AMD Polaris architecture – GCN 4.0

D.P. · 12 May 2016 at 23:10

Freddie1980 said:
You mean to I read a unsourced rumor with no validation and it wasn't true!! Whoh............This is heavy doc.

No idea what you are talking about :confused:

Freddie1980 · 12 May 2016 at 23:14

D.P. said:
No idea what you are talking about

Back to future reference.

D.P. · 12 May 2016 at 23:19

The back to the future refernce I get, just have no idea what the rest of it is.

bru · 12 May 2016 at 23:21

Just back from the pub and I ran into my friend who told me about Pascal using more power then Maxwell, (which it turns out he was right) so I asked him if he knew anything about Polaris, he shook his head, telling me "he cannot say but things could be going better."
Make of that what you will. I wouldn't put too much stock into it and yet he was the only source that mentioned Pascal using more power than Maxwell, which before the announcement seems like a really strange idea.

mattmos · 12 May 2016 at 23:26

bru said:
Just back from the pub and I ran into my friend who told me about Pascal using more power then Maxwell, (which it turns out he was right) so I asked him if he knew anything about Polaris, he shook his head, telling me "he cannot say but things could be going better."
Make of that what you will. I wouldn't put too much stock into it and yet he was the only source that mentioned Pascal using more power than Maxwell, which before the announcement seems like a really strange idea.

As your friend is completely wrong abut the power usage, I don't think he's a reliable source.

CAT-THE-FIFTH · 12 May 2016 at 23:55

AMD debunks recent negative rumours:

http://www.eteknix.com/amd-debunks-recent-negative-rumours/

Since the reveal of NVIDIA’s latest enthusiast graphics range, there have been numerous negative articles surrounding AMD which stems from unverified sources. We’ve reported on some of these in a similar vein to other media outlets but it’s always important to adopt a sceptical approach whenever information is leaked. As a result, the coverage we previously posted was incorrect and we’ve received clarification from AMD’s Northern Europe PR and Community Manager, Joe Cowell about a number of recent criticisms. Firstly, AMD will be attending Computex and this was never in doubt. It’s important to remember that AMD always lists their attendance for trade shows in a clear and transparent manner. Secondly, there are no production issues whatsoever with Polaris and it’s on schedule. This means our report suggesting AIB partners were unhappy with AMD is completely false and based on incorrect sources. Therefore, we were wrong in disclosing this information and wish to clarify the real truth after receiving an official word from AMD.

So what does this all mean? Well, AMD isn’t struggling to produce quick driver updates, the Polaris architecture or failing to attend Computex. Also, their Q1 financial results were actually better than most analysts predicted. Therefore, any claims predicting their future demise are completely inaccurate. Of course, the Q1 financials are going to be poor when they’re preparing product launches later in the year for both the graphics and CPU segments. Hopefully, this clarification helped to quash the rumours being published right now, and we openly admit our original articles were wrong.

Honestly, I’m excited to see what AMD has in store and looking forward to testing more DirectX 12 games to see if their strong performance continues.

Do you think AMD has been treated unfairly by media outlets? If so, let us know your views in the comments section below.

Mauller · 12 May 2016 at 23:56

My friends, girlfriends, cousins, uncle second removeds, dogs, hairdressers pet cat heard a rumour that 3DFX is coming back and being spun off from Nvidia. They will be taking up the low end and portable gpu market while nvidia differentiate into the car and server market. But I wouldn't trust that cat.

And in other news http://www.eteknix.com/amd-debunks-recent-negative-rumours/

bru · 13 May 2016 at 00:18

mattmos said:
As your friend is completely wrong abut the power usage, I don't think he's a reliable source.

GM104 165w TDP (980\970)

GP104 180w TDP (1080\1070)

Seems he was right.

bru · 13 May 2016 at 00:24

Well a clarification from AMD is good and shows us that Polaris is on schedule and not delayed as rumoured, which is great news.
Now all we need is for the last few weeks to tick by and we can actually get to see some cards.

mattmos · 13 May 2016 at 00:34

bru said:
GM104 165w TDP (980\970)

GP104 180w TDP (1080\1070)

Seems he was right.

C'mon, its got to be performance per watt that matters - the 1080 (and prob the 1070) outperform the 980ti, which is 250w part. They're clearly a much more efficient architecture.

Mauller · 13 May 2016 at 09:08

mattmos said:
C'mon, its got to be performance per watt that matters - the 1080 (and prob the 1070) outperform the 980ti, which is 250w part. They're clearly a much more efficient architecture.

Not the architecture, the node shrink and addition of finfets.

