Quantcast
[ 3 / biz / cgl / ck / diy / fa / g / ic / jp / lit / sci / tg / vr / vt ] [ index / top / reports / report a bug ] [ 4plebs / archived.moe / rbt ]

Due to resource constraints, /g/ and /tg/ will no longer be archived or available. Other archivers continue to archive these boards.Become a Patron!

/g/ - Technology


View post   

[ Toggle deleted replies ]
File: 120 KB, 1200x600, d1f5e54e0b476bc0b5514643f4c54588_XL.jpg [View same] [iqdb] [saucenao] [google] [report]
72079450 No.72079450 [Reply] [Original] [archived.moe] [rbt]

Prediction:
Intel's next gen will be 10nm. They're already capable of producing it, but hinting at that right now would be a bad business move, as it would decrease interest in current Intel products and give AMD an early warning.

You all owe me Domino's if i'm right.

>> No.72079458

cool blog

>> No.72079462

>>72079450
>They're already capable of producing it
We know that. Their discrete GPUs they plan to release in 2020 are on 10nm

>> No.72079468

>>72079462
I worded that poorly, that part is known of course.
What i meant is that they're capable of producing 10nm desktop CPUs already.

>> No.72079472

>>72079462
highly doubt. wasnt raya at samsung fabs a few times??

>> No.72079480

>>72079450
>they are capable of producing it
yeah with less than 20% yield rate

>> No.72079495

INTEL 7NM WILL BEAT AMD 7NM IN 2027

>> No.72079507

>>72079495
That's pretty backwards, i mean AMD 7nm barely beats out Intel 14nm and not even in all tasks.
If intel steps their manifacturing process up at all, AMD is back in the cuckshed.

>> No.72079542

>>72079507
new processing node doesn't mean better performance
by the time intel will be on 7nm amd will have already been on 3nm with and every year they will have increased their IPC by 20% or more

>> No.72079553

>>72079507
not to mention that intel's 10nm will bring a clock speed regression so the IPC increase will be useless for gaming

>> No.72079566

>>72079450
>what is yield rate %
Lol we won't see it for a while, definitely not this year anyway

>> No.72079584

There is no 10nm

>> No.72079604

why the fuck does it matter? it's not gonna make the CPUs cheaper, faster, having lower input nor any significant ratio of those. its not gonna fix all their architectural flaw, poor scaling for number of cores, decades of backward compatibility mess in ISA, nor fix all the creeping security bugs.

>> No.72079613

>>72079462
It would be neat if Intel enters the GPU market since then there would be red, green, blue.

>> No.72079621

>>72079566
intel's yield rate for 10nm is below 10%

>> No.72079645

>>72079621
>source: my ass

>> No.72079666

>>72079645
https://www.anandtech.com/show/13405/intel-10nm-cannon-lake-and-core-i3-8121u-deep-dive-review/2
>He reports that Intel’s yields were sub-10% for the Cannon Lake 10nm CPU, well below the 60% Intel had expected at this point

>> No.72079673
File: 66 KB, 960x928, 1563717784788.jpg [View same] [iqdb] [saucenao] [google] [report]
72079673

>>72079450
>10nm 2015
AHAHAHHAHAHAHHAHAHAHHAHAHAHHAHAHAHHAHAHAHHAHAHAHH

>> No.72079692
File: 765 KB, 3490x3124, 1561651641337.jpg [View same] [iqdb] [saucenao] [google] [report]
72079692

>>72079673
>7nm in 2017

>> No.72079697

Reminder once you stumble in the transistor race you lost forever.

>> No.72079702

>>72079542
>3nm ipc
What

>> No.72079710

>>72079566
this. why else pushing low core count mobile shits when 10nm would be soo amazing in servers?

>> No.72079717
File: 10 KB, 230x220, 1519067311778.jpg [View same] [iqdb] [saucenao] [google] [report]
72079717

>>72079666
devilish

>> No.72079724

>>72079468
No shit, really? They released 10nm laptop cpus last year. It doesn't mean the node is working. It's a pice of shit and will be forever.

