NVIDIA Pascal Thread

[moonbogg]

I know prices have increased, but the only way Nvidia was able to get away with it is because AMD cards haven't been able to compete, right? That's the only explanation. Think about it. Back in the day, could Nvidia have charged $1,000.00 for an FX 5800 Ultra with the 9700 pro competing right next to it at half the cost? Not a chance.
Its just timing as well. Titan X came out way ahead of Fury X so Nvidia charged people more because no competition. 980ti replaced it and sold at $650 and that's because no one would pay $1,000.00 any more with Fury X around, but no one would also pay for Fury X with 980ti around, lol. So, Nvidia is a bastard, but they are a smart bastard.

 

[tential]

I know prices have increased, but the only way Nvidia was able to get away with it is because AMD cards haven't been able to compete, right? That's the only explanation. Think about it. Back in the day, could Nvidia have charged $1,000.00 for an FX 5800 Ultra with the 9700 pro competing right next to it at half the cost? Not a chance.
Its just timing as well. Titan X came out way ahead of Fury X so Nvidia charged people more because no competition. 980ti replaced it and sold at $650 and that's because no one would pay $1,000.00 any more with Fury X around, but no one would also pay for Fury X with 980ti around, lol. So, Nvidia is a bastard, but they are a smart bastard.

A LOT of it has been timing. More than anything it's been timing and product releases/lack of product releases.

 

[swilli89]

I know prices have increased, but the only way Nvidia was able to get away with it is because AMD cards haven't been able to compete, right? That's the only explanation. Think about it. Back in the day, could Nvidia have charged $1,000.00 for an FX 5800 Ultra with the 9700 pro competing right next to it at half the cost? Not a chance.
Its just timing as well. Titan X came out way ahead of Fury X so Nvidia charged people more because no competition. 980ti replaced it and sold at $650 and that's because no one would pay $1,000.00 any more with Fury X around, but no one would also pay for Fury X with 980ti around, lol. So, Nvidia is a bastard, but they are a smart bastard.

Ha man those were the days. Traded my GeForce3 in for a 9700 Pro circa 2001. Had to run Everquest/Tribes 2 at those leet max settings yo!

 

[Headfoot]

I know prices have increased, but the only way Nvidia was able to get away with it is because AMD cards haven't been able to compete, right? That's the only explanation.

No, not at all. The other explanation, which is by all accounts what actually happened, is that AMD also increased prices.

 

[Bacon1]

No, not at all. The other explanation, which is by all accounts what actually happened, is that AMD also increased prices.

 

[Headfoot]

Excellent graph, really brings it into perspective. Both AMD and nVidia got more expensive on 28nm versus 40nm by repositioning an upper midrange card as the flagship for the initial release.

 

[Adored]

AMD's price increase was strategic. They were barely holding to 30%+ market share with $200 290's, people were still buying $300 970's instead.

Sooner or later that has to go or you end up with the "cheap" tag indefinitely. So they rebranded Hawaii with double VRAM and increased prices to where they fit compared to the competition.

That's why they dipped below 20% market share. Now however they've found their true market share around 20% and at "fair" prices, it can only go up from here with better cards.

...And I just realised that had nothing much to do with the graph, however it does help to understand Fury X's pricing, which has always been too high but still helps to shake off the "cheap" brand feeling. They don't care about Fury X sales, it's more important to look like it's worth the cost. In previous years AMD would have sold faster cards at lower prices (think 6950 vs 560 Ti) holding on to what is basically a fake 40% market share. That's just a bad strategy.

 

[moonbogg]

no, not at all. The other explanation, which is by all accounts what actually happened, is that amd also increased prices.

oh owned!

Also, pic not working? That sucks. Ownage isn't as fun like this, but it will have to do.

 

[antihelten]

Why are all the prices in that graph so weird? they certainly aren't MSRP, nor do they look inflation adjusted, so what's up?

oh owned!

Why are you celebrating being "owned", seems quite self-deprecating?

Headfoot said that AMD prices increased, which for the period in question (Kepler and onwards), the graph clearly shows to be correct.

 

[sontin]

I like how the price hike from 6970 to 7970 ends at $520 and not at $549 and that a GTX680 stands at the same price point instead of $499. :thumbsup:

 

[Grooveriding]

My guess, $399 for the 970 replacement.

No need to price it lower since its going to be a lot faster than the 970 and it's gonna have 8 (or 7.5GB) of vram.

