For all of the Vega rumors.
AMD Vega Architeture Speculation
- Thread starter Gandalfthewhite
- Start date
bittermann
Well-known member
Yeah, the latest word on the street is 2017. Seems like HBM2 is the hold up this time...AMD can't catch a break to move it up any earlier.
I saw a press release that Samsung started mass production of HBM2 in January 2016. How can HBM2 be the holdup?
[edit]
link:
https://news.samsung.com/global/sam...on-newest-high-bandwidth-memory-hbm-interface
[edit]
link:
https://news.samsung.com/global/sam...on-newest-high-bandwidth-memory-hbm-interface
bittermann
Well-known member
This is what I've found. No dates have been established.
http://arstechnica.com/gadgets/2016...gbs-bandwidth-samsung-begins-hbm2-production/
Faceless Rebel
New member
HBM2 is already available in very small limited quantities. The Tesla P100 uses HBM2 and the only way you can get that currently is the $129,000 Nvidia DGX-1 which is used for deep learning.
The speculation is that the Pascal Titan will be announced in August. We'll know for certain then what kind of shape HBM2 supplies are in.
The speculation is that the Pascal Titan will be announced in August. We'll know for certain then what kind of shape HBM2 supplies are in.
Most likely, Nvidia will launch a titan version of big Pascal first which given it's likely price point ( 1000 to 1500$ is my guess ), makes it an automatic moot point for 99% of users since that's way above the upper limit even by high end enthusiast standards when it comes to price points for a single video card.
Most don't pay that much for their CPU's, and they tend to hang on to them much longer than they do a video card.....
So it only becomes interesting on Nvidia's end once a TI version using big Pascal is released, and it having a more reasonable price tag but that might only happen when AMD is ready with it's high end offering that is meant to compete with big Pascal.....The largest Vega part is rumored to be a 6000 shader part.
Nagorak
Well-known member
With the size of the chips that are coming out, I am sort of wondering if both AMD and Nvidia are shooting their wad really early this generation, and then we're going to see few improvements for like the next 3-4 years.
What concerns me is that we got up to such large dies in the last gen (basically pushing the process to the absolute max), that to outdo those chips they needed to start at a fairly large size this time around.
The 1080's die was about 20% smaller than the 980s. So, let's say that 1080 Ti/Vega are 20% smaller than last gen's chips also, you get one more bump up in size and you're right back at the limits. Then it all comes down to efficiency, but I'm not sure there's that much more efficiency you can wring out of these chips, especially for Nvidia (AMD still has some catching up to do, although in DX12 the story could be different).
What concerns me is that we got up to such large dies in the last gen (basically pushing the process to the absolute max), that to outdo those chips they needed to start at a fairly large size this time around.
The 1080's die was about 20% smaller than the 980s. So, let's say that 1080 Ti/Vega are 20% smaller than last gen's chips also, you get one more bump up in size and you're right back at the limits. Then it all comes down to efficiency, but I'm not sure there's that much more efficiency you can wring out of these chips, especially for Nvidia (AMD still has some catching up to do, although in DX12 the story could be different).
noko
New member
By March of next year there will be many DX 12 titles out and maybe at a point where All AAA games will have DX 12. If AMD arch is built around a parallel design, it sure is, that is where it will shine in those applications.
Also GloFlo should get a lot of experience with the new process and will have time to make plenty of Zen chips coming up. GloFlo may not be able to produce enough chips for Polaris and Vega as well as Zen. For AMD, Zen is super critical and probably would be risking resources at the most critical time for Vega if everything was being piled up. If Zen is successful, it will make much more money then the RTG.
Then I expect a decision will be made if RTG will be sold or not.
Also in the very near future P11 should be making inroads and I am surprise not much has been put forth on that front. Is P11 really a viable gpu to put more GloFlo resources into it and try to take back the mobile market from Nvidia? Or would AMD be wasting on a loss cause against GP108? And taking away from a successful launch of Zen and future APU's?
For RTG, having cards available everywhere at the less then $300 mark is critical. Still no 470 or 460 and absolutely no word on mobile parts. That is not good. Forget about the 1060, I think the 480 will do just fine as long as you can get the cards out. The less then $200 cards AMD should do even better in. What is Nvidia going to do? Have a 3gb 1050 for $199, cut down 1060? The Rx480 would slaughter it.
So maybe more then just the availability of HBM2 coming into play. March of next year does not seem right for Vega but could be the case due to AMD other products particularly Zen. If AMD has plans for Samsung or TSMC is unknown as far as I can tell.
What will hurt AMD severely or critically is if Zen is not successful or a good viable cpu option.
bittermann
Well-known member
Except for Nvidia's specialized card I don't think they are going to get HBM2 consumer cards out any earlier than AMD will be able to. Ought to be interesting times the end of this year or early next.
Hapatingjaky
New member
The Titan Variant will probably use HBM2 whereas the Ti variant will probably be GDDR5X on a 384bit bus. I don't think they will give the Ti HBM2 because it will justify the cost of the Titan and performance difference.
What it will come down to is cost. I can see Nvidia releasing the Ti at the same price as the 1080 is now ( not FE ) and then dropping it's price on the 1080 down to 299/399 area which would decimate AMD'S lineup even more. The Titan will be around 1499/1599 area....
dream on nv has set it up with the 1080 FEe pricing of 699 for the ti to be 999 and titan 1500
I think the point of all this is releasing Vega as soon as possible. That 2x 980 power in 1080 is just too tempting. If I will get a 1080 this summer there's noway Ill be upgrading or sidegrading to a Vega, no matter if its 10 or 15% faster and I think this holds up for lot of people who buy top-end cards.
So timing is everything, 2017 is way too late.
yeah no, you're way off
It all comes down to how much of a pain in the ass will Vega 10 and 11 be in terms of performance and at what price will AMD set them at.......Even if the 2 rumored versions of Vega are competitive with the GTX1080 and GTX1080 TI / Titan but are decently cheaper, then Nvidia has no other choice but to drop prices......It's happened before in the past.
I really hope vega has at least 25% more performance than nvidia and launches at a 500 or less price tag. If AMD pulls another Fury X nonsense then we will never see these prices go down anymore. Nvidia already is raping with their 800+ 1080 parts that are getting sold out in seconds.