>Sunny Cove, built on 10nm, will come to market in 2019 and offer increased single-threaded performance, new instructions, and ‘improved scalability’. Intel went into more detail about the Sunny Cove microarchitecture, which is in the next part of this article. To avoid doubt, Sunny Cove will have AVX-512. We believe that these cores, when paired with Gen11 graphics, will be called Ice Lake.
> The upcoming microarchitecture for 2019 is called Tremont, which focuses on single threaded performance increases, battery life increases, and network server performance. Based on some of the designs later in this article, we think that this will be a 10nm design.
>From here we can deduce a few things. We were told that this configuration is the GT2 config, which will have 64 execution units, up from 24 in Gen9.5. These 64 EUs are split into four slices, with each slice being made of two sub-slices of 8 EUs a piece. Each sub-slice will have an instruction cache and a 3D sampler, while the bigger slice gets two media samplers, a PixelFE, and additional load/store hardware. Intel lists Gen11 targeting efficiency, performance, advanced 3D and media capabilities, and a better gaming experience.
>Intel didn’t go into too much detail regarding how the EUs are at higher performance, however the company did say that the FPU interfaces inside the EU are redesigned and it still has support for fast (2x) FP16 performance as seen in Gen9.5. Each EU will support seven threads as before, which means that the entire GT2 design will essentially have 512 concurrent pipelines. In order to help feed these pipes, Intel states that it has redesigned the memory interface, as well as increasing the L3 cache of the GPU to 3 MB, a 4x increase over Gen9.5, and it is now a separate block in the unslice section of the GPU.
>So here’s where Foveros fits in. Foveros is Intel’s new active interposer technology designed as a step above its own EMIB designs for small form-factor implementations, or those with extreme memory bandwidth requirements. For these designs the power per bit of data transferred is super low, however the packaging technology has to deal with the decreased bump pitch, the increased bump density, and also the chip stacking technology. Intel says that Foveros is ready for prime time, and they can produce it at scale.
>So this Foveros ‘3D’ packaging just sounds like a silicon interposer, such as what we’ve seen on AMD’s Fiji or NVIDIA’s high-end datacenter GPUs. However Intel is going above what those products are doing by actually making the interposer part of the design. The interposer contains the through-silicon vias and traces required to bring power and data to the chips on top, but the interposer also carries the PCH or IO of the platform. It is, in effect, a fully working PCH, but with vias to allow chips to be connected on top.
>The key part of this discussion however is this block diagram that was on one of the Intel slides. Here we see a single ‘Big CPU’ with 0.5 MB of private medium level cache, four ‘Small CPU’s with a shared 1.5 MB L2 cache, an uncore that has 4MB of last level cache, a quad-channel memory controller (4x16-bit) with support for LPDDR4, a 64 EU design with Gen11 graphics, the Gen 11.5 display controller, a new IPU, MIPI support with DisplayPort 1.4, and all of this in a tiny package.
>Seriously though, this has the potential to be a large revenue stream for Intel. They’ve made this chip, which allows the cores to enter C6 sleep states when not in use, that has a die size smaller than 12x12mm (144 mm2), and target the sub-7W fanless device market. That’s with a big Core, four Atom cores, and a GT2 64 EU design.
This will force amd to get better too in their design. However what really count is not the extreme high end but capabilities for price on middle end. Software may evolve to better multithreading use, becoming less dependent on single thread speed.
Landon White
ayymd drones are too busy jerking off to fake leaks
Meh, there isn't much difference between the two of them these days anyway. This still fails to make much difference honestly. Regardless of what you buy you'll get great performance.
Oh wait this is Jow Forums...
AMD rulez! Intel gonna get BTFO by ryzen2+++++!!!!!11!!!!!eleven!!!! Fuck jewtel amirite! Also gebtoo for life!!! So much cope from Intel fantards!
Jaxson Bell
nice cope
Anthony Adams
>Sunny Cove, built on 10nm (released at 14nm) > Tremont ... We think this will be a 10nm design (released on 14nm) etc
>AMD rulez! Intel gonna get BTFO by ryzen2+++++!!!!!11!!!!!eleven!!!! Fuck jewtel amirite! Also gebtoo for life!!! So much cope from Intel fantards! This but unironically.
