Nvidia GeForce GTX 1080 reviews and benchmarks

Why not? Custom 1080s might OC a lot. I have a 980 Ti and i need every last drop of performance for 4K. If the heavily OC'd 1080 beats the OC'd 980 Ti I will seriously consider the upgrade.

Seems to be even more of a lottery than before in securing a decent one. The GamersNexus test with the custom AIO here, shows 10% increases at 4K. That's with a clock speed of 2164MHz though, so quite a bump.

https://www.youtube.com/watch?v=w-3fi1ovAP0

http://www.gamersnexus.net/guides/2441-diy-gtx-1080-hybrid-thermals-100-percent-lower-higher-oc-room

It's not clear though, whether he is comparing it with a throttled FE card though.
 
So the 1080 went up for pre-order by surprise. That's annoying. But it's the Founders. So I don't care. When do the base models go up? I've heard next Friday. I'll be ready for that one for sure. How about in stores themselves? Will I be able to walk into Best Buy one day and grab one? My area is decently low income so I doubt anyone will even get one, heh. New PC buyer so this is all foreign to me. I want to assume nothing is different but assuming got me here.
 
Can someone please confirm that these reference cards are not nearly getting as much as they could be, and I should wait for AIB cards with more juice for OC
 
Can someone please confirm that these reference cards are not nearly getting as much as they could be, and I should wait for AIB cards with more juice for OC

If you are looking to push OC, I would wait. There will be cards that have an additional input for more power. Nvidia built in the area for the port on the board, it is just a matter of time a AIB partner makes a card that will be 8+6 pin or 8+8 pin that will deliver a lot more voltage for pushing the clocks.
 
If you are looking to push OC, I would wait. There will be cards that have an additional input for more power. Nvidia built in the area for the port on the board, it is just a matter of time a AIB partner makes a card that will be 8+6 pin or 8+8 pin that will deliver a lot more voltage for pushing the clocks.

Okay, this is EXACTLY what I wanted to hear

gamersnexus put a 1080 on watercooling and weren't able to squeeze any extra performance out of it, I'm assuming because the card could only take so much power

https://www.youtube.com/watch?v=w-3fi1ovAP0
 
Also, if you game in the same room as your PC is placed, never go for blower.
Unless you like your PC sounding like a vacuum cleaner mating with a taking off airplane.

I have 2 reference 980's in SLI and previously used a 680 with blower. I've never found this to be the case in any way. I don't even notice when they spin up from idle. I would guess if you used custom fan profiles for overclocking maybe?
 
Okay, this is EXACTLY what I wanted to hear

gamersnexus put a 1080 on watercooling and weren't able to squeeze any extra performance out of it, I'm assuming because the card could only take so much power

https://www.youtube.com/watch?v=w-3fi1ovAP0

One thing you have to remember, when AIB partners make boards that can deliver more power, the cooling method will need to be really good to keep the card from throttling at 82c. Also, when they add more power, they will add better voltage regulators that can take the added power. How much voltage the chip can take is unknown, but there has be a reason Nvidia designed the board with the ability to add another 8pin power port.

It definitely needs more voltage to see how far this chip can go, but until someone makes a board or someone mods an existing board to be able to deliver more power we wont know for sure. I think with more power, we should be able to hit 2500ghz or more with proper cooling.
 
Comparing die sizes divorced from the maturity of their respective manufacturing processes doesn't make all that much sense. A smaller die on a cutting edge process might be as expensive as or even more expensive than a larger die on a mature process.

I swear I read Nvidia saying die shrinkage has reached diminished returns and going smaller may end up costing more now. Or at least not really save anything anymore.
 
One thing you have to remember, when AIB partners make boards that can deliver more power, the cooling method will need to be really good to keep the card from throttling at 82c. Also, when they add more power, they will add better voltage regulators that can take the added power. How much voltage the chip can take is unknown, but there has be a reason Nvidia designed the board with the ability to add another 8pin power port.

It definitely needs more voltage to see how far this chip can go, but until someone makes a board or someone mods an existing board to be able to deliver more power we wont know for sure. I think with more power, we should be able to hit 2500ghz or more with proper cooling.

