Intel Courting Samsung to Manufacture Xe GPUs?

[mtcn77]

What if he just went there to see if he (Intel) can source memory chips now that Micron and Intel have a falling out over 3D.

Intel: Hey, Micron can you provide Vram for our upcoming GPU

Micron: Sure just pay the higher price

Intel: Eh, Samsung how much for Vram

Intel is at 2nd spot in gross semiconductor revenue whereas Samsung is 1. Intel is also struggling with volume which makes it all the more ironic.

 

[PhantomTaco]

He is just operating on outdated information, https://www.anandtech.com/show/1155...nm-plans-three-generations-700-mm-hvm-in-2018

That is what Zen2 would be on if Glofo hadn't dropped the ball, TSMC all the way now.

No, he was right, I was referring to TSMC, not Glofo. Sorry:

https://www.anandtech.com/show/13445/tsmc-first-7nm-euv-chips-taped-out-5nm-risk-in-q2
"TSMC initiated high-volume manufacturing of chips using its first generation 7 nm fabrication process (CLN7FF, N7) in April. N7 is based around deep ultraviolet (DUV) lithography with ArF excimer lasers. By contrast, TSMC’s second-generation 7 nm manufacturing technology (CLN7FF+, N7+) will use extreme ultraviolet lithography for four non-critical layers, mostly in a bid to speed up production and learn how to use ASML’s Twinscan NXE step-and-scan systems for HVM"

The first generation of "7nm" coming from TSMC uses DUV, which is essentially the same immersion lithography we are using today with ArF excimer lasers, and have used for the last several nodes and honestly is nothing special really. The second generation will use EUV selectively for only 4 "non-critical" layers. The fact that they're specifically only using it for non-critical layers is indication that they still haven't worked the kinks out of it yet and are only using it in limited scenarios to better understand ASML's new tooling.

A lot of people don't understand just how massive a shift EUV represents (this isn't directed at you in particular btw, this is just for people's general knowledge). You're going from using a 193nm laser with refractory optics and immersion lithography to a 13.5nm beam which is literally absorptive by just about everything. That means first and foremost we go from using refractory optics to only reflective optics that are composed of 20+ alternating layers of molybdenum and silicon just to make sure we're managing to reflect most of the light. Immersion lithography is out because water absorbs this wavelength as well. These new machines also have to have the entire process run in a vacuum for the same reason. It isn't so much that I'm making a jab at AMD and TSMC as I am trying to highlight that this shift is fundamentally the biggest we've had literally ever. Photolithography started with i-line mercury lamps at 350nm, then went to KrF excimers at 248, then we hit the current ArF excimers at 193nm and we've been there since

 

[BorgOvermind]

As I said in the comment of the previous news about intel making better competitive graphics, they are just trying to get closer to AMD and nV. There is no chance for them to compete with the top products any time soom.

 

[Totally]

Were you like, absent in the newsfeed then or something?

I know, had me puzzled the as the only thing that wasn't known about the 9900k was whether for certain there was TIM or solder under the IHS.

MCM Xeons were just odd because they came just after the Ryzen 'glued together' hoopla.

 

[mtcn77]

.
A lot of people don't understand just how massive a shift EUV represents (this isn't directed at you in particular btw, this is just for people's general knowledge). You're going from using a 193nm laser with refractory optics and immersion lithography to a 13.5nm beam which is literally absorptive by just about everything. That means first and foremost we go from using refractory optics to only reflective optics that are composed of 20+ alternating layers of molybdenum and silicon just to make sure we're managing to reflect most of the light. Immersion lithography is out because water absorbs this wavelength as well. These new machines also have to have the entire process run in a vacuum for the same reason.

It highlights a basic fundamental principle that you cannot have your cake and eat it, too. This is much like why we cannot have a full electron beam in lithography, same principle: the more charged the particle, the more interactive and decaying it is, the less half life and refractive properties it has. You spend your energy on friction when you need it on the substrate. Components would warp from excessive wear. Have to focus an electron microscope for calibration? By the time setup is complete, you just punch a hole in the looking glass. It has no refractory tolerance, its like harnessing laser illumination for photography, scouring the template in the process.

Personaly, I don't look forward to DUV to end. Look at the statistics,

1. We barely have enough photoscanners to run refractory lithography,
2. By the time EUV kicks in, against disbelief, proportionally, we will initially need an inversely high count of photoscanners by its volume production ramp standards because we would have now switched to reflective lithography and excess component wear will surge demand for the pellicles like flies drop near an electric fly killer.
3. Pellicle test stations, both used in DUV and EUV, will be much more important - their production units are already booked for the next 2.5 years - and with the rapid turnover of pellicles in EUV, rather than DUV, will highlight this exacerbation of EUV's inherent incompatibility to volume production even more.

Also, vacuum environment is a joke. Electron beams aren't light, per say, they're electromagnetic radiation. The moment they hit the pellicle, high energy radiation is reflected back from the layers, if only it was just as powerful and noninteractive. There is no vacuum for radiation. Maybe we can have quantum MR 'decapacitor' rings around the chamber stepping down the energy states within the cabinet in the future, though. I'm not discounting electron state deescalation, or its more general expression quantum magnetic effect - has been done in nuclear physics - however I have worries; eventhough suppressed electrons aren't helping to cool off the pellicle, that heat is still building up and will sooner or later warp the thing unless production slows. Now, the important question is: how do you cool, if you are in a vacuum?