>> No.72079726

>>72079702
by the time intel will be on 7nm amd will have already been on 3nm
amd will have increased their ipc every generation by 20% or more while intel will need to shrink their node again in order to get an ipc increase
>>72079710
who cares about 10nm 28 core server cpus when amd has 7nm 64 core server cpus with 256mbs of cache

>> No.72079732

>>72079702
generational uplift of IPC on top of process node advantage, be power or clockspeed

>> No.72079735

>>72079692
>1 atom graphene in 2020

>> No.72079749

>>72079666
devilishly checked

>> No.72079764
File: 30 KB, 298x279, 1541538409899.png [View same] [iqdb] [saucenao] [google] [report]
72079764

>>72079726
>amd has 7nm 64 core server cpus with 256mbs of cache
mfw

>>72079726
>amd will have increased their ipc every generation by 20%
i doubt but yet hope they'll deliver something like that, everything over 10% would be fucking great. last i heard was Norrad (norrod?? this Forest guy) said that Zen3 will be alot of power efficiency improvements but thats pretty old, if i remember correctly. i bet the I/O-die and IF still uses a shitton of power and could need some sort of a upgrade

>> No.72079767
File: 84 KB, 419x238, doubt.png [View same] [iqdb] [saucenao] [google] [report]
72079767

>>72079726
>3nm

>> No.72079800

>>72079767
i dont know dude if intel brings 7nm to desktop by 2025 then amd will be already on 3nm and beyond

>> No.72079809

>>72079767
>https://www.anandtech.com/show/14666/tsmc-3nm-euv-development-progress-going-well-early-customers-engaged
it will take years but its coming

>> No.72079884

>>72079450
There's no 10nm

They'll use 14nm+++++++ until 2021/2022...or until their own 7nm process shows up

Their Xe GPUs will be made by Samsung and depending on their reception and success, Intel will pay off gooks to build their new i7s

>> No.72079915

>>72079884
>Intel will pay off gooks to build their new i7s
better sell all intel shares

>> No.72079949

>>72079884
>>72079915
kek this

>> No.72079985

>>72079450
You're going to summon the carbon nanotube retards.

>> No.72081131

>>72079450
What if they found out how to make graphine tubes at scale and just said "Fuck it, we'll take a soaking this year" but then they come out with this 1nm fucking insanity
Then i woke up, but still a cool dream

>> No.72081232

They can barely get out low volume 10nm mobile chips that perform moderately better than the last generation. Desktop parts are end of 2020 at the absolute earliest.

>> No.72081255

>>72079450
Intel isn't gonna produce 10nm CPUs until 2021/2022 at the earliest, and they'll probably be even more expensive than the i9900k

>> No.72081332

>>72081255
Don't forget to buy your 14nm+++ 10 core with no ht, 5.2ghz 350 watt i9 for 600 dollars later this year

>> No.72081350

>>72079450
There might be some truth to that. Intel 14nm already is slightly ahead of AMD 7nm...10nm will be a slaughter so why rush things?

Why do you think Nvidia kept the 1000 series going on for so long?

>> No.72081381

>>72081350
>10nm will be a slaughter
At lower than 4ghz?

>> No.72081433

>>72079450
>Prediction:
>Intel's next gen will be 10nm.
That isn't a prediction, anon. Intel has a public roadmap showing exactly when they're releasing what parts. They've already started shipping their 10nm mobile parts to OEM, and they're ramping up more. They'll be shipping SKUs between 6w and 25w, up to 4c/8t, with a big "Iris Plus" IGP.
High performance desktop parts however are still going to be 14nm+++ for a short while until intel tweaks the process enough to get better yields with larger parts and higher clocks.
>They're already capable of producing it,
They're capable pf producing small low TDP parts. Anything else would be losing money on a per die basis just like the OG desktop Broadwell which was made end of life after being on the market for 2 months.

This is nothing but retard fanfiction on your part, OP.

>>72081350
Architecture and process are entirely different things.
Per clock 7nm Zen2 consumes significantly lower power than intel's latest 14nm Skylake refresh.

>> No.72081556
File: 234 KB, 768x1024, 1544980774770.jpg [View same] [iqdb] [saucenao] [google] [report]
72081556

>>72079613
>Raja Koduri GPUs
Member Vega?

>> No.72081840

>>72081131
That would be plausible but
>Intel

>> No.72081950

>>72079724
>It's a pice of shit and will be forever.

This isn't a thread about Zen 2

>> No.72082149

>>72081950
Exactly, anon.

