My wallet is not ready.
>>54992803
>16384 MB HBM2
>>54992803
non of this is confirmed, but sorry for your wallet anyway
>AMD
am I doing this right?
>>54992803
They'll release it for <$500. AMD know they're the poorfag brand.
>>54992803
9.8gfops wow its fucking nothing, by the times they release this nvidia gonna release 1080ti which will rape that into the ground.
sad but true, dead on realse if they don´t make it cheap as fuck
I'm curious, how will 16GB of HBM help VR? Honest question, I know VR and standard gaming on a monitor have different requirements but will VR really benefit from that much vRAM?
If this is real then it's gonna be expensive as fuck though. Maybe like $800.
>>54992868
Since Nvidia likes to jew customers, they went with the poor vRAM, leaving a massive supply for AMD to use on higher end cards. So, they got to buy in bulk, and say fuck it, more vRAM for our high end cards, because we aren't jews! Also, 1 screen per eye, so if you mod a VR game, you could possibly use over 8gb of vRAM, depending how pretty/heavy textures are.
HOUSEFIRE
>>54992860
The 1080ti would likely be just under 10tflops
And it would be $1200
>>54992961
this is an amd thread not an nvidia thread no housefires to be seen
>>54992949
nvidia put 12gb on their card
> OMG SO USELESS NVIDIA LEL 12GB
AMD put 16gb on their card
> OMG AMD IS SO BASED TOTALY NOT LIKE JEWIDIA
ok champs
>>54992803
>4096bit bus
muh dick !
>>54992868
I read somewhere that since they can only stack more hbm on itself by it's own value, instead of going up by a factor of 2 it needs to go up by a factor of 4
Thus it's either 4gb or 16gb, nothing in between
>>54992995
They are not fucking capable of cooling HBM on the shitty fury and fury X, you believe those imbeciles will be able to cool down HBM2 lol ?
>>54992984
The 1080 is already at 9tflops and thats just with 20SM, a 1080ti with 40SM should double that and lets not go with 60SM which is the full chipset.
>>54993041
Yes
Also they are able to cool it down just fine
>>54992803
>4096 bit bus
o my god. and i thought 512 was alot
>>54993060
> implying a 1080ti would be double
This is shillvidia we're talking about
They want to milk as much as they can out of their new 1080
>>54993039
>tfw you'll see a 128gb GPU in the next 5 years
>>54992803
NVIDIOTS ON SUICIDE WATCH
>>54993083
You are right, without any company to compete against them they would milk the 1080 as much as they can, releasing something like a 1080ti would be only something someone incredible stupid would do as it would cannibalize their own sales.
>>54993041
My Fury never goes above 70C at load. Don't know what you're talking about faggot.
>>54992824
>This
I'm really unsure if this thing can compete 1080Tis, and I'm not willing to spend 1200 € when I can get double the performance with 2x Vega
Also the 1080Ti will only feature 600 GB/S GDDR5X
>>54993092
i thought of this 10 years ago
>>54993092
what the fuck do you even do with 128gb?
>>54993204
>what the fuck do you even do with 128gb?
Ai
>>54993204
this >>54993287
128gb could load large NN
>>54992995
Who ever said anything about that? Nvidiots tried to cry over the 8GB of vRAM on the 290, and 390 series. Beats 3.5 GB of vRAM.
>>54992995
It's useless when the 12gb goes at snails pace
>>54992868
>64x MSAA
>>54993373
If you have a Maxwell Nvidia card you can already do 64x MSAA while using less bandwidth than 8x MSAA
https://research.nvidia.com/sites/default/files/publications/hpg15_DCAA.pdf
The problem isn't bandwidth nor memory, the problem is having the dedicated hardware to do 64 samples per pixel, current GPU only can do 8 samples per pixel.
>>54993373
If you have a Nvidia Maxwell card you can already do 64x MSAA while using less memory and bandwidth than 8x MSAA.
The problem isn't memory neither bandwidth the problem is the lack of dedicated hardware to do 64 samples per pixel, current GPUs can only do 8 samples per pixel at hardware level.
https://research.nvidia.com/sites/default/files/publications/hpg15_DCAA.pdf
>>54993435
>>54993497
Fucking 4chan ate my first post.
>>54992995
Well AMD 16gigs require little to no compression while nvidia 12Gigs has compression up the ass in it.