AMD Polaris Annoucment

Be cool if they show something other than PowerPoint slides for the announcement.
 
they did this with Fiji, and look how it turned out :D, seriously just slides, the GPU won't be ready for another 2Q's though, so the info is nice, hopefully the full deck will have something to chew on other than what we kinda already expected in those slides.
 
Slides leak again before NDA was supposed to expire. People will do anything for page hits these days
 
Slides leak again before NDA was supposed to expire. People will do anything for page hits these days

If they weren't under NDA, then you can't blame them. Blame whoever leaked the slides to them. After all people reading your site (page hits) is what it's all about. And the way you do that is by getting news out first.
 
AMD is just trying some things on the PR front :)
The news itself is ok , still would have to wait, then again if you are looking for instant gratification on AMD side of things that won't happen :)
 
Heh I immediately thought of "woodscrews" as well

Mockups are fine, but that Fermi mockup was just, wtf was Jen-Hsun thinking?

Just had too (sorry not trying to de-rail)
post-5165-0-15709700-1372193887.jpg
 
AMD is just trying some things on the PR front :)
The news itself is ok , still would have to wait, then again if you are looking for instant gratification on AMD side of things that won't happen :)

I think you mean marketing, PR and marketing are two different departments ;)
 
Hehe, nvidia will never get to rid themselves of the fermi-woodscrew-fail, lol. Especially Charlie, man, he will never pass up a chance to bring that situation up, lol.
 
Okay, I saw "console caliber performance" and pretty much lost interest.
 
Okay, I saw "console caliber performance" and pretty much lost interest.
Why make a worthless post like this?

The full quote was "Console quality performance in a thin and light notebook".
 
Why make a worthless post like this?

The full quote was "Console quality performance in a thin and light notebook".

It's similar to some of the comments on the article over at AnandTech; one person said, "The GTX950 doesn't consume 140w"; when the article clearly stated that system power consumption was being measured.

I still wonder why people only skim articles and react instead of actually reading the entire thing.
 
Why make a worthless post like this?

The full quote was "Console quality performance in a thin and light notebook".


Ironically they forgot about the 970m thin and light notebooks out there lol
 
I, for one, look forward to seeing what Polaris can deliver at the top end. It'd be nice to have a viable product for Freesync ultrawides so I can avoid the Gsync tax when it comes time to make that move.
 
Well, they acknowledge it's Q2-16 launch so that's good. Perf/watt improvements look great, so if the high end card is a 250-300W it should deliver a massive improvement from current gen.
 
Yeah I highly doubt it will get pushed back that far. AMD really needs to beat nVidia to the punch here, like they did on 28nm. If they release after nVidia .. well then a lot of people will have already upgraded and well ... it won't be so good for AMD. I am sure they are aware of this.
 
On the same token they love to undercut nvidia as little as possible (if at all). You can bet your ass if the next titan retails for $1100 then they'll come out at $950 instead of $650 like they did last time;) Dam this should be a bad ass year for new cards and hopefully cpu's as well
 
Hrmmm...AMD in the $1000 price bracket ... I wouldn't hold your breath, but it may happen.
 
I hope it will perform better than the 290s. Not getting the performance they boasted. Good, just not crazy
 
I hope it will perform better than the 290s. Not getting the performance they boasted. Good, just not crazy

?

The only takeaway worth getting, specs wise, from that "performance test" is that AMD is (at least to a level of adaptive v-sync) doing a pretty darn good job pushing the power envelope down. Reading any more into that demo is a fool's errand. That, unto itself, is great news (but largely expected with the node shrink, although the magnitude will be interesting to see).
 
I'm looking forward to the possibility of an R9 Nano successor that can push 4K at 60 fps. Hopefully it'll motivate NVidia to come out with a high end itx competitor (I know there's the 970 mini, but the R9 Nano is way ahead).
AMD used to be a leader in power consumption, good to see them getting back to that.
 
I hope it will perform better than the 290s. Not getting the performance they boasted. Good, just not crazy
Well I mean the 290s were released what, over 2 years ago? What were you expecting? I definitely got the power they promised when I bought mine at release.
 
they did this with Fiji, and look how it turned out :D, seriously just slides, the GPU won't be ready for another 2Q's though, so the info is nice, hopefully the full deck will have something to chew on other than what we kinda already expected in those slides.

Be cool if they show something other than PowerPoint slides for the announcement.

They did show something else. They demoed a functioning Polaris video card.

