NVIDIA Announces GP102-based TITAN X with 3,584 CUDA cores

[Lays]

5 minutes ago, laminutederire said:

It is a hell of a good card anyway, most people don't have that

 

Yeah, they're paying themselves a lot too much (10 billion is a lot, and they're not that many being employed there, it isn't a company like Volkswagen in terms of labor force) and you underestimate the money they spend in marketing.

I know there are hidden costs, and those were included in the 500$. I could be less pessimistic and say it costs them 800$, but that's still an enormous margin..

(By the way nvidia is the typical American model company if you ask me, they spend a loooooot to get results and they invest even more in marketing. Just like the  US does with the CIA by the way. (Anyway, now I'll stop the hate :p))

No offense but unless you work there and you're in charge of their accounting things, you or myself have no clue what goes on there and how much they should/shouldn't charge or how much they should/shouldn't pay their employees.

[dtaflorida]

5 minutes ago, Sparviero said:

Just think of the margins on satisfying our epeen compensation al needs.

I'm not even in the market lol. I like watching ppl get their hopes up about how much bigger they think they'll be before even using the pump. Even goading them on and faking how impressed I'd be, wow it's gonna be huge, as big as 1080! 

 

And then laugh at them when they when they get the results. Polaris was hilarious. I'm kinda sad NVIDIA came so early with the new titan. At least they coulda waited till it was closer to Vega releasing, or just after.

[patrickjp93]

4 minutes ago, dtaflorida said:

I'm not even in the market lol. I like watching ppl get their hopes up about how much bigger they think they'll be before even using the pump. Even goading them on and faking how impressed I'd be, wow it's gonna be huge, as big as 1080! 

 

And then laugh at them when they when they get the results. Polaris was hilarious. I'm kinda sad NVIDIA came so early with the new titan. At least they coulda waited till it was closer to Vega releasing, or just after.

Volta will be ready for dealing with Vega.

[SteveGrabowski0]

$1200 for a fucking card that's not even the flagship of Pascal? 

 

You knew these fuckers were going to jack the price way up. Five years ago this class of card was $500.

[Prysin]

4 minutes ago, SteveGrabowski0 said:

$1200 for a fucking card that's not even the flagship of Pascal? 

 

You knew these fuckers were going to jack the price way up. Five years ago this class of card was $500.

That was "last" year and this is this year.

[Badly Browned]

$1200 for Titan X, so probably $900 MSRP for 1080ti (which means it will actually sell for $1000), bleh.

 

I guess Nvidia taking advantage not having any competition on the high end right now...

[-BirdiE-]

9 minutes ago, SteveGrabowski0 said:

$1200 for a fucking card that's not even the flagship of Pascal? 

 

You knew these fuckers were going to jack the price way up. Five years ago this class of card was $500.

The NEW Titan X is definitely the flaship of Pascal.

 

There's a certain misconception that the GP102 is lesser than the GP100. The GP102 is basically the full GP100 chip, except made for PCIe rather than NVLink (only for enterprise ATM), without the 64x compute units (useless for gaming), and with GDDR5X instead of HBM2 (only one that matters slightly). 

 

I believe they've steered clear of HBM2 because of cost and availability. They'd rather put that HBM2 into their P100s which they can sell for $12,500, rather than the Titan X which sells for $1,200. But TBH, I don't think it will make a very big difference in performance in the near future.

[VerticalDiscussions]

So, in Portugal (1600€), right? Lawl, this card is definetly not for me!

[Briggsy]

15 minutes ago, -BirdiE- said:

The NEW Titan X is definitely the flaship of Pascal.

 

There's a certain misconception that the GP102 is lesser than the GP100. The GP102 is basically the full GP100 chip, except made for PCIe rather than NVLink (only for enterprise ATM), without the 64x compute units (useless for gaming), and with GDDR5X instead of HBM2 (only one that matters slightly). 

 

I believe they've steered clear of HBM2 because of cost and availability. They'd rather put that HBM2 into their P100s which they can sell for $12,500, rather than the Titan X which sells for $1,200. But TBH, I don't think it will make a very big difference in performance in the near future.

Pretty much this ^^

 

Also, the original Titan had less Cuda cores than the 780ti. The 3840 cores of the P100 may very well end up in a Pascal TitanX Black Edition, or the 1080ti (less likely). 

 

People saying this new TitanX isn't the Pascal Flagship forget history too easily.

