I'm actually looking more forward to what they do APU wise with this after. Intels big little stuff... sandwiched with a real APU could be much more interesting. I think over the next 2-4 years things like the Deck are going to be what drive gaming.
It might sound crazy... but I am more looking forward to a Intel Nuc type device with Say 4 big cores 2 little and a decent Arc (I guess) GPU running some fast system ram and Steam OS. To would be a super interesting product to me... more then a third player in a GPU space where everyone is using TMSC anyway so stock issues and price gouging to go with don't change anyway.
Can Intel build a better console, that’s basically what that would come down too.I think this is where it's going too, and I really like the idea of a nuc-like PC running steamos with an advanced APU in it. Something 2-3x the power of the Steam Deck's processor could be a real game changer for small form factor gaming systems, or "living room" systems running steamOS in big picture mode as a viable "PC gaming console".
I like that idea a lot.
Can Intel build a better console, that’s basically what that would come down too.
A big part of that being a possibility is going to really depend on what their Linux GPU drivers look like.
It is, pretty close but that of course included, console, mobile, and PC gamers and there is undoubtedly some double entries in there.
ex. You use an outlook.com email for your Xbox, but use gmail for steam, and a secret Apple.com for your mobile. Unless you’ve got them all linked somewhere public that one user counts as 3 for their studies.
The description of the report I linked says that they are not counting duplicates. It is supposed to be a truly unique count. I'm not paying $3k to find out, though.Exactly, if you have a cell phone, a console, and a PC, you are counting as 3 gamers in their numbers.
Unless they do something insane like price them at 199 msrp that is.
These cards will probably be a good option on Linux as long as they're decent. Intel is quite active in the Linux kernel community and there is no doubt the cards will be supported.Intel's drivers are pretty solid, I kinda hope they lay off any creature features and just put out stable kick-ass drivers, preferably for Linux too.
The cards they put out for Aurora were supposedly better than what AMD and NVidia were offering up at the time so that’s at least remotely promising. The GPU’s they’ve tossed into the 11th gen Mobile parts aren’t too bad either even at 10nm I would imagine that on their 7nm and TSMC 5 things only get better. I’m hopeful, I love that AMD is pushing them, I just hope they can continue to do so for the foreseeable future.These cards will probably be a good option on Linux as long as they're decent. Intel is quite active in the Linux kernel community and there is no doubt the cards will be supported.
Or something insane like… have them available at launch….I think they are shooting themselves in the foot by waiting until q1 2022 and targeting specifically the 3070 for performance. The likelihood of AMD or Nvidia paper launching a new product during that quarter are high and it wont exactly make the run of the mill consumer excited when performance is compared to "old" products.
Unless they do something insane like price them at 199 msrp that is.
Intel is supposedly upgrading the following fab's over to 7nm:Or something insane like… have them available at launch….
It's also a semiconductor manufacturer with nearly 3x the revenue of NVIDIA and AMD combined. And one that has been making (integrated) GPUs for quite a while. Of course be hesitant but more competition is something to look forward to.Some of the suggestions here sound downright ridiculous considering the circumstances. A company with no track record of producing good GPUs or supporting said GPUs is going to have a lot to prove before demanding Nvidia or AMD pricing. Until we see actual reviews there is absolutely no reason for excitement here IMO.
I'm a bit confused, is this consumer Arc GPU (Xe-HPG DG2 codenamed Alchemist) being made on TSMC 7nm, 5nm, or 3nm? Or do we not know? I found conflicting reports but some were a year old.
It can suck just fine as long as it runs 1080P, exists, and is cheap.Please don't Suck.
Please don't Suck.
Please don't Suck.
Unless Intel scrapped a generation these will be 7nm(+) TMSC.I'm a bit confused, is this consumer Arc GPU (Xe-HPG DG2 codenamed Alchemist) being made on TSMC 7nm, 5nm, or 3nm? Or do we not know? I found conflicting reports but some were a year old.