AnandTech said:
For their brief demonstration, RTG set up a pair of otherwise identical Core i7 systems running Star Wars Battlefront. The first system contained an early engineering sample Polaris card, while the other system had a GeForce GTX 950 installed (specific model unknown)

http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture
 
Lol.... they compared next gen hardware to a $150 old tech card? I wonder which was faster.
 
Lol.... they compared next gen hardware to a $150 old tech card? I wonder which was faster.

The purpose of the demo was twofold. First, it proved that AMD already has functioning 14nm GPUs. Secondly, it showcased the power efficiency of the new 14nm process and AMD's new Polaris video cards:

In the live press demonstration we saw the Polaris system average 88.1W while the GTX 950 system averaged 150W. Meanwhile in RTG’s own official lab tests (and used in the slide above) they measured 86W and 140W respectively.
Both systems were running at 1080p Medium settings – about right for a GTX 950 on the X-Wing map RTG used – and generally hitting the 60fps V-sync limit.

Given the fact that power efficiency has suddenly become the most important metric for purchasing a video card, I expect a lot of current Nvidia owners to jump ship to AMD once Polaris cards hit the store shelves. :D
 
They did show something else. They demoed a functioning Polaris video card.



http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture


I'm not going to get into this with you, but did that show us anyting about the Polaris architecture?

Nope, I will say this though, at least they won't be 9 months late on getting something out ......

I was talking about performance, not about anything else......


By capping the frame rates, they might was well not showed anything this, it was pointless. I'm just going to copy and paste my own post from Rage3d.

Power targets and frequency go hand in hand as does architectural design. Lets take the g80 and ilk for example since that has a wide variety of clock frequences *more so* than previous gen GPU's and GPU's after, its longer pipelines allowed for higher frequency's in the shader array, but the rest of the chip (outside of the memory bus), wasn't able to clock as high. This was a design choice nV made.

Also going to finfet and 3d transistors in general take up more space than a regular transistor. So in fact even going to 14nm doesn't really give the increase of 4 fold for transistors as one would normally expect, jumping from 28nm, since the metal layers being similar to 20nm lithography you still get the end result of similar transistor amounts to 20nm without finfet, since the transistor density is less with finfet, this too adds the size of the silicon.

What the Polaris demo showed me was, great they have a chip up and going, abit its a low end chip doesn't really matter, they will be getting something out in q2 of this year, that is a great start, they aren't going to be late to the party like last gen. It didn't show me anything other than that. We can presume the small Polaris is using around 50watts total minus the system. Which puts it at around half the power usage of the 950ti with capped frames, but this doesn't give us power usage of when its uncapped, AMD's frame targeting power reduction is close to 25%

If uncapped at those settings on that game, the 950 ti averages around 90fps as shown by reviewer benchmarks at those settings for that specific level, not very scientific but we have nothing else to go by. So if nV does no architectual changes to their 950 ti and just down sizes it on 14 or 16 nm finfet they will get to just 50% of the 50-70% power savings at the same performance level as per what Fabs have said about their process and it is reiterated by AMD with this annoucement. This is conjecture but 950ti on 16nm should hit that 50watts usage fairly easily. Of course we don't even know if this Polaris chip is low low end or mid low end or high low end. So it might not even be in the 950 ti bracket and could be lower which is good for performance not so for performance per watt. Also have to factor in there is an optimal range of performance and watts used. The 950 ti is not the most performance per watt effecient out of the 9x0 line, I think that is held by the gtx 970 or 980, even the 750ti seemed to be better too.

Then we have to factor in that this game does well on AMD architecture where it seems to have a sizable advantage of 15-20%.

Am I reading into it too much. yeah most likely, but given all these things, its a little worrisome.

So I don't see this card having any extra power efficiency over a 950ti which is a generation older and is on 28nm!.
 
So what if they didn't reveal the innermost secrets of the Polaris architecture? Nobody expected them to. As I stated above, the Polaris card they demoed was simply to prove that AMD has working 14nm silicon and that it resulted in greatly reduced power consumption vs 28nm.

If AMD had simply released a press statement claiming to have produced a working Polaris based video card without showing it, most people here would have scoffed at them. In fact, some clueless people still assumed that's what happened!

Be cool if they show something other than PowerPoint slides for the announcement.
"Demonstrates" does not mean "talks about".
This was my first thought too when reading this AMD PR release was wondering when the definition of "demonstrate" changed. Think I'll go "demonstrate" i.e. talk about how I might at some poing go about the job, to my wife how I've fixed the leak in the sink -- that should go over well :D
In my dictionary demonstrates means showing a working example, and not speaking about what they hope it would be.

