WYP
News Guru
According to recent AMD presentation to investors the companies Vega GPUs will launch in the first half of 2017.
Read more on AMD's Vega GPUs coming in 2017.
Read more on AMD's Vega GPUs coming in 2017.
AMD's Vega GPUs will launch in 2017
- Thread starter WYP
- Start date
NeverBackDown
AMD Enthusiast
But which Vega? 10 could be considered the high mainstream whereas 11 would be enthusiast? Could go either way tbh
WYP
News Guru
I really hope that AMD surprises us all with a GPU with a 384-bit memory bus to make an RX 490. I would love to see a competitor to Nvidia's GTX 1070 or perhaps higher.
NeverBackDown
AMD Enthusiast
Well Vega is basically gaurenteed to be a 384bit as anything with more CUs would add more to the bus width. But that alone doesn't mean anything. We would need more IPC gains, more advanced Asynchronous Compute performance, and better tessellation performance to really blow everyone away instead of just simply adding more cores.
But considering GCN up to now has been IP8 based architecture for AMD, Vega is IP9 for them so we should see something amazing
But considering GCN up to now has been IP8 based architecture for AMD, Vega is IP9 for them so we should see something amazing
NeverBackDown
AMD Enthusiast
Probably Vega 11. Otherwise it would be too many Fury's for two different chips. Wouldn't make sense
tolagarf
Member
And once they do that, nvidia will just release optimized versions of their chips and it all starts over again. The game is on.
AlienALX
Well-known member
I heard a rumour that they could be using the Rage name again.
I'm not sure they are playing any games tbh. Raja tweeted a couple of months back to say that he was in Shanghai discussing Vega. IE - talking business about getting it into production. That all takes time sadly, and he did point out that it would not be ready for quite a while.
At that point orders had been seen for custom PCBs. Two custom PCBs. So if you take that as gospel they were still testing two samples about two months ago.
Saying that if they launch both Zen and Vega in January they could do really well. I bet they would sell more units than launching stuff in the summer when hardly any one is around.
Last edited:
Kaapstad
Active member
The problem is NVidia has got the Pascal Titan out already, it is blisteringly quick and worse still for AMD there is a potential GP100 Titan with HBM2 memory lurking that could launch very quickly if needed.
AngryGoldfish
Old N Gold
I wonder when Volta is coming out, whether it's almost two years away yet or whether it's only around the corner.
Kleptobot
Member
Are you sure about that because re engineering a die to work with HBM is not something that can be done quickly... New pcb, new bios, new memory controller if I'm not mistaken. Or did they always plan on making one and have just kept it very hush hush?
Kaapstad
Active member
Pascal Titan uses GP102
NVidia professional cards use GP100 with HBM2
This gives NVidia the option to drop a GP100 Titan.
AlienALX
Well-known member
TBH I am pretty sure that AMD have a more than clear enough picture of what GP100 can do now, what with the Titan XP being out.
AMD do not need to beat the Titan XP. That's a £1100 GPU. All they need to do is release something that will rival the 1080 IMO and bring prices down on GPUs around that area of the market.
The 480 didn't beat anything Nvidia had but seems to be flying off of the shelves.
Kaapstad
Active member
You are right
There is no point in targeting the Titans as even if AMD can beat it they get nothing.
AMD need to target the 1060/70/80 as this is where the money is.
Did you mention RX 480, I have got 4 of them and they are fantastic cards.
AngryGoldfish
Old N Gold
Yeah, 1080 performance is what I'm waiting for from AMD. Titan XP performance is pointless for someone on a 1440p screen playing single player games. I can wait until 2017. It just better be worth the wait.
Thelosouvlakia
Classy-Fried-Chicken
I don't think that amd is interested to hit the titan in the consumer market but in the professional/research/scietific market.
They need a card that does openCL better than what the latest titan can do with CUDA, and that's about it.
Imagine recieving emails from a universities ordering 10 Titans or 10 Firepros at once for their new deep learning workstations... That's a lot of money in one place
They need a card that does openCL better than what the latest titan can do with CUDA, and that's about it.
Imagine recieving emails from a universities ordering 10 Titans or 10 Firepros at once for their new deep learning workstations... That's a lot of money in one place
Similar threads