[-BirdiE-]

Just now, Briggsy said:

Pretty much this ^^

 

Also, the original Titan had less Cuda cores than the 780ti. The 3840 cores of the P100 may very well end up in a Pascal TitanX Black Edition, or the 1080ti (less likely). 

 

People saying this isn't the Full Pascal Flagship die forget history too easily.

I believe the P100 also has 3584 cuda cores. I don't think there's a chip out there with the full 3840 yet... Unless they've released a new version of the P100 I don't know about.

[Notional]

1 hour ago, -BirdiE- said:

Yeah. It's HBM... I mean, it has effective 4GB of VRAM, draws way more power, gets outperformed, and costs a lot more...

 

But it's better because it's HBM!

 

LAWL

 

$1200 is a lot, and you're not going to get the same price/performance as something like the lower cards.... But $1200 is pretty reasonable for this card, and it's far from unprecedented. 

Agreed, 1200$ for 1 graphics card is insane, but I don't think it's fair to compare a generation old card, 2 nodes older with a new one.

 

There is nothing reasonable about the price. These Titan cards are a straight out rip off.

1 hour ago, dtaflorida said:

I watched the fanboys being overly optimistic about the first Polaris offerings before the actual release too. And then AMD released something that got compared to the previous generation from NVIDIA. 

 

LOL I even purposely hyped it for them to see their despair when it finally came out. Saw a lot of immediate switch to getting a 1080. 

 

It's all the same to me. /sitback /popcorn

The 480 delivered on everything it promised though. Nothing beats it price to performance on the market. It equals a 980 in DX12/Vulkan for about half the price. I don't see a problem with it.

If people got a 1080 instead of a 480, that are to vastly different market segments, they must not have understood what Polaris was.

[Notional]

1 hour ago, patrickjp93 said:

Volta will be ready for dealing with Vega.

Vega will be out by Q4/16 to Q1-17. Volta is like 2018. It will be Volta vs Navi.

1 hour ago, SteveGrabowski0 said:

$1200 for a fucking card that's not even the flagship of Pascal? 

 

You knew these fuckers were going to jack the price way up. Five years ago this class of card was $500.

It is the flagship. It's the highest end GP chip. P100 is a compute card for servers.

 

But yeah, NVidia has been pumping up the prices for years.

[VagabondWraith]

1 hour ago, patrickjp93 said:

Volta will be ready for dealing with Vega.

Yep, methinks too. I'll hold out till income tax time for a new G-Sync 1440/4K monitor with the hopes Volta is ready by spring/summer 2017. Titan is intriguing but the 1080 is plenty enough in the meantime.

[VagabondWraith]

1 hour ago, Badly Browned said:

$1200 for Titan X, so probably $900 MSRP for 1080ti (which means it will actually sell for $1000), bleh.

 

I guess Nvidia taking advantage not having any competition on the high end right now...

Jen-Hsun once said that competition determines the price. We can see the effect of a beaten and bruised AMD does to the industry. They deserve blame. If they didn't have a clown show making their shitty business decisions we may just have better prices.

 

Damn I wish AMD would dissolve over to Intel and give nVidia a real swift kick in the johnson...

[-BirdiE-]

1 minute ago, Notional said:

Agreed, 1200$ for 1 graphics card is insane, but I don't think it's fair to compare a generation old card, 2 nodes older with a new one.

 

There is nothing reasonable about the price. These Titan cards are a straight out rip off.

I don't think $1,200 is unreasonable for that card. As I mentioned before (in another comment), when you get up into the absolute top consumer GPUs, you should be expecting significant diminishing returns on performance per dollar. If anything this one is priced under the curve you'd expect. About the same going from a 1080 to this (200% price, 150% performance), as a 1070 to the 1080 (157% price, 120% performance).

 

Normally I wouldn't bring up an older generation card... The only reason I bring up the Pro Duo is because they released it only a couple months before Pascal and Polaris, and priced it at $1500. This looks like a serious bargain compared to that... And don't even get me started on the Titan Z... Stupidest card ever made.

 

With all that being said... I'm not saying it makes any sense for most consumers to buy it... I'm just saying the pricing makes sense, economically, for what you get. And the pricing is far from unprecedented from either vendor, AMD or Nvidia.

[Notional]

1 minute ago, -BirdiE- said:

I don't think $1,200 is unreasonable for that card. As I mentioned before (in another comment), when you get up into the absolute top consumer GPUs, you should be expecting significant diminishing returns on performance per dollar. If anything this one is priced under the curve you'd expect. About the same going from a 1080 to this (200% price, 150% performance), as a 1070 to the 1080 (157% price, 120% performance).