What kind of fps returns would you expect to see at 2500ghz? It'd definitely be worth the price at that point, i'd imagine.
 
I swear I read Nvidia saying die shrinkage has reached diminished returns and going smaller may end up costing more now. Or at least not really save anything anymore.

Yes, it's also been said that as fabrication nodes shrink, the ability to overclock shrinks.
 
What kind of fps returns would you expect to see at 2500ghz? It'd definitely be worth the price at that point, i'd imagine.

I have no idea...lol. We need more OC on the current card to try to figure out what kind of gains can be had if it can achieve 2500ghz.
 
Yes, it's also been said that as fabrication nodes shrink, the ability to overclock shrinks.

No? Intel has been doing fine there. The reason Haswell and Skylake hit OC walls is because of thermal limits imposed by Intel being cheap and using thermal paste under the heatspreader instead of solder. People who pull the lids off and put them underwater get crazy OCs.

Okay, this is EXACTLY what I wanted to hear

gamersnexus put a 1080 on watercooling and weren't able to squeeze any extra performance out of it, I'm assuming because the card could only take so much power

https://www.youtube.com/watch?v=w-3fi1ovAP0

Custom boards will have 2 PCIe power connectors to break the limit of 225 W that can be supplied by PCIe slot + 1 connector.

GPU Boost is annoying and throttles cards back to prevent exceeding the BIOS locked power limit. You need to mod your BIOS to supply more voltage to the core and break over-voltage throttling. I've been modding my BIOS since my GTX 680 because you can't break through GPU Boost otherwise.

There is a lot of stuff that can be done to these cards. We aren't even seeing actual OCs with the full software tools right now, these people can only use EVGA Precision X because Rivatuner/MSI Afterburner obviously don't support Pascal yet. When they do we'll have much better control of OCs.

People are looking at voltage locked, thermal locked, single PCIe power connector reference cards and concluding that OCing these things sucks. At least wait for the custom cards. If you can already hit 2+ ghz on a reference card with all these limitations, I can't begin to guess what will happen when you lift these limits.
 
Someone on another forum found this

evga.com/precisionxoc

Cards confirmed:

Super clocked
FTW
Classified
Hybrid
Hydro Copper
Delicious. Wonder what all the differences are. Hybrid = AIO + Air cooling, Hydro Copper = Ready for custom loops?, SC = slight OC? But the rest...

I hope they go up for pre order once the timer ends.
 
I swear I read Nvidia saying die shrinkage has reached diminished returns and going smaller may end up costing more now. Or at least not really save anything anymore.
That's a bit misleading as with time the costs do go down and the price per transistor is eventually lower on newer processes. It may cost more during the start though hence why we've seen smaller chips in high end at the beginning of 28nm and are seeing them again now, at the beginning of 16nm.

14/16nm FF seems to be the reverse, clocking pretty well.
That's because of FinFETs locking down on leakage, one time thing basically, won't happen again on 10nm. And we don't really know how 14nm is doing here yet.
 
I'm very much interested in the 1080 but I have absolutely no interest in OC,I just want a nice looking (I do love the reference cooler actually) card with high performance (planning to game on 1440P)and generally quiet and efficient use. Can I go for the founders edition or should I wait. I do need a graphics card very very soon.

Thanks a lot GAF!
 
I keep reading from everyone that they are all going to sell out fast, yet everyone is also saying that the founders edition is a rip off. So who here is actually going to buy a founders edition?
 
I keep reading from everyone that they are all going to sell out fast, yet everyone is also saying that the founders edition is a rip off. So who here is actually going to buy a founders edition?
People who want the early adopter bragging rights or don't care for better cooling and OC I guess.
 
I keep reading from everyone that they are all going to sell out fast, yet everyone is also saying that the founders edition is a rip off. So who here is actually going to buy a founders edition?

Waiting for a custom job.. But i never buy anything before i can read reviews of the product plus reviews of alternatives.
 
I have no idea...lol. We need more OC on the current card to try to figure out what kind of gains can be had if it can achieve 2500ghz.

