Overclock.net banner

1 - 11 of 11 Posts

·
Registered
Joined
·
154 Posts
Discussion Starter #1 (Edited by Moderator)
Fudzilla wants to make something clear. Ampere is not a mining GPU, it is not a gaming, and it is not an AI-focused Tensor Core GPU, it is all of the above, and comes after Turing 2018 GPU.

Nvidia will launch a GPU that everyone calls Turing this year for both AI and gaming applications. That GPU will appear before the end of the year, and most likely at some point in Q3 2018. Nvidia wants to be ready for the back to school market and this timeframe is what we repeatedly heard from a few independent sources.
Source
 

·
RIP OCN 2018
Joined
·
3,119 Posts
Before we are asked about our source, we are the source. This is an example of investigative journalism, and we keep our sources confidential.
Erm...
 

·
Premium Member
Joined
·
6,681 Posts
Fudzilla is known to just make stuff up out of the blue. They are worse than wccf, because at least wccf reposts things as news that they see forum users wishing for, rather than making it entirely up themselves.
 

·
Registered
Joined
·
1,713 Posts
Is the naming pattern nVidia has used not enough to know that Ampere and Volta both worked with electricity but Pascal and Turing worked with logical systems?

Ampere is the Tesla chip, Turing is the PC/Gaming chip.
 

·
Registered
Joined
·
193 Posts
Is the naming pattern nVidia has used not enough to know that Ampere and Volta both worked with electricity but Pascal and Turing worked with logical systems?

Ampere is the Tesla chip, Turing is the PC/Gaming chip.
Why would that imply that Ampere is a Tesla chip?
 

·
Hey I get one of these!
Joined
·
7,577 Posts
Why would that imply that Ampere is a Tesla chip?
Code names generally follow themes. To have two that fit clearly into existing themes is a fair indicator all on it's own.

Bulldozer
Piledriver
Steamroller
Excevator

Callisto
Heka
Deneb
Zosma
Thiban

Sandy Bridge
Ivy Bridge

Haswell
Broadwell

Skylake
KabyLake
CoffeeLake
CannonLake

Redwood
Juniper
Cypress

Cape Verde
Pitcairn
Tahiti
Hawaii
Fiji

etc
 

·
Registered
Joined
·
193 Posts
Code names generally follow themes. To have two that fit clearly into existing themes is a fair indicator all on it's own.

Bulldozer
Piledriver
Steamroller
Excevator

Callisto
Heka
Deneb
Zosma
Thiban

Sandy Bridge
Ivy Bridge

Haswell
Broadwell

Skylake
KabyLake
CoffeeLake
CannonLake

Redwood
Juniper
Cypress

Cape Verde
Pitcairn
Tahiti
Hawaii
Fiji

etc
That doesn't mean that Ampere = Tesla. There are Tesla cards with pascal chips.
Volta is also in a Titan card.
 

·
Registered
Joined
·
5 Posts
This actually makes sense if you look at it from the node perspective

The almost certainly announced at Computex GPU, which I'll call Turing, will not be on TSMC 7nm. And yet, TSMC is running out a 7nm node in the near future. It would make a lot of sense for NVIDIA to both do a Computex GPU because the notebook makers demand one on schedule, and a GPU on the 7nm node, hence Ampere.
 

·
Dream with your eyes open
Joined
·
3,837 Posts

·
Premium Member
Joined
·
6,681 Posts
This actually makes sense if you look at it from the node perspective

The almost certainly announced at Computex GPU, which I'll call Turing, will not be on TSMC 7nm. And yet, TSMC is running out a 7nm node in the near future. It would make a lot of sense for NVIDIA to both do a Computex GPU because the notebook makers demand one on schedule, and a GPU on the 7nm node, hence Ampere.

TSMC has already been through validation testing, risk production, and is officially now into high volume mass production of 7nm FF node for GPUs and CPUs. If this will launch July, we could easily see it on 7nm given Nvidia would have already been able to tape it out on 7nm months ago. If Nvidias mass production is now starting next month, they would have hundreds of thousands of wafers at OEM by June, with final soldering assembly and packaging done and going out to retailers by the end of June for a hard launch in July. So the timetabel for 7nm fits even for hard launch let alone soft launch. Doesnt mean they will use 7nm, but it entirely possible for them to be doing so, especially sine they had to wait anyway for GDDR6 that had the same timeframe. It is already in mass production at Samsung and Hynix mass production will start in June.
 
1 - 11 of 11 Posts
Top