Discussion in 'Frontpage news' started by Hilbert Hagedoorn, Jan 20, 2021.
the machine lines running 7nm are not the ones running 6nm, different part of the factory.
Intel has never made a GPU from start to finish before, all their igp's are derived from aquisitions and larabee never made it out of RND (as a graphics card)
You do know smartphone companies like Apple buy a big chunk of the best process available? And with Nvidia breathing down AMD neck for process nodes, AMD will want some of that 6nm goodness too. Everyone will compete and some will be left out.
apple is on 5nm and 7nm, not 6nm.
That's not really true. Between process nodes, the only things that differ are the photolithograpy machines (steppers) and sometimes the photoresist material application machines because that material is usually different. All of the other three or so dozen types of machines are shared - things like furnaces, ion implantation, deposition machines, diffusers, wetbenches, etc. from one node to the next because the processes are still the essentially the same.
The photoresist appliers are one of my favorite machines in the fab. I used to stop and watch those things often. They use little, tiny Puma robots that are about two feet tall and make some great twisting moves as they take wafers out of the cassette and put them on the stage and then load them back into other cassettes. My machines were much more boring.
That'd be a problem, because then three GPU manufacturers will be bidding higher and higher for a larger cut of the pie, raising prices.
ooo whas dis?
I cant wait for @Hilbert Hagedoorn to make a Intel GPU/Drivers section here on Guru3D. Now crazy is that?
Why isn’t there more competition in the foundry business? Why are all the chipmakers dependent on these few foundries, for YEARS
Raytracing with 1FPS at best . Support is support
Here's to hoping so.
I'm looking for some serious competition in the coming year or so. Something to taper off this disgusting price scale that's taken effect as of late.
That and some more availability of hardware of course.
How can it be better with another hand in the pot and nothing inside to grab onto?
Foundry business is an extremely expensive and risky one, where companies have to spend billions to advance to better nodes.
There are very good reasons why 90% of top companies are fabless and use the foundries of others.
There are at least two reasons. Among them are the cost of entry is massive. A modern, state-of-the-art fab will cost in the $10B range - that's ten gigabucks. Second is the technology. As we can see, very few companies are capable of large scale manufacture of chips at ultra-small nodes. They can be counted on one hand with fingers to spare. Intel has failed spectacularly for a few years now and have been surpassed by Samsung and TSMC and you can be guaranteed others are also in close pursuit.
What I am surprised by is the lack of joint ventures. There are still some now, mostly involving memory, but not nearly as many as there used to be. This would be a great time for Intel to align themselves with someone like TSMC or Samsung and build a new fab somewhere. They might find more eagerness with one of the second-tier companies like Global Foundries, LG, or Philips. On the other hand, Gellsinger might pull some heads out of some backsides and get Intel back on course.
all true except TSMC actually has capacity and they're spending $60+ billion on *new* fabs @ 3n
the funny thing is Intel has already "spun" the production node but those fabs for the improved 7n are already in place and not at capacity.
i have the same p.o.v., but Intel actually has the infrastructure and logistical supply chains to fulfill demand. Intel doesn't need AIB's either. i don't expect the cards to be anything special but in the current market they don't have to be.
a completely mediocre but widely available card would be the Tortoise against the Nvidia/AMD Hare.
If there was free capacity at TSMC. Why would Sony/MS/AMD/nVidia have troubles with supply?
One can blame crypto-something on GPUs, but on consoles?
Wonder if Intel will use their new capacity to compete EDIT: (and) or build more to scale
Sony and M$ supply problems are a fraction of the issue with enthusiasts.
there are more next gen consoles being shipped to market than enthusiast product. period
the simple fact is that consoles are a far larger market. show me one gpu being given away in a national television marketing campaign you cannot because there are none.
No AIB is the size or has the market penetration of either Sony or M$, or even more importantly remember they are the goose laying the golden eggs for AMD and profit from the earliest iteration plus custom silicon. they also are not as dependent on gddr6 supply as their contracts predate the current situation or 30xx/rx5800/5900.
the situation for the consoles is a good thing for them with more pent up demand than for next gen gpus combined from all manufacturers. and the pace of delivery is far outstripping ANY gpu release.
it is truly apples and oranges
it's a holding action to remain relevant while new regime is doing what the old should've done two years ago. this simply keeps DG2 on time
Yeah I am happy about the new management... and I agree. Far as holding action... hmm I mean they’re still building more fabs why people say things like Intel is only leasing space from TSMC or whatever makes no sense to me. It’s no secret there’s yet another fab just about ready in Chandler