Will Nvidia be selling 7nm graphic cards in 2019?

[D.P.]

The architecture might be capable of it but the node is (was - not sure if it is still true) around 20% worse than the Samsung one for range of clocks at a given power level.

What was ironic was how much certain people were hyping up the GF process and proclaiming how Nvidia were screwed because TSMC was failing etc.

 

[Rroff]

What was ironic was how much certain people were hyping up the GF process and proclaiming how Nvidia were screwed because TSMC was failing etc.

Then they trot out the AMD/other has bought up all the capacity line. They are just repeating (sometimes verbatim) what is going around on a certain other forum by 1-2 pro-AMD people who in former times had quite a bit of credibility but these days most of what they say is just plain wrong.

(Funny thing is I used to be a moderator there before it became an AMD centric place (I mean the whole site) and can still see the private forums hah - 1-2 people here have slagged off my posting here in them which is too funny).

 

[D.P.]

TSMC has already confirmed Nvidia is a major cudtomer of the 7nm node.

Nvidia have a very close relationship with TSMC, to the extent that the 12nm process used for Volts was a customized process made for Nvidia in order to fullfil Oakridge commitments.


Anyone know how AMD's contractual obligations to buy a certain wafer order from GF works out when GF seem to be abandoning 7nm?

 

[Gregster]

TSMC has already confirmed Nvidia is a major cudtomer of the 7nm node.

Nvidia have a very close relationship with TSMC, to the extent that the 12nm process used for Volts was a customized process made for Nvidia in order to fullfil Oakridge commitments.


Anyone know how AMD's contractual obligations to buy a certain wafer order from GF works out when GF seem to be abandoning 7nm?

They were tied into wafers and any order from GloFo that AMD no longer wanted would cost them. No idea how much but that was how it was.

 

[Mauller]

Anyone know how AMD's contractual obligations to buy a certain wafer order from GF works out when GF seem to be abandoning 7nm?

Obligations like that hold only if the contractor is capable of providing for the customer, which they won't anymore since they have decided to shelve 7nm for the time being.

If demand starts to surge, then Glofo might end up putting the investment back into bringing a 7nm production line up. But since AMD was their only customer for the time being, they decided that the 2-4Billion needed for the investment might not be worth the return yet.

What was ironic was how much certain people were hyping up the GF process and proclaiming how Nvidia were screwed because TSMC was failing etc.

The glofo process is/was a bit ahead of TSMC's, Glofo actually relaxed some of the features of their original design to make cross sourcing easier between Glofo, TSMC and Samsung.

 

[KillBoY_UK]

But you can also tweak Pascal clocks and voltages to get even more performance per watt.

Peoplr making this argument compare Vega's power with the 1080ti but the performance with the 1080.

In reality the 1080 used less power and I'd still5-8% faster. You can down clock and under volt the 1080 a little to get similsr performance to Vega at a fraction of the power

yep I think I got down to like .93v and still boosted into the low 1800's with a 35-40w off in power. In percentage terms that's huge

 

[TheRealDeal]

yep I think I got down to like .93v and still boosted into the low 1800's with a 35-40w off in power. In percentage terms that's huge

The difference with Vega is you can take serious amounts of power off and actually gain a lot of performance. To get extra performance on Pascal you need to increase volts or stay at stock volts. It's 2 difference scenario's i.e. with my Vega 64 i can drop from 1200mv to 1000mv and still run 1630 which is well above stock Vega. I can probably push it more but that's enough for me atm. Power usage goes down where as performance goes up. What you are doing is sticking around stock boost and reducing power usage. If you want to go higher you will need to increase the volts again unlike with Vega which can go faster with reduced volts.

 

[KillBoY_UK]

The difference with Vega is you can take serious amounts of power off and actually gain a lot of performance. To get extra performance on Pascal you need to increase volts or stay at stock volts. It's 2 difference scenario's i.e. with my Vega 64 i can drop from 1200mv to 1000mv and still run 1630 which is well above stock Vega. I can probably push it more but that's enough for me atm. Power usage goes down where as performance goes up. What you are doing is sticking around stock boost and reducing power usage. If you want to go higher you will need to increase the volts again unlike with Vega which can go faster with reduced volts.

I get what you saying but increasing volts on Pascal unless you are into LN2 is pointless, peeps where talking about efficiency. I doubt Vega64 will even hit stock 1080ti efficiency aka Performance per watt , let alone a under volted one

Some peps will get a performance boost a undervolted 1080ti on air as it be hitting against the boost limits thermally far less.

 

[Nasher]

Undervolted to 1000mv it probably will be.