Tuesday, March 15th 2016

AMD Unveils GPU Architecture Roadmap, "Polaris" to Skip HBM2 Memory?

Alongside its big Radeon Pro Duo flagship graphics card launch, AMD unveiled its GPU architecture roadmap that looks as far into the future as early-2018. By then, AMD will have launched as many as three new GPU architectures. It begins with the launch of its 4th generation Graphics CoreNext architecture, codenamed "Polaris," in mid-2016. Built on the 14 nm FinFET process, "Polaris" is expected to offer a whopping 2.5x increase in performance-per-Watt for AMD, compared to its current GCN 1.2 architecture on 28 nm.

Hot on Polaris' heels, in early-2017, AMD plans to launch the "Vega" GPU architecture. While this appears to offer a 50% increase in performance-per-Watt over Polaris, its highlight is HBM2 memory. Does this mean that AMD plans to skip HBM2 on Polaris, and stick to GDDR5X? Could AMD be opting for a similar approach to NVIDIA, by launching its performance-segment GPU first as an enthusiast product, giving it a free run on the markets till early-2017, and then launching a Vega-based big-chip with HBM2 memory, taking over as the enthusiast-segment product? Some time in early-2018, AMD will launch the "Navi" architecture, which appears to offer a 2.5x performance-per-Watt lead over Polaris, taking advantage of an even newer memory standard.
Add your own comment

46 Comments on AMD Unveils GPU Architecture Roadmap, "Polaris" to Skip HBM2 Memory?

#1
Chaitanya
Aint HBM2 expected to enter production in 2nd half of this year? Still both Amd and Nvidia cant afford to skip summer for new gpu release.
Posted on Reply
#2
arbiter
Chaitanya said:
Aint HBM2 expected to enter production in 2nd half of this year? Still both Amd and Nvidia cant afford to skip summer for new gpu release.
Gddr5x which is looking like could be using then they could be fine with it.
Posted on Reply
#3
Naito
In before the crushing tide of flamers from the Pascal GGDR5X news article. I wonder how their opinions may change after this speculation?
Posted on Reply
#4
ZoneDymo
Naito said:
In before the crushing tide of flamers from the Pascal GGDR5X news article. I wonder how their opinions may change after this speculation?
not changed in the least, sorely disappointed in both companies if the leave HBM2 on the shelf for longer then needed purely to sell more incremental upgrades because moar money/f technological progression.

Still more disappointed with Nvidia though considering their much larger market share.
Posted on Reply
#5
Ferrum Master
As expected.

It seems simply there wont be enough HBM2 memory by the end of the year. As shippings start only in Q3. And that is too late.

First chips will be like 7970 and 680 repeating history. Like start from the middle end. It smells even like conspiracy...

Question is, how wide will the bus for both of them. If it is 384Bit+ then who cares HBM2 or not, it will be enough.
Posted on Reply
#6
the54thvoid
ZoneDymo said:
not changed in the least, sorely disappointed in both companies if the leave HBM2 on the shelf for longer then needed purely to sell more incremental upgrades because moar money/f technological progression.

Still more disappointed with Nvidia though considering their much larger market share.
Limited quantity of HMB2 and that much larger market share. Use logic please.

As for AMD, presumption is continued use of HBM1? or GDDR5X if they require higher memory usage beyond 4Gb? Either way, HBM's primary benefit is lower power consumption.
Posted on Reply
#7
Semel
Nothing to worry about;.

Even non-HBM memory bandwidth is more than enough.

The main purpose of putting HBM1 on fury cards was to

-reduce power consumption

-reduce the cards size and use this space

-to add more stream processors and shit without making a card huge and too power hungry.

Polaris is gonna be built on 14nm FinFET (vs 28nm fury).
Posted on Reply
#8
Naito
Semel said:
Polaris is gonna be built on 14nm FinFET (vs 28nm fury).
This and Nvidia's 16nm is what people should be looking forward to the most. 28nm has been available for just over 4 years and I'm sure most of us will be glad to see it go.
Posted on Reply
#9
bug
So basically they've said nothing except this generation will forego HBM2.
Posted on Reply
#10
ZoneDymo
the54thvoid said:
Limited quantity of HMB2 and that much larger market share. Use logic please.

