[Various] NVIDIA GeForce GTX 1080 Review Thread

[railven]

How much would a non reference GTX 1080 cost? More than Nvidia Founders Edition?

No one knows. EDIT: Personally hoping they are cheaper than Founder's (woof my broken English showed!)

Find out hopefully soon and not "months" from now.

 

[Qwertilot]

Suspect they'll want HBM2 for big Pascal. Those things will want to really push the VR/4k performance and it'll be a big boost in those terms/also some competition from Vega of course.

Maybe they'll manage to keep it until Volta though

 

[thesmokingman]

http://www.overclock-and-game.com/news/pc-gaming/46-gtx-1080-what-s-not-being-discussed

Interesting read on the GTX1080 hype and benchmarks.

Wasn't Nvidia claiming it was faster then SLI 980s?

Well they didn't waste time mincing words cut right to the chase. :hmm:

 

[n0x1ous]

GP100 will never make it into a consumer product. Clock 1080 higher and you get about the same SP FLOPS in half the die area.

We've heard this before about Nvidias big die. Eventually they end up releasing a consumer version. They ALWAYS have; when that happens, probably depends on AMD

 

[AnandThenMan]

We've heard this before about Nvidias big die. Eventually they end up releasing a consumer version. They ALWAYS have; when that happens, probably depends on AMD

Nvidia will definitely put big Pascal into a consumer card, like you said depends on AMD when it happens.

 

[Arachnotronic]

Nvidia will definitely put big Pascal into a consumer card, like you said depends on AMD when it happens.

Not GP100. GP102.

 

[Mopetar]

Nvidia will definitely put big Pascal into a consumer card, like you said depends on AMD when it happens.

I think it depends more on their yields. GP100 is a massive chip and they already announced that P100 is disabling 4 of the SMs so there is no uncut GP100 at this point. Even if yields are good as long as NVidia can sell P100's for $10K+ a crack, they have no reason to sell a consumer version unless they're saving GP100 dies that could still work with 8-12 disabled SMs.

A few people have stated that they don't think GP100 will be a consumer chip at all due to it being designed for HPC applications and having a lot of space used for FP64 performance that isn't useful for gaming. Those people tend to believe that the big Pascal for the consumer space will be GP102.

Nvidia might even be waiting on that decision until they see what AMD is doing in the market.

 

[know of fence]

The reviews need to compare the 1080 to the 680, because both are the first Geforce GPUs on a new node, and both are about the same die size ~300mm². Except it's verboten, because the 680 launching after AMDs 7970 had a competitive MSRP of 499 USD.
With the strong dollar asking to pay 600 or 700 of those for a mid range video game card is pretty outrageous. Or in other words paying so much for a 20-30% performance bump over the 980 ti (650$) would be pretty [redacted] stupid.

http://www.anandtech.com/bench/product/1714?vs=1715


No profanity in VC&G. Keep it professional.

-Elfear

 

[Arachnotronic]

The reviews need to compare the 1080 to the 680, because both are the first Geforce GPUs on a new node, and both are about the same die size ~300mm². Except it's verboten, because the 680 launching after AMDs 7970 had a competitive MSRP of 499 USD.
With the strong dollar asking to pay 600 or 700 of those for a mid range video game card is pretty outrageous. Or in other words paying so much for a 20-30% performance bump over the 980 ti (650$) would be pretty [redacted] stupid.

http://www.anandtech.com/bench/product/1714?vs=1715

A 20-30% bump for the same price or possibly 7.7% more price is a bad deal? Strange...

 

[Timmah!]

A 20-30% bump for the same price or possibly 7.7% more price is a bad deal? Strange...

Honestly, it is. We are talking here about something with a 700 USD/800 EUROs price right now. Granted, the value of money is different to various people, if you have enough them to burn, then probably it would be a good deal to you, even if the performance bump was 5 percent, just based on the fact its newest and fastest thing, which you want. To me, and majority people though around the world, as i suppose, its still quite a lot and difficult to justify to spend (when already owning 980Ti).

The reasons are simple. If you look at those gaming benchmarks, what are those cca 25 additional percent of performance bring to you? Something like Doom is now 125 FPS instead of 100 on 980Ti, what difference does that make? Unless very specific case, when 120FPS is some theoretical limit, you want to get over, cause of 120MHz display or VR, none. Ashes of Singularity is then 50FPS instead of 40 - once again, how does that even help? At 40 FPS, its already 10FPS above the 30FPS minimum, yet still 10FPS far away from at least 60FPS optimum....while its better, it does not really solve anything IMO.

