>be amd >glue two cpus together >actually werks and

>>60660516
Their architecture is a monolithic low-yield ringbus turd.

you got shitwrecked

>>60660516
Don't worry, they'll copy IF 5 years later just like they copied HyperTransport and K8's IMC.

>>60660700
And x86-64.

>>60660744
Too bad AMD's train is not going to stop any time soon.
>we're going to throw more money into R&D
>t. Dr. Lisa Su

>>60660795
amd should learn how to use the loo first

>>60660795
based Lisa

>>60660838
Hello IMG_xxxx. No, mr. Brian Krzanich is not going to pay you for damage controlling SKL-X abortions on gook cartoon trading forums.

>>60660516
they are already EMIB-ing cpus with fpgas

>>60660894
That's about as niche as it gets. We're talking about scalable designs.

>>60660516
that's what they did with the Q6600 over 10 years ago, and that was a LEGEND. Better question is WHY they don't do it.

>>60660940
Intel hates NUMA
AMD was always more comfortable with NUMA designs.

>>60660940
My dad still uses my Q9650
Works like a motherfucking charm

>>60660956
>Intel hates NUMA
They don't just hate it, their NUMA implementations suck. See 4S Xeon scaling.

>>60660969
Well, they're gonna have to deal with it, because there's nothing stopping AMD from going even crazier with 300mm^2+ dies and gluing them together, meanwhile their monolithic die maxes out at what... 700mm^2? AMD would have almost twice the total amount of silicon against them, their highend parts would compete with AMD's mid range parts.

>>60660838
Speak for yourself Rajesh

>>60660935
emib will likely see some implementations outside that niche.

ref. http://www.anandtech.com/show/11115/intel-confirms-8th-gen-core-on-14nm-data-center-first-to-new-nodes

it's weird how different these threads are to /v/'s.
they're actually touting the X-series as destroying Ryzen despite Intel just following suit with moar corez and having a higher TDP and RRP.

They're also still posting day 1 Ryzen 7 gaming benchmarks.
To be fair we've been taught you need Intel for gaymen for over 5 years now.

>>60660986
7nm IBM is what, 30% denser? They can put 4 more coarz and one more memory channel on Zeppelin 2.0.
>>60661002
>data center first to new nodes
>Cannonlake-EP/EX cancelled a few weeks late
LMAO.

File: 122523494643.gif (25KB, 300x291px) Image search: [Google]

25KB, 300x291px

>>60660956
>Intel hates NUMA
HERESY
https://www.youtube.com/watch?v=OE8WzYNRPNU

>>60661012
>/v/
/g/ isn't much better, but at least people are aware physics exists and you're not getting 10, 12, 14, 16 or 18 cores over 3.5GHz without blowing up the motherboard VRM and the PSU

File: 1472443306956.png (241KB, 500x375px) Image search: [Google]

241KB, 500x375px

>>60661028
>that video
a simpler time.

>>60660940
i thought i had q6600 but i have t6600

>>60661012
Every time I think /g/'s tech threads are retarded, I'm reminded /v/ exists. But they're still over there fighting over console exclusives, so it's obviously a bunch of underage b& that can't actually afford more than one console or even a single mid-range gayman PC.

EPYC houses 780mm^2 of 14nm silicon split between 4 dies. I wonder how big 28c SKL-EP will be.

>>60661092
consolewar threads look good compared to the SJW threads.
Far Cry 5 got so many threads over the fake controversy it was pathetic.

>>60661094
We should know when Skylake-X die sizes come out.

Skylake-EP is XCC(triple ringbus or KNL mesh), Skylake-X is HCC(dual ringbus)

>>60661105
It's KNL mesh.

>>60661105
Can confirm this, LCC Skylake-X die is up to 12 cores, HCC is either 18 or 20 cores.
Everything over that is XCC.

>>60661012
>>60661058
>>60661092
Don't even get why /v/ would care about HEDT platforms as it's massively outside of their price range (of mom's allowance), and also is clearly not for "muh gaymes" unless you're a complete retard.

File: Lisa-Su-Ryzen-CPU.jpg (387KB, 1920x1080px) Image search: [Google]

387KB, 1920x1080px

>>60660795
ALL HAIL THE QUEEN!

