AMD Polaris architecture – GCN 4.0

[drunkenmaster]

So what does it all mean? Absolutely nothing. NVIDIA was just not ready to show Pascal, but they failed to inform us that the Drive PX 2 prototype had Maxwell installed.

You do get that is precisely the point right, they weren't ready to show Pascal.

Time line of events, getting silicon back > being ready to show the silicon to people > being ready to show a working demo > being ready to send out samples to reviewers > release.

Many companies have somewhat misled people over how close they were to release something before, most companies don't hold something up pretending it's one thing when it's something else.

Tell me what Nvidia had to gain by showing a fake product, the ONLY gain they have is making people believe they were closer than they really were. No one cares if you don't hold up something. It's common to talk about a preproduction product and not have something to show, it's common to hold up a finished product to say "hey, this will be available really soon". Most of the time the reason for holding up a fake version of a product is to mislead people.

Should we mention wood screws? Nvidia held up a fake product and said hey, this WILL be out before the end of the year. IT launched what, 6 months later, well after they said it was. The only reason to hold up the fake was to reinforce that lie that it would be released soon... it turned out to be a lie and it also wasn't released as soon as they said.

 

[pmc25]

Pretty sure AMD announced they are using TSMC for some products.

14nm and 16nm, are actually the same process size BTW, just different marketing names. For a long time the process size name has just been a marketing name and has very little basis in physical reality. Both TSMC and Sasung process are actually more heavily based on their 20nm processes. They added FinFet technology and gave the processes a new marketing name to make it sound like a 1 node improvement. Thsi will be more and more common since the costs just go up andf up and the technology harder and harder to progress.

Polaris is 14nm only. I.e. no TSMC. AMD are listed as a partner / customer for TSMC at 16nm, but so were both AMD and NVIDIA on 20nm (latter were belatedly but the 20nm SHP process which was made specifically for them they pulled out of causing it to be canned entirely).

I see zero reason to respin Polaris for 16nm when any performance benefit of the higher power process is likely hugely outweighed by lower costs and vastly higher capacity at Samsung / GF, with slightly higher density to boot.

 

[Simon Belmont]

They want to wait until GTC. Showing it early gives nVidia no benefit

 

[Ayahuasca]

No as it does overclock and you can dream about overclocking it. The card they had in hand may have been an overclockers dream.

lmao

Thanks for some Tuesday chuckles.

Both camps should have been hung, drawn and quartered years ago then.

Exactly.

 

[Mauller]

All they had to do was hold up the demo board and say 'This is what the final product may look like, currently we have two Maxwell cards connected for testing and to give you an idea of how it will look. but the final product will come with Pascal boards' queue long talk about performance figures and estimates.

 

[humbug]

If Nvidia have Pascal cores they would not have needed to use Maxwell cores to spin a fib.

They don't have any, which is a whole nother argument.

 

[TheRealDeal]

They want to wait until GTC. Showing it early gives nVidia no benefit

Nvidia clearly think otherwise as they showed the product albeit with Maxwell.

 

[Orangey]

 

[pmc25]

As the Samsung deal has been done, only TSMC product that may make sense now is big server zen chips whigh might benefit more from the HP process. But by then Samsung's HP process is likely ready to go.

 

[drunkenmaster]

So should AMD be in big trouble as well then for the Fury X overclockers dream moment?

Lying is lying after all...

It's very possible it does overclock like a dream, but they also found out overclocking kills HBM so they locked off memory controller voltage or something to prevent it which prevents it from overclocking. It's entirely possible for something to change so that it turned out not to be able to overclock. It's not possible to hold up something with woodscrews through it and claim it's a real card at the time.

It's entirely possible AMD lied or were misleading, it's also possible something happened and what they said at the time was the truth. It isn't possible Nvidia wasn't directly and intentionally lying about the 480gtx.

I think Nvidia implied as strongly as possible that the Drive PX2 was real, I went through parts of the conference and in a flat out lying situation he could potentially get away with saying he never claimed it was Pascal. It kept refering to the drive px2, talked about it, showed slides, shows specs then when it was brought on stage he did say "and this is what it looks like".

You could argue that is true, the final product would look like that, but he also did a build up talking about Pascal only then held it aloft. It's the same words you would use say at the Fury X release have all the info then hold one up and say "and this is what it looks like".

He exceptionally strongly implied it was Pascal, they also made it sound like the demo being run 'in real time' behind the scenes was being run on a working Pascal version, then he brought one out to be seen.