I didn't expect you to have an exact answer, but I figured since you were already extrapolating why not quantify it in a way I can comprehend. Guess you're only a partial Nostradamus, lol.
 
14/16nm FF seems to be the reverse, clocking pretty well.
Agreed. The clocks are crazy high on 16 nm FF. Try running a Maxwell GPU at 2.1 Ghz. The room to overclock depends entirely on what Nvidia sets the base/boost clock at for a certain card and what the actual limit will be. We will found out next month with the partner cards. Kinda excited to see how far Pascal can be pushed.
 
I'm very much interested in the 1080 but I have absolutely no interest in OC,I just want a nice looking (I do love the reference cooler actually) card with high performance (planning to game on 1440P)and generally quiet and efficient use. Can I go for the founders edition or should I wait. I do need a graphics card very very soon.

Thanks a lot GAF!

You want quiet you should wait. With some of the temps posted with hacked together aftermarket coolers, the 1080 custom cards should be really quiet, and likely some will be passively cooled at lower loads.

Also I'm confused why people would have absolutely no interest in OC. It isn't complicated anymore, it is literally sliding a couple of sliders in MSI afterburner while checking things stay stable, then you can leave it alone. Seems a no-brainer to get an extra 10%+ out of your expensive GPU.

Don't think if it as overclocking - think of it as the manufacturer underclocking the card to be conservative and you're just putting it back where it wants to be :)
 
Delicious. Wonder what all the differences are. Hybrid = AIO + Air cooling, Hydro Copper = Ready for custom loops?, SC = slight OC? But the rest...

I hope they go up for pre order once the timer ends.

If the past is anything to go by:

SuperClocked is clocked higher over vanilla (up to 100Mhz or so).
Hybrid is an AIO cooler clocked slightly higher over SC
Hydro Copper (same as above but with waterblock for custom loop)
FTW clocked higher again
Classified, same clock as FTW but with the ability to feed it more power for even more overclocking
 
Also I'm confused why people would have absolutely no interest in OC. It isn't complicated anymore, it is literally sliding a couple of sliders in MSI afterburner while checking things stay stable, then you can leave it alone. Seems a no-brainer to get an extra 10%+ out of your expensive GPU.

Don't think if it as overclocking - think of it as the manufacturer underclocking the card to be conservative and you're just putting it back where it wants to be :)

I've never done it because I'm too scared to touch the power slider.. I used the auto OC thing on the motherboard for the CPU as well because I've got no idea what I'm doing with voltages
 
I've never done it because I'm too scared to touch the power slider.. I used the auto OC thing on the motherboard for the CPU as well because I've got no idea what I'm doing with voltages

I overclocked my 970 using just the three basic sliders - the power max % (?) which tops out at about 110% or something quite low, then the boost and memory clocks a bit at a time while running unengine to monitor for glitches or instability. I didn't touch the voltage directly as I wasn't comfortable either. Got a nice boost from that
 
Overclocking is super-limited by default starting with Kepler. You literally cannot damage your card by OCing using just the default settings, because they are crazy conservative and they are very strict about not letting you do anything beyond the defaults.

This is why people are forced to do BIOS modding these days to OC Nvidia cards.
 
Agreed. The clocks are crazy high on 16 nm FF. Try running a Maxwell GPU at 2.1 Ghz. The room to overclock depends entirely on what Nvidia sets the base/boost clock at for a certain card and what the actual limit will be. We will found out next month with the partner cards. Kinda excited to see how far Pascal can be pushed.

The clocks seem to be high, but the fps increase isnt.
 
Who usually has the best warranty/RMA service besides EVGA? Was thinking of going for them initially but all this talk about inferior coolers puts me off.
 
I have 2 reference 980's in SLI and previously used a 680 with blower. I've never found this to be the case in any way. I don't even notice when they spin up from idle. I would guess if you used custom fan profiles for overclocking maybe?

You 'may not notice' but the decibel output of a reference 980 compared to, say, an EVGA SC ACX 2.0 cooler is huge, with the EVGA 980 cooler being much quieter and totally silent at idle because of idle fan-off mode.
 