>>60661116
For HCC and LCC anand confirms it's still ring bus

http://www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999/2

XCC would naturally require a third one for the extra 2 memory channels, so maybe that's KNL mesh, but that's a large arch difference so I don't see it happening so easy.

>>60661138
if you think the average poster gave it anymore more thought than "i9? that's more than i7!" then you've got some misplaced faith in their intelligence.

>>60661146
Oh fuck, 5 more years of bingbuses before Intel finally copies IF.

File: 1466635978292 amd lisa su ceo.jpg (7KB, 125x132px) Image search: [Google]

7KB, 125x132px

>>60661143

>>60661164
>BUY AMD

>>60661153
Looks like our friend IMG_xxxx-kun worked overtime there.

>>60661153
Even the dumbest normalfag will stop once he sees the prices.

>>60660940
You are thinking of the Pentium D. 2x the housefire for 1.2x performance. C2D was the proper multi core CPU.

File: 1496141494918.jpg (148KB, 1192x702px) Image search: [Google]

148KB, 1192x702px

>>60660516
>>60660516
64 lanes

>>60661238
>40% more lanes than Intel's most powerful Xeon

Oh God I'm dying over here.

>>60661215

That's not what he means. Iirc C2Q was two C2D's glued together

>>60661238
>x399
>ONE HUNDRED MORE than intel x299
intel finished and bankrupt

File: The_Best_I_Can_Do_2_i9_Boogaloo.png (2MB, 1920x1200px) Image search: [Google]

2MB, 1920x1200px

>>60660516
Enjoy waiting until 2021 for a new uarch, faggot :^)

File: 1494767421045.jpg (43KB, 630x630px) Image search: [Google]

43KB, 630x630px

>>60660838
Yours tears are DELICIOUS!

File: apprentice and second firing door in foreground.jpg (331KB, 1067x1600px) Image search: [Google]

331KB, 1067x1600px

>>60661012
>MFW we actually made them go back to /v/eddit

>>60661276
huh, so c2q is 2 dies on mcm? never knew that.

>>60661188
not any more than it stops them from purchasing top priced 2 titans for sli

File: 1485636474297.jpg (171KB, 960x1280px) Image search: [Google]

171KB, 960x1280px

>>60661188
I think you underestimate them

>>60661012
Because they don't know how computers work, so shills easily take control of the conversation. And the ones who do get into the argument fuck off 20 seconds later to boot up skyrim, because it's almost bedtime.

File: POWER5 MCM.jpg (458KB, 1488x1488px) Image search: [Google]

458KB, 1488x1488px

>>60660516
They have, it's called the Pentium D. I believe the Core2 Quad was also an MCM.

>>60661437
that image makes me angry.

>>60661437
Okay I'm sorry kill it with fire

>>60661437
>>60661460
>>60661465
x86 deserves to burn to the ground, desu.

>>60661472
Intel's helping.

>>60661472
in favor of what? ARM?

>>60661492
POWER or MIPS

>>60661143
M..mommy?? Me milky mommy!

>>60661492
MIPS is being sold out and POWER is simply too high power.

>>60660838
Seems they're doing pretty well shitting on Intel instead.

>>60661514
>MIPS is being sold out
What?

What if AMD decided it gives no fucks and puts 16 cores at $799?

>>60661524
Apple will stop using Imagination GPUs so they are going like almost bankrupt.
>>60661535
12 cores.

>>60661499
in the consumer sphere?

>>60661524
not that anon but here: http://www.anandtech.com/show/11334/imagination-to-sell-mips-and-ensigma

>>60661499
>housefires
>dead

>>60661524
http://www.anandtech.com/show/11334/imagination-to-sell-mips-and-ensigma

>>60661535
Doubt they would, but the shoah would be legendary.

>>60661535
I've heard from some sources than threadripper prices will be really disruptive, considering a Zeppelin die costs like $40 to make, and two cost $80, they make mad margins already, so taking a big fat dump on Intel isn't too strange.

>>60661567
>Zeppelin die costs like $40 to make, and two cost $80
And they have FUCKING 80% YIELDS.

>>60661575
Yes, impressive, if they don't give a fuck about margins too much they could drop prices down by half and still make good money from them.

Shareholders won't like it, but shareholders have only been fucking AMD in the ass, revenue is more important.
Nobody bitches to Apple about 37% gross margins


So I'm actually expecting

>ŧhreadripper launches
>Ryzen 7 and 5 prices fall
>12 core threadripper starts at $500
>16 core at $800 with a lower TDP variant at $650

>>60661551
Memes and falsities? Great reasons.

