IMO, That confirms that the QCOM exec is a bit clueless then during this interview.
Qualcomm CEO: Intel is not an option today. We would like Intel to be an option
- Thread starter Daniel Nenni
- Start date
IMO, That confirms that the QCOM exec is a bit clueless then during this interview.
Artificer60
Well-known member
Not my strong suit, but according to my good friend google. "semiconductor cell libraries with different heights offer distinct power, performance, and area (PPA) trade-offs. A taller standard cell can increase circuit performance at the expense of area and power, while a shorter cell sacrifices performance for improved density and lower power consumption."
It is my understanding that Intel 18A only offers cell libraries with two heights one for high density and another for high performance. A cell library for low power is not offered. I believe that is being addressed in 18AP and am almost certain it will be address in 14A. So the comment by the Qualcom CEO is correct. 18A does not offer the capability they need. It is not a condemnation of 18A overall, it just isn't designed to meet Qualcom's needs.
Another oversight in Pat Gelsinger's "if you build it they will" come approach to foundry.
No? I don't think the leakage current is that bad considering the original purpose of Intel products.
It's not that Intel's technology or processes are poorly developed.
Since it is originally a process for our own products, it is tailored to the purpose of the product.
In the first place, mobile devices other than laptops (or below) were outside of Intel's scope, so it can be said that they have little experience with processes and libraries suitable for them.
MKWVentures
Moderator
18A was announced in 2021 as undisputed leader for foundry. Pat thought every customer would sign up and be happy to use Intel. Everyone ran a test chip. No one has taped out. Qualcomm seems completely disinterested in what Intel does.
Intel needs to get some back up sourcing and 2nd waver sourcing commits ASAP.
Note: Internal is not the same as foundry..... what made Intel great on IDM (Manufacturing, TD, Design, marketing working together ... no IP arguments) is the opposite of what they need for foundry (deliver a stable product and low cost for multiple customers). LBT has already reset the "dont share info with customers, customers are potential enemies" culture... so that is good
Intel needs to get some back up sourcing and 2nd waver sourcing commits ASAP.
Note: Internal is not the same as foundry..... what made Intel great on IDM (Manufacturing, TD, Design, marketing working together ... no IP arguments) is the opposite of what they need for foundry (deliver a stable product and low cost for multiple customers). LBT has already reset the "dont share info with customers, customers are potential enemies" culture... so that is good
siliconbruh999
Well-known member
This exists for every customer? I will not be surprised for AMD/Nvidia though
hist78
Well-known member
He just wants to be polite, and for good reasons.
Was the intended market for 18A in 2021.. "everyone", or was it more HPC focused - where Pat was expecting/hoping to win over customers like Nvidia/AMD instead of Qualcomm?
That's one way to look at it.
I think the deeper game is it's in Qualcomms interest for Intel to fail and lose it's potential IDM advantages -- so the playing field is more level between Qualcomm and Intel for x86 Windows PCs. Qualcomm has been trying for a while to break into x86 server and PC markets - so the weaker Intel becomes, the better for them.. so they have little incentive to 'help' Intel by using their nodes.
hist78
Well-known member
I’ve always felt that Intel’s IDM business model is too complicated to manage. There are many serious internal and external conflicts of interest that make Intel difficult to manage, let alone win over customers who are also competing with Intel.
siliconbruh999
Well-known member
Samsung is an IDM as well and even more complex than Intel
MKWVentures
Moderator
excellent point.
Both Intel and Samsung were dominant leaders, very aggressive and very scary. Both are struggling now in this new world. Is it a coincidence?
And Samsung has fabbed Qualcomm chips though I think that's also due to Samsung buying silicon from Qualcomm for its phones..
siliconbruh999
Well-known member
so answer me if Samsung hasn't screwed over the foundry part and 10nm didn't happen do you think they would still struggle? It's all about decisions you need to make the correct decision. For Intel that include not screwing up AI/Optics/GPU and stuff
MKWVentures
Moderator
In 2021 Pat said Intel had 100 customers in the pipeline including 2 signed customers. They said Qualcomm was a customer. Then qualcomm clarified that they like the idea of another foundry but had no product plans. 18A was going to be the leadership node with many external customers. They built it and people did not come. So did 100 customers look at it and all decided no? was it the PDK issues and timing?
If you were fabless, would you sign up for 14A as a primary source for a volume product. would you spend the money to develop a sku on 14A?
If you were fabless, would you sign up for 14A as a primary source for a volume product. would you spend the money to develop a sku on 14A?
MKWVentures
Moderator
Personally, I think the issues is that both were arrogant and had a business model based on being the dominant leader and people had to work with them as they delivered and you had to work with them (reluctant customers). When they made the mistakes, the dominance changed, the "no one got fired for choosing IBM" changed, customers got screwed.
They need to both do a complete change based on not being the dominant force anymore and determine what their actual competitive advantage is. It is not being the Alpha Dog. I think LBT will fix this for product group.
If I were Nvidia or Apple sized, I'd be wary of being required to buy from a monopoly permanently going forward
But they did come, although they also left... The issue was, at least if we are to believe Zinsner, that 18A yields were simply rubbish and the process was not ready. So in reality Intel did not really build it. Going by what Dave said, yields on PTL are still not good enough and this is a small die.
Artificer60
Well-known member
Completely agree, Intel missed the PDK window for 18A. By the time they were where they needed to be the TSMC ship had sailed. If Intel foundry is going to work 14A PDK 0.5 must be out by end of Q4 or early Q1'26.