AMD Working on a "Polaris" Chip with 232 mm² Die Area

[btarunr]

A former AMD employee who was with the company till July 2015, disclosed vague details of the various chip projects he was involved in. Two of those projects, labeled "A" and "B" were core-logic (southbridge). Project "F" drew the attention of the press to a graphics chip with a die-area of 232 mm², 430 function blocks, built on the 14 nm LPP process. A function block can be any differentiated or unique structure on a silicon die. 3DCenter speculates that this could be a GPU based on the company's upcoming "Polaris" (GCN 4.0) architecture; and likely a performance-segment chip from the next-gen GPU family.



View at TechPowerUp Main Site

 

[john_]

There is also a project E in that list that I think we never seen it. So this list could say something or just nothing. companies create many prototypes that never find their way to the market.

 

[the54thvoid]

This isn't useful. Given the processes involved, we know they will have working samples as we speak. A 'leak' from a guy who worked on it in Summer last year is just like yelling us grass is green.
We want constructive operating capacity and performance leaks. And even then, we'll fight like old married couples about the merits of it.
Pah...

 

[dj-electric]

Also, there's project H. Its a GPU that actually confirmes Polaris. We know nothing about it, but its polaris related.

Project H was created mostly to be published on tech websites and creat a hype over nothing

 

[Casecutter]

In reality this is just this...

 

[happita]

Not like this is a reputable article, but why are there all these Polaris leaks when I haven't seen a peep from Nvidia's Pascal? Are the people at AMD that loose-lipped that they can't keep it together until proper announcements start coming around? It's generally not a good idea to give hints to the company you are competing against about your next best product.

 

[Ferrum Master]

Monday... with AMD news that are not news but nothing

 

[Enterprise24]

RIP "Pitcairn".

 

[alucasa]

Wouldn't he have signed NDA?

 

[Nejc]

that would be a huge jump in transistor count, if only surface area is taken into account

 

[HisDivineOrder]

Not like this is a reputable article, but why are there all these Polaris leaks when I haven't seen a peep from Nvidia's Pascal? Are the people at AMD that loose-lipped that they can't keep it together until proper announcements start coming around? It's generally not a good idea to give hints to the company you are competing against about your next best product.

AMD needs to build positive buzz. nVidia is still trying to sell their current product before making people hunger for what comes next. As a result, AMD is willing to lose some sales now to build a crazed, frenzied fanbase because one thing they are lacking is marketshare and in today's head to head, they lose. So why not try to get everyone looking toward the future rather than take stock of what's happening right now.

 

[HD64G]

Not like this is a reputable article, but why are there all these Polaris leaks when I haven't seen a peep from Nvidia's Pascal? Are the people at AMD that loose-lipped that they can't keep it together until proper announcements start coming around? It's generally not a good idea to give hints to the company you are competing against about your next best product.

Article's source is an ex-AMD employee as it clearly mentions that, so those info are not derived from AMD itself.

 

[HumanSmoke]

This isn't useful. Given the processes involved, we know they will have working samples as we speak. A 'leak' from a guy who worked on it in Summer last year is just like yelling us grass is green.
We want constructive operating capacity and performance leaks. And even then, we'll fight like old married couples about the merits of it.
Pah...

There actually seems to be something "off" about including such level of detail in an online CV. IF it is actually true, every tech company should steer clear of hiring this guy - he clearly doesn't know the meaning of confidentiality. Hard to imagine that anyone working on unreleased silicon didn't have to sign an NDA.

Not like this is a reputable article, but why are there all these Polaris leaks when I haven't seen a peep from Nvidia's Pascal? Are the people at AMD that loose-lipped that they can't keep it together until proper announcements start coming around? It's generally not a good idea to give hints to the company you are competing against about your next best product.

AMD's information dissemination is usually only second to the Vatican, so this seems definitely to be an anomoly. But then, AMD seem to be hitting the publicity trail early with the 400 series. Might be a change in strategy with the RTG split and Koduri's influence. Also tends to point to Gemini (Fury X2) being primarily a system integrator OEM part rather than a consumer product, because they are effectively Osborning the thing by talking up Polaris.