>>60661539
>>60661547
Well shit, maybe whoever buys it will actually do something with it.

>>60661567
$330 for 8c/16t would be considered disruptive, and it is, but we can see that intel continues selling products unscathed. their grip on the mentality of consumers (and also most professionals) remains airtight and some of the early pro intel reaction to skl-x only emphasizes that

>>60661606
They can price 12core extremely aggressively.

>>60660795
it wont matter since keller is working at tesla, should have put a ring on him tbqh

>>60661630
>that intel continues selling products unscathed
>-14% desktop Q1 revenue with full KL quarter
Lmao.

>>60661644
Keller managed the team. It's Clark who made Zen.

>>60661606
Won't the X399 board cost more than the X299 board?

>>60661672
Maybe a bit.

>>60661672
That's mostly down to chipset pricing, and Intel asks for $60 for their X chipsets.


So no, at worst they'll be equal with similar features.

>>60661058
I pray everyday for an aftermath report with video of nascent housefire as an 18 core oced to 5ghz trys to pull 800w
>vrms pop
>standoffs melt
>cpu melts
>socket combusts
>heatsink droops a little
>molten tin melts through the bottom of the case and sets desk on fire
>psu cap explodes, colors invert
>camera fumbling noise as video ends

>>60661651
are you comparing to q4? semiconductor industry has seasonally strong q4 vs weaker q1.

but no you are on the right track. i should say 'relatively' unscathed. their gross margin remains ridiculous so i am hoping for epyc to put the screws on them.

but we'll see.

>>60661606
I think $999 for the 16-core would be a plenty low enough pricepoint to inflict a total single socket Holocaust on Intel. Remember it has 64 PCIe lanes and ECC support, it would throw most of Intel's single socket Xeon in the bin at those kinds of price points.

>>60661716
> gross margin remains ridiculous
Cool, but volume of sales will drop hilariously low.

>>60661729
i want to believe, anon. i really do.

File: core_i9.jpg (283KB, 1920x1200px) Image search: [Google]

283KB, 1920x1200px

>>60660516
Housefire(tm) experience guaranteed.

>>60661683
More pins and other stuff?

>>60661723
AMD needs to murder Intel, not kill them, they have to tell people that Intel is no longer the premium brand, this is extremely important.

They can do this with Threadripper and Epyc(consumer edition)

And if they had any sense, they would.

>>60661773
> And if they had any sense
Don't worry, AMD's CEO is among the most competent ones in the industry.

>>60661764
>other stuff
Is mostly the same, AMD has more controllers directly on the CPU than Intel so there's a slight advantage there.

Yes, there's more pins, but socket pin costs are small, the retention bracket is more expensive.

>>60660516
this cpu gluing technology might be revolution in gpu actually, make bunch of shitty small dies for pennies glue them all together suddenly you have beast gpu, just gotta make sure they all appear as single gpu.

>>60660940
i had Q6600 for years, before i bought 3570k 5years ago... srsly we have been stagnating in one place for so long.

>>60662135
that is the goal, yes.

ref. http://semiaccurate.com/2017/01/19/amd-infinity-fabric-underpins-everything-will-make/

>>60660516
>Simlutaneous Multy-Track drifting.jpg
Oy vey

>>60661138
according intel i9 is for ultra high end gaming, so clearly its for gaming? Intel said it themself you want the 18core cpu for 12k gaming

>>60662198
You're so beyond GPU bottlenecked at those resolutions you might as well get a i5 or R5

something people didnt notice..

it has 64 lanes out of the box from the cpu....ONLY

but
it has x3 m2
and 24 more from the chipset itself

>>60662135
That's their plan for Navi. It'd probably have 1 base dies with ACEs+command processor and smaller dies with Shader Engines.

Why would Intel copy an inferior Rypoo Pooripper garbage with bad performance, poor AVX2 performance and no AVX-512 performance at all?

You buy Intel to do real work, you buy AMD if you're a poorfag

>>60662210
so what, intel still is recommending there i9 10-18cores for ultra high end gaming, so where that, i9 isnt for gaming came from? When manufacturer themself say its for gaming?

>>60662242
Marketing usually has no relation to reality.