The clocks seem to be high, but the fps increase isnt.
It's a pretty small chip with less transistors and still beats a flagship card released not even a year ago handily. If you expected it to be two times better than a 980 Ti you simply have unreasonable expectations. Performance wise it's absolutely a great card, price wise I personally think it's too much for a small chip.
Who usually has the best warranty/RMA service besides EVGA? Was thinking of going for them initially but all this talk about inferior coolers puts me off.
MSI has a 3 year warranty, granted I only had to use it once for a card no longer in production. Got my money back basicly.
 
It's a pretty small chip with less transistors and still beats a flagship card released not even a year ago handily. If you expected it to be two times better than a 980 Ti you simply have unreasonable expectations. Performance wise it's absolutely a great card, price wise I personally think it's too much for a small chip.

MSI has a 3 year warranty, granted I only had to use it once for a card no longer in production. Got my money back basicly.

I am not saying, that the performance is bad, its just that performance increase from overclocking might not be as good as Maxwell.
 
If the past is anything to go by:

SuperClocked is clocked higher over vanilla (up to 100Mhz or so).
Hybrid is an AIO cooler clocked slightly higher over SC
Hydro Copper (same as above but with waterblock for custom loop)
FTW clocked higher again
Classified, same clock as FTW but with the ability to feed it more power for even more overclocking
Yeah, I pulled some data from EVGAs website and wrapped up this tiny excel sheet. Unfortunately the pricing history of the 980 Ti models is super whack and I couldn't find any official USD or EUR list prices so...

aebPzMZ.png

I still think my guesstimate prices are insane considering the 980 started at 540ish € (reference design) but hell, considering the 789€ for the FE I feel these guesses are pretty optimistic.

I guess we'll see how wrong I was soon enough :P
 
Comparing die sizes divorced from the maturity of their respective manufacturing processes doesn't make all that much sense. A smaller die on a cutting edge process might be as expensive as or even more expensive than a larger die on a mature process.

The post I replied to was comparing to prices of similar cards of previous gens.

So by your reasoning that logic is invalid as well
28nm was cutting edge as too when the 680 released, it's apples to apples.

Which is why we got only a 300mm² die with the 680, and as yields improved we got a 400mm² die with the 980. Guess what is going to happen on 16nm? pascal launch is a carbon copy of the 28nm kepler launch.
-launch with a 300mm² die for x80-x70 card
-have a big fp64 card for workstations/servers on the new node

The 28 nm template will be upheld :
next will be titan based on the workstation card and an entry level chip for the x50-x60 cards
Then there'll be another ti card based on gp100

Then for their next architecture (if it's still on 16nm) we'll get a slightly bigger die for the x80 again (like maxwell gm204 @398mm²) and probably another pure fp32 card as there is no point for nvidia to design another workstation card on the same node.


And if you're really going to argue that each time a new process comes out it 'might be more expensive' then you're just suggesting to rationalize any price as being fair. How long are you going to keep that up? We're already completely off the deep end here... Tell anyone in 2011 you want them to pay 800 euros for a 560ti spec card and they'd laugh in your face, because it is ridiculous.

edit: and a good hint that it isn't is the fact that they stopped making maxwell cards a while ago, which means it's cheaper for them to sell you 16nm gpus than to rebrand a 28nm one for the lower end.

The fact is that nvidia stopped releasing BoM information after the 5xx cards so we will never know if it's more expensive or how much. Pluto might be filled with potatoes, can't prove otherwise so let's start taking those potatoes into consideration.

What we are left with without Bom info is a performance/dollar proposition and whether it's increasing enough to justify buying a new card after X years.

In this case , with the 1080, the value proposition is not much better than the 980ti, which has been available for ages.



It's really obnoxious how the early adaptors on this forum go out of their way to rationalize their spending every time some new crap comes out. So much so that the rest of us can't have a normal discussion about it.


Gpu prices keep inflating rapidly every generation and it has to stop. You can rationalize paying 800 euros for the same card that would have cost 300 euros back in 2011, but I don't have to.