And then some GPGPU performance, at those 125 percent of 980Ti, it would take me to render single picture cca 105 minutes instead of 2 hours (since at 200 percent of 980Ti it would take one hour). Under assumption thats 2 hours is the time needed to render that single pic. So i save 15 minutes. I need to render 8 pictures to save 2 hours. If i already own 980Ti (i dont), are those 2 hours of my life saved worth 800 EUROs? The answer is no, it would be, if only i produced those pictures en masse. Obviously in such case i would earn enough money to make the price more acceptable from my POV. But as you can see, even if the card is making you the money, its not that straight-forward YES!, thats good deal.

That said, as someone with gtx580 in my case right now, which i expect to be about 3x slower for my needs, i am probably gonna buy it. Unless some Titan coming already this year kind of rumors pop-up, which would make me reconsider.

My 2 cents.

 

[sontin]

The GTX1080 is not the successor of the GTX980TI. The card is marketed towards people who own a GTX980 or GTX780TI.

 

[Det0x]

The GTX1080 is not the successor of the GTX980TI. The card is market towards people who own a GTX980 or GTX780TI.

But this is a enthusiast forum.. Iam pretty sure there are more 980ti owners then 980 owners around here
For us this card is simply not worth it or anything special, especially with that "craftsmanship" cooler on the founder edition cards.

 

[SimianR]

The GTX1080 is not the successor of the GTX980TI. The card is market towards people who own a GTX980 or GTX780TI.

The GTX 1080 might not be the true "successor" to the 980 Ti, but that doesn't mean they aren't targeting 980 Ti owners, since it will offer a modest performance bump. And then in 6 months they'll release the 1080 Ti and target the 1080 (non-ti) owners (again). I'm sure that must feel great.

 

[Bacon1]

The GTX1080 is not the successor of the GTX980TI. The card is marketed towards people who own a GTX980 or GTX780TI.

Funny, they priced it at 980 TI levels not 980...


980 = $550 @ launch and has been $500 for a long time now.

1080 = $700 @ launch and will maybe get to $600 in a few months. I bet most custom cooled cards will stick around $700 or even higher.

So why is the successor to your card priced so much higher than it?

 

[sontin]

Funny, they priced it at 980 TI levels not 980...


980 = $550 @ launch and has been $500 for a long time now.

1080 = $700 @ launch and will maybe get to $600 in a few months. I bet most custom cooled cards will stick around $700 or even higher.

So why is the successor to your card priced so much higher than it?

Why shouldnt it? The GTX1080FE offers >70% more performance for a 27% higher MSRP. It is the same deal nVidia had offered with the GTX980 which was 60% faster for a 10% increase over the GTX680.

 

[know of fence]

A 20-30% bump for the same price or possibly 7.7% more price is a bad deal? Strange...

Well it's a fair question how much should a performance increase cost. Across a the same generation the prices pretty much are exactly proportional to the number of shader cores. Basically a card with twice as powerful specs (shader cores) card costs twice as much. In practice there are diminishing returns, both objective like (bandwidth limit, refresh rate, visual acuity) and subjective like perceiving no noticeable improvement from added resolution or frame rate.
Over time however thanks to Moore's law improvements are basically free, in return companies get to sell new hardware to customers every so often because it became twice or 4 times as fast as before. It's not unreasonable to expect a mid range card to stay mid range, maybe adjust for inflation.

If price per transistor keeps decreasing thanks to node shrinks and node age, hardware sellers need either to increase their core count for the same price or lower the price for the 7 billion xtor chips.

It's possible that the 700$ price for a 500$ card is a result of an expensive process, but it may very well be a result of a deceptive and exploitative strategy, a lack of competition and an absolute killer app in VR, which guarantees almost unlimited demand.

 

[Bacon1]

Why shouldnt it? The GTX1080FE offers >70% more performance for a 27% higher MSRP. It is the same deal nVidia had offered with the GTX980 which was 60% faster for a 10% increase over the GTX680.

By your own numbers, 10% increase in price for 60% gain.

So a 20% increase should give 120% more fps?

27% increase should then give 162% more fps?

My point was, they are charging more than the current 980 TI, so why would you not compare it to the 980 TI?

 

[wingman04]

But this is a enthusiast forum.. Iam pretty sure there are more 980ti owners then 980 owners around here
For us this card is simply not worth it or anything special, especially with that "craftsmanship" cooler on the founder edition cards.