>>60662238
>AVX2 performance and no AVX-512 performance

Housefire central.

File: Core_i666.png (2MB, 1920x1200px) Image search: [Google]

2MB, 1920x1200px

>>60662238
Cry more you bootyblasted shekelchaser :^)

>>60662290
t. constipated poojet

File: help.png (6KB, 350x314px) Image search: [Google]

6KB, 350x314px

>>60662238
>AVX-512
LMAO they can't even give away Xeon Phis, literally no one gives a fuck about AVX-512

>>60661515
haha nice

>>60660940
Q6600 is love Q6600 is life

Intel Core i7 7800X @ 5755MHz (1918 cb )
http://hwbot.org/submission/3563200_shimizu_cinebench___r15_core_i7_7800x_1918_cb

AMD Ryzen 5 1600X @ 5148MHz (1784 cb )
http://hwbot.org/submission/3561222_johan45_cinebench___r15_ryzen_5_1600x_1784_cb/

File: tears_of_rage.jpg (68KB, 700x700px) Image search: [Google]

68KB, 700x700px

>>60662304
Your tears are DELICIOUS :^)

File: 1494787256743.png (180KB, 370x544px) Image search: [Google]

180KB, 370x544px

>>60662461
>Model: Intel Core i7 7800X 'Skylake-X'
Cooling: Liquid Nitrogen
Speed: 5,755MHz (+64.43%)
Partner Intel

>>60662461
That IPC difference

>>60662521
the i7 has 600mhz on it, it's not all ipc

>>60662153
exact same story here friend

>>60662539
You mean it's not at all IPC, right?

Try doing the math yourself.

>>60662539
It's clocked significantly higher and is only 7.5% faster... That's lower IPC

>>60661164

>does this look like the face of mercy to you!?

How long until Intel releases ©Ultra Threading™ enabling a full 3 threads / core in order to successfully fake MOAR KORS?

>>60662662
Never. SMT4/SMT8 requiers very specific ultrawide core design, see POWER.

>>60662553
1918 - 100
1784 - 98.01

5755 - 100
5148 - 89.45%

>>60662662
Not without major increases in core width, so never.
Maybe their upcoming Itanium in 2021

>>60662553
>>60662563
then we're all saying the same thing. i misinterpreted anon to mean cb = ipc measurement

>>60662694
i thought itanium was tabled for good (http://www.anandtech.com/show/11372/intels-itanium-takes-one-last-breath-9700-series-released)

>>60662752
He's talking about new Intel's uarch, which is supposed to break BC in some way. Yes, again.

>>60662752
Bad working.


Their "Itanium v2 return of the EPIC housefire"

http://sqlblog.com/blogs/joe_chang/archive/2016/12/27/intel-processor-architecture-2020.aspx

>>60662692
>1784 - 98.01
it's actually 93%

>>60662662
Knight's Landing has four threads per core, so 72 cores/288 threads

>>60662799
Knights Landing also uses fucking silvermont cores.

>>60662799
It uses wider duct taped Silvermont (Atom) cores.

>>60662760
i thought they learned their lesson the last time. it couldn't be any worse timing with the mess that is <10nm nodes, a competitive AMD, and ARM nipping at the ankles

>>60662899
You forgot POWER being ultra-competetive at HPC again. Yes, the situation is dire for Intel.

>>60662769
interesting article. it's curious that he basically suggests psuedo-HSA/big.little as possible paths forward for not-itanium

>>60662927
my only reservations with power is that, iirc, GloFo has to fab them since that was part of the agreement of taking over IBM's foundries (correct me if wrong); is power9 actually out now?

>>60663066
>is power9 actually out now?
It was oficially announced, but is not shipped yet.

File: 1494272913148.jpg (273KB, 1032x774px) Image search: [Google]

273KB, 1032x774px

>>60660516
IT'S NOT FAIR

I'm a little confused. Will 2018 be the year of the 14nm+ AMD APUs and 2019 the year of the new 7nm Ryzen CPUs? Also will they use indium gallium arsenide?

>>60663540
>indium gallium arsenide?
Meme.
>Will 2018 be the year of the 14nm+ AMD APUs
And Zen+ refresh and maybe some risk production of 7nm APUs.
>2019 the year of the new 7nm Ryzen CPUs
Yes, since GloFo lists 7nm with EUV on their roadmaps for Q1 2019.