>> No.72082156
File: 42 KB, 426x510, 1560441878880.jpg [View same] [iqdb] [saucenao] [google] [report]
72082156

>>72081950
>t.

>> No.72082176

>7nm
>2017
O NO NO LOOK AT THIS DUDE

>> No.72082190

>>72079915
>sell all intel shares
>when the only thing holding intel back is process node
(lol

the moment they touch 7nm is the time they will annihilate amd, even with old arch

>> No.72082202

>>72082190
keep dreaming kike shill

>> No.72082237

>>72082190
INTEL 7NM WILL ANNIHILATE AMD 2NM IN 2027

>> No.72082357

>>72082190
For high performance SKUs intel has nothing but 14nm silicon coming for all of 2020.
Glacial Falls, Basin Falls, Comet Lake, and a refesh of Coffee Lake S cover intel's near term releases from now until Q4 2020. That is their entire product stack from low TDP 35w SFF chips to 150w Prosumer market segment chips.

AMD will be releasing Zen3 parts on 7nm EUV next year that will be competing with intel's 14nm high end SKUs.
Come 2021 AMD will be releasing Zen 4 which is being fabbed on TSMC's 5nm EUV process. This is the time frame when intel might start releasing high performance 10nm silicon, non mobile parts. Its 5nm TSMC vs 10nm from intel.
We know from intel's own public figures that the OG 10nm process had up to 100mil xtors per mm2. This is the original node which no longer exists. What they're offering now is 10nm+ which has relaxed BEOL pitches to decrease density. However from TSMC we know their 5nm node has a maximum density of 173mil xtors per mm2. TSMC already had a significant advantage over intel's 10nm at 7nm with denser SRAM, but now its an entirely different ballpark. intel is falling majorly behind across the board.

By the time intel actually starts producing 7nm Trigate parts they'll probably be facing off with something AMD has fabbed on TSMC's 3nm EUV GAA node. intel isn't ever catching up here unless TSMC has a major fab fire.

>> No.72082541
File: 2.97 MB, 720x404, 1560209932977.webm [View same] [iqdb] [saucenao] [google] [report]
72082541

>>72079666

>> No.72082644

>>72082541
based

>> No.72082746

>>72082541
based
fuck intel and their damage control shills

>> No.72082954

>>72079666
>>72079717
>>72079749
>>72082541
>>72082644
>>72082746
Thats why the original 10nm node no longer exists, and now you have "10nm+" for which intel has disclosed no details publicly. Its not going to have anywhere near the 100Mt/mm2 they originally targeted.

They're 5 years late by their own original roadmap, and only shipping low TDP mobile chips now. It'll be 6-7 years late for top binned desktop SKUs.

>> No.72082983
File: 241 KB, 720x960, poo.jpg [View same] [iqdb] [saucenao] [google] [report]
72082983

>>72079472
He is

AMDrones are so afraid of Intel paying off Samsung to build their shit

AMD is only leading because Intel is sitting on old technology and 14nm++++++++++++, give them Samsung's 7nm and they'll send AMD back to irrelevance

Notice Nvidia is also moving to Samsung, both Intel and Nvidia will be shitting on TSMC

>> No.72082996
File: 48 KB, 800x729, 8nRqoXW.jpg.png [View same] [iqdb] [saucenao] [google] [report]
72082996

It's amazing how autistic this board has become over the size of transistors acting like it's all that's needed to make fast CPUs. AMD has been ahead of Intel for over 3 years now and only now barely beating Intel at the top levels.

You people have ruined this board with this retarded ass spam 24/7

>> No.72083034

>>72082983
Do Intel shills really believe this?
Inb4 wait 5 years for intel 7nm it will definitely beat amd's 3nm with its 20% better IPC over skylake and 333% less clockspeed

>> No.72083198

>>72083034
intel will be a 7nm super power in 2027!

>> No.72083237

>>72079450
Desktop lineup will remain 14nm and there will be another 10nm laptop paper launch. After that we'll see.

>> No.72083240

>>72079702
AMD will most likely be on zen 5 or 6 by the time they get to 3nm, so yes ipc will be higher by then

>> No.72083252

>>72082996
AMD has been ahead of Intel on process technology for exactly 3 weeks.

>> No.72083277

>>72083252
AMD has had 7nm products since last august, this is just the release of 7nm cpus for consumers

>> No.72083306

>>72083277
Intel has 10nm parts in their labs too, does that mean they are equal in process technology now?

