Deleted member 209350
D
Deleted member 209350
Exactly! Has to be one of the 2, and either way it looks bad for intel
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Intel kills 10nm ?? oO
- Thread starter Zeed
- Start date
Exactly! Has to be one of the 2, and either way it looks bad for intel
Soldato
I am pretty sure if they had been sitting on this new 40+% IPC gain architecture then they would have released it before losing fairly large amounts of market share to AMD. Especially in the server market which is really beginning to open up to Epyc now.
Well, not exactly. Zen brought 52% IPC improvement, Zen+ brought 2-3% IPC improvement, Zen 2 brings ~15% IPC betterment, Zen 3 and Zen 4 will also bring something, and with them, no matter what SunnyCove brings, it''ll be mostly cancelled.
The IPC 'leak' is most likely completely made up. They've been scraping the bottom of the barrel with recent Core revisions. Potential hardware mitigations of their various problems will help, but the figures stated? Nowhere close. I'd be very pleasantly surprised if it was 5-7%.
I don't think there's any chance of Desktop Ice Lake anymore. Either Enthusiast or what ever Lake they'd call it for HEDT.
It's been in the news today that the 10nm plant in Israel (one of only two) has just been delayed another 6 months and will probably only start production in Q2 2020.
Their 10nm process is screwed. Yields are abysmal, clocks are literally 1Ghz or more lower on the low power chips (can't imagine how bad it'd be on high power chips), EUV on 10nm has been completely scrapped (the node was meant to be EUV quickly after launch as a + originally), and raw capacity (even excluding yields) has been a minute fraction of what they hoped.
I don't see how they could do high power parts on it. Very small chance of small low power server chips, but I'd say that would be for late 2020 *if* yields drastically improve.
What ever Intel do on desktop in 2020 will be 14nm++++. They can't do 10nm.
What they really have to hope is that their 7nm EUV isn't totally borked, yields are ok, and clocks are great. Because as far as we know, 2022 is the earliest release candidate for their post Core architecture.
I don't think there's any chance of Desktop Ice Lake anymore. Either Enthusiast or what ever Lake they'd call it for HEDT.
It's been in the news today that the 10nm plant in Israel (one of only two) has just been delayed another 6 months and will probably only start production in Q2 2020.
Their 10nm process is screwed. Yields are abysmal, clocks are literally 1Ghz or more lower on the low power chips (can't imagine how bad it'd be on high power chips), EUV on 10nm has been completely scrapped (the node was meant to be EUV quickly after launch as a + originally), and raw capacity (even excluding yields) has been a minute fraction of what they hoped.
I don't see how they could do high power parts on it. Very small chance of small low power server chips, but I'd say that would be for late 2020 *if* yields drastically improve.
What ever Intel do on desktop in 2020 will be 14nm++++. They can't do 10nm.
What they really have to hope is that their 7nm EUV isn't totally borked, yields are ok, and clocks are great. Because as far as we know, 2022 is the earliest release candidate for their post Core architecture.
Unclear if it'll start in 2020 or 2021. But it's basically an admission that their 10nm is done for, 7nm isn't ready, and they don't have enough capacity for 14nm++++++++++++++++++++++.
Also, I think it's now pretty unlikely their 10nm GPUs are on their own process next year. Not sure Samsung's 10nm will be suitable for GPUs, but maybe they'll do a special one for them, or put it on 7nm.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Construction has been delayed 6-12 months, such projects take years rather than months. IF construction doesn't start till into 2020 then production won't start till late 2021 at the earliest but depending on the scale of the production project, could be a couple of years after that.
If (I didn't think it was the case) they are still in major construction phase then definitely a couple of years after that - could take a year alone for qualification.
Caporegime
You’ve got me convinced.
Soldato
I wonder when/if they will swallow their pride and eventually go to Samsung/TSMC for help for their cpus.
Judging from how Intel have done things so far with their 10nm development, then no I don't think they will, at least not with their 10nm process.
Now that Intel's new 10nm plant has been delayed by six to 12 months, maybe Intel will instead try and accelerate their 7nm development.
Now that Intel's new 10nm plant has been delayed by six to 12 months, maybe Intel will instead try and accelerate their 7nm development.
Soldato
- Joined
- 1 Dec 2015
- Posts
- 18,512
prob not, their fab is 'meant' to of been higher quality . again node sizes arent true to their sizing now , though intels is still the closest .
Venders rep comment on here in ryzen post
though have a feeling its 10 core Intel then 10nm intel - but yeah- intel needs to get a move on !
Soldato
I know the fabs are all worked differently and the nm size of the node doesn't actually mean too much now, but they clearly have to do something and have been stuck on their 10nm node for 6 or so years now. Eventually they will have to do something.
There's no way they can do desktop on their own 10nm process. Apart from yields and no wafer availability, the clocks are an absolute disaster.
Intel feels the Ryzen pressure, cuts the price of 9th gen CPU's by 15%
https://www.techpowerup.com/256700/...op-processors-by-15-in-response-to-ryzen-3000
https://www.techpowerup.com/256700/...op-processors-by-15-in-response-to-ryzen-3000
When was the last time this happened?
i9-9900K from 488$ down to 410$. It's better.
Not sure where you got your facts from, the issue Intel is having is validation, the passing criteria is intentionally very strict. The other foundries are shipping willy nilly because the parts are consumer validated not server/datacenter, automotive etc...
Im not even going to mention node sizes since every foundry measures differently and passes numbers to marketing, if you look at the PDK’s of some of the nodes the sizes don’t actually change (example Samsung 14nm vs 11nm).
Last edited:
Side note: automotive's passing criteria is also willy nilly.