Nvidia GeForce GTX 1080 announced: Faster than two GTX 980s in SLI

Faster at what? can we stop throwing numbers willy nilly?
I would imagine, since the P in GPU stands for processing, it runs through processes faster.

He's referring to specific rendering tasks, render benchmarks etc etc

FurMark based stress test (OpenGL 2.1 or 3.2).
TessMark based tessellation test (OpenGL 4.0).
GiMark: geometry instancing test (OpenGL 3.3).
PixMark Piano pixel shader test (OpenGL 2.1 or 3.2).
PixMark Volplosion pixel shader test (OpenGL 2.1 or 3.2).
Plot3D vertex shader test (OpenGL 2.1 or 3.2).
Triangle one of the most simple 3D scene ever made... (OpenGL 2.1 or 3.2).
 
Faster at what? can we stop throwing numbers willy nilly?

I know, right? Saying it is faster than 2 980s in SLI actually contradicts the "substantially faster than a TitanX, because that's the same as stating that the 1080 is only 20-25% faster than a single 980, on the optimistic side. Why? Because two 980's in SLI is not 200% the power, vs, a single 980...in fact, most SLI games you are lucky to get 25% better performance, if SLI is supported at all...why is this? Because instead of like the old days of SLI, one card is not drawing one portion of the scene, while any second, third or fourth card draws the rest. Instead, one card draws one scene, and the others draw the next. So, when one card is outdated an unable to run games at full settings(doesn't take but less than a year), all the rest of your cards are outdated, too.
 
"This technology has two huge benefits: no more fish-eye lens effect in multi-monitor gaming environments"
WOW, did -I- ever miss a memo..

"and $699 for "Founder's Edition", although it's not completely clear what benefits.."
Unique and "satisfying" easter-eggs as well as a unique 11.2B color prismatic LED logo that shows "FOUNDER" in an orgasmic font through your clear case panel, and bright enough that it projects on a wall up to 4 meters distant.

I am -amazed- that they didn't turn this into a Titan-line product. It ain't cheap, but NI-ice upgrade in one generation (if it lives up to, an' all that...)
 
...in fact, most SLI games you are lucky to get 25% better performance, if SLI is supported at all...why is this?
You complain about throwing numbers out - where did you get this one? Most of the games I've played supported SLI and in those I can't think of any that only gave a 25% boost. Most often I get 50-80% if I'm not already bumping up against some sort of frame cap.
 
Of course it's faster, SLI is often broken.

That's what I was thinking. Too many games using effects that don't work on alternate frame rendering. I think that's the worst comparison Nvidia could have made because they could have simply have taken their numbers from a game with little to no sli scaling. I'd much prefer straightforward comparisons.

It looks like I was right about Pascal using more power overall than Maxwell though.

Can anyone explain Nvidia's overclocking claims? A card consuming more wattage produces more heat, thus a pascal card will produce more than it's maxwell equivalent resulting in an overclocking bottleneck. For all we know this could very well be just like when AMD claimed the R9 290x could overclock to a crazy high level but when we got the cards no one could get there due to heat issues. I wouldn't be surprised if Nvidia only got those results using liquid nitrogen with a cherry picked card.
 
...in fact, most SLI games you are lucky to get 25% better performance, if SLI is supported at all...why is this?
You complain about throwing numbers out - where did you get this one? Most of the games I've played supported SLI and in those I can't think of any that only gave a 25% boost. Most often I get 50-80% if I'm not already bumping up against some sort of frame cap.


Nvidia says otherwise

http://www.geforce.com/games-applic...itle=&sort_bef_combine= &sort_order=&sort_by=

Nvidia's official list of SLI games. Some of those are tech demos, not even games and that list is every SLI game ever.

He's not wrong about the performance numbers either. Many effects in modern games simply don't work well with alternate frame rendering. Only games that Nvidia partners with seem to get decent SLI performance with. Most other games get it enabled in a patch wherein the performance is 50% at best. A good chunk of modern games simply don't support it, even at the engine level like Just Cause 3.
 
That's what I was thinking. Too many games using effects that don't work on alternate frame rendering. I think that's the worst comparison Nvidia could have made because they could have simply have taken their numbers from a game with little to no sli scaling. I'd much prefer straightforward comparisons.
It's marketing. It is convenient to leave out inconvenient details. It is also par for the course that this is an industry wide issue - why this is suddenly a deal breaker when AMD touted framerate as the centerpiece of its VR Radeon Pro Duo a couple of weeks ago only to conveniently forget that frame pacing has taken a turn for the worse lately is quite frankly beyond me
GTAV_3840x2160_PLOT.png


It looks like I was right about Pascal using more power overall than Maxwell though.
I would hold off judgement if I were you until the actual numbers are in. You've just pilloried Nvidia's nebulous claims, and are now taking their 180W board power limit as being indicative of actual usage. An example of things not necessarily being created equal might be Nvidia's board power for the GTX 780 Ti and 980 Ti is the same at 250W - or the fact that the GTX 970 has a board power of 165W while the GTX 980 has a board power of 180W, yet the 970 generally consumes as much if not more than the 980 thanks to the card holding boost clocks more consistently
power_peak.png