 

Normally I wouldn't bring up an older generation card... The only reason I bring up the Pro Duo is because they released it only a couple months before Pascal and Polaris, and priced it at $1500. This looks like a serious bargain compared to that... And don't even get me started on the Titan Z... Stupidest card ever made.

 

With all that being said... I'm not saying it makes any sense for most consumers to buy it... I'm just saying the pricing makes sense, economically, for what you get. And the pricing is far from unprecedented from either vendor, AMD or Nvidia.

Sure price to performance falls as you move up the performance ladder. But cards costing over 700€ should not be a thing. There simply is not any reason for it, other than ripoff. Why do you think NVidia sells these Titan cards before launching the x80ti version, that instantly makes the Titan card pointless? Blatant cash grab.

 

Yeah but the Pro Duo was never made a gaming card, but rather a professional card for devs. Compared to a Firepro card, it's a steal. Titan Z... even at half price they couldn't get anyone to buy them. Well it had the Titan name for a reason.

 

NVidia are excellent at pricing and market manipulation, I'll give them that. Just sad that people are so easily manipulated.

[Daegun]

2 hours ago, -BirdiE- said:

 

 

Yeah. It's HBM... I mean, it has effective 4GB of VRAM, draws way more power, gets outperformed, and costs a lot more...

 

But it's better because it's HBM!

 

LAWL

 

 

$1200 is a lot, and you're not going to get the same price/performance as something like the lower cards.... But $1200 is pretty reasonable for this card, and it's far from unprecedented. 

Wait a minute here. Not jumping on the amd fanboy train here, but the only thing you got right about hbm there was the cost. It both draws less power and performs better than GDDR5.

[dtaflorida]

17 minutes ago, VagabondWraith said:

I'll hold out till income tax time for a new G-Sync 1440/4K monitor with the hopes Volta is ready by spring/summer 2017

You can't lose by waiting. Right after 480 released I saw a lot of ppl wishing they hadn't gotten a freesync monitor so they could get a 1080. Now they wait for Vega, or try to sell their monitor and buy a 1080. 

[-BirdiE-]

2 minutes ago, Notional said:

But cards costing over 700€ should not be a thing. There simply is not any reason for it, other than ripoff.

I feel like you're just taking your use case and applying it to everyone.

 

I actually need quite a bit of GPU power to run my 3440x1440 monitor at 144Hz. Not to mention anyone spending a bunch on fancy 4K panels in the near future, or wants to make a SFF 4K gaming PC.

 

And as for the 1080ti, I showed in a previous comment that the price of the new Titan X actually makes sense when you factor in the cost of graphics over time.

 

And can the new Titan X not serve the same purpose as the Pro Duo did? The Titan series has always been marketed at a prosumer GPU. One that can be used for professional purposes, but also be gamed on.

 

Please try and see things objectively, instead of making excuses for AMD, and trying to shit on Nvidia.

[-BirdiE-]

7 minutes ago, Daegun said:

Wait a minute here. Not jumping on the amd fanboy train here, but the only thing you got right about hbm there was the cost. It both draws less power and performs better than GDDR5.

I was talking about the card as a whole.

 

HBM does not make or break a card. I'd rather have a better performing card, than a slower card just because it has HBM... is the point I was trying to make.

 

All I care about is the performance I get when I use it.

[Marinatall_Ironside]

5 minutes ago, -BirdiE- said:

And can the new Titan X not serve the same purpose as the Pro Duo did? The Titan series has always been marketed at a prosumer GPU. One that can be used for professional purposes, but also be gamed on.

True! I do gaming (would love to do 4K gaming), and 3D rendering, animation, and modeling as part of my game development major. So me needing this much GPU horsepower is justifiable when I subject it to GPU-based/accelerated rendering (Lookin' at Nvidia Iray here)

 

With two of these obscene bastards, we're looking at 7,168 CUDA cores. That would translate to extremely, extremely fast GPU rendering in Nvidia Iray. I am not so sure if adding 12 threads out of my 5930K CPU will even accelerate the rendering any faster. I would have to investigate further into that.

[laminutederire]

2 hours ago, Lays said:

No offense but unless you work there and you're in charge of their accounting things, you or myself have no clue what goes on there and how much they should/shouldn't charge or how much they should/shouldn't pay their employees.

It's pretty obvious they won't pay everyone 50 000$/month

[Notional]

Just now, -BirdiE- said:

I feel like you're just taking your use case and applying it to everyone.