The current GPU’s for the Aurora Supercomputer is made jointly on TSMC 5 and Intel 7nm. Given this is the consumer version of those cards I would expect them to keep the same processing node.I'm a bit confused, is this consumer Arc GPU (Xe-HPG DG2 codenamed Alchemist) being made on TSMC 7nm, 5nm, or 3nm? Or do we not know? I found conflicting reports but some were a year old.
I was under the impression that Aurora is actually using Ponte Vecchio GPUs mfged at 7nm at Intel.The current GPU’s for the Aurora Supercomputer is made jointly on TSMC 5 and Intel 7nm. Given this is the consumer version of those cards I would expect them to keep the same processing node.
Intel ran into troubles with yields and fell behind, the DoE threatened to sue Intel for breach of contract for failing to meet their deliverables. Intel worked out a deal with TSMC for space on their 5nm to pick up the slack. So now the chips are made with both processes, they use different SKU’s but perform nearly identically.I was under the impression that Aurora is actually using Ponte Vecchio GPUs mfged at 7nm at Intel.
Unless something has changed the Gov lists;
Xe arch-based “Ponte Vecchio” GPU; Tile-based chiplets, HBM stack, Foveros 3D integration, 7nm
I was under the impression that Intel was still planning to fab their own XE thing for Aurora the type of thing that makes zero sense to try and mass produce but that their fabs may be able to handle in smaller numbers for Aurora... with TMSC getting the DG2 contract. (which is a more sane design without the chiplet foveros stuff) Intel for sure has a 7nm+ contract with TMSC. Who knows though Intel is such a mess they probably have 4 different GPU projects going at once. lol
I’m hoping it competes with the 1660 super and is available at truly insane quantity. I want to see like weekly restocks in the 100s at the local microcenter. That would basically cover the reasonable lower middle tier of gaming card and at large enough quantity bring the price to something reasonable. Right now the low end at microcenter is basically $189 for a 1050ti.It will be absolutely destroy the i740.
Beyond that, well, wait for reviews, don't get hopes too high.
Intel’s roadmap has 7 facilities running 7nm for 2022, TSMC currently has 1 fab running 5nm, with a second being built in Arizona.Thank you ChadD and Lakados. What is the total fab capacity of Intel 7nm vs TSMC 5nm? How do they compare today? Is intel at, say, 5% of TSMC? I'm a bit out of the loop.
Intel’s roadmap has 7 facilities running 7nm for 2022, TSMC currently has 1 fab running 5nm, with a second being built in Arizona.
Output wise the two aren’t comparable, Intel operates 4x more facilities than TSMC does. That’s not a bad thing though, TSMC gets to be relatively agile in these respects, they get to do and try things Intel just can’t outside of a lab simply because their size allows for it and they have a far FAR better management structure.
God only knows, neither Intel nor TSMC tend to talk about yields overly much. We do know their yields weren't great with their initial 7nm processes when they started producing the Ponte Vecchio chips but was that due to chip design or process or both who knows. Intel's Q4 earnings disclosure did state the 7nm process was back on track, for full delivery in 2023, not sure if that means fiscal 2023 which would be Q3 of 2022, or actual 2023 but Pat Gelsinger spoke on the call saying "he has personally reviewed progress on the company's 7nm process over the last week and that he is pleased with the "health and recovery of the 7nm program."Does that take into account yields? Are intel's yields comparable?
As part of today’s architecture day announcements, Intel is finally answering the burning question and disclosing the fab and node being used. As many have expected, Intel is indeed turning to TSMC to fab their gaming GPU, and they will be using TSMC’s N6 process to do it.
The VCZ link I posted earlier said 6nm. I don't know why people were assuming 7nm or 5nm when TSMC's production has been tapped out on those fabs by other companies forever and a day now.https://www.anandtech.com/show/16895/a-sneak-peek-at-intels-xe-hpg-gpu-architecture
I unfortunately saw no mention of concrete plans to move the design to Intel fabs for Xe-HPG, but surely it's in the works if we know Xe-HPC is already taped out or in early production on Intel 7nm. Like many of you I just want enthusiast GPU prices to go down and I hope intel can move some production away from TSMC for this. Though even just at TSMC, with Intel's new GPU the total GPU production at TSMC ought to increase, I assume.