 

[TheRealDeal]

There should be more than 3 tears as his drivers have killed a lot more cards over the years

 

[bru]

For gods sake all of you, you all know that it is standard industry practice to show non complete or non working samples at trade shows. Why are you making this such a big deal.

In my mind AMD had the right idea, show off the thing but don't let anyone take pictures, that way nobody will ever know if the sample you hold up in your hand is what you actually say it is. Show off the product in a closed case that nobody can see so you can show it working, but again nobody has any idea just what you've shown off. AMD certainly seem to have the upper hand at the moment, but for once it is because they have been sensible in their marketing/PR for the new upcoming product. Maybe they have learnt from their past mistakes and will be much better at it from now on.

Bottom line the new stuff is coming, hopefully sooner than we think.

 

[Simon Belmont]

For gods sake all of you, you all know that it is standard industry practice to show non complete or non working samples at trade shows. Why are you making this such a big deal.

People are bored. A little drama goes a long way..

 

[drunkenmaster]

For gods sake all of you, you all know that it is standard industry practice to show non complete or non working samples at trade shows. Why are you making this such a big deal.

In my mind AMD had the right idea, show off the thing but don't let anyone take pictures, that way nobody will ever know if the sample you hold up in your hand is what you actually say it is. Show off the product in a closed case that nobody can see so you can show it working, but again nobody has any idea just what you've shown off. AMD certainly seem to have the upper hand at the moment, but for once it is because they have been sensible in their marketing/PR for the new upcoming product. Maybe they have learnt from their past mistakes and will be much better at it from now on.

It's standard to show mock ups or fake parts AND SAY THEY ARE SUCH. It's NOT standard and NOT acceptable to show a mock up and heavily, heavily imply it's real. That is not standard and no one but Nvidia does it.

AMD showed a working demo to multiple journalists who all confirm exactly what it is. Unless every tech journalist is lying, there is nothing to make out about what they did. LIkewise the demo they showed was impossible to achieve on 28nm. Roughly 35W gpu vs 90W Nvidia GPU with the same performance. It was clear as day a working 14nm gpu in there, now unless they nabbed an Nvidia Pascal card it's going to be Polaris.

 

[Mtom]

We will see how the cards work when they will be released, until we have some official everything is just theory.
The years first media battle is won by AMD thanks to NVs mishap, and the fact that first time in long years they pulled the right PR stuff.

 

[Gregster]

As someone who doesn't really care what is in his machine, it is fun to watch these threads Keep up the good work guys. My boredom is subsiding

 

[TheRealDeal]

As someone who doesn't really care what is in his machine, it is fun to watch these threads Keep up the good work guys. My boredom is subsiding

Keep the laughs coming.

 

[bru]

It's standard to show mock ups or fake parts AND SAY THEY ARE SUCH. It's NOT standard and NOT acceptable to show a mock up and heavily, heavily imply it's real. That is not standard and no one but Nvidia does it.

AMD showed a working demo to multiple journalists who all confirm exactly what it is. Unless every tech journalist is lying, there is nothing to make out about what they did. LIkewise the demo they showed was impossible to achieve on 28nm. Roughly 35W gpu vs 90W Nvidia GPU with the same performance. It was clear as day a working 14nm gpu in there, now unless they nabbed an Nvidia Pascal card it's going to be Polaris.

Really..... taken from the Anardtech article.

As a quick preface here, while RTG demonstrated a Polaris based card in action we the press were not allowed to see the physical card or take pictures of the demonstration. Similarly, while Raja Koduri held up an unsoldered version of the GPU used in the demonstration, again we were not allowed to take any pictures. So while we can talk about what we saw, at this time it’s all we can do.

Did you even read my post, I was saying that AMD did it right, that hopefully they have learned from past mistakes and seem to be ahead at the moment. But of course you just have to jump to AMD's defence even though they were being praised and people take the mickey out of DP for NVidia damage control.

 

[CAT-THE-FIFTH]

I don't know why some are so surprised AMD MIGHT be out of the gate faster at a new process node. Ever since 55NM they have been the first to a new process node,and even the shortest time between Nvidia and AMD has been the HD7970 and GTX680 which was a few months. I would be more surprised if AMD and Nvidia launched new cards on the same(or similar) process node at the same time!!

 

[Gregster]

Keep the laughs coming.

I have gone from Titan to 290X to 970 to Titan X to Fury X, so I think I have a fair shout and do love a wind up. Depending on what comes first and what looks like the better option all round, quite possibly my next card will be AMD. By all means though, feel free to think what you like, all good to me

No as it does overclock and you can dream about overclocking it. The card they had in hand may have been an overclockers dream.

Now that's funny