Discussion: Polaris, AMD's 4th Gen GCN Architecture

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
The thing is Roy Taylor himself mention about pascal being high end while amd targeting polaris towards mainstream user. And before VCZ post that 3D mark score for polaris on previous article they mentioned amd ask board partner not to show polaris on computex. Not even the prototype board. And Vega launching in october still considered as rumor at this point. Looking at the spec and theorical performance nvidia put for GP100 AMD probably did not expect nvidia was very aggressive with GP104. An overclocked GP104 could be very well much more faster than GP100 in FP32 performance. So they push Vega into this year to compete with GP104. Vega could end up quite faster than GP104 but nvidia big gun already waiting on the wing. They already on volume production for their 610mm2 chip so it is expected nvidia have does not habe much trouble with their big chip. Though right now i'm incline to believe that we will not going to see GP100 but GP102 instead for 1080ti (or whatever it's name is)
 
One issue is the naming conventions that get confusing. The FuryX is the enthusiast level card and the 390x is performance level. Yet I've seen reports of Polaris 10, potentially a 490x, hitting GTX 980 Ti levels and yet being called a mainstream card. So is Polaris 10 mainstream or performance segment? Does it include a replacement up to and including the 390x, or are we looking at something truly more mainstream?
http://www.guru3d.com/news-story/amd-polaris-10-gpu-to-offer-near-980-ti-performance-for-299-usd.html
 


If that does go mainstream, wow. AMD will have hit the jackpot! :)

I think it will go mainstream, finfet and lower nm definantly makes this a huge possibility.

So, if polaris 10 is that powerful...who know's how fast polaris 11 is going to be. :ouch:
 
Polaris 11 is the less powerful chip, mostly slated for the laptop market. According to the article, that Polaris 10 performing "near GTX 980 Ti levels" should be the 490x and will probably represent the top of the line until Vega is released. If so, that means the GTX 1080 may be sitting atop the performance crown for nearly half a year.
 

Polaris 10 performs near 980ti in Hitman dx12 benchmark in 4K. I don't think its data enough for generalizing polaris 10 performance. So many articles on net is just hypothesis. 1080 will definitely have the crown performance for quite some time. That's why Nvidia is asking a premium.
 
In case of hitman the game engine itself are more likely favored GCN architecture to begin with. Hence the strong performamce for radeon evencin DX11 vs nvidia card. DX12 barely lift the performance in case of hitman. Worse they could have less FPS than DX11 (even on radeon) version and have far more stabilities issue to boot.
 


Duu. Forgot about that. :)
 


there is no numbers just that back when AMD show prototype of polaris running hitman under DX12 with the FPS most likely locked to 60FPS. but there is some conflicting info about it. some said they run the game on ultra setting and some said they were not.
 


This is indeed THE big question that we need to find an answer for.
 
folks are clearly hung up on the name like there is some rule that says a certain name has to be a certain product. they change names and series descriptions all the time even if it's the same old gpu. so i put little value into the numbers they use and look to see how they perform. the article i linked to said a card would be named the 480 with the polaris gpu. how it performs or what the name means to others i beyond my concern. only linked the info :)

same with the new nvidia cards. i never assumed the 1080 was a DIRECT replacement for the 980. only that it was a new card they chose to name the 1080. now that i see it's performance it is placed in my mental hierarchy list of what card can do what. the name is meaningless. they could just have easily called it the New GPU 1 and followed it up with the New GPU 2 and so on. the name is meaningless, the performance is what is important. don't let the naming of it distract you from what is actually important with the new releases
 
MERGED QUESTION
Question from hypervyper : "Price of AMD polaris GPUs?"





then I thought I was reading AMD is not going after the top end card range there just going to do low to mid range cards only ?? as said above time will tell and unknown

maybe ''next year'' that' seems like all amd got going for it anymore . funny how next year ends up next year when next year runs out ??

https://www.yahoo.com/news/amd-polaris-graphics-cards-targeting-175629413.html
 
yah that's what i have read as well. around $300 for 980ti performance and the 480 (supposedly) benchmarks above are rigth about that performance range. that is pretty much what the 1070 seems to be headed for as well at $479 so nvidia gonna have to do some adjusting if it all ends up true.

for the consumer i sure hope it all ends up true cause $300 for 980ti performance is a game changer for sure. amd did say they wanted to bring VR to the masses and make it much cheaper to get into and this would fit that bill dead on.
 


If those benches are true and the 480X is just under the Fury, that's pretty decent for a 'mainstream' card. I'm looking forward to the next few months when we'll hopefully see a 490/490X.
 


And this is where I have to disagree with you. I know you have a mental hierarchy because you know more about videocards so you can place them based on specs. However, there are loads of people who know little more than the naming conventions and (rightly) presume dat a 490 is the successor of the 390 and they will cost rougly the same for roughly the same or a little better performance. If its rebrand they probably wont know, because you have to be into the chip names for that.

And that is why, for marketing reasons, its unwise to change names all the time - and thats why they DON'T (where did you get that? Some minor changes are there but AMDs line-up shows a clear logic for years now) So its all very nice to talk about performance instead of names when you're among well informed people, but a lot of people are simply not.

