** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[Jono8]

+1

In practice I think the first mid range Pascal cards that turn up will be struggling to beat a GTX 980 Ti by a small amount.

Nah.

If we are talking about the 980 equivalent for Pascal, I reckon it will decently beat a 980Ti (20% or more).

Otherwise the 1080Ti (or whatever it will be called) won't be that much of a jump over a 980Ti.

 

[Telecaster]

Nah.

If we are talking about the 980 equivalent for Pascal, I reckon it will decently beat a 980Ti (20% or more).

Otherwise the 1080Ti (or whatever it will be called) won't be that much of a jump over a 980Ti.

I think some people's expectations are a bit high. People are getting carried away a bit by the potential performance level of the next gen.

 

[Simon Belmont]

Nah.

If we are talking about the 980 equivalent for Pascal, I reckon it will decently beat a 980Ti (20% or more).

Otherwise the 1080Ti (or whatever it will be called) won't be that much of a jump over a 980Ti.

This^

Exactly mate

I think some people's expectations are a bit high. People are getting carried away a bit by the potential performance level of the next gen.

We'll find out soon enough.

 

[Telecaster]

We'll find out soon enough.

If soon enough is about 12 months, then yes, we'll find out soon

 

[Kaapstad]

Nah.

If we are talking about the 980 equivalent for Pascal, I reckon it will decently beat a 980Ti (20% or more).

Otherwise the 1080Ti (or whatever it will be called) won't be that much of a jump over a 980Ti.

You won't even notice 20% unless you are running FRAPs.

I often don't notice 300% when I am using 4 cards instead of 1 card.

For gaming I would not base an upgrade on a 20% increase, it would be a total waste of money.

I would not wait for months for a next gen card (that could be up to 20% faster) if I needed a high end card like a GTX 980 Ti now.

 

[Tommy_Here]

I'm expecting small pascal to be a slightly bigger jump due to HBM2 than 580>680 small kepler vs big fermi
small pascal > big maxwell should offer a good 30% jump.... watch this space

 

[Simon Belmont]

If soon enough is about 12 months, then yes, we'll find out soon

12 months? I doubt that. Come early summer or end of. With VR releasing shortly I would think AMD and Nvidia will have something ready for that,.

 

[Telecaster]

12 months? I doubt that. Come early summer or end of. With VR releasing shortly I would think AMD and Nvidia will have something ready for that,.

Yes, probably a 970/390 replacement to clean up the middle market where all the money is, no full fat Pascal for a year I reckon.

 

[Jono8]

You won't even notice 20% unless you are running FRAPs.

I often don't notice 300% when I am using 4 cards instead of 1 card.

For gaming I would not base an upgrade on a 20% increase, it would be a total waste of money.

I would not wait for months for a next gen card (that could be up to 20% faster) if I needed a high end card like a GTX 980 Ti now.

What an odd thing to say

Big Maxwell (980Ti) is only 25-35% faster than small Maxwell (980) so why all the hubbub about waiting for big pascal when you might not even notice the difference according to you (or is 30% enough ).

 

[Jono8]

, no full fat Pascal for a year I reckon.

Why wait? You wont notice the difference over small pascal anyway according to kaapstad.

 

[Kaapstad]

What an odd thing to say

Big Maxwell (980Ti) is only 25-35% faster than small Maxwell (980) so why all the hubbub about waiting for big pascal when you might not even notice the difference according to you (or is 30% enough ).

When Pascal arrives you will have several choices over time.

Medium Pascal is likely to turn up first and if you want one you won't get much change out of £500 for about 20% better performance than a 980 Ti.

Big Pascal is likely going to turn up later and will cost you a little over £500 for 80% or 90% increase over a 980 Ti.

If I had around £500 to spend on a Pascal GPU in the next year I know which option I would go for.

Anyone buying a mid range Pascal or Polaris is going to feel very sore when the big variants turn up. It will be like comparing a GTX 970 to a TitanX.

 

[robfosters]

When Pascal arrives you will have several choices over time.

Medium Pascal is likely to turn up first and if you want one you won't get much change out of £500 for about 20% better performance than a 980 Ti.

