI chips are all manufactured by TSMC. While Nvidia grow revenue crazily, lntel's revenue shrunk. With the same availability of TSMC manufacturing capabilities to Intel, AMD, and Nvidia, it's the product design, product vision, and execution made the huge difference.
Intel urgently needs killer products to grow or at least maintain its revenue. IFS' revenue is too small to solve the much bigger Intel product problems.
Developing IFS and designing great products for the merchant chip markets are orthogonal. Intel's chip design problems, like in GPUs, have nothing to do with Intel fab investments, and never have. Intel is not under-investing in chip design, if anything they overspend. Intel design team management has been misguided and complacent for a long time, and a fab process advantage often saved them from disaster. Of course, that advantage is gone. An additional problem is that Gelsinger is probably the least likely CEO to drive fixing Intel's design mess. We should note that both superstar tech leaders Raj Koduri and Jim Keller worked for Intel and quit, which tells you a lot I think about the dysfunction in Intel's design teams.
Developing IFS and designing great products for the merchant chip markets are orthogonal. Intel's chip design problems, like in GPUs, have nothing to do with Intel fab investments, and never have. Intel is not under-investing in chip design, if anything they overspend. Intel design team management has been misguided and complacent for a long time, and a fab process advantage often saved them from disaster. Of course, that advantage is gone. An additional problem is that Gelsinger is probably the least likely CEO to drive fixing Intel's design mess. We should note that both superstar tech leaders Raj Koduri and Jim Keller worked for Intel and quit, which tells you a lot I think about the dysfunction in Intel's design teams.
If Intel design teams are “dysfunctional” and if design is getting commoditized as evidenced by cloud companies like Amazon, MSFT, Google and others starting to design silicon, then doesn’t it make sense for Intel to invest more in fabs as a backup business plan?
If Intel design teams are “dysfunctional” and if design is getting commoditized as evidenced by cloud companies like Amazon, MSFT, Google and others starting to design silicon, then doesn’t it make sense for Intel to invest more in fabs as a backup business plan?
We should note that both superstar tech leaders Raj Koduri and Jim Keller worked for Intel and quit, which tells you a lot I think about the dysfunction in Intel's design teams.
There have been a couple of good interviews with Jim where he described that the mentality of everyone inside “The Moores Law company” was that “Moore’s Law was dead”. It’s very hard to re-energize everyone when the culture is at that point, though he did leave on the watch of the previous CEO.
I know Raja has a reputation as a superstar, and while I know he’s certainly a far superior engineer than I’ll ever be, I’m not so sure he “quit” Intel. I hope someday we get the full story on the birth and initial delivery of ARC graphics (with some related insight on the GCN/Vega/RDNA years at AMD).
There have been a couple of good interviews with Jim where he described that the mentality of everyone inside “The Moores Law company” was that “Moore’s Law was dead”. It’s very hard to re-energize everyone when the culture is at that point, though he did leave on the watch of the previous CEO.
I know Raja has a reputation as a superstar, and while I know he’s certainly a far superior engineer than I’ll ever be, I’m not so sure he “quit” Intel. I hope someday we get the full story on the birth and initial delivery of ARC graphics (with some related insight on the GCN/Vega/RDNA years at AMD).
I don't get it. As an architecture & design guy, I'd relish the end of Moore's Law, if and when it ever happens. Imagine how important innovative chip architecture and design would have to become to keep adding value over time. If that happened I might actually have to go back to work. (BTW, I'm not trying to be humorous. I'm serious.) But, alas, Moore's Law is not dead, and even though it has slowed other forms of chip innovation are emerging.
I agree Intel needs killer products soon, but we really need to see what the Intel products do on 3, 20A, and 18A.
The Intel 3 server products close the gap substantially with AMD, Meteor Lake on Intel 4 is pretty good competition for AMD on mobile despite being a first generation product in many ways. Lastly on the desktop - Intel is mostly holding ground now despite being on an older process. AMD won’t be able to launch Zen 5 until 2H 2024 and it will be a TSMC N4 product facing Arrow Lake on Intel 20A. Intel should flip the table on at least 1 of those 3 markets by year end, maybe 2.
I'm not sure how much Granite Rapids is going to close the gap with AMD's Turin. It has the same Redwood Cove core found in Meteor lake, which basically has no IPC improvement at all: https://www.tomshardware.com/pc-com...enchmarks-show-ipc-regressions-vs-raptor-lake
Meanwhile, Turin has Zen5 which is a major upgrade over Zen4 - at least 15% IPC improvement.
Also, most Arrow Lake SKUs are using TSMC N3 - only the low end SKUs use Intel 20A according to rumours.
I'm not sure how much Granite Rapids is going to close the gap with AMD's Turin. It has the same Redwood Cove core found in Meteor lake, which basically has no IPC improvement at all: https://www.tomshardware.com/pc-com...enchmarks-show-ipc-regressions-vs-raptor-lake
Meanwhile, Turin has Zen5 which is a major upgrade over Zen4 - at least 15% IPC improvement.
Also, most Arrow Lake SKUs are using TSMC N3 - only the low end SKUs use Intel 20A according to rumours.
Fair - My comments were thinking Intel has a chance to lock desktop and mobile with 20A/18A (not guaranteed, but possible); Server is definitely a different beast. Arrow Lake is supposedly a major architecture change from Raptor Lake, and is a pretty drastic process change (Intel 7 —> 20A). Intel is already beating AMD on desktop application performance on a ‘7nm’ class node vs ‘5nm’ (though losing heavily in efficiency). Even if a substantial portion of ARL ends up on TSMC N3 that’s still a node ahead of Zen 5.
Intel is way out of the box on server, their next generation may put them in the ballgame enough to stop market erosion at least, and probably win in some smaller markets (sizeable niches that depend on certain accelerators that Intel will have that AMD doesn’t).
I’ll admit this is a little optimistic. Either way we’ll really know at the end of the year how it’s going. AMD Zen 6 is a 2026 product..