Wyatt Edwards
>AMD does Chiplets >Intel wants that shit for their 14nm++++(10nm but actually 14nm) new line of shit
Elijah Brown
How many more times are you going to let the kikes pull the wool over your eyes?
Nathaniel Bennett
>Want to emulate >Want to get DRM cracks that actually work The answer will always be Intel Aviv until AMD can be competitive in either of these. I don't care if AMD have a 1nm 10Ghz 500 core ultra instinct single core perf based chip running at 1W TDP. If it can't do the above, it's shit.
Carter Stewart
He's the same mentally ill intel shitter who bought the 9900k and got buttmad when zen2 leaks came out. Now he shitposts on Jow Forums pro-intel and anti-amd shit because he's a mentally ill manchild with buyer's remorse.
Thomas Sanders
Emulation hasn't been a problem for AMD for years unless you're trying to do PS3 emulation, which is still in alpha stages anyway.
As for DRM cracks, I have no clue what you're talking about and have never heard of a DRM crack that somehow didn't work on AMD.
Chase Anderson
RPCS3 is a thing and it needs Intel because of TSX
Kevin Campbell
>Buying Intel after having been jewed with sticking to 4 cores 8 threads during 9 years Good goys
Jack Taylor
It doesn't "need" Intel, it merely offers a performance boost. Also TSX is disabled in newest microcode release due to Spectre/Meltdown using it as an exploit, so to get the RPCS3 performance boost you need to downgrade your microcode.
Dylan Stewart
I like what I'm hearing about Zen 2 and 7nm, but if Sunny Cove delivers on big single thread increases, I may go Intel again.
Just wait for 7nm in 2077 bro, it will be worth it
Joseph Russell
We infinity fabric now Even Jim was on board presenting Like always Intel will do it right what amd started
Logan Nelson
poo in the joo shitkike
Isaiah Hall
>sunny cope >willow cope >golden cope
Nathaniel Richardson
Too bad these performance increases come with 10 more Meltdown class security flaws.
Nicholas Mitchell
It's not really BTFO as much as Intel admitting to their faults on a very top level point of view. The BTFO will come when they have to talk about manufacturing at a later date.
The good news is that Jim Keller and Raja Koduri is unfucking Intel to some extent already and playing catch up where they can to AMD's approach, and removing manufacturing technology from the architecture is probably the most important thing as all the *Lake refreshes and rehashing of graphics kind of damaged Intel's image to tech enthusiasts. They are also doing a ton of experiments with their Atom and Core lines along with the Foveros interposer stuff, and they seemed to have indicated they will try and start taking a similar approach to AMD with chiplets on the CPU side and GPU designs from Raja designed to not only game but to do more compute heavy stuff from the rearrangement configuration for Gen11 graphics and beyond.
The only problem is the time. Intel is only releasing Sunny Cove and etc. sometime late 2019 and possibly early 2020 if shit goes wrong, but Lisa Su at CES is going to announce consumer Zen 2 chips which will probably ship on a timeline like the original Ryzen at worse which means Intel is lagging around half a year of time or more and can't respond to AMD and I doubt the 9900k will be enough to fend off Ryzen 3000 series top of the line, since I think the gaming performance gap will shrink even further or become near 0 which means AMD will have another Athlon moment again. So we'll see if AMD can get somewhere near the 30-40% of the market at that point. But yeah, Intel definitely needs to prepare themselves for hemorrhaging cash.
Logan Wright
i'd fug this so hard for an HTPC, would also love a cut down version for a NAS or router.
Yer rite il believe it when I see it. Still housefire thermals and stupid power plus 10nm nowhere to be seen memebingbus crap still and now 7nm? Gimmi a break shills
Nathan Rodriguez
>TSX is disabled in newest microcode release due to Spectre/Meltdown using it as an exploit NOOOOOOOOO
Wyatt Hughes
30% ipc bump on a unfucked 10nm process. Ryzen about to get BTFO JIM SHITWRECKER KELLER
>Sunny Cove, built on 10nm, will come to market in 2019
late 2019 i suppose
>> The upcoming microarchitecture for 2019 is called Tremont
atom has always been the only intel product ive been instered in during these years. but since they left the mobile game, where will they place it? netbook and low specs notebooks are also dead.