As for AMD, presumption is continued use of HBM1? or GDDR5X if they require higher memory usage beyond 4Gb? Either way, HBM's primary benefit is lower power consumption.
Not sure what you are having problems with understanding there.
Nvidia has a much larger market share, aka much more money to spend on development.
They could much easier then AMD just make the jump for all their future cards to HBM2 if they wanted to but nope.avi, rather just have another incremental update then giving consumers the full beans right away because MOAR MONEYZZZ

Company could give us A B and C in one new generation but they rather give us A first, then A and B, then A B and C because again, moar moneyz.
Posted on Reply
#11
ZoneDymo
Semel said:
Nothing to worry about;.

Even non-HBM memory bandwidth is more than enough.

The main purpose of putting HBM1 on fury cards was to

-reduce power consumption

-reduce the cards size and use this space

-to add more stream processors and shit without making a card huge and too power hungry.

Polaris is gonna be built on 14nm FinFET (vs 28nm fury).
and now we wont have any of those benefits....
Posted on Reply
#12
64K
ZoneDymo said:
Not sure what you are having problems with understanding there.
Nvidia has a much larger market share, aka much more money to spend on development.
They could much easier then AMD just make the jump for all their future cards to HBM2 if they wanted to but nope.avi, rather just have another incremental update then giving consumers the full beans right away because MOAR MONEYZZZ

Company could give us A B and C in one new generation but they rather give us A first, then A and B, then A B and C because again, moar moneyz.
Nvidia can't make the jump to HBM2 on all of their cards if there aren't enough chips to go around and it doesn't really make sense to drive up the price of entry level and midrange cards by going with HBM2 for those. Also Nvidia isn't a charity. Just because they are a profitable company doesn't mean they are obligated to take on the burden of pushing HBM2 on every card.

It looks like AMD may be seeing the same thing and reserving HBM2 for their high end card designs later on down the line.
Posted on Reply
#13
PP Mguire
Nvidia can't make the jump because mass production of HBM2 won't start until Q3 which was outlined in a previous news article and GDDR5X will be available this summer when both companies want to push out cards. Not real sure why this is so hard to understand guys. This is also why we will be seeing GP104 with GDDR5X instead of HBM2 instead of big Pascal and I'm willing to bet AMD is doing the same thing.
Posted on Reply
#14
ZoneDymo
64K said:
Nvidia can't make the jump to HBM2 on all of their cards if there aren't enough chips to go around and it doesn't really make sense to drive up the price of entry level and midrange cards by going with HBM2 for those. Also Nvidia isn't a charity. Just because they are a profitable company doesn't mean they are obligated to take on the burden of pushing HBM2 on every card.

It looks like AMD may be seeing the same thing and reserving HBM2 for their high end card designs later on down the line.
"it doesn't really make sense to drive up the price of entry level and midrange cards by going with HBM2 for those"

Who said anything about driving up the price, how about instead of going for moar profits cuz moar moneyz they just make less profit and push tech.
No they are not obligated to do anything, but we as consumers are also not obligated to purchase crappy incremental upgrades.
Id much rather have us all as consumers make a stand and demand better.
Posted on Reply
#15
Semel
ZoneDymo said:
and now we wont have any of those benefits....
They are not "needed" coz it will be on 14nm FinFET (vs 28nm fury).
Posted on Reply
#16
EarthDog
MEH... this really effects so few people (read, those at 4K or maybe 3x 2560x1440), it doens't bother me in the least that either company will skip it on their new line of cards in the mid-range segment.

ZoneDymo said:
and now we wont have any of those benefits....
Which are........................... hooking up a fire hose to a house spicket and expecting firehose type performance? Doesn't work that way. Look how good the Fury products are at 4K..... then go down in resolution (where 99.9% of people are).
Posted on Reply
#17
PP Mguire
EarthDog said:
MEH... this really effects so few people (read, those at 4K or maybe 3x 2560x1440), it doens't bother me in the least that either company will skip it on their new line of cards in the mid-range segment.