As for Nvidia, they seem to be playing their cards close to their chests (as AMD have done in the past), and just like AMD of old, shipping manifests seem to be the primary indicators of production. With the first parts supposedly HPC orientated Tesla's (likely being showcased in April at GTC) it is probably understandable that performance leaks - if the silicon is ready - aren't doing the rounds.

 

[TheGuruStud]

Not like this is a reputable article, but why are there all these Polaris leaks when I haven't seen a peep from Nvidia's Pascal? Are the people at AMD that loose-lipped that they can't keep it together until proper announcements start coming around? It's generally not a good idea to give hints to the company you are competing against about your next best product.

Like it matters. Nvidia had multiple spies at AMD. I wouldn't doubt that they have more that didn't get caught. And of course nothing happened when they got caught.

 

[Fluffmeister]

Who needs spies when AMD have multiple ex-employees.

 

[xenocide]

Like it matters. Nvidia had multiple spies at AMD. I wouldn't doubt that they have more that didn't get caught. And of course nothing happened when they got caught.

You'd have to be pretty naive to think AMD doesn't have their own insiders at Nvidia\Intel. This is corporate America, these companies know exactly what one another are doing.

 

[TheGuruStud]

You'd have to be pretty naive to think AMD doesn't have their own insiders at Nvidia\Intel. This is corporate America, these companies know exactly what one another are doing.

Sure, with all that extra cash floating around.

 

[xenocide]

Sure, with all that extra cash floating around.

Not knowing what your competition is up to is literally suicide in the business world.

 

[Captain_Tom]

AMD needs to build positive buzz. nVidia is still trying to sell their current product before making people hunger for what comes next. As a result, AMD is willing to lose some sales now to build a crazed, frenzied fanbase because one thing they are lacking is marketshare and in today's head to head, they lose. So why not try to get everyone looking toward the future rather than take stock of what's happening right now.

Sure but there are also a lot of rumors going around that point to AMD being able to launch on GloFlo's 14nm process before Nvida/TSMC's 16nm process is ready.

I don't think it is just marketing. AMD will likely (And arguably needs to) launch Polaris a good few months before Pascal comes out.

 

[xfia]

Sure but there are also a lot of rumors going around that point to AMD being able to launch on GloFlo's 14nm process before Nvida/TSMC's 16nm process is ready.

I don't think it is just marketing. AMD will likely (And arguably needs to) launch Polaris a good few months before Pascal comes out.

i could see that.. the architecture is very different when switching to hbm and amd has no doubt been working on it much longer than nv and making better friendships with other companies that want to push for higher resolutions and advance computing as a whole.
its not really good for nv to be keeping so many doors locked and using proprietary stuff where microsoft already has very well optimized stuff in place for them to use like direct compute.
i would have to ask nv if they even want to be in the windows ecosystem and if its a no i would tell them to go elsewhere. im sure another company out there has what it takes to make gpu's and actually work together with microsoft and amd without a bunch of bs.

 

[Captain_Tom]

i could see that.. the architecture is very different when switching to hbm and amd has no doubt been working on it much longer than nv and making better friendships with other companies that want to push for higher resolutions and advance computing as a whole.
its not really good for nv to be keeping so many doors locked and using proprietary stuff where microsoft already has very well optimized stuff in place for them to use like direct compute.
i would have to ask nv if they even want to be in the windows ecosystem and if its a no i would tell them to go elsewhere. im sure another company out there has what it takes to make gpu's and actually work together with microsoft and amd without a bunch of bs.

Preaching to the coir my friend. AMD has put all of its eggs in the 2016 basket (Because they didn't have the resources to do anything else). Sure Nvidia made a boatload for 2 years now, but they did so with very short-sighted methods.

Then again we will see how this all pans out. I hope AMD doesn't just f*ck up for another year. They can't afford to.