I don't see this. I think we'll see higher prices. They dropped the 970 with this huge price/perf marketing push. Snuck the 3.5GB real VRAM speed past to position it as a full speed 4GB card against their competition's same VRAM amount. I'm not seeing how they'll need a huge price/perf push with a new series on a new node with a decent speed increase. If the 1070 is a good 30% faster than a 970, I could see a return to $499 on that card and $699 on the 1080. They haven't always done an affordable second string card like they did at the 970/980 launch.

I guess it depends on how much performance they can get out of these small dies on the new node. If the 1080 is only going to be 10-20% faster than a 980ti, they'll probably have to be more conservative on pricing, making your 1070 price prediction more possible if it turns out to be equal to a 980ti with better power use. Could be possible as I'm expecting some disappointment on performance from the new cards, the dies are too small to give a good boost, imo. Still hoping for the potential to be surprised like I was with the 980ti.

 

[3DVagabond]

I know prices have increased, but the only way Nvidia was able to get away with it is because AMD cards haven't been able to compete, right? That's the only explanation. Think about it. Back in the day, could Nvidia have charged $1,000.00 for an FX 5800 Ultra with the 9700 pro competing right next to it at half the cost? Not a chance.
Its just timing as well. Titan X came out way ahead of Fury X so Nvidia charged people more because no competition. 980ti replaced it and sold at $650 and that's because no one would pay $1,000.00 any more with Fury X around, but no one would also pay for Fury X with 980ti around, lol. So, Nvidia is a bastard, but they are a smart bastard.

The only way they got away with it is people were stupid enough to pay it.

 

[SolMiester]

People were too impatient to wait for the competition or price drops, and I imagine like most of those people, are happy paying the premium rather than wait.

 

[exar333]

Why are all the prices in that graph so weird? they certainly aren't MSRP, nor do they look inflation adjusted, so what's up?



Why are you celebrating being "owned", seems quite self-deprecating?

Headfoot said that AMD prices increased, which for the period in question (Kepler and onwards), the graph clearly shows to be correct.

Maybe prices bottomed-out due to a once-in-a-lifetime financial crisis and has been recovering since then?

 

[antihelten]

Maybe prices bottomed-out due to a once-in-a-lifetime financial crisis and has been recovering since then?

A once-in-a-lifetime financial crisis that somehow only really affected AMD?

We don't really have to try and guess why prices bottomed out for AMD, since we know the reason perfectly well. They produced a horribly uncompetitive chip in the R600, and then switched over to the small-die strategy which worked decently for getting market share, but somewhat poorly for generating profits.

You can read the story here

The question being discussed isn't why prices bottomed out, but rather why they spiked up again (starting with GCN and Kepler).

 

[C@mM!]

The question being discussed isn't why prices bottomed out, but rather why they spiked up again (starting with GCN and Kepler).

Driver rewrite helped with performance and ability to get drivers out for new games, allowing AMD to be more competitive with Nvidia.

By this time, Nvidia had increased pricing on small die GPU's due to the performance lead at the time. AMD simply matched Nvidia in pricepoints (usually with a 10% undercut) with a similar performing product.

 

[AtenRa]

With GCN at 28nm and Tahiti (353mm2), AMD started with a big die relatively to its 40nm Cayman HD6970 (389mm2).
Since 28nm was more expensive and having lower yields at the time vs 40nm and because HD7970 was the fastest GPU on the market made AMD to increase the price.

On the other hand, NVIDIA with Kepler and GK104 (294mm2) started with a smaller die than GF110 GTX580 (520mm2). They also had to increase the price because of the same reasons described above for AMD.
But what actually really made the big difference was the Titan. It was released one year (Feb 2013) after the GK104 for an out of the charts price of $999.
That also made the GTX 780 (May 2013) at $649 look attractive and then when AMD released Hawaii at $549* on October 2013, NV released the GTX780Ti one month later for $699. :sneaky:

GTX 780Ti was released almost two years after GTX 680. At the time NVIDIA could easily decrease its price to $500-550 and return to the GTX 580 era but it was evident they wanted to keep the high prices and that was confirmed later with the GTX 980 and 980Ti.

* AMD didnt increase the price of Hawaii above the $549 they released Tahiti in 2012

 

[antihelten]

Driver rewrite helped with performance and ability to get drivers out for new games, allowing AMD to be more competitive with Nvidia.

By this time, Nvidia had increased pricing on small die GPU's due to the performance lead at the time. AMD simply matched Nvidia in pricepoints (usually with a 10% undercut) with a similar performing product.