>>60663580
Considering Zen 2 has already taped-out, I think AMD might skip Zen+ altogether and release Zen 2 in 2018.

>>60663832
Zen2 tapeouts are scheduled for late 2017, anon.

>>60663879
Whoops. You are correct. Still, that means they could get them out the door in 2H 2018 if they really want to.

>>60663940
GloFo 7nm DUV is only Q3 2018 at best and it's fucking 80 masks quad patterning retardation. Aka you still have to wait for 2019 for EUV.

>>60661661
His title was VP and chief architect. He definitely had hands on with the design.

>>60663580
>>60663832
There is no Zen+. Zen on 14nm+ is the APUs with Vega (also on the same 14nm+)

>>60663981
Clark was Zen's chief architect though. Keller was chief architect of ye olde K7. He was team lead for Zen/K12 team. Also where's K12?
>>60663995
>what is Pinnacle Ridge
Literally on freshest AMD roadmaps.

>>60664006
>Keller was chief architect of ye olde K7
When they said he was joining AMD his title was Corporate VP and Chief Architect of Microprocessors. AMD was working on Zen and K12, sister uarchs but implementing different ISAs. It's basically a certainty he played a key role in high level design. Team lead isn't a real title you know, you're a lead x. My department's "team lead" is the principal sotware architect, he still makes big decisions on the actual design of the software and touches code all the time. My team lead is the lead programmer of our team, still touches code all the time and so forth. Being a team lead isn't some misty manager position.

>>what is Pinnacle Ridge
A mystery. Also, that roadmap said Raven Ridge was 2018 but it's apparently moved up by an astounding half year. It's clear the roadmap was intentionally vague about any details, including the timeline.

File: 1489510222325.png (1MB, 1361x810px) Image search: [Google]

1MB, 1361x810px

>>60664367

>>60661363
>>60661747
>>60662290

But why is it in a graphics card?

Why doesn't AMD make processors for laptops?

Is it because the more cores strategy doesn't work there?

>>60664767
They do.

>>60660516
Because Intel believed that MOAR COARS was just a meme and would never actually go anywhere, so they never bothered to actually improve their architectures on a basic level - but instead just kept relying on node shrinks and minor optimisation.

>>60664767
APUs are supposed to drop later this year. I suspect AMD will enter the laptop market with them then.

>>60664767
The power efficiency of Ryzen makes it an excellent choice for laptops. They're already on their way. http://hexus.net/tech/news/laptop/105937-asus-rog-teases-amd-ryzen-powered-gaming-laptop/

>>60664776
>>60664800
>>60664802
They haven't had a good laptop CPU since Duron.

>>60664870
I'd just like to say that you're one of my favourite tripfags. During all my years here, your tripcode is one I like to see pop up the most!

File: LRM_EXPORT_20170411_145818.jpg (1MB, 1293x1730px) Image search: [Google]

1MB, 1293x1730px

>>60664767
Kill yourself you perma-BTFO'd tripfaggot

>>60664870
>>60664896
Nice samefag. Kill yourself, btw.

Holy shit Duron was 16 years ago.

I don't think my local Best Buy even carries laptops with AMD processors.

>>60664870
fagget

>>60660524

I don't think you know what most of those words mean.

>>60662662
AMD is the only company that does that. My i7 had 4cores on the box.

>>60664962
He's right, expect LCC uses KNL mesh instead of bingbuses now.

File: xeon_e5_v4_hcc_rings.jpg (183KB, 1200x706px) Image search: [Google]

183KB, 1200x706px

>>60664962
I want to get off Intel's wild ringbus ride.

>>60664984
>KNL mesh instead of bingbuses now
Isn't it bascially the same shit?

>>60665006
It's mesh of smaller bingbuses AFAIK.

>>60664995
Mr Intel wild bingbus ride never ends

File: bipeline.png (973KB, 801x1500px) Image search: [Google]

973KB, 801x1500px

>>60661363
>>60661747
>>60662290
>>60663297
Anyone care to do an edit of this?

File: i9 ebin.png (986KB, 801x1500px) Image search: [Google]

986KB, 801x1500px

>>60665150
Already had one from a few weeks back when the i9 naming scheme was "confirmed"

>>60665150
There's the chipset edit floating around. One for threadripper/i9 chips should be done.