This is a joke right? Is this supposed to be funny and actually fool people? Since when would Intel actually manufacture AyMD gpus and GloFo would just sit back and let it happen?
>50 Tflops
>HBM4
>wccftech
>2.4GHz
wccftech.com
This is a joke right? Is this supposed to be funny and actually fool people...
Brandon Martin
Brayden Barnes
>wccftech.com
stop linking to this trash website.
Jason Ramirez
It was expected.
Nvida has no competition so someone had to join forces.
AMD needs money and Intel need half capable GPU division.
John Reyes
>Usman Pirzada
why is it that pajeets always hate nvidia so much? is it because they're drivers aren't made in java and python?
Camden Parker
how does this article differ from their other regular articles?
also, don't link pajeettech.com here.
Levi Ward
the date
todays date.
this day is know for something
i wonder what tho...
Wyatt Martin
that's why it was quoted in the OP, also the issue is the info
the first world is still in March, but I know what you mean
Brandon Baker
easter
Nolan Foster
>why is it that pajeets always hate nvidia so much?
let's see
>whole site is an intel/novidia shill place
>the pajeet copy/pasta writers own nothing else but intel/novidia
>desperate attempt at pajeet tier april fool joke
>why pajeets hate novidia so much.
Luis Bell
that's one of reasons behind this post, is something like that even believable?
Also can't AMD sue for something like this? I'm pretty sure that goes beyond fair use of a copyrighted logo
Alexander Ortiz
Europe is already in April. It's all that matters.
Kayden Rogers
Its April first in Europe.
John Lee
>350W
350W
>350W
350W
>350W
Cameron Parker
>This is a joke right?
>HBM FOUR
>RELEASE DATE H2 2018
Wikipedia:
HBM 4
For the future of Exascale high-performance computers HPE, predicts on the OPGHC HBM3+ and HBM4 to be released 2022 till 2024. More stacking and higher capacity should bring a lot of more memory per socket and higher speed. HBM 3+ is planned with 6 TB/s and 384 GB per socket. (Comparison to actual (2018) high-end AMDs EPYC, which has 150 GB/s and 2000 GB RAM per socket.)[17]
What kind of retarded nigger do you have to be to not realize that it sounds weird that a HBM4 product would come out when HBM3 products dont exist and HBM2 is still rare and very expensive?
delete this shitpost
Charles Collins
Wccftech had good 1 April jokes
Hudson Williams
>fooled by a wccftech april's fool
op is a fag, as usual
Logan Fisher
>only 3x 8-pin
weak ass April fools attempt
Andrew Perez
>Jews and pajeets are teaming up to fuck the Taiwanese.
>Golden Shit GPU or FemBoy GPU...
Well fuck
Jaxon Lewis
For 50 Tflops.
1080 ti has ~11 Tflops
Vega 64 has ~13 Tflops (~25 Tflops for half precision)
So This new one will probably be at ~25 Tflops.
Still might be twice as powerful as Vega64.
Titan V reaches ~15 Tflops with boost (12 with base).
Adrian Perez
can't delete it anymore, but thanks for your googling
Jack Bell
>ITT: Jow Forums gets trolled by Pajeet's April Fools joke
Jace Morgan
>wccftech.com
>50 Tflops
>HBM4
>wccftech
>2.4GHz
>350watt TDP
First off all my speculation is meaningless due to this being from the curry nigger site
but ill speculate anyway
>50 tflops
so double the raw compute of vega. You can do this by simply making an 8,000 shader gpu with similar specs however it would be 700watts tdp
to cram 50 tflops into the same 350 watt envelope you either have to massively lower clocks since power consumption is on a exponentially increasing curve OR you need a new process node.
HBM4 memory combined with a raw base clock of 2,400 mhz in a 4,096 shader GPU could theoretically close in on that 50 tflops mark since faster memory would increase the tflops output.
but to achieve this you need a small a crazy advanced process. Im very skeptical that 7nm is this good performance wise.
UNLESS!!! You go with an MCM design.
IF AMD does a 1,000-1,500 shader die with a complete redesign on the piepline to boost the maximum clocks on a 7nm process node to get that 2.4ghz core clock and then stitched those dies together to get a 4,000-6,000 shader gpu with HBM4 then we can see this happening.
as a single die? fuck no. An MCM design on 7nm? maybe. Im guessing Intel is paying for the R&D in exchange for profit share and licensing rights.
Carson Hill
>wccftech.com
>HBM4
We aren't even on HBM3 wtf.
Sounds like something aimed for like 2022.
Oh right it's April 1st.