Worlds first 7nm CPU - Rome

Even without a chiller the new Intel 48-core chip completely BTFO Threadripper and EPYC.

>t. InTel-aviv kike

>faggot doesn't know where Rajah is right now.

AMD is in its core an Indian company. Don't try to deny that
Jow Forums stands with Trump and Israel.

Would you look at this juden.

Attached: 1541545242451.jpg (1536x2048, 470K)

Wenn das der Führer wüsste, er würde ryzen aus seinem Grab

The A literally stands for American.

It stands for Advanced you dumb nigger

Citation needed.

>Modern Datacenters require new thinking.
Intel is owned!

Attached: new_CPU_GPU_architecture.jpg (453x604, 56K)

6 trillion jews!

>NEW: more cores doesn't matter!

apologize

Please give me more info on this. I'm a die hard team blue fanatic and need facts to fight back my AMDrone friends. Please. Please.

I heard 7nm cpus release unhealthy radiation so I'll be staying with Intel, thank you.

Haven't Apple done it? Or is this a different kind of 7nm or something?

>new Intel 48-core chip
Sure, but I'll take my extra 30% performance.

Attached: 30_off_cores.jpg (1024x600, 166K)

Cannot wait to take the AMD pill.

Attached: drop down menu.gif (600x338, 2.18M)

>DIE SIZE DOESN'T MATTER

Attached: 1494162098217.png (653x726, 84K)

is this a macOS interface meme? very cool

Google servers are switching to Rome.
Its over.
youtube.com/watch?v=DaUy880vtRM
9:00

BTFO your wallet and electric bill fer sure

Man AMD really kicking goals, good to see!

>Google servers are switching to Rome.
Yikes!

> being this cheap and not springing for the full package

Attached: dual socket option.jpg (1464x926, 342K)

> 14 (nm I/O), 8 * 8 (core)

Attached: 1481045697987.png (349x438, 215K)

Y-y-you DOUBLE HITLER!

Dual socket intel CPUs worth $13k each outperformed by a single EPYC CPU.........

AMD invents everything first. Seriously they have. Other companies keep competing because they focus on gaming performance.

>"S-single core performance d-doesn't matter!!"

>7nm
NOOOOOOO *PFFFFFFFFFSSSSSSCCCHHHHHHTTTTTTTTT*

Attached: 1527629778452.jpg (679x758, 54K)

Attached: baking_xeon.jpg (960x720, 116K)

HEIL HITLER

POO IN THE JOO

Attached: untitled-3.png (682x798, 38K)

Intel always has to copy AMD like a bunch of retarded monkeys.

>those little Chiplets
jesus the yields must be amazing

Probably above 85% tbhh

>mfw

Attached: cutress.png (593x477, 319K)

Amazing? No. 7nm uses incredibly complex multi patterning all the fuck over the place. It isn't mature either.
It is smaller than the couple ARM SoCs being run on it though, so thats something.

Holy fucking guess again kid.

And no socket change.

just repeat his words, do you doubt the logic?

intard NPC motto, just mindlessly repeat everything your intard friends say

3 more years until intlel new uarch, hang in there :)

BASED
A
S
E
D

>14nm
>2x7
>double performance

Can't speak german, but read i can, and this made me giggle

BASED AND REDPILLED

Attached: 1519591123265.jpg (1616x931, 292K)

Well for one Intel is so sure of their performance they're gonna turn off hyper threading just to make it a fair fight. AMD is gonna be really embarrassed when they need a 128 thread processor to beat one with only 48.

Rabbi, I...

Attached: AMD-EPYC-2-Rome-v.-Intel-Xeon-Scalable-Cascade-Lake-SP-November-2018-Disclosures.jpg (1076x316, 100K)

>Die size dosen't matter
>n of cores dosen't matter
>soon clock speeds dosen't matter
What will intel fanboys in 2022 say that dosen't matter?
How many more "Israeli Security Companies" will be funded by intel trying to fuck amd?
Linus said "Fuck you Nvidia"
I say "Fuck you Nvidia, fuck you intel, fuck you fucking micro-piece-of-soft-shit"

kek

this is the first x86 chip.

Just yield

NPC can't yield.

Attached: 1538773645623.png (490x586, 44K)

EPYCflaws.com when?

Not so sure about that just yet, cause there's more things to consider than just that

Adored is right again

It is confirmed.
Meltdown and ForeShadow does not affect AMD ZEN2!

>4.4ghz to beat vanila 8600

how many cores and how good will it be for gaming?

just ran this bench. it says 550 for 8700k single core perf. my 2700x still caned it on multi tho.

Hopefully with 7nm ayymd unfucks their consumer gpus, too.

