September 17th launch for GTX 900 series? Inno3d teases!

How do you end up with 138 TMUs when you have 1664 Cores? That makes no sense at all as its should be 104 TMUs based on 1664/16. I think that GPUZ can not properly read the specs and thinks this is like Kepler and trying to approximate because 1664/12 would be 138.66.
 
I'm in for 3, but I really hope they have stock at launch and I don't have to wait forever (or get gouged on ebay).
 
970 should be neck to neck with 780 at $399 price range and 980 should be neck to neck with 780ti at $499 price range.
I think that will be impressive if you ask me.
780 currently costs 420-440$ (with MIR) so please explain why a new card with a similar performance for $399 impresses you?
If it was $299 or way faster than 780 (or both ;) ) then that would be impressive. But right now that is usual nvidia price gouging when they feel no pressure from the competition.

I sure hope that's not for your sig pc as even just one 980 would be held back quite a bit.
Held back by CPU? Then what CPU is need so that 980 is not held back?
 
Last edited:
I'd call the 780 price drops vendor clearance deals, not price gouging. Retailers are trying to make room for the new product by selling off remaining inventory at a reduced cost - it lowers their (the retailers) margins (nVidia already has their money from the retailers) but it beats a product with a $500 price tag taking up space on the shelf/in the warehouse that is beaten by or very close in performance to a newer and cheaper $400 product.

Basically its to sucker the chumps into buying soon to be old tech for what is perceived as a deal in the eyes of the uninformed. It's a well known fact, people see bigger red numbers and are more inclined to buy the product.
 
Definitely not a Titan successor in 1H '15. They need 16FinFet for that and we would be lucky to see it by Q4 '15.

I'm expecting a quick shrink of GM204 to 16FinFet in 1H '15 to try and fight against Fiji.

Depends on how fast it's going to be. All they have to do is make a 2nd gen Maxwell that's similar in size to Kepler (561mm²), clock it at 1.2GHz and you are looking at 40-50% above a GTX Titan. You don't need 16nm/20nm FinFet for that. Also the are shipping manifests to Nvidia's internal testing facility at Bangalore for a GM200 came out only a month after the GM204 manifests which means they taped out at about the same time. The card is probably already ready to go now, they are just waiting for yields to rise or waiting for a timed release.
 
Sometimes we don't even get the specs until launch morning. While it has been a while, I remember waking up at the crack of dawn and reading reviews/specs for the 8800 before work. I ended up making a b-line to Microcenter instead.
 
Sometimes we don't even get the specs until launch morning. While it has been a while, I remember waking up at the crack of dawn and reading reviews/specs for the 8800 before work. I ended up making a b-line to Microcenter instead.

You and me both... :D
 
Sometimes we don't even get the specs until launch morning. While it has been a while, I remember waking up at the crack of dawn and reading reviews/specs for the 8800 before work. I ended up making a b-line to Microcenter instead.

LOL yep... I scrambled to buy one online :D. As I said, "if even that early" to the day before :eek: .
 
780 currently costs 420-440$ (with MIR) so please explain why a new card with a similar performance for $399 impresses you?

its $20 to $40 less without a MIR and its likely you could get it less than msrp not much after launch.
 
Well im hoping to get another 780GTX SC and go Quad SLI and then wait for Pascal if the rumors of the performance are true.
 
Depends on how fast it's going to be. All they have to do is make a 2nd gen Maxwell that's similar in size to Kepler (561mm²), clock it at 1.2GHz and you are looking at 40-50% above a GTX Titan. You don't need 16nm/20nm FinFet for that. Also the are shipping manifests to Nvidia's internal testing facility at Bangalore for a GM200 came out only a month after the GM204 manifests which means they taped out at about the same time. The card is probably already ready to go now, they are just waiting for yields to rise or waiting for a timed release.

You are basing that off GM107... which doesn't need a high DP/compute output nor other compute orientated features, has a much more simple bus and simply had the room to grow, in transistor count and die size

It is also harder to hit high clocks on larger and more complex ASICs. Add in having to increase transistor density, which negatively affects clocks/leakage, you come to the conclusion that a ~560mm2 28nm Maxwell wouldn't be as big as an improvement as you think.
 
You are basing that off GM107... which doesn't need a high DP/compute output nor other compute orientated features, has a much more simple bus and simply had the room to grow, in transistor count and die size

It is also harder to hit high clocks on larger and more complex ASICs. Add in having to increase transistor density, which negatively affects clocks/leakage, you come to the conclusion that a ~560mm2 28nm Maxwell wouldn't be as big as an improvement as you think.

Agreed..The yields on a 1.2Ghz+ clocked die that size would be terribly poor, to say the least, which would drive the cost up due to all those dies with 75~80% functional shaders etc that just get piled up in the corner..

The 980 is likely going to be the best we get until 16mm FinFet ramps up next year, unless Nvidia launches an extremely low production Halo product.
 
You are basing that off GM107... which doesn't need a high DP/compute output nor other compute orientated features, has a much more simple bus and simply had the room to grow, in transistor count and die size

It is also harder to hit high clocks on larger and more complex ASICs. Add in having to increase transistor density, which negatively affects clocks/leakage, you come to the conclusion that a ~560mm2 28nm Maxwell wouldn't be as big as an improvement as you think.

Agreed..The yields on a 1.2Ghz+ clocked die that size would be terribly poor, to say the least, which would drive the cost up due to all those dies with 75~80% functional shaders etc that just get piled up in the corner..

The 980 is likely going to be the best we get until 16mm FinFet ramps up next year, unless Nvidia launches an extremely low production Halo product.

I was thinking the way you were after jotting some numbers and calculations on a napkin until I saw a glowing coming from my closet. I went there and opened the door and at that moment I saw it... my forum famous crystal ball was glowing, when I looked inside I saw fire, brimstone and GTX 980 Pwning ALL!!

It will be over 45% faster than GTX 780Ti in many/most scenarios. Later Nvidia will launch a 980Ti (releasing around Xmas) since as they mentioned the GTX 980 doesn't replace the 780Ti which is at the top of their stack. (ridiculous $1000+ cards excluded)

I also saw overclocking performance than can have a GTX 980 reaching almost 75% faster than 780 Ti & 980 Ti doubling GTX 780 Ti in performance when overclocked.

I can play your napkin play with you and I come to a single logic. There is no logic. Now Please forgive me, I need to rest for a day to recover from the vision it drains me. :D

My forum famous crystal ball has spoken

2013-real-estate-crystal-ball.jpg
 
I'm living vicariously through you guys and the reviews this round. I'm on 680's and normally skip a generation so I would typically jump on these but I just haven't gamed as much as I'd like to in the last year or two... Still excited to see the results/improvements though!
 
I currently have a 760 but if these new cards can give me at least 20-25% more performance I might upgrade.
 
I was thinking the way you were after jotting some numbers and calculations on a napkin until I saw a glowing coming from my closet. I went there and opened the door and at that moment I saw it... my forum famous crystal ball was glowing, when I looked inside I saw fire, brimstone and GTX 980 Pwning ALL!!

LOL! Fun to see that the crystal ball is back. Missed it the last couple of launches. :D
 
Back
Top