Sorry but this is just completely wrong. When 7970 Launched it had a price of $550 and it most certainly did not have good drivers at launch, the significant driver improvements AMD saw, didn't come until quite a bit later.

Furthermore Nvidia pricing their GK104 die (GTX 680) at $500 also came before AMDs driver improvements.

So AMDs drivers has nothing to do with the increased prices we saw for GCN and Kepler.

Lastly AMD didn't match Nvidias prices, Nvidia matched AMDs prices, seeing as 7970 launched first and 680 second. And at launch the 680 was the faster product.

With GCN at 28nm and Tahiti (353mm2), AMD started with a big die relatively to its 40nm Cayman HD6970 (389mm2).
Since 28nm was more expensive and having lower yields at the time vs 40nm and because HD7970 was the fastest GPU on the market made AMD to increase the price.

On the other hand, NVIDIA with Kepler and GK104 (294mm2) started with a smaller die than GF110 GTX580 (520mm2). They also had to increase the price because of the same reasons described above for AMD.
But what actually really made the big difference was the Titan. It was released one year (Feb 2013) after the GK104 for an out of the charts price of $999.
That also made the GTX 780 (May 2013) at $649 look attractive and then when AMD released Hawaii at $549* on October 2013, NV released the GTX780Ti one month later for $699. :sneaky:

GTX 780Ti was released almost two years after GTX 680. At the time NVIDIA could easily decrease its price to $500-550 and return to the GTX 580 era but it was evident they wanted to keep the high prices and that was confirmed later with the GTX 980 and 980Ti.

* AMD didnt increase the price of Hawaii above the $549 they released Tahiti in 2012

28nm being more expensive, might explain it to some degree, Personally I think that AMD prioritizing profits over market share played a bigger role.

Anyway because the graph Bacon1 posted annoyed me a bit (due to inexplicable errors in prices, and way too rough of a granularity in the dates, which made it look like some cards that released months apart were simultaneous), I made a new one which uses launch prices adjusted for inflation (to 2016 dollars):

 

[antihelten]

dp

 

[airfathaaaaa]

but where is titan z/x and 295x2?

 

[antihelten]

but where is titan z/x and 295x2?

I only included the "normal" flagship cards, if that makes any sense. The point of the graph is to see if there's any difference in the way Nvidia/AMD has been pricing said cards, as such including cards which have always belonged to different price segments don't make all that much sense.

 

[mohit9206]

Any idea when we might see sub 75W Pascal cards not requiring pcie connection? I doubt its going to be anytime sooner than early to mid 2017.Till then the 750ti and gtx 950 will be the only sub 75W performance cards from Nvidia till next year.

 

[Adored]

I don't think you can pick and choose which cards are "normal" flagships tbh.

5 years ago and earlier the flagships were the 580, 480, 280. Now they are the Titan, 780 Ti and Titan X, basically following a similar pattern. The 980 Ti would have been the 570, 470, 260 in years past (based on the fact that there was a faster, full GPU card costing more) and should be compared to them, not the true flagship cards.

 

[FatherMurphy]

http://videocardz.com/59009/rumor-nvidia-plans-three-gp104-skus-in-june

The latest rumor suggests that Nvidia will launch three GP104 SKUs in June that would replace the 980ti, 980, and 970. If true (grain of salt), I wonder if Nvidia would launch:

1080ti: full die, higher clocks, with GDDR5x
1080: full die, slightly lower clocks, GDDR5
1070: cut die, slightly lower clocks, GDDR5

And price the 1080ti at $650 to ensure volume is low enough to allow GDDR5x supply to build up. That would also keep up the trend of Nvidia pushing up "mid-range" die prices.

/speculation

 

[poofyhairguy]

http://videocardz.com/59009/rumor-nvidia-plans-three-gp104-skus-in-june

The latest rumor suggests that Nvidia will launch three GP104 SKUs in June that would replace the 980ti, 980, and 970. If true (grain of salt), I wonder if Nvidia would launch:

1080ti: full die, higher clocks, with GDDR5x
1080: full die, slightly lower clocks, GDDR5
1070: cut die, slightly lower clocks, GDDR5

And price the 1080ti at $650 to ensure volume is low enough to allow GDDR5x supply to build up. That would also keep up the trend of Nvidia pushing up "mid-range" die prices.

/speculation

I think it's more going to be:

1080 -full die, maybe DDR5X
1070 - cut die, DRR5
1060 ti - really cut die, DDR5