Oh, almost forgot an important point. Naming is tied to price, not only performance. So if AMD goes for 480 with the most powerful Polaris 10 part they can't ask 500 dollars for it. Its not what people expect, and its not what they are going to pay (unless it of course performs like a 500 dollar part, but as i said, it would cause unclarity). Therefore, its a nice sign if AMD does go with 480 since then we have a better chance of guessing the price before they actually put the tag on themselves.
 
can understand what you mean but i see little overall consistency in how they name cards. each series may follow a pattern but the performance over the years did not follow suit.

even these new amd cards would not follow your logic. a 380 is about a gtx 960 in performance and $200 but a 480 would be around a 980ti in performance if these benchmarks are to be believed and cost around $300. i don't see the consistency in that at all. sure the name 480 will bring to mind the 280/380 but the performance is not even in the same league. people are going to have to learn the card's abilities anyway to know if the price is worth it just like they always have.

a new 1080 is $250 more than a 980 is and 60% better performing. how is that somehow following a pattern with the naming. the 80 part says where it falls in the overall nvidia line-up (we expect a 1080ti and a 1070 and 1060 and maybe a 1050) and we'd know which is better between them by the number but it has no relation to the current 50/60/70/80..... cards in what it can do.

that's my point with what i said. the name tells me where it falls in the current series if they keep it going but does little to let me know how it compares to other series and generations. that is something you have to just know or know where to look for such rankings. fortunately i do know where to look if i am unsure and it is here http://www.tomshardware.com/reviews/gpu-hierarchy,4388.html not perfect but let's me know roughly where one card fits compared to another. 😀
 


For VR specifically though, AMD would have to come up with similar VR-specific performance improvements to what Nvidia achieved with Pascal. It's entirely possible they've made some gains in this area, but I can't recall hearing anything about it. If they haven't done anything, Nvidia will kinda rule VR for the time being.
 
@mpc007

I disagree that if the fastest polaris is named as 480 then AMD cannot priced it at $500. Regardless of naming amd will priced card at what ever price they want (if they come out first) or according to competition. Take this simple example that taken from AMD very own history of pricing. the MSRP for 6970 is $390. by your logic AMD can't price 7970 significantly higher than that because 6970 was set at 390. then at what price 7970 debut at?
 
Just, wow...
http://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility#.V0hTH5ErKUk

Second paragraph explains a lot imo:
"Let’s start with where we are currently. Full disclosure - HardOCP was not invited to this weekend’s launch in Macau as AMD PR has made a decision to no longer brief this site with the rest of the industry. That’s well within AMD’s rights to do, but that is telling as well."

Then it goes on the explain how 'industry sources' have told them how AMD are going to fail, how Raja is leading a rebellion against Lisa Su and is trying to sell RTG off to Intel, and how AMD are going to utterly fail because Polaris 10 isn't going to be as fast as the recently announced GTX 1080. Some seriously good tin foil hat work there imo.

If the info we have on Polaris is legit (along with AMD's own comments about it) then it is going to be slower than the 1080, with no small part of that stemming from the fact it's a *much smaller gpu* (approx 25 - 30% less transistors / die area by recent estimates). He then concludes how it's a 'polished turd' and they can 'only compete on price'. I guess it's just me that thinks selling a smaller gpu for less money than a larger one is kinda the point?

Update: Also for those questioning if it's possible to sell well without the 'GPU crown'.... the old 48XX series was one of AMD / ATi's most successful product ranges. The interesting thing is back then nVidia held a significant performance lead with their top end part, however the 4850 and 4870 were still pretty fast and priced really keenly and sold well. I don't see why AMD can't do something similar with Polaris if it's fairly small (so cheap to produce) and performance competitive within it's class compared to NV. I mean Pascal when all said and done is 30% faster than Maxwell on average, it's a pretty standard performance jump when considering a new node.
 
this is exactly what amd has stated they are gonna do right now. they know they are not winning at the top end and have said they are focused on the mid and lower cards right now. why is it getting lost in the mix STILL??

they do not have to compete with the 1080 at all if they don't/can't want to. the mid range cards are what makes the money for both amd and nvidia. they only have to offer great performance at the lower prices to make an impact. the focus on the top end is misguided and absolutely neglects the reality and history of the gpu market.
 
While the discrete GPU market is declining in terms of volume, the enthusiast portion of the market is thriving. In 2015, shipments of enthusiast graphics cards doubled to 5.9 million units. Much of this increase, I suspect, was due to NVIDIA's GTX 970, a graphics card launched in late 2014 for $329 that has been a massive success for the company. Despite its high price, the GTX 970 is by far the most popular discrete graphics card in use on the Steam PC gaming platform.

AMD's plan to focus on the mainstream GPU market comes at a time when the market is clearly shifting toward the high end. NVIDIA was able to pull PC gamers to higher price points in 2015 with the GTX 970, and the company is looking to do it again with the GTX 1070 this year. If all of AMD's Polaris graphics cards are priced below the GTX 1070, the company will need to offer compelling performance in order to prevent the same thing from happening this time around.

AMD's bet on mainstream graphics could backfire if Polaris falls short of expectations.
http://www.fool.com/investing/general/2016/05/18/amds-bet-on-mainstream-graphics-could-backfire.aspx
Click to expand...
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom