Seriously don't give a fuck about the new amd cpu, when

also pic related, this is what amd needs to take over nvidia. and price should be ~$800

>>57997822
This, poorfag Nvidiots can't afford it then.

>>57997822
>Dual GPU card
>taking over anything
'no'

>AMD can't compete with a single die

jej

>>57997796
>>57997822
>>57998057
Dual GPU card... J U S T

Besides it was stuttering on Battlefront @ 4k on yesterdays stream....

>>57998093
That wasn't a pro duo or any dual GPU card and it wasn't stuttering either. Don't shitpost

>>57998057
>buys AMD
>Calls nvidia buyers poorfags

Holy shit you are one retarded cunt. Also dual cards? so they cant keep up just using one?

>>57998072
>>57998079
AMD is actually trying to get Lego block designs working properly, starting with MCM Zen for enterprise, then Zen+Vega/HBM MCMs for HPC in particular.

For a while Navi was described as having "scalability", so they might actually be trying to get multi-die GPUs to work properly.

It could certainly be a lot cheaper to just make and validate one or two smaller chips then chain them together arbitrarily rather than just make big chips at low yields and low sales volumes.

>>57998079
Multi-die seems to be a great fit for most of the applications GPUs are used for, to be honest. If it decreases cost while increasing performance I'm all for it.

>>57998133
Dont misunderstand me, i want amd to succeed in this, so the competition can actually do something other than releasing crap year after year.

And if you did not see stutters, then you need a new pair of eyes or just kys.

>>57998365
it's trickier than you might at first suspect.

A general renderer structure today might go:
> G-/Z-buffer generation, shadow map updating, sceen-space ambient occlusion, scene shading, then post-processing.
The G/Z/shadow buffer generation uses mostly fixed function units and should be done in parallel as much as possible with the other ALU-centric stages, which means trying to overlap the post-processing at the end of one frame with the g/z-buf generation at the beginning of the next one.

The biggest challenge for multi-die is probably trying to figure out how to effectively split the G/Z-buffer generation across GPUs (splitting single-scene geometry processing is a fundamentally hard problem) and then getting the necessary pieces of the rather huge G-buffer across the interconnect ASAP for SSAO processing or whatever.

>>57998133
Im not anon but it was stuttering.
They mentioned the framerates too and claimed they were on the screen but by chance they were not visible to us, what a suprise.
Another dud from AMD

>>57998601
>>57998998
I saw dropped frames immediately after the game loaded in, which is completely typical of most systems. A few seconds into game play, it looked fine

>>57999213
Except when tyrones bitch was changing direction of the fighter...

>>57999308
you mean AMD's Red Team left titty cam?

>>57999323
when
>she
was
>playing
the game for a moment and actually moved the fighter in other direction other than just following a straight line.

>>57997822
dual gpus suck majorly. Vegas 590 needs to be better than the 1080 and when nvidia drops the ti model, amd should answer with a fury x II.

>>57999213
what was the presentation streamed at? 30fps?

>>57999820
vega is 400 series I'm pretty sure

it will be sold in tandem with polaris

File: AMD-VEGA-CUBE.jpg (104KB, 800x698px) Image search: [Google]

104KB, 800x698px

raja on the vega tech review told the jurnalist that this cube will house 4 vega mi25 (100tflops ) and will use 150 watts all together..
amd literally is using borg tech

>>57999343
come on to pick on renee she needs a hug :(

>>57999922
more like whipping and a chain

>>57997796
march most likely, optimistically february

all chips will go to enterprise first, we get scraps

>>57999906
> Pajeet said 4 vega mi25 (100tflops) will use 150W

Kek, 1500W would be far more likely to be true for 50 TFL(32)OPS/100 TFL(16)OPS.