 

I actually need quite a bit of GPU power to run my 3440x1440 monitor at 144Hz. Not to mention anyone spending a bunch on fancy 4K panels in the near future, or wants to make a SFF 4K gaming PC.

 

And as for the 1080ti, I showed in a previous comment that the price of the new Titan X actually makes sense when you factor in the cost of graphics over time.

 

And can the new Titan X not serve the same purpose as the Pro Duo did? The Titan series has always been marketed at a prosumer GPU. One that can be used for professional purposes, but also be gamed on.

 

Please try and see things objectively, instead of making excuses for AMD, and trying to shit on Nvidia.

No ones talking about performance. There is no upper limit for performance needed. But price does not necessarily reflect on the cost of the product. Don't fool yourself, the profit on these GP100 and GP102 chips are very very high, even with low yields.

 

The cost of graphics over time? What does that even mean? Those prices are dictated by AMD and NVidia, with little influence on foundry cost. NVidia has been constantly pumping up prices for years now, 300$ for an x60 NVidia card? Wtf.

 

Idk if it can. Not sure about the Fiji's double precision capablities, but the Titans are GTX cards, so they are branded as gaming cards, and use gaming drivers. Duo Pro can also use FirePro drivers, as well as gaming drivers, which makes them capable of things a Titan cannot.

 

I am being very objective. I'm just looking at things from a consumer perspective instead of a companies perspective.

[i_build_nanosuits]

6 hours ago, DXMember said:

you've been here for like 2 hours already arguing with some kid on the web about who's having a better gaming experience with his hardware....

 

http://url.hwbot.org/1wnhwS4

https://www.overclockers.at/downloads/GPUPI 2.3.4.zip

@Pohernori @DXMember

http://www.3dmark.com/3dm/13521837?

how about that? 1440p

(putting money aside for an MSI GTX 1080ti Gaming Z)

[-BirdiE-]

13 minutes ago, Notional said:

No ones talking about performance. There is no upper limit for performance needed. But price does not necessarily reflect on the cost of the product. Don't fool yourself, the profit on these GP100 and GP102 chips are very very high, even with low yields.

Please. Enlighten us with your knowledge of their cost to produce... Don't forget to include all of manufacturing, R&D, and distribution.

Based on market prices of GPUs, the price for that card makes sense.

 

13 minutes ago, Notional said:

The cost of graphics over time? What does that even mean? Those prices are dictated by AMD and NVidia, with little influence on foundry cost. NVidia has been constantly pumping up prices for years now, 300$ for an x60 NVidia card? Wtf.

This was in reference to your comment about them later releasing the 1080ti for less. This is what I wrote earlier to someone...

4 hours ago, -BirdiE- said:

Let's look at a scenario. We'll assume that Volta comes out 1 year after Pascal, and the 1080ti comes out in 4 months at $899.

 

So I buy the NEW Titan X for $1199. Four months later a similar performing 1080ti comes out for $899. That means I spent $300 to have that performance 4 months earlier (75/mo). Then, the similarly performing 1170 comes out 6 months later for $449. That means you just spent $450 to have that performance 6 months early (75/mo).

 

13 minutes ago, Notional said:

Idk if it can. Not sure about the Fiji's double precision capablities, but the Titans are GTX cards, so they are branded as gaming cards, and use gaming drivers. Duo Pro can also use FirePro drivers, as well as gaming drivers, which makes them capable of things a Titan cannot.

So if you don't know, why are you speaking to it? Does the Titan need different drivers to be able to do these things? They are branded under the Geforce logo because it's intended for regular consumers, not enterprise. "Prosumers". If you'll look above, you'll see someone give a perfect example of a use case.

18 minutes ago, JurunceNK said:

True! I do gaming (would love to do 4K gaming), and 3D rendering, animation, and modeling as part of my game development major. So me needing this much GPU horsepower is justifiable when I subject it to GPU-based/accelerated rendering (Lookin' at Nvidia Iray here)

 

With two of these obscene bastards, we're looking at 7,168 CUDA cores. That would translate to extremely, extremely fast GPU rendering in Nvidia Iray. I am not so sure if adding 12 threads out of my 5930K CPU will even accelerate the rendering any faster. I would have to investigate further into that.

 

13 minutes ago, Notional said:

I am being very objective. I'm just looking at things from a consumer perspective instead of a companies perspective.

You're not... You're giving baseless opinions to support your pro AMD stance. If you have any actual numbers or factual arguments, let me know... But as it stands right now I've shown that it does have a valid target market, the price is not unprecedented, and based on the GPU market the pricing is actually lower than would be expected for that performance.