[TweakTown] HBM2 could be delayed.

[dacostafilipe]

And what about 8-hi HBM1 ?

 

[ShintaiDK]

And what about 8-hi HBM1 ?

Doesn't exist.

 

[ShintaiDK]

So, no new cards next year?

If we assume the sources are telling the truth. Then it could seem so. I think we will see lower end cards tho no matter if its true or not.

 

[AtenRa]

IF there is a HBM 2.0 problem we could see 16nm middle-end GPUs with GDDR-5 first in 2016 and HBM 2.0 refresh in 2017.

For example, GP204 at 350mm2 with GDDR-5 could be 15-30% faster than 600mm2 GM200 (GTX980Ti) at half the power but at $750

or

GP207 at 160mm2 with GDDR-5 could be 15-30% faster than GM206 (GTX 960) at half the power but at $250

 

[desprado]

IF there is a HBM 2.0 problem we could see 16nm middle-end GPUs with GDDR-5 first in 2016 and HBM 2.0 refresh in 2017.

For example, GP204 at 350mm2 with GDDR-5 could be 15-30% faster than 600mm2 GM200 (GTX980Ti) at half the power but at $750

or

GP207 at 160mm2 with GDDR-5 could be 15-30% faster than GM206 (GTX 960) at half the power but at $250

You are right.It has a good chance to be in that way.

 

[dacostafilipe]

Doesn't exist.

It's on Hynix slides/site ...

 

[ShintaiDK]

It's on Hynix slides/site ...

Was. Its a cancelled product. I wouldn't be surprised if 8Hi HBM2 gets axed too.

 

[littleg]

If the chips have been designed for HBM it's not going to be that easy to change to GDDR though.

 

[dacostafilipe]

Was. Its a cancelled product. I wouldn't be surprised if 8Hi HBM2 gets axed too.

Alright, good to know, thanks.

 

[AtenRa]

If the chips have been designed for HBM it's not going to be that easy to change to GDDR though.

yeap, but i believe AMD and NV would know long time ago if HBM 2.0 would be available for mass production by 2016 and they would have designed their chips for GDDR-5 or HBM.

 

[Tuna-Fish]

yeap, but i believe AMD and NV would know long time ago if HBM 2.0 would be available for mass production by 2016 and they would have designed their chips for GDDR-5 or HBM.

They would know what the other vendors plans for the future are. In top-tier semiconductors, unplanned product delays do happen.

HBM being functional depends on more independent process steps working well than traditional ram. They need the memory chips themselves, the signaling die below the memory chips, the bonding steps for the ram chips, the interposer, the bonding step for ram to interposer and the bonding step for GPU to interposer all to work well enough for HBM2 to be usable. Too bad yields in any of those steps means no go, and many of those steps have not been in mass production before Fiji.

 

[Despoiler]

If we count the current roadmaps HBM2 will be available in Q1 2016. But I think this is a very aggressive plan, and I may say it will only available at the middle of the year.
The harder part is the interposer. For HBM2 it must be clocked to at least 800MHz, which will be really hard for the bigger chips. Even if the HBM2 will available in time, there is a huge chance that the interposers won't.

In addition Q2 2016 for a volume target.

The interposer is the only part that doesn't have logic on it. It's actually the easiest piece of HBM.

 

[readers]

This HBM2 rumor is BS, if there is going to be a delay it will affect both N&A, there is no way only nvidia going to get them and AMD not.

 

[stahlhart]

Thread cleaned. Review the OP, and stay on topic.
-- stahlhart

 

[ShintaiDK]

This HBM2 rumor is BS, if there is going to be a delay it will affect both N&A, there is no way only nvidia going to get them and AMD not.

That just seems to be the article writer going crazy to increase the drama. His quoted source is this:

"wouldn't count on using HBM2 next year"

So yes, equally affect both if true.

 

[Kenmitch]

From trouble sourcing the amount of HB2 AMD wants to HB2 is delayed? How'd that happen?

Looks like another BS click bait article too me.