I agree with you, I'm a enthusiast and will probably wait till the price goes down on the GTX 1070.

 

[Arachnotronic]

By your own numbers, 10% increase in price for 60% gain.

So a 20% increase should give 120% more fps?

27% increase should then give 162% more fps?

My point was, they are charging more than the current 980 TI, so why would you not compare it to the 980 TI?

I would say that the GTX 980 does not have a true "replacement" nor does the 980 Ti. The product stack has been fundamentally changed.

GTX 1080 MSRP is $599, which is exactly $50 away from where the 980 launched as well as $50 away from where the 980 Ti launched. It is a "replacement" for both of these products in the marketplace.

As far as the upgrades go, I don't see a lot of 980 Ti owners dropping their cards for the 1080, especially since most of the models are aftermarket and faster than stock.

1080 will appeal to those running 980s, 780 Ti/OG Titan/Titan Black class cards, and I think the upgrade proposition there is quite compelling.

To get the 980 Ti users to upgrade, NVIDIA will need a "GM200" style Pascal card, likely GP102.

 

[sirmo]

1080 will appeal to those running 980s, 780 Ti/OG Titan/Titan Black class cards, and I think the upgrade proposition there is quite compelling.

I run a 980 currently. I like to keep my office cool and when it first came out I was mostly impressed by perf/watt. I do feel like I overpaid for it though, but oh well it served me well for a year and a half now.

1080 doesn't excite me in the slightest, I am going back to the team red. FreeSync support, and Async Compute is where it's at. Even if 1080 is 30% faster than anything else AMD comes out FreeSync alone makes up for it.

And I get a feeling Vega is going to crush 1080 especially in newer games leveraging new APIs.

 

[antihelten]

By your own numbers, 10% increase in price for 60% gain.

So a 20% increase should give 120% more fps?

27% increase should then give 162% more fps?

My point was, they are charging more than the current 980 TI, so why would you not compare it to the 980 TI?

Thats not how the math works out at all.

You want to maintain a constant increase in perf/$. 10% increase in price with a 60% gain in performance equals a 45% improvement in perf/$ (1.6/1.1=1.45).

To maintain the 45% improvement in perf/$ with a 20% price hike you would need a 75% increase in performance (1.75/1.2=1.45).

To maintain the 45% improvement in perf/$ with a 27% price hike you would need an 85% increase in performance (1.85/1.27=1.45).

For what it's worth, at the $600 price point the 1080 brings a 48% increase in perf/$, so slightly better than the 680->980 jump.

As long as you don't go for the $700 founders edition, the 1080 is a decent enough jump in perf/$

 

[CFP]

A 20-30% bump for the same price or possibly 7.7% more price is a bad deal? Strange...

Except it's not really a 20~30% bump when you consider that most 980ti's hit 1400mhz easy. That's like, what, a 7% ish performance bump, especially when you take the 1080 throttling into account.

 

[AtenRa]

Why shouldnt it? The GTX1080FE offers >70% more performance for a 27% higher MSRP. It is the same deal nVidia had offered with the GTX980 which was 60% faster for a 10% increase over the GTX680.

And GTX1080Ti will be 50-60% faster than TITAN X and will be priced 27% higher at $1300. Why shouldn't ???

Then GTX 2080 will be 20-30% faster than GTX 1080Ti at 27-30% higher MSRP at $1690.

Do you want me to continue ???

 

[sam_816]

Except it's not really a 20~30% bump when you consider that most 980ti's hit 1400mhz easy. That's like, what, a 7% ish performance bump, especially when you take the 1080 throttling into account.

And used 980ti xtreme by gigabyte & 980ti hybrid by evga are going for about 500-530 on ebay now.

Xtreme series cards can hit 1500 clocks has 8+8+6 power delivery system, quite/effective cooler n looks as well.

For 699 FE is not so attractive.

This 599 Msrp is a sham & unless the reaction of internet forces them to launch 599 version I don't think there will be any non-gimped 599 version before AMD 's Polaris cards launch.

I think 1070 coming at its msrp when 480x launches & if it is competitive. Nvidia will quickly drop prices to cut AMD sales and anyone who buys the card before that will pay premium.

 

[raghu78]

https://www.youtube.com/watch?v=bNCfn4y8dBw

nvidia guys being questioned by the press on FE pricing. fun stuff.