Gregster · 13 May 2016 at 09:22

bru said:
Just back from the pub and I ran into my friend who told me about Pascal using more power then Maxwell, (which it turns out he was right) so I asked him if he knew anything about Polaris, he shook his head, telling me "he cannot say but things could be going better."
Make of that what you will. I wouldn't put too much stock into it and yet he was the only source that mentioned Pascal using more power than Maxwell, which before the announcement seems like a really strange idea.

That's weird, as I was down the pub and sat there minding my own business when 2 guys wearing NVidia shirts said that Pascal could be powered by a potato and they had finally invented a Portal gun (whatever one of those is). In my haste to get back and spread this news, I bumped into Dot Cotton who had some juicy gossip and I completely forgot about the NV guys but I did let Dot cook me breakfast in the morning

TheRealDeal · 13 May 2016 at 09:25

Gregster said:
That's weird, as I was down the pub and sat there minding my own business when 2 guys wearing NVidia shirts said that Pascal could be powered by a potato and they had finally invented a Portal gun (whatever one of those is). In my haste to get back and spread this news, I bumped into Dot Cotton who had some juicy gossip and I completely forgot about the NV guys but I did let Dot cook me breakfast in the morning

You also have lung Cancer due to inhaling all Dots second hand smoke.

Gregster · 13 May 2016 at 09:27

Harsh!

flopper · 13 May 2016 at 10:40

mattmos said:
C'mon, its got to be performance per watt that matters - the 1080 (and prob the 1070) outperform the 980ti, which is 250w part. They're clearly a much more efficient architecture.

If you draw 15 more watts, you draw more power how can you deny that?
Its like trying to deny your alive and that your are a zombie.
We will all believe your a zombie btw in regards to the sentence and remark but I digress.

If you OC it also draw more power.
anyone saying they buy hardware for low power and OC seems like the lie to themselves and many does as it makes them feel, impotent.

Anyhow, its a good wait, a couple of weeks until AMD present the line up.
serving brains last I checked.

Lokken86 · 13 May 2016 at 11:06

flopper said:
If you draw 15 more watts, you draw more power how can you deny that?
Its like trying to deny your alive and that your are a zombie.
We will all believe your a zombie btw in regards to the sentence and remark but I digress.
If you OC it also draw more power.
anyone saying they buy hardware for low power and OC seems like the lie to themselves and many does as it makes them feel, impotent.

Anyhow, its a good wait, a couple of weeks until AMD present the line up.
serving brains last I checked.

If it used 165w it wouldn't look so good next to a 980ti...

Rroff · 13 May 2016 at 11:25

Lokken86 said:
If it used 165w it wouldn't look so good next to a 980ti...

I think people might be underestimating a little the performance versus previous generation AMD will be getting from optimisations mind - if the stuff about improvements in triangle setup, etc. are even half true then we should expect to see around 15% overall increase in like for like performance just from that.

mahius · 13 May 2016 at 11:26

CAT-THE-FIFTH said:
AMD debunks recent negative rumours:

http://www.eteknix.com/amd-debunks-recent-negative-rumours/

Gah beat me to it.

I guess this whole Polaris 10 Pro and Polaris 10 XT seems likely, but I'm stil very skeptical about the performance of the pricier one being closer to 980ti/FuryX/1070. The pricing difference between the 380 and 380X (at each of their launches) was only about $30 (more in GBP though). Even if one costs $100 more than the other, I still think it's unlikely that one will perform at 390/970 levels and the other at 980ti/Fury X levels. More likely the 2nd will be 390x/980 levels of performance if there will be a 480X.

Not to mention that Hitman 1440p benchmark. With no indication as to the settings, we can't tell if it was FuryX or 390 levels of performance. Though I bet if it was max settings, someone at AMD would have mentioned it. 980ti is irrelevant there since it doesn't perform as well as AMD cards in DX12 currently.

Plus I don't think too many folks would be happy if an x80X card was priced way above £200. Nvidia seem to be upping their prices from previous gen launch prices a tad, folks are hoping that AMD don't follow suit, after all they did make a fuss about price-to-performance.

AMD's Launch event/Computex can't come sooner enough.

TheRealDeal · 13 May 2016 at 11:35

Gregster said:
Harsh!

You ever seen her without a smoke in her hand

. Standing next to her must be like putting your face in a coal fire Chimney.

Lokken86 · 13 May 2016 at 11:37

Rroff said:
I think people might be underestimating a little the performance versus previous generation AMD will be getting from optimisations mind - if the stuff about improvements in triangle setup, etc. are even half true then we should expect to see around 15% overall increase in like for like performance just from that.

I get that. What I'm getting at is the 1080 is a replacement for the 980, well in my eyes, so it could have used the same power and still outperformed it. NVIDIA are just cranking it up a bit to shade out the 980ti as a sorta added bonus of being the new highest performing card, albeit only for a short time until the ti version is out.