>> No.72083323

>>72083034
intel 7nm superpower 2077 just wait

>> No.72083367

>>72083306
AMD's 7nm were available for servers since last August and now for regular PCs as well. Intel's 10nm doesn't exist as a product.

>> No.72083383

>>72083367
There are zero 7nm epyc parts outside of testing labs.

>> No.72083386

>>72083306
AMD released Mi cards last august retard, unlike intel that still has no real 10nm products

>> No.72083387

>>72083034
>pay off Samsung
>wait for 5 years
AMDrones can't even read properly

If they pay off gooks they won't have to wait for 5 years, they'll have a new CPU in a year or two

>> No.72083402

>>72083383
He is talking about 7nm gpus

>> No.72083403

>>72083383
You have brain damage pajeet

>> No.72083414

>>72083386
Tell me how Radeon Instinct competes with Intel CPUs.

>> No.72083416

>>72079450
I work at the cpu_jew. we have 10nm working.

>> No.72083423

>>72083414
Just face it retard AMD has been on 7nm for a year and intel is never catching up

>> No.72083466

>>72079495
intels 7nm = tsmc's (amd) 5nm

intels 10nm = tsmc's 7nm

>> No.72083525

>>72083466
This is nowhere near true

Intel's 10nm is not the original 10nm they initially advertised, its a watered down version of the original

>> No.72083559
File: 21 KB, 600x400, 1528452813082.jpg [View same] [iqdb] [saucenao] [google] [report]
72083559

>>72082190
>when the only thing holding intel back is process node
thats not the /g/ humour thread, summerfag

>> No.72083570

>>72083387
If they pay Samsung to do it then their 10nm transistor density will be worse than amd's 7nm

>> No.72083584

>>72083416

and how many backdoors?

>> No.72083601

>>72083416
Your dick doesn't count.

>> No.72083608

>>72083584
39 at the moment

>> No.72083668

>>72083387
MAGIC SEKRIT SAUCE 10NM ARCHITECTURE 2020 SUPERPOWER

>> No.72083698

>>72079462
>>72079472
>>72082983
It's more likely for intel to outsource the GPUs to some other fab simply because they need to make them fairly cheap to gain marketshare, and that sure ain't gonna happen on their problem-ridden 10nm.

>> No.72083730

>>72083525
where are the proofs

>> No.72083742

>>72083387
I hope you like massive mark up on top of Intels already enormous 60% profit margin. Samsung would be basically Intel admitting 10nm and 7nm are shit and so far behind on development that their fabs are now an albatross of worthlessness even worse than when AMD owned Global Foundries.

>> No.72083783

>>72079542
>new processing node doesn't mean better performance
It does, just indirectly. Efficiency gains allow for "wider" cores at same power as previous node which results in better IPC.
Saying that node shrink doesn't do anything just for the sake of saying that is stupid, because in reality it virtually always does, just indirectly.

>> No.72083980
File: 172 KB, 710x559, Droning.png [View same] [iqdb] [saucenao] [google] [report]
72083980

>>72083730
Shh don't ask AMDrones for evidence.
REMEMBER:
INTEL BAD
AMD GOOD
now repeat after me:
INTEL BAD
AMD GOOD

>> No.72084065

>>72083980
Repeat with me rabbi
INTEL GOOD
AMD BAD

>> No.72084081

>>72079450
>graphene
biggest meme of the 21st century

>> No.72084106

>>72084081
I think the problem is they still haven't figured out a viable method for large volume production, and that's necessary for it to not be a meme

>> No.72084147

>>72083466
Maximum density quoted at each node:
Original 10nm Trigate = 100Mt/mm2
Current "10nm+" unstated, but lower density than original
TSMC non EUV 7nm =96.5Mt/mm2
TSMC 7nm EUV 113Mt/mm2
TSMC also has an advantage over intel with SRAM cell density.
TSMC 5nm EUV 173Mt/mm2.

intel can't compete on any front right now. TSMC already has 5nm EUV in risk production btw.

