Clues Gather Regarding Possible New AMD Polaris (Re)Revision Launch

[eidairaman1]

The reason Fury and Vega was coupled to HBM is that AMD, being an underdog with next to no money for R&D had to take the risk and bet on something not traditional to have chances to win.

It worked quite well with Zen (Infinity Fabric), it didn't work with HBM, which nVidai also had project to develop with, just in case and because they sit on heaps of money, (thank you very much for buying overpriced cards), but unlike AMD, had resources for alternative, conventional projects


TI is unlikely, but 1080 levels are realistic to expect.

Hbm actually works well on vega, fury was a prototype.

Rtx is struggling in powerdraw against vega56 and 1080 in idle.

Just remember, AMD didn't go with gddr5x, they only had gddr5 and hbm 1/2.

 

[londiste]

The reason Fury and Vega was coupled to HBM is that AMD, being an underdog with next to no money for R&D had to take the risk and bet on something not traditional to have chances to win.

AMD needed bandwidth and for multiple reasons they decided on HBM. I would not say they did not have other options, Hawaii had 512-bit bus and had comparable bandwidth to the newer 980Ti that was using very fast memory chips for the time. They gambled on a riskier solution that did not pay off this time.

Hbm actually works well on vega, fury was a prototype.
Rtx is struggling in powerdraw against vega56 and 1080 in idle.
Just remember, AMD didn't go with gddr5x, they only had gddr5 and hbm 1/2.

What do you mean HBM works well? Of course it does, there were no technical problems with HBM and how it works. The problems are manufacturing and price.
RTX idle power draw sounds like something Nvidia should be able to fix in drivers. God knows they fuck up the idle modes every couple years :/

Strangely, it has been Nvidia going with newer memory types lately. Running first to both GDDR5X and GDDR6.

 

[eidairaman1]

AMD needed bandwidth and for multiple reasons they decided on HBM. I would not say they did not have other options, Hawaii had 512-bit bus and had comparable bandwidth to the newer 980Ti that was using very fast memory chips for the time. They gambled on a riskier solution that did not pay off this time.

What do you mean HBM works well? Of course it does, there were no technical problems with HBM and how it works. The problems are manufacturing and price.
RTX idle power draw sounds like something Nvidia should be able to fix in drivers. God knows they fuck up the idle modes every couple years :/

Strangely, it has been Nvidia going with newer memory types lately. Running first to both GDDR5X and GDDR6.

Hbm was a new memory type.

 

[londiste]

Hbm was a new memory type.

Yup. But there have been two new ones since then.
I remember ATI/AMD almost always being the one jumping to the new memory type first. DDR, GDDR2, GDDR3, GDDR5

 

[DeathtoGnomes]

Yup. But there have been two new ones since then.
I remember ATI/AMD almost always being the one jumping to the new memory type first. DDR, GDDR2, GDDR3, GDDR5

Nvidia waits for the "new" memory to be stable and for prices to come down. New Memory is always higher priced.

 

[medi01]

They gambled on a riskier solution

Exactly my point.

At the same time, competitor, bathing in money, didn't have to gamble and simply tried both (or even more options, who knows).

 

[londiste]

Nvidia waits for the "new" memory to be stable and for prices to come down. New Memory is always higher priced.

They didn't for neither GDDR5X or GDDR6. There was considerable doubt if GDDR5X would be able to compete with HBM at the time. GDDR6 is new and coming out right now.

 

[DeathtoGnomes]

They didn't for neither GDDR5X or GDDR6. There was considerable doubt if GDDR5X would be able to compete with HBM at the time. GDDR6 is new and coming out right now.

GDDR was/is an established architecture.

 

[Casecutter]

Let me ask this even though the leak was in regard to "Polaris 30", does have to be actual resign to that of the old Polaris GCN 4.0?

AMD has "Vega" architecture (GCN 5.0) for APU's with 11CU, then there's the Vega that Intel has been working with, while finally there was the announcement of that semi-custom chip commissioned by Chinese firm Zhongshan Subor gaming console. Both are said to be 24CU parts. Could AMD just bump that to a 32CU part and GDDR6 make it on a 12nm and have a mid-range offering? The major design work for the base architecture in done and I'd think it's modular/scalable given how it used so far. Perhaps the project name is here to throw us just as done in this forum... into chaos like it has. As what's in a name?