Nvidia Ampere might launch as GeForce GTX 2070 and 2080 on April 12th

chroniclard · 26 Feb 2018 at 15:59

At any one point in time there is a top end chip. When/If these are released it will be the 2080 or whatever they call it.

Rroff · 26 Feb 2018 at 16:06

Its like calling the GTX460 a top end chip

if consumers want to do that to themselves so be it I guess but no one should simply allow nVidia to mug them off like that.

Rroff · 26 Feb 2018 at 16:10

subbytna said:
GTX780 - Cut down GK110 - Again not the full fat chip but again Nvidia gave it the *80 to denote flagship high end.

I can forgive that one as it was close to the full spec core - slightly cut down from full spec but using the same 561mm2 core as the top end card - but it should have been the GTX670 or GTX670ti or something really.

bru · 26 Feb 2018 at 16:48

On the one hand.
Ok so lets all moan at NVidia for launching the midrange cards as the top end.
don't do it again you naughty company....

On the other hand.

Why do we have to wait twice as long for the new cards to arrive, they should release something new every year.

andybird123 · 26 Feb 2018 at 16:52

bru said:
On the one hand.
Ok so lets all moan at NVidia for launching the midrange cards as the top end.
don't do it again you naughty company....

On the other hand.

Why do we have to wait twice as long for the new cards to arrive, they should release something new every year.

because there isn't a new node worth moving to on a yearly basis any more (or the mobile phone industry eats up all the early availability) - mobile phone chips are tiny compared with GPU's and with mobile phones pushing equally silly prices even without mining as a motivator, mobile phone companies can afford to pay a premium to ride rough shod over the GPU industry, same reason memory for PC's is shooting up in price, because the mobile phone industry is eating ever square inch of silicon it can lay its grubby hands on

bru · 26 Feb 2018 at 17:00

andybird123 said:
because there isn't a new node worth moving to on a yearly basis any more (or the mobile phone industry eats up all the early availability) - mobile phone chips are tiny compared with GPU's and with mobile phones pushing equally silly prices even without mining as a motivator, mobile phone companies can afford to pay a premium to ride rough shod over the GPU industry, same reason memory for PC's is shooting up in price, because the mobile phone industry is eating ever square inch of silicon it can lay its grubby hands on

So your argument is, that there isn't a new node to move too?

You do realise that since the 680 ( the first of the midrange cards to be branded at the top end), there has only been one node shrink from 28nm to 16nm. No new node didn't really stop them before, did it.

andybird123 · 26 Feb 2018 at 17:09

bru said:
So your argument is, that there isn't a new node to move too?

You do realise that since the 680 ( the first of the midrange cards to be branded at the top end), there has only been one node shrink from 28nm to 16nm. No new node didn't really stop them before, did it.

exactly, we had maxwell in 2014/15 that introduced 16nm, pascal in 2016/17, so they were on 16nm for 4 years with only 2 architectures, they weren't releasing new arch's every year they drip fed the new cards over the course of two years
Pascal was remarked upon as surprising that they did manage to get such an improvement without a new node, as it hadn't really been done before

even 12nm isn't really a new node as such, its still an incremental improvement to 20nm tech just as 16nm was

unless you are saying that a 1080 in 2016 and a 1080ti in 2017 qualifies as every year, in which case a 2080 in 2018 still IS "every year"

FoxEye · 26 Feb 2018 at 17:23

Rroff said:
Its like calling the GTX460 a top end chip if consumers want to do that to themselves so be it I guess but no one should simply allow nVidia to mug them off like that.

Agreed. We shouldn't give AMD a free ride tho (not that I care), because the Fury was exactly the same situation. 290->390, then a new "class" (aka pricing tier) and a silly name for the real top-end card, priced waaay above it. The 390 even got a small price bump above the then 290 prices when it was released. But forget the 390, the take-away is that the Fury wasn't called the 390. It was AMD's Titan and they tried to sell it for Titan prices.

bru · 26 Feb 2018 at 17:25

andybird123 said:
exactly, we had maxwell in 2014/15 that introduced 16nm, pascal in 2016/17, so they were on 16nm for 4 years with only 2 architectures, they weren't releasing new arch's every year they drip fed the new cards over the course of two years
Pascal was remarked upon as surprising that they did manage to get such an improvement without a new node, as it hadn't really been done before

even 12nm isn't really a new node as such, its still an incremental improvement to 20nm tech just as 16nm was

unless you are saying that a 1080 in 2016 and a 1080ti in 2017 qualifies as every year, in which case a 2080 in 2018 still IS "every year"

Maxwell was entirely 28nm.