:rolleyes:

GPU makers typically do not take a brand new process and attempt to use it with their flagship designs. Instead, they use a "pipe cleaner" lower end card to work out the kinks before starting production of more complex cards. The engineering sample Polaris card is undoubtedly a pipe cleaner and thus it is far too early to reliably infer production performance levels, which is why none were revealed. Final silicon may differ from what was shown at CES. Drivers will undoubtedly improve between now and release as well.
 
Last edited:
So what if they didn't reveal the innermost secrets of the Polaris architecture? Nobody expected them to. As I stated above, the Polaris card they demoed was simply to prove that AMD has working 14nm silicon and that it resulted in greatly reduced power consumption vs 28nm.

If AMD had simply released a press statement claiming to have produced a working Polaris based video card without showing it, most people here would have scoffed at them. In fact, some clueless people still assumed that's what happened!






:rolleyes:

GPU makers typically do not take a brand new process and attempt to use it with their flagship designs. Instead, they use a "pipe cleaner" lower end card to work out the kinks before starting production of more complex cards. The engineering sample Polaris card is undoubtedly a pipe cleaner and thus it is far too early to reliably infer production performance levels, which is why none were revealed. Final silicon may differ from what was shown at CES. Drivers will undoubtedly improve between now and release as well.

People just like bashing radeon tech/amd for no reason + trolling its no use trying to explain something here just read and move on.
 
So what if they didn't reveal the innermost secrets of the Polaris architecture? Nobody expected them to. As I stated above, the Polaris card they demoed was simply to prove that AMD has working 14nm silicon and that it resulted in greatly reduced power consumption vs 28nm.

If AMD had simply released a press statement claiming to have produced a working Polaris based video card without showing it, most people here would have scoffed at them. In fact, some clueless people still assumed that's what happened!






:rolleyes:

GPU makers typically do not take a brand new process and attempt to use it with their flagship designs. Instead, they use a "pipe cleaner" lower end card to work out the kinks before starting production of more complex cards. The engineering sample Polaris card is undoubtedly a pipe cleaner and thus it is far too early to reliably infer production performance levels, which is why none were revealed. Final silicon may differ from what was shown at CES. Drivers will undoubtedly improve between now and release as well.


how old are you? You don't remember prior to the fx line form nV, did they have major issues with big GPU transitions to smaller newer nodes?

Creating smaller chips on a new smaller node, has very little do with "pipe cleaning" as you so aptly put. It has to do with smaller chips having better yields. During the manufacturing and growing of silicon crystals, errors tend to propagate at the edge of the wafer, so as you start producing larger chips, you get more chances of errors because there is more area per chip and of course more errors as you get closer to the outer circumference of the wafer per chip. Thus the chances of a fully functional chip is less. Now they will have to make changes in design, in libraries, in masks, in metal layers to decrease growth errors. This is why you have respins. This is also why chip designers make redundancies with the chip based on how the errors propagate.

Its not a simple as ok the small chip is good to go and we are getting sufficient yields so now we can do a big chip on the same process with out worrying about it.

Smaller nodes have less tolerance as well, so this is also a reason maturity takes longer.

People just like bashing radeon tech/amd for no reason + trolling its no use trying to explain something here just read and move on.


No when people don't understand something they try to make things up from what they do know, and often its wrong, they should take the time and understand what they are saying and where that information comes from.

And reading into things and making comments on how power efficient Polaris is based on what AMD has showed, if you read into it, the opposite is what you get, it seems to match current Maxwell 2's power efficiency if Maxwell 2 was put on 16nm. And this is why I stated, I'm reading into it too much, because making wide statements like that, are just grand speculation.
 
Last edited:
how old are you? You don't remember prior to the fx line form nV, did they have major issues with big GPU transitions to smaller newer nodes?

Creating smaller chips on a new smaller node, has very little do with "pipe cleaning" as you so aptly put. It has to do with smaller chips having better yields. During the manufacturing and growing of silicon crystals, errors tend to propagate at the edge of the wafer, so as you start producing larger chips, you get more chances of errors because there is more area per chip and of course more errors as you get closer to the outer circumference of the wafer per chip. Thus the chances of a fully functional chip is less. Now they will have to make changes in design, in libraries, in masks, in metal layers to decrease growth errors. This is why you have respins. This is also why chip designers make redundancies with the chip based on how the errors propagate.

Its not a simple as ok the small chip is good to go and we are getting sufficient yields so now we can do a big chip on the same process with out worrying about it.

Smaller nodes have less tolerance as well, so this is also a reason maturity takes longer.
In fairness it is both. AMD launched the 4770 on the 40nm process while the rest of the 4000 series was on 55nm to get a jump start for their next GPU revision.
 
Back
Top