Comet lake rumors aren't certain might be very late 2019 or early 2020. Chances of Rocket lake being a sunny cove backport is slim because of FIVR, if that's the case Intel is stuck on Skylake on desktop until 2022.
Keller's new nextgen arch should be ocean cove around 2023 on desktop 7nm which can be competing with Zen6 on 3nm
Intel's roadmap is a series of paper launches and rebrands. Don't bother as long as they keep calling their architecture "core", it's going to be the same shit.
>Desktop >------- >2020 Comet Lake (skylake) 14nm+++ *10cores >2021 Rocket Lake (skylake? / sunny cove backport?) 14nm+++? *10 cores >2022 Alder Lake? (golden cove)? 10nm++? >2023 Meteor Lake? (ocean cove)? 7nm The fuck, where's my sekrit sauce architecture?
Parker Nguyen
There won't be any good availability of a desktop 10 core this year even if they rush it, they have very recently started sampling it
Leo Lee
Rocket lake is 12 cores skylake with a improved chipset. And that's the limit of the ring bus.
Anything after that they'll have to move to 2D Mesh, but that also means losing the intra-core latency advantage they had over AMD
Christian Wright
Rocket lake was listed as 10cores in leaked Dell slide
Aaron Powell
Then better pray that's a mistake because it's not a backport, even the dumb French nigger says they didn't have plans (or time now) to do backports.
Grayson Edwards
They are already pushing the thermal envelope with 8c, they can probably get away with 10core with the next +++ but I have doubts if they can do a 12core with high clocks in a "sane" TDP
Thomas Garcia
As if Intel cares about TDP, shit will be 95-105W TDP at 2.0GHz base clocks and 5.4 Turbo.
Because durr, TDP is calculated at base clock (for Intel at least)
Luis Perry
they're going to be fuck expensive too
Michael Brown
kek
Evan Lee
>2020 Cooper Lake (skylake) 14nm+++ >2020 Icelake (sunny cove) 10nm+ That doesn't seem right.
Julian Campbell
From the 2019 investor meeting. I pity the souls that are buying into cooper lake-sp
Cope lake is three die (2 compute + pcie4 IO) monstrosity Intel came with in a month, needless to say it's about as a failure as their other AP line. Icy Cope tops out at 26 cores, the only thing it has over Cascade is Pcie4 and 8 memory channels
Intel's roadmap will take a detour down along the line when some small branch comes up with something better than core and/or they buy some company that does something better than core.
Colton Johnson
They should at least use skylake server cores with AVX512 in mainstream desktop. It's beyond stupid that 2-4 core chromebooks and NUCs are getting this now with icelake but desktop is seemingly neglected.
Cooper Sanders
I wouldn't give a single fuck about desktop if I were Intel tbqh
Noah Johnson
nobody really gives a fuck about intel. even when bulldozer came out and couldn't compete in regular workloads with intel, companies like cray who built gpgpu behemoths, bought massive amounts of bulldozer opterons and used them due to their massive I/O capabilities. Intel's only market right now, is the laptop market, where amd hasn't released anything that competes against intel, as in the desktop segment.
Carson Davis
There's a chance that they wanted it this year. The 8700k and 9900k chips released months before most of their generations. Reason I don't think the 10900k is coming this year is because of the 9900kys hasn't released yet. It's a stop gap for the delayed 10cores.
William Johnson
I think the leak said rocket lake was a chiplet, so it's a 10 core chiplet with igpu chiplet on 10nm potentially. For marketing they've missed the chance for 10th gen 10core 10nm.
Charles Price
>Intel's only market right now, is the laptop market A market they're losing to ARM.
Yes but intel made him work on ultra energy efficient low power chips that are closer to a raspberry pi than a cheomebook cpu then moved him for some gpu shit interconnect instead of using him for their desktop/server high end architectures. Intel pen pusher CEO and high ranking managers are brilliant.
John Cruz
Intel really is not that diverse of a company. Sad.
keller wasn't even responsible for most of zen's design. he helped out on the soc business iirc
Elijah Ross
tech companies like to codename stuff after places in the real world.
there were only supposed to be a handful of lakes, then intel became so delayed they started making them up. its all the same arch just slightly tweaked.
thats why sunny cove isn't a -lake cpu as its a new arch.
amd are currently using city names for their server chips. rome, milan etc.
Isaac Bell
Some jew over at Intel probably has a lake/cove fetish. These people are from the middle east so water is pretty rare for them and a lake is probably seen as a gift from god himself.
Owen Hernandez
CQPE
Hudson Brooks
He designed Infinity fabric and oversaw general management like performance targets, interconnects and integration is his specialty
Joseph Butler
They probably don't know themselves. If they indeed have a new architecture they should at least change the fucking naming scheme. They're going full retard now.
Benjamin Green
They should name their next architecture after supernovas.
Chase Hall
thanks based user
Charles Evans
>thats why sunny cove isn't a -lake cpu as its a new arch.
>2019 Ice lake (sunny cove) 10nm+
Julian Nguyen
Or black holes.
Kevin James
what makes him so special? Why are companies constantly poaching him? Aren't there thousands like him? Or is he just Chip Rain Man?
Ayden Foster
intel should just rename their 10nm to 7nm, problem solved
Isaac Wright
Why can't companies just do a simple naming scheme like AMD 2019, Intel 2020. These fucks only release updated archs once a year at best.
Adrian Howard
I wish Itanium didn't fail. Here's to hoping RISC succeeds.
Lincoln Jackson
because normies would see a laptop at a store and go >oh, this has an Intel 2018, I probably shouldn't buy it
tech has to be cryptically named so that the old stock can still be sold to retards
Liam Wright
>Keller “was involved in the early days of Zen, we worked together on the arch and he made me lead architect for it because he was running the whole [processor design] group,” said Clark. “The engineering team loved him because he’s an engineer at heart and you felt you had a champion,” he said. Is what Mike Clark said about him. He is a veteran engineer that keeps delivering but he wasn't really a rockstar before his Apple stint. Intel engineers traditionally weren't usually welcoming to outsiders but he and Raja crashed in as VPs so we will see how it goes.
Charles Walker
Naming schemes aren't about being consistent they're about marketing and perception.
AMD changed Navi from 6xx series to 57xx to try to move the perception of Navi away from being just another GCN. It was also to create the perception that these are high end cards, not mid range. 5700 sounds large, therefore consumers have a better impression of the performance when they hear it.
Similarly, AMD named their cpus in a similar manner to Intels so they would be compared in the mind of consumers in that way. I.E. 3700x equivalent to 9700k, r7 equivalent to i7 etc.
Xbox One was supposed to "the One" console you needed, that did everything you could need it to do. Of course, it backfired and Xbone got memed all over the place.
RTX is another one, it was an attempt to differentiate the cards from GTX by NVidia, that it was the next evolution of gaming. It also meant that if the cards flopped the RTX brand could be dropped and GTX wouldn't be harmed for it.
This all sounds retarded but companies spent millions on the perception of their products and their brand in order to get you to continue buying shit you don't need. If it didn't work they wouldn't do it.
Adam Moore
they should both go back to Pentium and Athlon.
IT'S THEIR DESTINY.
Zachary Clark
If you name your arch after a place you also don't need to worry about trademarks since place names can't be trademarked
make fun of them , they make shitton of money by playing vydias. i wish i was a zoomer esport player
Angel Murphy
I don't like the new architecture naming.
I like it when it was stuff like Lynnfield, Haswell, Prescott, etc.
Owen Anderson
I'm looking forward to what Jim Keller does for upcoming Intel chips. Sunny Cove claims a 18% ipc gain and Keller said the new 10nm won't have clock regression from 14nm++ But it'll take a while for Jim to un-Fuck Intel's mistakes.
Intel has 2.4x scaling from 10nm to 7nm (~241.8 MTr/mm^2), TSMC 5nm which AMD will be on in 2021 has 1.8x scaling from TSMC "7nm" (~182.4 MTr/mm^2). AMD's "7nm" is a marketing term.
Owen Cruz
>he wasn't really a rockstar before his Apple stint I don't know user... Keller is responsible for some of the best Alpha CPUs, a good bit of the AMD64 arch, and the AMD K8 microarch. He was already a rockstar when Jobs invited him to join Apple. In fact that's likely why he was invited, because Jobs wanted all the A players he could get his hands on.
Yeaah but the thing is the 14/10/7/5/3nm are just marketing names and there's no actual features so small in there. So even when we're down to "3"nm we'll still have some way to go down.
Jack Long
>tHey arE JuSt mArKeTiNg nAmEs aNd ThErE's No aCtUaK FeAtUrEs sO sMaLl iN tHeRe
The only interesting part that intel has to offer in the near future is Tremont. If they can keep it passively cooled it's gonna make great low end homeservers, NAS, set top boxes etc.
Andrew Bennett
C2D wasn't in any planned slide
Conroe was just thought of as another P4 and look what happened
If Intel's GPU launch is successful they will move desktops to Samsung to shit on AMD and TSMC
Juan Gray
So is cooper still 14nm? Why even release it if ice lake is going to come out at the same time?
Noah Wood
Because diversity?
Camden Harris
Here's intel's roadmap plan and my face reacting to it.
There is this bear pride faggot boomer near where I live in San Fran that looks just like Keller but fatter. I need to leave this place.
James Russell
>Chances of Rocket lake being a sunny cove backport is slim because of FIVR I don't really see why the microarchitecture would need to be bound to a certain method of power delivery. I find it more likely that the reason Sunny Cove won't be backported to 14 nm is simply because the increased OoO machinery won't run efficiently on larger transistors. I'm pretty sure the reason increases in these structures are historically bound to node shrink is simply because the increases represent the power-optimal size for that node.