>> No.72084181
File: 54 KB, 679x758, 1527629778452.jpg [View same] [iqdb] [saucenao] [google] [report]
72084181

>>72084065
ORANGE LOGO BAD

>> No.72084182

>>72084147
I kinda wonder if AMD will go with 5nm or take 6nm instead which, as I understand, basically a 7nm EUV+, more density but also backward design compatibility. I guess that depends on whether TSMC was offering 6nm back when Zen 3 was being designed.

>> No.72084189

>>72083730
intel had to relax BEOL pitches on 14nm to improve yields after the launch of Broadwell Core M was a huge disaster. Broadwell to Skylake saw a significant reduction in back end scaling.
Intel themselves have publicly stated that yields on their first OG 10nm part were below 10%. That part was a pitiful dual core mobile i3 that didn't even have a functional IGP. Sub 10% yield on that tiny low clocking part.

10nm+ is not the same process. They aren't going to keep trying for 100Mt/mm2 when they can't even produce a dual core die with double digit yields.

>> No.72084200

Honestly I don't really follow the hardware tends anymore. I've pretty much stopped caring. There's zero point to it. My computer does all I need so fuck it. If the PSU shits, I'll just order a new one and replace it.

The only thing I upgrade anymore is my storage. For my servers and for backup purposes. So long as ATX and SATA don't go anywhere in the next 10+ years I'll be ok.

>> No.72084211

>>72084182
The 6nm process is another stop gap for ARM vendors. Companies like MediaTek who want more performance, but already have existing designs built around 7nm design rules. Its not a high performance node.

Enterprise Genoa has allegedly already taped out on 5nm. Charlie from S|A says so anyway.

>> No.72084231

>>72084147
>intel can't compete on any front right now. TSMC already has 5nm EUV in risk production btw.
Hmm we don't really know that though. Information about TSMC or Samsung progress is much easier to obtain due to how many customers they're serving. Intel on the other hand mostly does intel stuff, so aside from the bare minimum they tell their shareholders, we really don't know much on what's being cooked in their fabs

>> No.72084282

>>72084189
>hasty generalization from another node and no direct evidence
lol

>> No.72084288

>>72084231
If intel had the capacity they'd be using it. They're still launching only mobile chips for late 2019, which won't have any real presence in the market until sometime in Q1 next year. For all of 2020 they are only releasing 14nm chips for all high performance segments. Everything not mobile is still 14nm.

Thats a pretty damn clear indication that they cannot afford to fab high performance large die 10nm parts. They're no longer the industry leading foundry, and they're not going to magically jump back ahead.

>> No.72084422

>>72084282
Sure, little kid. intel couldn't produce a gimped dual core mobile i3 die, but they're totally going to ship volume levels of fully enabled quad core HT chips with a huge IGP no problem.
Definitely going to have the exact same MMP and CPP.

>> No.72084456

>>72082983
Even without the process issue, AMD has better IPC than Intel.

>> No.72084517

>>72079450
>They're already capable of producing it

a halo product that it wont even go to mass production is intel version of capable so far

>> No.72084578

>>72084288
>If intel had the capacity they'd be using it
I'm not talking about 10nm though, I'm talking about what's beyond. It's not like 10nm is the only thing they're working on, just like TSMC has 5, 3 and so on in their pipeline.

>> No.72084595

>>72084578
Process tech is iterative, built upon prior generations. At 7nm intel is still using their Trigate FinFETs. I don't believe they ever committed to moving away from Trigates at 5nm either.
Speculating on what they have waiting in the wings when they're still floundering on 10nm is pretty pointless.

>> No.72084668

>>72084422
>My personal incredulity is too strong so I'll just make up my own baseless reality instead

>> No.72084691

>>72084668
>I don't like the real world so I'll let my blind fanboyism delude me
Its no loss to anyone that you choose to be retarded, kid. The facts of the matter are pretty straight forward.

>> No.72084729
File: 13 KB, 400x224, projector projecting.jpg [View same] [iqdb] [saucenao] [google] [report]
72084729

>>72084691

>> No.72084779

>>72084729
>I'm not a retarded fanboy blatantly ignoring reality because it contradicts my narrative
>y-you're projecting!
Nice COPE, little kid.

sub 10% yields with a gimped dual core mobile i3 sans functioning IGP.
Thats reality for intels initial 10nm line. That doesn't just magically go away. Loosening BEOL pitches to decrease logic density and improve defect density is how you remedy that. Its what intel did with 14nm Trigage. Its what they did with 10nm Trigate too.
When a foundry encounters a problem that makes it impossible for them to generate positive revenue they don't keep repeating the failure. They fix it.