>>From here we can deduce a few things. We were told that this configuration is the GT2 config,[...] and it is now a separate block in the unslice section of the GPU.
rubbish
>Foveros
is this an infinity fabric equivalent, or HBM equivalent? It's not clear.
intersting, but if this is a response to an amd zen apu, it will be shit no matter what since their graphics department is still lagging years behing radeon
The man who made Zen, Zen+, Zen2, and Zen3 is now head of product development at Intel. 2019/2020 will be when the products he has had a direct hand in developing will be launching. Kiss your AMD revival goodbye.
Mason Reed
>10nm in 2019 >uhhhhhhh, 2021 >no wait, 2024 >definitely 2030
meanwhile: >competition at 7nm in 2018
And all my devices have Intel. It's laughable how incompetent they've become by being at the top all these years.
Mason Anderson
just get a 775 C2Q, you only browse Jow Forums and play old games from 2008 and earlier.
Hunter Gonzalez
>now head of product development at Intel False. Keller was hired to work on developing an infinity fabric equivalent, not to develop a brand new uarch, which Intel has literally never done successfully in the last 20 years.
>2019/2020 will be when the products he has had a direct hand in developing will be launching False. The development cycle on a brand new uarch is 4-5 years. It is totally impossible that Keller had a significant hand in 2019-2020 products.
>Kiss your AMD revival goodbye. >JUST WAIT!(TM) lmoa
Adrian Howard
He only made Zen. he left AMD after he was done with it, other AMD engineers have been improving on his design.
Ryan Rogers
>False. Keller was hired to work on developing an infinity fabric equivalent, not to develop a brand new uarch I never said he was developing a new uarch, but you're out of your mind if you think he isn't dabbling in EVERYTHING.
> I was hired to be Senior Vice President of Silicon Engineering, so Murthy had a plan and a structure in place. So generally a large number of the SoC teams and the CPU execution team work for me, so that's a pretty big space, it's quite a large group and they have a lot going on.
When he first left AMD rumors were he had finished Zen2 design and Zen3 and left roadmaps for everyone to follow.
Nathan Walker
3D packaging. De-facto it's just active interposer.
Blake Lee
He never did anything Zen. Just K12 and some minor IP. He did zzap AMD's design teams into shape, though.
Ethan Phillips
read the articles.
>So this Foveros ‘3D’ packaging just sounds like a silicon interposer, such as what we’ve seen on AMD’s Fiji or NVIDIA’s high-end datacenter GPUs. However Intel is going above what those products are doing by actually making the interposer part of the design. The interposer contains the through-silicon vias and traces required to bring power and data to the chips on top, but the interposer also carries the PCH or IO of the platform. It is, in effect, a fully working PCH, but with vias to allow chips to be connected on top.
>The first iteration of this technology is less complicated that the slide above, just using a set of CPU cores attached to the PCH below, but the idea is that a large interposer can have select functions on it and those can be removed from the chips above to save space. This also lets Intel use the different transistor types in different chips – the example we were given uses an interposer built on the 22FFL process node, with a 10nm set of CPUs on the top die. Above this, DRAM is provided in a POP package. Sounds cool, right?
Zachary Brooks
2019 kek, try 2022-2023, ravij
Oliver Brown
I wouldn't buy Intel pozzed garbage even if it was infinitely better for just $1.
Fucking this, Intel's 10nm has turned into a running joke. I'm suddenly supposed to believe them this time? And it'll probably be gimped as hell compared to what they originally planned, anyway.
reminder that a poojeet is in charge of everything at intel right now
Zachary Barnes
>7nm this time for sure guys! >paper launch >low end only >massively overpriced Pick three.
Parker Carter
>Also TSX is disabled in newest microcode release due to Spectre/Meltdown using it as an exploit Kek, soon they'll have to disable the whole CPU and emulate it on GPU to mitigate all their security issues.