>Worlds first x86 7nm CPU
Fixed that for you. Kirin 980 and Apples A12 Bionic came first, but they are ARM chips.

>just ran this bench.
Was it a benchmark on paper like that 48core paper release?

cucks need not to apply.
Speak englando or gas you are self.

tod den Anglos.

(*) with SMT off
(**) without all mitigations installed
(***) code compiled with an intel compiler and optimised only for intel

and you get in one benchmark 2x48 = 96 cores beating 2x32 = 64 cores (that's 50% more cores) but only with a 30% increase in performance.

What will you get if you enable SMT, and optimise the code for the EPYC processors?

>What will you get if you enable SMT, and optimise the code for the EPYC processors?
a letter from their (((lawyers))) that you caused holocaust flashbacks

no

Attached: uwotm8.png (2077x1311, 275K)

>"polished" software ahahaha

God it's almost sad seeing these shills try so hard only to be BTFO at every turn

>sad: read absolutely hilarious

Attached: d3f5066f.jpg (630x630, 60K)

NOOOOOOO STOP STOP STOP STOP STOP STOP STOP!!

Attached: 1515221586890.png (682x792, 339K)

Threadly reminder 16 cores on AM4 won't happen.

PCIe lanes remain in the CCX, not in the I/O die, so a 16 core part would have twice as many lanes requiring a socket change.

AMD could fuse the lanes off on one of the CCX dies, but then you have latency issues. Even if they could fuse half the lanes off on each die it still wouldn't be problem free.

The Rome sample demo'd was a beta sample. It's still not even the final EPYC 64 product and it STILL beats Intels best 2s Xeon product by 7.47% in a C-Ray benchmark. All the while consuming less power and no fancy cooling solutions required.
They did not reveal the frequency. The final product will no doubt be faster than the sample demo'd. My guess would conservatively be 10% on average than the Intel system.

Intel announced a 48c 2s product for next year but AMD will be refining EPYC and are already roadmapped to 7nm+ and 5nm not long after. Intel's product will no doubt be faster but at what cost in power usage?

Interesting times ahead.

Oh and of course EPYC will cost substantially less. AMD will make bank on the fact that the chiplet design means EPYC yields! I am guessing that the combination of 14nm chip and 7nm chiplets will bring them to over 85% yields if not higher.

>xeon was great
>so was coal
kek
AMD has gotta be a fantastic place to work

Call me when they remove the PSP botnet

>PCIe lanes remain in the CCX, not in the I/O die, so a 16 core part would have twice as many lanes requiring a socket change.
Are you retarded?

yea...but it loses per core...amd still behind...sad...guess my datacenter will keep using intel...

The desperation in the blue camp right now warms my heart.

Attached: 1528987497987.jpg (960x878, 125K)

Technically they could fit 16 cores in an AM4 socket, I could see them segment the ryzen 7 lineup to have either a 3800x with 16 cores, 3700x with 12 and 3700 with 8.
Alternatively they might have a different name for the 16 core part, imagine a 16c CPU for maybe $600, it'll be like another shoa

>hang in there
On the end of a rope.

Release when? I want to upgrade from my i7-3770k, I want to see whether to get a r7 3700x or something like a i7 9700k.

Are you?

youtu.be/YmPimQp7xLE?t=138

>"So while the memory controllers now reside in the I/O die, PCIe remains directly attached to the CPU cores."

128 lanes in Epyc / 8 = 16 lanes per CCX, same as Zen 1. Having 2 CCX chiplets on AM4 would mean 32 PCIe lanes, and the socket wasn't designed for that, requiring a pinout change and breaking compatibility.

Now they could just fuse the lanes off and make it a compute-only die ala the 2990WX, but it remains to be seen what effect that may have on heavy gaming workloads since the compute die would have to go though the other ccx (and maybe even the I/O die) before accessing the GPU.

I am going to go out ona limb and also assume Portsmash does not work on Zen 2. Mark Papermaster pretty much confirmed they have addressed any side channel attacks.

Meanwhile over at Intel...

This post bought to you by Goldman Sachs.

DELIVERING AS PROMISED

Attached: big_amd-papermaster-zen-roadmap.jpg.ashx.jpg (1170x658, 133K)

>r7 3700x or
>i7 9700k
Choose wisely.

Attached: amd-or-chiller.jpg (960x960, 101K)

It's too close to TR2. They need market segmentation after all. My guess will be a 10 or perhaps 12 core part. Not sure how they would make it but it would definitely fill a gap.

Nobody but gaming babbies choose Intel. real content creators and businesses choose whatever saves them money. Gaymen fritter money away on worthless trash.