Fact. Non reference coolers aren't great.

xxniBPj.png


This is with a evga 970 cooler at load.

You cherrypicked the trash gtx 970 evga cooler, where the heatpipes don't even make contact with the die... it's not representative of what aftermarket coolers are like.

My asus strix never goes over 40 (99 percent of the time at 35 percent) percent fanspeed and temps stay around 65-68C


I mean my post you replied to already shows what a good aftermarket cooler does for the 1080, there is no guesswork or speculation here the data is there.

Can we PLEASE stop pretending that it's difficult to cool a 180W gpu? Seriously that is some emperor's new clothes shit , nvidia's 'premium' 'founders edition' marketing is clearly effective if that is where we are at now.

That's because of FinFETs locking down on leakage, one time thing basically, won't happen again on 10nm. And we don't really know how 14nm is doing here yet.
It's a one time thing in that it's a one time thing for finfet.

(also love the little stab at amd , you couldn't resist could you :p)

There's 2 possibilities in the near future:

-new transistor design that further reduces leakage similar to finfet over planar, giving another lowering in operating voltage and another clockspeed boost

-stuck on finfet which means there simply be no more node shrinks for gpus after the next one (10nm, due in 2017) , it'll be like 32 and 20nm on planar: pointless so noone will bother for anything but low power chips

It's going to be either of these ,at least until they stop using silicon.
 
My reference 970 is OC at 1500 mhz base clock with +400 to memory (never noticed issues so probably could go higher). Never increased the voltages. At highest load around 40-50% fan. Temps never go above 70c usually around 65-68.
 
I think the biggest takeaway here is the fact that paying $100 over MSRP used to mean getting hardware on the level of EVGA Classified or MSI Lightning: custom PCB, cooler capable of dissipating 400W+ of heat and robust power delivery.

Instead, this time around, we get a GPU that barely maintains its default boost clocks and is only capable of overclocking once the fans are cranked up to near 100% (which according to HardOCP is unbearably loud). So what exactly does this mean for the 1080's at MSRP? An even shittier cooler? Lower base/boost clocks?

I don't know why people are expecting the aftermarket boards with better coolers/components to be cheaper than the Founder Edition, but I'd love to be wrong.
 
Also I'm confused why people would have absolutely no interest in OC. It isn't complicated anymore, it is literally sliding a couple of sliders in MSI afterburner while checking things stay stable, then you can leave it alone. Seems a no-brainer to get an extra 10%+ out of your expensive GPU.

Doesn't the card get LOUDER when it's overclocked?
 
Instead, this time around, we get a GPU that barely maintains its default boost clocks and is only capable of overclocking once the fans are cranked up to near 100% (which according to HardOCP is unbearably loud). So what exactly does this mean for the 1080's at MSRP? An even shittier cooler? Lower base/boost clocks?

I don't know why people are expecting the aftermarket boards with better coolers/components to be cheaper than the Founder Edition, but I'd love to be wrong.

Because that's how market works - as long as Nvidia set partners price with 600$ price in mind then you can be sure that one of companies will try to be more aggressive with pricing in order to increase their market share. Then everyone else will follow to defend their market share.

And that's how it will work as long as we have x corporations targeting same market with sum of sales targets exceeding that market value :)
 
I think the biggest takeaway here is the fact that paying $100 over MSRP used to mean getting hardware on the level of EVGA Classified or MSI Lightning: custom PCB, cooler capable of dissipating 400W+ of heat and robust power delivery.

Instead, this time around, we get a GPU that barely maintains its default boost clocks and is only capable of overclocking once the fans are cranked up to near 100% (which according to HardOCP is unbearably loud). So what exactly does this mean for the 1080's at MSRP? An even shittier cooler? Lower base/boost clocks?

I don't know why people are expecting the aftermarket boards with better coolers/components to be cheaper than the Founder Edition, but I'd love to be wrong.

That's the 'genius' of the founders edition marketing I guess?
Convincing people that good cooling somehow costs 100 dollars.
While in the past the good aftermarket aircoolers have never cost more than 20-25 dollars extra over the crappiest zotac potato cooler you could get.