It is only from Pascal onwards that has been on 16/14nm.

Rroff · 26 Feb 2018 at 17:45

bru said:
Why do we have to wait twice as long for the new cards to arrive, they should release something new every year.

While fabricating something towards the top end of the possible core size with realistic yields is more complex than smaller parts it isn't massively so unless there are significant teething issues with a node. Once you are able to make ~300mm2 parts its generally just a case of the economy of the yields from there.

melmac · 26 Feb 2018 at 17:53

FoxEye said:
Agreed. We shouldn't give AMD a free ride tho (not that I care), because the Fury was exactly the same situation. 290->390, then a new "class" (aka pricing tier) and a silly name for the real top-end card, priced waaay above it. The 390 even got a small price bump above the then 290 prices when it was released. But forget the 390, the take-away is that the Fury wasn't called the 390. It was AMD's Titan and they tried to sell it for Titan prices.

I see what you are trying to say, but, it wasn't exactly the same situation, Figi was a new architecture. And it wasn't near titan prices. It was the same price as the 980ti.

Besides this is a thread about Nvidia's GPUs, so why should AMD's naming even be introduced?

andybird123 · 26 Feb 2018 at 18:06

bru said:
Maxwell was entirely 28nm.

It is only from Pascal onwards that has been on 16/14nm.

Sorry, i got mixed up - maxwell was commented on for being a surprising improvement without a node change

FoxEye · 26 Feb 2018 at 19:22

melmac said:
I see what you are trying to say, but, it wasn't exactly the same situation, Figi was a new architecture. And it wasn't near titan prices. It was the same price as the 980ti.

Besides this is a thread about Nvidia's GPUs, so why should AMD's naming even be introduced?

When there's only two vendors comparisons are inevitable. You literally cannot avoid them

And why should you?

Also the fact that Fiji was a new architecture it 100% irrelevant. The 290 was AMD's best card in that generation. The 390 was the 2nd best card. It doesn't matter that the 390 and the Fury were different architectures, that is entirely incidental to the point being made. They could have called the 390 the 380, and Fury the 390. It would have made sense. It was a choice AMD made; the same choice nV made to make the xx04 chips the new xx80 cards.

JediFragger · 26 Feb 2018 at 19:32

subbytna said:
Man that's mining for you, mid range card for £600, that's mental. You used to get the top end for that cash.

1080 released at £530-£550 so not much change there tbh ($599). Dropped down a little when the Ti was released.

melmac · 26 Feb 2018 at 19:38

FoxEye said:
When there's only two vendors comparisons are inevitable. You literally cannot avoid them And why should you?

Also the fact that Fiji was a new architecture it 100% irrelevant. The 290 was AMD's best card in that generation. The 390 was the 2nd best card. It doesn't matter that the 390 and the Fury were different architectures, that is entirely incidental to the point being made. They could have called the 390 the 380, and Fury the 390. It would have made sense. It was a choice AMD made; the same choice nV made to make the xx04 chips the new xx80 cards.

This is a thread discussing Nvidia's cards and prices and discussing the way they changed their tier levels. Why bring AMD even into it?

And if you can't see how different the two situations are, I am not going to waste my time explaining it to you.

Silent_Scone · 26 Feb 2018 at 20:45

FoxEye said:
When there's only two vendors comparisons are inevitable. You literally cannot avoid them And why should you?

When the performance difference makes the comparison both uninteresting and irrelevant.

C-Roy · 26 Feb 2018 at 20:47

Nasher said:
Yea, we need digital Thatcher

C-Roy · 26 Feb 2018 at 20:48

It'll be interesting to see how much of a performance boost we get this time.

FoxEye · 26 Feb 2018 at 22:07

Silent_Scone said:
When the performance difference makes the comparison both uninteresting and irrelevant.

It wasn't a performance comparison tho? :confused:

LOL. Was only saying that nV are not the only ones to invent new price tiers for their top end cards. People sure do take things oddly in this forum.

Silent_Scone · 27 Feb 2018 at 09:16

FoxEye said:
It wasn't a performance comparison tho?

LOL. Was only saying that nV are not the only ones to invent new price tiers for their top end cards. People sure do take things oddly in this forum.

Which, has everything to do with performance. You can't create a new halo tier if your competition is beating you. lol