Which are........................... hooking up a fire hose to a house spicket and expecting firehose type performance? Doesn't work that way. Look how good the Fury products are at 4K..... then go down in resolution (where 99.9% of people are).
Even at 4k we're still bottlenecked at GPU power. We might be getting closer to memory limitations of GDDR5 but GDDR5x will widen the gap greatly and HBM2 will just make that even further.
Posted on Reply
#18
wiak
well 28nm GPUs says is not using HBM2 yet the Fiji does..
Posted on Reply
#19
bug
Semel said:
They are not "needed" coz it will be on 14nm FinFET (vs 28nm fury).
What do you mean? Already Fury (X) has shown somewhat better scalability when moving towards 4k. They were held back by the 28nm process, but at 14/16nm I expect cards will be better equipped to use the added bandwidth. What remains to be determined is whether they'll be able to use the whole bandwidth HBM2 offers or not. If not, GDDR5X may be good enough.
Posted on Reply
#20
Fx
I could've swore we hashed this HBM2 business out yesterday... :rolleyes:
Posted on Reply
#21
the54thvoid
wiak said:
well 28nm GPUs says is not using HBM2 yet the Fiji does..
Fiji uses HBM1
Posted on Reply
#22
arbiter
ZoneDymo said:
not changed in the least, sorely disappointed in both companies if the leave HBM2 on the shelf for longer then needed purely to sell more incremental upgrades because moar money/f technological progression.

Still more disappointed with Nvidia though considering their much larger market share.
GDDR5x may be slower then HBM but it can be up to 2x faster then what GDDR5 is. Per chip GDDR5 runs about 28GB/s, GDDR5x push's that number up to 56GB/s. So pretty much can run up to 2x faster then gddr5. Maybe not as fast as HBM but Just having a ton of bandwidth isn't always a benefit if the gpu can't use it completely. its not real good to compare with but my laptop has a gtx 960m with 4gb dedicated gddr5 memory, It is only limited to around 80GB/s. Overclocking memory only barely even bumps the score in testing. So just cause you have stupid amount of memory bandwidth. GDDR5x still could be enough to feed the card data fast enough to let it run at its best. You will see if that is cast when do memory overclock's on them and how much it helps if any.
Posted on Reply
#23
iO
So finally a replacement for the ancient rebrands but no new Fury till next year..
Posted on Reply
#24
Chaitanya
arbiter said:
Gddr5x which is looking like could be using then they could be fine with it.
Lets not forget due to new process node, both AMD and Nvidia might have too many chips in hand that don't make cut for Top end GPU. And majority of people will be willing to spend anywhere between 200-400$ on new GPU which might mean they can make good profits on these chips for now.
Posted on Reply
#25
Stefan Payne
1. There is only one Polaris chip on the roadmap.
AMD statet that there will be two chips. So nothing changed, just some inaccuracy of the roadmap...
So nothing new here...


2. nVidia can NOT just go for HBM, because HBM is a completely different beast than every other memory technology so far!
The manufacturing process is stil pretty damn complicated and for low long has AMD worked on HBM?! Do you really think that they just let nVdia use this technology?!
Besides:
AMD has already an HBM chip out and knows of the downsides - nVidia has not!

So isn't it safe to assume that nVidia may need some more R&D sessions than AMD because they don't know shit about HBM?!

Just having the spec is worthless this time!
Because how do you mount the memory onto the interposer!

You need some experimenting for that...

And also you need the machienery for that - and you can bet that AMD has exclusive concracts with everyone able to place the chips onto the interposer - especially since AMD is one of the HBM pioneers!!!
And nVidia is just again the parasite that comes around when everything is done - like they did with GDDR4 and GDDR5...


Buttom line:
There's a Polaris chip missing on the roadmap, the one that will have HBM.
And HBM insn't as easy to implement this time because it's completely different.

And have you any infos that AMD will use GDDR-5X?! I haven't herad anything about it...
And why should AMD use GDDR-5X?! There's no need for that - with the smaller chip they don't need it, that's something for notebooks and something like that. And with the bigger one they'll use HBM...
Posted on Reply
Add your own comment