Can anyone explain Nvidia's overclocking claims? A card consuming more wattage produces more heat, thus a pascal card will produce more than it's maxwell equivalent resulting in an overclocking bottleneck.
So you again assuming that actual consumption is higher for Pascal, and that the new cooler has the same cooling characteristics (including airflow and fan profile) as the Maxwell cards, and that the GPU workload is identical under what are very likely different workload (game) scenarios? Sounds faulty to me. By all indications, the GTX 1080 (180W) is faster than the 980 Ti (250W), and even if you were comparing apples-to-apples, do you really think that a GTX 1080 will use more power than a GTX 980 in an identical situation. I think I'll take that bet since it will probably pay off as soon as a comparison benchmark is run in a CPU/framerate limited scenario.
I wouldn't be surprised if Nvidia only got those results using liquid nitrogen
...and there it is! he shoots...he trolls!. The demo system was clearly shown during both the presentation wrap up, the Doom Vulkan demo, and the in-depth seminar today.......Does this look like liquid nitrogen?
oX7mGZt.jpg

How about a friendly wager? If the GTX 1080 requires LN2 to reach 2.1GHz I'll donate a token $US100 via PayPal to the registered charity of your nomination. If the GTX 1080 is capable of reaching 2.1GHz with conventional (non sub-zero) cooling, you donate the same amount to my nominated registered charity.
with a cherry picked card.
No doubt it wasn't a dud, but I seem to remember the same hopefulness and derision regarding Nvidia's Titan X and 980 Ti demonstration about how there was no way retail cards would hit 1400MHz+......yet there seems to a be veritable swathe of owners pushing that and more with air cooled cards.
If you need something to hold on to desperately ameliorate the impact - and you do seem desperate to find something to denigrate , I suggest Pascal might be like Maxwell in that Nvidia will make the board power limit hardware locked, so air and water cooling will probably have a hard limit (+/- a small variance) unless the owner sources a sub-zero BIOS.
 
Last edited:
Oh you gotta be kidding me!!!!! I gotta pay more attention to new gpu news. I just shelled out $650 for a EVGA 980Ti. Darnit.

Don't worry. People will rush out and buy the 1080, then a month later the Titan Whatever will come out and then the 1080Ti. Like they always do.
 
Either way I'm anxiously waiting for benchmarks of the 1060 as well as the 1070 and 1080. More so than I normally am during a card release.
 
Either way I'm anxiously waiting for benchmarks of the 1060 as well as the 1070 and 1080. More so than I normally am during a card release.
If the rumour mill is to be believed, you might get your GTX 1080 questions answered on 17th May.
It actually wouldn't surprise me if Nvidia carved a third SKU out of GP104 ( GTX 1060 Ti) to combat AMD's Polaris since it looks likely that AMD's offerings might fall short of the 1080 (and maybe the 1070 for the markets AMD is pushing - multi-monitor/4K/VR).
Still a little surprised the GP106 hasn't made an appearance. It would seem like an ideal lead product for gaming on a more modest budget. Maybe they are waiting on the OEMs laptop refresh cycle and some firm info on AMD's Polaris 11.
 
So wait 5 months? That way you will be getting the card that has the bugs ironed out, the full vram rather than the card missing a GB at the top end.

No one has patience anymore. Definitely would consider updating my 970 to one of these once I figure out what monitor will go nicely with it
Believe me I was thinking the same thing but I'll wait to see the price point and actual real world usage
 
So I know this conversation is all about GTX 1080, but wouldn't this be a good time to look at cards from last year to see if you can pick one up for a steep discount? I'm thinking 2 x 980ti for like the price of 1??? Just a thought.
 
Oh you gotta be kidding me!!!!! I gotta pay more attention to new gpu news. I just shelled out $650 for a EVGA 980Ti. Darnit.

Every year they bring out a new card like 30% faster. Every now and then this happens, a card that is double (100%) faster, in this case twice as fast as a 980. I shelled out like $800 canadian last year. If I had wait one year I would have gotten a 1080 which is twice as fast. Moore's Law at it's finest. They invested billions into this venture I hear.
 
So I know this conversation is all about GTX 1080, but wouldn't this be a good time to look at cards from last year to see if you can pick one up for a steep discount? I'm thinking 2 x 980ti for like the price of 1??? Just a thought.
Oh, don't worry about that, there are many people waiting for exactly the right time for prices to bottom out on the 980 Ti and Titan X. eBay is usually a feeding frenzy a week before new cards drop - once the auctions close and the buyers have their Paypal funds they can immediately divert the cash to their pre-ordered new card.

In this particular instance, new 980Ti prices haven't really moved too much as of yet ( Zotac's AMP is $540 at Newegg but seems more like the exception than the rule) and I'm not sure that prices will plunge when the card goes on sale on the 27th May - I think many people will wait a couple of weeks until the high clocked AIB cards arrive - Gigabyte is teasing the Xtreme Gaming (and I don't think the other large OEM/ODMs are far behind
GALAXY-GeForce-GTX-1080-XTREME-GAMING.jpg
 
Back