Big Pascal is likely going to turn up later and will cost you a little over £500 for 80% or 90% increase over a 980 Ti.

If I had around £500 to spend on a Pascal GPU in the next year I know which option I would go for.

Anyone buying a mid range Pascal or Polaris is going to feel very sore when the big variants turn up. It will be like comparing a GTX 970 to a TitanX.

You think it will really be almost double the speed of the 980Ti? That's a mega upgrade!

 

[Kaapstad]

You think it will really be almost double the speed of the 980Ti? That's a mega upgrade!

TitanX = 8.1 billion transistors

Big Pascal (and Polaris) around 17 billion transistors.

It will be a much bigger upgrade than going from Kepler to Maxwell.

 

[bru]

I'm expecting small pascal to be a slightly bigger jump due to HBM2 than 580>680 small kepler vs big fermi
small pascal > big maxwell should offer a good 30% jump.... watch this space

I'm fully expecting small/medium Pascal to be GDDR5 (nonX), hope fully 384bit but I wouldn't be surprised to see another 256bit part.

When Pascal arrives you will have several choices over time.

Medium Pascal is likely to turn up first and if you want one you won't get much change out of £500 for about 20% better performance than a 980 Ti.

Big Pascal is likely going to turn up later and will cost you a little over £500 for 80% or 90% increase over a 980 Ti.

If I had around £500 to spend on a Pascal GPU in the next year I know which option I would go for.

Anyone buying a mid range Pascal or Polaris is going to feel very sore when the big variants turn up. It will be like comparing a GTX 970 to a TitanX.

For me I would be looking at Medium Pascal for the second tier card, the effective 970 replacement. coming in at about the £300 mark for something maybe 10% faster than a 980ti. People will snatch them up like crazy. Then as you suggest the 980 replacement will probably be £450/500 and another 10% or so in performance. Of course I would love to see more performance, but I think these figure are probably more realistic.

 

[james.miller]

I'm fully expecting small/medium Pascal to be GDDR5 (nonX), hope fully 384bit but I wouldn't be surprised to see another 256bit part.




For me I would be looking at Medium Pascal for the second tier card, the effective 970 replacement. coming in at about the £300 mark for something maybe 10% faster than a 980ti. People will snatch them up like crazy. Then as you suggest the 980 replacement will probably be £450/500 and another 10% or so in performance. Of course I would love to see more performance, but I think these figure are probably more realistic.

I hope not, that's barely an upgrade from a 970. Having to spend 500 quid just to see a half decent upgrade would be depressing to say the least.

 

[VincentHanna]

It's all crystal ball gazing, but you don't think a mid tier card being 10% quicker than one of the fastest cards available for half the price of said card, would be much of an upgrade?

 

[straxusii]

580 to 780 was the last time we saw a full fat upgrade over a die shrink. That was 3B transistors going to 7.1B. Similar to the increase rumored for 980Ti to 1080Ti.

Add in HBM2 and increases in performance in the range of 80-100% seem reasonable based on the 580 -> 780 performance increase.

 

[straxusii]

It's all crystal ball gazing, but you don't think a mid tier card being 10% quicker than one of the fastest cards available for half the price of said card, would be much of an upgrade?

Drop £500 to gain +10%? Doesn't sound that great to me, depends how much £500 is worth to you I guess (minus what you get for the 980Ti)

 

[straxusii]

Sadly history has proven that unless AMD have something up their sleeve Nvidia will do their trick of launching the mid range (2/3 die size) part first as the flagship. (680/980)

 

[Kaapstad]

It's all crystal ball gazing, but you don't think a mid tier card being 10% quicker than one of the fastest cards available for half the price of said card, would be much of an upgrade?

Trouble is using a GTX 970 as an example it does not work that way.

If you want to do 2160p an old 6gb Titan is a far better bet than a GTX 970.

What I am saying is you are still getting more with the older top tier cards than you get with the new mid range card that is supposed to replace them.

I don't think there will be anything in the mid range Polaris/Pascal ranges that will be as good as 12gb TitanXs in SLI @2160p for example. If anyone mentions the price of the TXs, the old TXs will outlast the newer mid range cards that are supposed to replace them.