>30% ipc bump on a unfucked 10nm process. They'd have to make human sacrifices to satan for their 10nm to get unfucked
Carter Murphy
Agreed, even as an intel fan, they'll skip over 10nm as quick as possible and move to EUV 7nm
Andrew Brooks
Whoops, meant >10nm but I was just so used to typing 7nm as products on it will genuinely be coming out at affordable price points next year.
Wyatt Price
>but since they left the mobile game, where will they place it? netbook and low specs notebooks are also dead. Chromebooks, NUCs, IoT, general embedded. Something like an x86 atom combined with FPGA fabric might be a popular embedded part.
No we haven't raja only joined in 2017. Nothing he has done will see the light until at least the next generation, and any major design influence wont be seen for another generation or two after that.
John Stewart
Do you even know what you're looking at?
Hunter Sanders
a NUC with Vega M
Oliver Mitchell
>new instructions
Fuck OFF! x86 is already too high level
Eli Robinson
Who pitched and sold it to Intel? Who developed it for Intel?
There is a reason they call it Hades Canyon.
Kayden Sullivan
Who cares? What raja will be doing internally with intel is entirely different, get your head out of your ass.
Liam Gonzalez
I'm dumb, I still don't understand
Will this help intlel deplyong a chiplet strategy like amd did thanks to infinity fabric? Because this would be the only thing that could possibly help them coming out of the massive pile of shit they've diven into
Camden Carter
Its for package on package, something cellphone SoC makers have been doing for a while. Image sensors have been doing it for longer. Intel might have stacked logic dies on one package, but not chiplets like AMD is using.
Joseph Jones
He will be pushing snake oil and Indian outsourcing miracles.
Have fun hoping for that faggot to do anything but fail.
Lmao, when he wore a red shirt he was your savior and you couldn't wait to see what he was bringing, now that he wears blue, suddenly he's just another poojeet. Kek
Eli Harris
If ourguy Jim is working for Intel, we might be screwed bros.
Christian Hughes
So Intel employs stacked design with TSV interconnects now. It sure is going to be a great win, like HBM was... Right?
The fact that they had to have a week long conference to bring this to the table and their other projects is telling me to avoid them, atom isn't that great of a chip to begin with and the market will be taken over by ARM really quickly and parts of the market already have, who are they trying to get to buy this?
I still don't see ARM systems capable enough to replace low-power x86 servers or routers. They have a locked bootloader and don't offer enough peripherals, like 6xSATA or a couple of PCIEx8 ports. > it will be much more expensive! How much? x86 boards with a soldered CPU are in $50-$100 range, plus throw in some peanuts for a RAM stick. Something tells me ARM boards won't cost $100, but twice as much, at least.
Carter Long
microcontrollertips.com/router-processor-based-on-64-bit-quad-core-arm-chip/ They are moving that way right now, I couldn't say about price, also, are these new intel chips eec capable? if not, then there's no point of running servers on them over arm, considering tham arm has already taken over IoT and POS machines.
>It's laughable how incompetent they've become by being at the top all these years. Can't really blame them, you guys were throwing money at them while they served you the rehashed diarrhea plate of the past year.
Parker Carter
tl;dr intel damage control, moving goalpost yet again, and paid journalists making positive speculations
Cameron Wood
[Screams in 14nm]
More fake benchmarks
Xavier Bell
still waiting for a 10w 4 cores skylake tier ipc to replace my cluster
No, dumbfuck, after the core features he promised for Vega turned out to be pure vaporware that were never even put into software development that he bullshitted about to conceal that RTG had spent 3 years overclocking Fury and hadn't even started the drivers until 3 months before hardlaunch, anyone with two brain cells to rub together realized he went full poojeet snake oil salesmen.
The last thing any tech firm should ever want as a department head is an asshole who *lies* about the state of development to avoid scrutiny until it's too close to hardlaunch to fix anything. Raja has shown himself to be that with Vega, and now Intel is going to suffer at his hands.
Jaxson Price
AMD´s ¨¨7nm¨¨ are actually 10nm, just a reminder for you shitposter.
Gavin Brooks
>TSX is disabled in newest microcode release AFAIK only for haswell
