Read more.And they shine a light onto continuing 10nm, emerging 7nm woes.
Read more.And they shine a light onto continuing 10nm, emerging 7nm woes.
oh dear...
It never rains but it pours...
Old puter - still good enuff till I save some pennies!
New CEO is probably a good move... as long as they can find one that it’s an engineering background, not some marketing suit.
But damn! That last paragraph had my jaw on the floor! Not catching up till 2025 basically means (IMO) they don’t know when they’ll catch up...2025 is just when their intel (pun alert) on upcoming AMD designs runs dry...
Could make Alan Sugar an offer. Can see that blue logo with " Amstrad Inside " catching on.
INTEL can 'simply' use TSMC nodes under a contract just like Global foundries.
Do you mean licensing the technology like GloFo did with Samsung, or actually contracting TSMC for wafers? Either way, there's nothing simple about it.
It would not be easy to port TSMC's process to Intel's fabs even if TSMC had a motivation to license it out. And even if they did, it would not be a quick fix - the deal between Samsung and GloFo was in place for a long time before production started. You can't just copy and paste things between fabs.
Even when it comes to contracting TSMC to produce wafers for them, it's something of a last resort and has many drawbacks. Nonetheless they are apparently still considering this.
Drawbacks include very limited supply, and TSMC are very unlikely to give any sort of priority to Intel, who are likely to be a short-term customer, over long-term customers like Qualcomm, Apple, AMD, Nvidia, etc who are already competing for capacity.
Another issue is it means porting cores to another node (something Intel have done before but only for cores like Atom), and that's not guaranteed to go smoothly - nowadays CPU cores are designed with a target node in mind and it can require significant reworking to get it to just work, let alone compete with other IP targeted at those nodes.
Another is cost - outsourcing to another company means paying them and losing one of the main advantages Intel have had by keeping their own fabs in-house (though that's arguably offset if it means having a competitive product vs not).
I would also say there's a risk of leaving their own fab lines idle which would be bad news, but unless it lasts many more years, they'll probably still be producing plenty on 14nm anyway.
I'm available for work. I don't know much about processor design but I am perfectly capable of blowing the dust off a 4C8T 14nm design and saying "change the pin count slightly and boost the clock by 10MHz" so I should fit right in.
DanceswithUnix (18-12-2020),Gentle Viking (18-12-2020),Output (19-12-2020)
I've said this for a long time. They knew they were caught sleeping and couldn't retaliate, the constant underhanded tactics and fake reveals were just attempts at delaying the inevitable.
They've obviously realised, a CEO dragging things out to save his own skin is not doing anyone any favours. As mentioned, they REALLY need someone like Lisa as CEO, business minded, but more so, someone who understands the business they are in, someone who with technical experience, passionate, not another suit.
Shares will tumble, while they still have cash, they're years off CPU wise, but like AMD were not too long ago, the have to fund R&D for GPU's too.
Atleast they aren't broke like AMD were. The real issue is restrictions placed on the company by the board and shareholders, no "free reign". Things could easily turn ugly very quickly (but I don't expect that).
I feel AMD will continue taking small portions of market share at a time, whilst Nvidia will doing what they do.
Intel will continue with spin and try to get GPU upto speed to compensate for the years taken to bring CPU side back up to task.
Is ANY of this real, it's so ironic, it's truly poetic. Seems made up it fits so well, lol
Intel already is using external foundries for around 20% of product, and Swan has repeatedly emphasized that this percentage can go up or down.
Intel already sampling 10ESF products ... Xe-HP, Alder Lake, Sapphire Rapids Server.
Xe-HP demoed with 42TFlop FP32 ... anyone matching that?
Sapphire Rapids CXL and PCIE5 operation verified by PLA and Synopsys ... anyone matching that?
I'm going to just mostly copy and paste my last replies to you because you never answered and I suspect you never will.
Yep, and industry insiders are concerned about the actual availability of Alder Lake because it's being indicated as being on 10nmSF, where are you getting 10nmESF? I expect you won't provide a source like last time.
Sapphire Rapids is already a year delayed and has missed it's window for the Aurora supercomputer (oops) which means it's going to be releasing between Milan and Genoa from AMD (big oops, Sapphire was meant to be a leapfrog!).
It's not installed yet, Intel dropped the ball with their 7nm, we have no idea when it's coming and whether the Xe-HP will be even the same. It is also a claim and not a reality, yet. Additioanlly, in the previous quote you said Xe-HP was 10ESF but it's actually 10nmSF for the main tile and only a small portion (the cache) for 10nmESF while we still don't know what node is being used for the compute. Sooooo, with all the lack of and confliction of information, do we really believe in it except for being a nice number on a piece of paper?
Marketing piece that didn't really need to be shown, you can bet bottom dollar that both AMD and IBM with Power10 have CXL and PCIe 5.0 being tested right now. Just neither of them need to do a marketing piece to keep themselves relevant in the market.
I am being (somewhat) unnecessarily harsh on Intel, but when people like yourself come in going Nyerrrr, they deserve it.
more pro-intel posts from a new member. It's almost like they only arrive here to post pro-intel comments. Are you paid by or affiliated to intel JayN?
Alder Lake is 10esf . See anandtech, "The AnandTech Decoder Ring for Intel 10nm", Dr. Cutress
Sapphire Rapids is broadly sampling. PCIE5, CXL, DDR5, amx tiled matrix operations, bfloat16, avx512
Xe-HP is 10esf .see anandtech. Already sampling, available on Intel's devcloud, installed at Argonne.
Intel server chips supporting Optane, dual avx512, dlboost, bfloat16, DDIO for networking ... AMD has long way to catch up.
There are currently 1 users browsing this thread. (0 members and 1 guests)