Nvidia Ampere Discussion [2020-05-14]

[Kaotik]

You've linked to NV's 3070 page with availability date mentioning that they are still saying 29th. This was about reviews too then?

Yes, because it's the only date NVIDIA has currently communicated to public and press. If some sites are given priority access to new dates, everything is possible, but I know for a fact that not even all those who have gotten RTX 30 FE's straight from NVIDIA for reviews have heard any other dates than the publicly mentioned 15th and 29th.

 

[DegustatoR]

Yes, because it's the only date NVIDIA has currently communicated to public and press. If some sites are given priority access to new dates, everything is possible, but I know for a fact that not even all those who have gotten RTX 30 FE's straight from NVIDIA for reviews have heard any other dates than the publicly mentioned 15th and 29th.

And as I've already said on that - okay.

 

[T2098]

https://videocardz.com/newz/nvidia-allegedly-cancels-geforce-rtx-3080-20gb-and-rtx-3070-16gb

Big move if true. Sounds like a significant hole in the product stack that would be (mostly?) filled by various flavours of a GA103...

"NVIDIA allegedly cancels its December launch of GeForce RTX 3080 20GB and RTX 3070 16GB. This still very fresh information comes from two independent sources.

Technically GeForce RTX 3080 20GB and RTX 3070 16GB could launch at a later time, but the information that we have clearly stated that those SKUs have been canceled, not postponed. NVIDIA has already canceled its RTX 3070 Ti model (PG141 SKU 0), so the RTX 3070 16GB (PG141 SKU5) and RTX 3080 20GB (PG132 SKU20) will be joining the list."

 

[SimBy]

I can’t take videocardz seriously lately.

 

[Jawed]

3080Ti 12GB for $700?

 

[SimBy]

3080Ti 12GB for $700?

So basically a 3090 with 1/2 VRAM for 1/2 the price.

 

[Picao84]

I can see two reasons for this:
- They already know AMD performance and are not impressed, so no need to launch this on 8nm Samsung when they are reportedly planning to move to TSMC.
- Yields of 8nm Samsung are so bad that they are accelerating the move to TSMC.

 

[trinibwoy]

I can see two reasons for this:
- They already know AMD performance and are not impressed, so no need to launch this on 8nm Samsung when they are reportedly planning to move to TSMC.
- Yields of 8nm Samsung are so bad that they are accelerating the move to TSMC.

The whole thing reeks of reactionary chaos. You don't plan and subsequently cancel SKUs left and right when things are in your favor. Of course, this is assuming that the rumors aren't bunk.

 

[SimBy]

I can see two reasons for this:
- They already know AMD performance and are not impressed, so no need to launch this on 8nm Samsung when they are reportedly planning to move to TSMC.
- Yields of 8nm Samsung are so bad that they are accelerating the move to TSMC.

According to Komachi it’s the SS8. Nvidia might actually pull the plug.

 

[Erinyes]

Also seems unlikely that TSMC has the capacity to supply enough 7nm wafers to Nvidia. I could see them going to Samsung 7nm for a refresh next year though,

 

[Qesa]

Apple moving to 5nm should free up a lot of capacity on 7?

 

[Digidi]

@Qesa I think not because machines from 7nm will be reworked to 5nm machines.

 

[Erinyes]

@Qesa I think not because machines from 7nm will be reworked to 5nm machines.

No, that was for 10nm and 7nm. For 5nm they have a new dedicated fab. TSMC 7nm is in demand though and has several large customers. There have been reports that AMD has placed orders for almost double their current capacity for Q4 onward. Nvidia would probably use any additional TSMC capacity they are able to procure towards fulfilling A100 demand. Samsung 7nm should be sufficient for the consumer parts.

 

[Deleted member 13524]

The whole thing reeks of reactionary chaos. You don't plan and subsequently cancel SKUs left and right when things are in your favor. Of course, this is assuming that the rumors aren't bunk.

Could also be that nvidia was planning for the 3070 16GB and 3080 20GB to be higher margin cards considering the memory amount, but they're now looking at stronger than expected competition from 12GB Navi 22 and 16GB Navi 21 which makes the placement of these new catds more difficult.

 

[Deleted member 13524]

According to Komachi it’s the SS8. Nvidia might actually pull the plug.

Why would nvidia pull the plug on Samsung 8nm?
They're getting ridiculously high transistor density on what is effectively an optimized 10nm node. GA104 has almost the same amount of transistors as the TU102 and it's half the size. And clocks didn't even hurt on that process.

 

[SimBy]

Why would nvidia pull the plug on Samsung 8nm?
They're getting ridiculously high transistor density on what is effectively an optimized 10nm node. GA104 has almost the same amount of transistors as the TU102 and it's half the size. And clocks didn't even hurt on that process.

I don't know, yields? I mean why are they on SS8 in the 1st place, when they literally said it's gonna be mainly on TSMC.

https://www.tweaktown.com/news/69424/tsmc-handle-nvidias-next-gen-7nm-ampere-gpus/index.html

 

[Kyyla]

Power consumption is terrible for one.

 

[DegustatoR]

Where is the idea that Nv had to launch a full Ampere production ramp to know yields coming from?
Do these people even know how chips are produced?

 

[Jubei]

I don't know, yields? I mean why are they on SS8 in the 1st place, when they literally said it's gonna be mainly on TSMC.

https://www.tweaktown.com/news/69424/tsmc-handle-nvidias-next-gen-7nm-ampere-gpus/index.html

No they didnt?

”Huang said that most of the 7nm GPU production will be done by TSMC while Samsung will only handling a small portion of 7nm GPU production for NVIDIA”

 

[trinibwoy]

Could also be that nvidia was planning for the 3070 16GB and 3080 20GB to be higher margin cards considering the memory amount, but they're now looking at stronger than expected competition from 12GB Navi 22 and 16GB Navi 21 which makes the placement of these new catds more difficult.

Considering the Radeon VII had 16GB of HBM2 there’s no way Nvidia was surprised by 16GB of GDDR6. There’s more to it than that.

Where is the idea that Nv had to launch a full Ampere production ramp to know yields coming from?
Do these people even know how chips are produced?

Yeah yields wouldn’t be new information at this point. Question is what could Nvidia have learned recently to have them scrambling. It doesn’t look good either way. I definitely don’t buy the story that Nvidia is backing off because they’re not impressed by Big Navi. That would be a silly reason to leave money on the table.