THANK YOU BASED INTEL

I dont think thats how gpus work

>2019
>still no 12 nm
just buy it goyim

Attached: Intel roadmap Working-on-7nm-and-5-nm-Manufacturing-Technologies-3.jpg (1688x795, 342K)

G-POO-U SUPER POOWER 2020

Attached: Retarded.png (361x334, 87K)

Another Raja failure.

Attached: AMD-GPU-Roadmap-2016.jpg (1360x622, 263K)

No, it doesn't look like one at all. Idiot. And no that's not what it's for. Idiot

Reminder that he put next-gen on the roadmap too before he left. Radeon is dead on arrival. Intel can at least afford engineers

>That aggressive market segmentation

Jesus chirst. When will these scam corps be punished for this anti consumerist bullshit?

Every decision we're witnessing now is the work of at least 5 years. This means that both Polaris and Vega is his doing, and their failure of arrival is too. He's good engineer, he just can't release products.

Imagine being at the 14nm technology node in 2018

moar supersized SIMD units

Am I pissing something or this plot shows the performance/Intel relation?

The xeon segmentation is a thing of beauty to see.

Don't say shit like its fact if you don't know what the hell you're talking about.

this

>He's good engineer
eteknix.com/rx-480-exceeding-pci-express-power-specifications/

Kind of but it never worked good enough at graphics, so it was scrapped for this project

Why does the growth look logarithmic instead of something faster

With their CPIs its a nightmare. No hyperthreading on anything sub i5, then hyperthreading becomes i9 only, then they have a seperate line of i-series cards with high-bandwidth memory and more cores. A K-Series line for both so you can overclock, a xeon line for workstations, a xeon line for entry servers, a xeon line for midrange servers, xeons woth lowcore/highperf xeons with highcore/lowperf, etc etc.

It's no worse than nvidia gpus

Not to mention mayo thermal paste, having to delid, nuclear powered cooling requirements, blowing fuses, funding the military-industrial complex of Israel (toda goyim), 14nm+7 tock/tock/cluck

Raja isn't an engineer you tards... Engineer is a professional (job) title. He was CTO of ATI and "Chief architect and VP of GPU" at AMD after coming back. That means he makes high level decisions on what problems technology should tackle, what markets they'll go for, and do high level management of teams and group strategy.

It's a technical manager position so he may have the skill but it'll mostly be for the sake of making business and product decisions. For example, he would be the one deciding how closely RTG will work with semi-custom partners like MS, Sony, and Intel; how many new dies they'll produce for an architecture; starting the FOSS initiative like drivers and rocm and HIP; what workloads/markets Vega should tackle and (with consulting from his actual engineering team) how to hit those goals (Vega targeted professional visualization and film editing, so they put TBR in vega and produced the Vega Radeon pro SSG). Small stuff like the 480 overdrawing power a little would be like a 10 minute meeting in his daily schedule to yell at some idiots to get it fixed ASAP

That's what I mean my dude. The high level decision was all his. Vega isn't a failure as isn't Polaris, but they were released in a bad state due to his decisions.

High RX 480 power draw was a high-level corporate decision stemming from HW architectural and driver optimization (also, JewWorks industry capture) deficits relative to Maxwell/Pascal.
How hot/efficient a chip is perceived to run is largely down to stock clocks and the fact that power draw grows quadratically or worse near absolute frequency limits. I.e., AMD chose to sell factory overclocks, and it was not some sort of design or fabrication SNAFU.

Rajeet is a sham and a good-for-nothing. AMD did the right thing getting rid of him.

It was going to be, until management walked in close to project completion, announced it was canceled, and then announced it was to be turned into Xeon Phi.

xeon phi isn't a GPU.
Where GPUs are a bunch of "stream kernels" smashed together, a xeon phi is a bunch of Pentium cores smashed together.
The difference is that "stream kernels" are stripped-down versions of CPUs that can only do floating point operations so you can fit more of them in the same space.