>> No.72084847

>>72084779
Yeah, that's the only remedy. You definitely know what you're talking about. Keep going. It's always fun to watch you pseudo-intellectual redditor teenagers squirm.

>> No.72084885

>>72079666
>Charlie over at SemiAccurate
You mean the manchild blogger that talks out of his ass all day long and charges >$1000/year to read his insane ramblings?

>> No.72084930

>>72084847
>I-I-I-I'm not retarded
>y-y-you're reddit!
Nothing but highly informed quality posts out of you, little kid. Very impressive.
Yes, that is the easiest solution, which is why it was employed with 14nm Trigate as well. The denser you make logic the higher defect density will be. You can't just readily reduce lith masks and exposure steps with a process, you can't just change fin profiles to alter electrostatic characteristics without affecting performance, but you can change MMP and CCP to reduce density.

Keep shitposting, little kid. Your COPE is all you have.

>> No.72084997

>>72084147
Intel sat on their ass too much in the past years. They can't compete with R&D with all these mobile shekels TSMC (and Samsung) have gotten in recent years. They'll have to play catch-up for the foreseeable future - unless they fold and let other's do their latest chips while their own fabs still push out 14nm crap for the masses to put their own fabs to good use.

The only realistic way Intel could catch up is the competitiors being stuck on 2 or 3 nm because it's the physical limit.

>> No.72085310

>>72084997
Intel had problems from the onset when they first switched away from planar gates. The Trigate structure requires more lith masks, more layers than comparable bulk FinFETs from other foundries.
They haven't been sitting on their asses, they've just consistently been biting off more than they could chew for going on a decade now. They've developed processes which have more steps, aimed at absurd densities, and they keep having massive delays and launch issues because of it.

>> No.72085430

>>72084930
So post your hard data on ice lake density. Should be easy.

>> No.72085942

>>72084106
that's always the problem
diamonds for example can be made artificially
but only very small ones and with low quality

>> No.72086127

>>72081332
you forgot a new motherboard

>> No.72086438

>>72079450
>4 cores - 2006
>8 cores - 2014
>12 cores - 2022+
Jesus.

>> No.72086759

>earlier this year:
7nm 5ghz is coming!
>now:
3nm IPC will surely beat intel this time! just wait some more!

I've been told to just wait from AMD since 2011. what the fuck?

>> No.72087564

>>72086759
I've been waiting for Intel to not be the corelet choice since 2010, how long do I have to wait?

>> No.72087632

>>72087564
Intel's had 8S 28C for a while now, that's more than AMD's 2S 64C rome which hasn't even been released yet.

>> No.72087714

ITT: Brainlets who doesn't know that 7nm/10nm means jack shit.

>https://en.wikichip.org/wiki/10_nm_lithography_process

>> No.72087775

>>72079450
>They're already capable of producing it,
then why haven't they? its been 3 years since the slated 7nm release

>> No.72088237

>>72087714
Nanometers don't matter

>> No.72088509

>>72084885
lmao. Eat shit, shill. https://semiaccurate.com/2019/06/05/a-look-at-intels-ice-lake-and-sunny-cove/

>> No.72088510

>>72081556
Nice PT Cruiser

>> No.72088637

>>72087775
Just wait...

>> No.72088923

>>72088509
Way to confirm my point.

>> No.72089092

>>72088237
That's what I said. Intel 10nm will have more than 1.5x the density of TSMC 7nm

>> No.72089175

>>72088923
Even Intel's gimped 10nm-- is low yield garbage. Kill yourself.

>> No.72089632

>>72089175
Yeah, keep repeating your mudslinging over and over again. I'm sure it'll become true eventually.

>> No.72089720

>>72089632
>https://semiaccurate.com/2019/06/05/a-look-at-intels-ice-lake-and-sunny-cove/
Holy crap that site is literal hot garbage.

>> No.72090117

>>72079450
core i7 1065G7 will win against all Ryzen laptop CPUs

>> No.72090674

>>72079450
Microsoft buys intel and the ceo cashes out while he's ahead.

>>
Name (leave empty)
Comment (leave empty)
Name
E-mail
Subject
Comment
Password [?]Password used for file deletion.
Captcha
Action