This is what nvidia has now convinced people ITT of:

-that cooling a 180w gpu is hard (lol)

-that a cooler adds a shitload of cost to the BoM of a gpu (yeah let's forget that you can buy an enormous cpu cooler with 3x more material and 3x more fin area for 25 dollars, gpu heatsinks are made of unicorn poop and cost way more somehow)

Which then makes people conclude that a 1080 can only cost 700 dollars with a decent aftermarket cooler
 
That's the 'genius' of the founders edition marketing I guess?
Convincing people that good cooling somehow costs 100 dollars.
While in the past the good aftermarket aircoolers have never cost more than 20-25 dollars extra over the crappiest zotac potato cooler you could get.



This is what nvidia has now convinced people ITT of:

-that cooling a 180w gpu is hard (lol)

-that a cooler adds a shitload of cost to the BoM of a gpu (yeah let's forget that you can buy an enormous cpu cooler with 3x more material and 3x more fin area for 25 dollars, gpu heatsinks are made of unicorn poop and cost way more somehow)

Which then makes people conclude that a 1080 can only cost 700 dollars with a decent aftermarket cooler

We are entering in dangerous territory, Intel has the monopoly on the CPU business, every new line of mainstream cpus add 10% increase in IPC at best, an overclocked i7 920 still gets the job done playing the latest games, on q4 2016 this cpu will be 8 years old. Can you imagine in 2008 playing the latest games with a P4 1.5ghz? Nope

Same is happening with the GPU market, but even worse since we are already paying almost 800€ in Europe for a GPU that is not a huge upgrade compared to a GPU released 1 year ago and people are going crazy with how awesome the 1080 FE is, a GPU with a "premium cooler" (as stated by nvidia) that cant even cool 180W!

Does anyone believe that a proper custom 1080 will be cheaper without any kind of competitor? Anyone can dream. Be prepared to pay Titan prices for 1080ti, and titan will get a new high record price for consumer GPU.

So every one that trashed AMD all this must time be really happy with the results, the fault is on AMD, not saying the contrary, but seeing some users trash AMD even if it´s dissimulated are the ones that are mostly happy with this "beast" called 1080 Reference Card, sorry "Founders Edition".

If Vega\Polaris\Zen fail and if you dont have deep pockets you better forget High End PC gaming.
 
You can rationalize paying 800 euros for the same card that would have cost 300 euros back in 2011, but I don't have to.
I've never spent more than 400€ on a GPU in my entire live, but unlike you seem to I don't begrudge those who do.

So every one that trashed AMD all this must time be really happy with the results
I'd be exceedingly happy if AMD actually managed to compete on their own terms rather than requiring the charity of people concerned about what other companies are doing.
 
I've never spent more than 400€ on a GPU in my entire live, but unlike you seem to I don't begrudge those who do.

I'd be exceedingly happy if AMD actually managed to compete on their own terms rather than requiring the charity of people concerned about what other companies are doing.

Of course im concerned, im seeing the results. I want to buy the best for the least money possible and that will only happen if competition exists. I dont care if Nvidia or AMD wins, im not a shareholder of either company.

When i see people justifying the premium price of 1080FE even if it´s nothing more than a reference card for 800€ (in europe) i get the impression that we do deserve to pay this prices.

I usually buy high end GPU´s, so of course im concerned, from paying 500+€ 5 years ago to 800€ leaves a bad taste in my mouth, i know im not obliged to pay but if i want the best available that´s the current price i have to pay so i can only hope that AMD rises from the grave in order to have a price battle again. Just a recent exemple was the 290x launch, before that the 780 was almost 200€ higher, when 290x was released the 780 falled back to 500€ and we got a new high end gpu, the 780ti.

Everybody wins if we have at least two competitors on the market. Just my opinion.
 
So the new AMD cards are supposedly going to be brought out at Computex?

I need the 400 series or whatever it's gonna be called to smack Nvidia across the face for having such absurd prices on their blower cards.
 
Top Bottom