Nvidia GeForce GTX 1080 reviews and benchmarks

I expect the 1070 to cost around 550 euros here. And I'll very suprised if a partner sells it for less than 450. Definitely never gonna cost 370 here in Europe.
With the 1080 costing €780 here I would be really surprised to see the 1070 even dip below €500 in Europe.

Europe am cry.
 
I expect the 1070 to cost around 550 euros here. And I'll very suprised if a partner sells it for less than 450. Definitely never gonna cost 370 here in Europe.
Some of you are just spreading completely baseless FUD.

Why would the 1070 be relatively more expensive than a 1080?
 
What are you talking about? The GTX 1080 costs around 780-800 euros here. And the 1070 will definitely cost 500-550 euros.
A 699 dollar card will cost 780 euros but somehow a 449 card will cost 550 euros. This make zero sense.

Btw. founders edition prices are bullshit in general but because everyone loves quoting those prices I ran with them.
 
We'll see about that. A 980ti with a standard overclock is pretty much on par with a stock 1080. Not sure where that leaves the 1070.

This, I don't believe the hype for a second, I've been checking a lot of benchmarks and the 1080 is not leaps and bounds ahead of the 980ti as Nvidia wants us to believe, in fact they're very close, specially at high resolutions.
 
It's so entertaining to go read the comments of people who only a few days ago were shitting on the 1070. They act surprised now. :D Anyway we need some real benchmarks.
 
Do they ever do game bundles with reference cards? Like any chance they could add a game with the cards on Friday?


The only game I could think of that would make sense is Doom since they have marketing with nvidia and those drivers thing.
 
Do they ever do game bundles with reference cards? Like any chance they could add a game with the cards on Friday?


The only game I could think of that would make sense is Doom since they have marketing with nvidia and those drivers thing.

Those sweeten the pot later, doesn't need anything now.
 
Is this due to software or hardware limitations? If software, wouldn't a custom vBIOS be able to fix this?

its due to nvidia only supplying 1 8pin connector to save on manufacturing costs. im sure in the future you will be able to flash the bios which will allow you to draw more power than pci and 8pin specs are designed for but its not ideal and voids your warranty. better to just wait for custom cards that will have 2 power connectors. nvidias cooler isnt sufficient anyway unless you dont mind lots of noise from 100% fan speed

Ive read this tons of places. Are the non founders/reference going to be able to draw more power so OCing is more viable?

historically the good custom boards have always done so. it would be a first if they didnt.
 
the problem is once you start playing games the clocks throttle down because the card cant draw enough power. doesnt matter what you set in afterburner.

14636166464UR8bKmzfk_1_5_l.gif


Did you read the article I linked? Average clock during BF4 test was 2061mhz.

Am I missing something, or does the card overclock just fine?

Because they set the fan to 100%, which might be too loud for some.

14636166464UR8bKmzfk_1_6_l.gif


The fan is at 100% but the temperatures are also ~20 C below the thermal "throttling" limit at its highest. Would it not follow, then, that the fan curve could be scaled back?

It's incorrect to state the card CANNOT overclock properly. It can. The fan is designed to work up to and including 100% of its possible speed.

I've been trying to figure out whether I should pick one of these up and the misinformation being thrown around in this thread about the overclocking capabilities of the card has made it amply difficult to understand what exactly the fuck is going on. Glad HardOCP actually did some tests. Card seems fine, if anything the stock cooler just isn't all that great.
 
its due to nvidia only supplying 1 8pin connector to save on manufacturing costs. im sure in the future you will be able to flash the bios which will allow you to draw more power than pci and 8pin specs are designed for but its not ideal and voids your warranty. better to just wait for custom cards that will have 2 power connectors. nvidias cooler isnt sufficient anyway unless you dont mind lots of noise from 100% fan speed



historically the good custom boards have always done so. it would be a first if they didnt.

Yeesh that's dire. Hopefully custom cards will allow extra juice.
 
Is this due to software or hardware limitations? If software, wouldn't a custom vBIOS be able to fix this?

You can only draw so much power from the pci lane and single power connector, potentially less than it needs even with a bios flash.

Iirc max from pci is 75 watts and max from single 8 pin is 120w, depending on your psu rail set up, for total of Max 195 draw.

I imagine a high boosting after market card with 2x 8 pin would pull more than that to sustain boost.
 
14636166464UR8bKmzfk_1_5_l.gif


Did you read the article I linked? Average clock during BF4 test was 2061mhz.

Am I missing something, or does the card overclock just fine?



Then it's incorrect to state the card CANNOT overclock properly. It can. The fan is designe to work up to and including 100% of its possible speed.

I've been trying to figure out whether I should pick one of these up and the misinformation being thrown around in this thread about the overclocking capabilities of the card has made it amply difficult to understand what exactly the fuck is going on. Glad HardOCP actually did some tests. Card seems fine, if anything the stock cooler just isn't all that great.

power loads vary depending on the game and even the scene. i tested this a lot when i had a reference titan X. hardocp tests bf4 on shanghai. that map does not stress the power target. i was able to sustain clocks in the 1400s on that map. when i turned on TRSSAA and tested on caspian border, clocks would constantly throttle down to the high 1200s and low 1300s. same for the heavy forested maps of evolve. i think it is heavily influenced by how hard bandwidth and rops are being hit. those seem to be heavy hitters for power use. witcher 3 was another game that would power choke my titan x. when i switched to a g1 980ti with a better power system, ive never seen throttling. the sillicon lottery will also play a much bigger role in reference cards. maybe you will get lucky and get a 1080 that has a very high asic and wont be as prone to throttling. then again, maybe you wont. custom boards get rid of this variance.

id suggest you experiment with how much power is used in x game during x scenario by monitoring the power readout in afterburner. i think youll find that anything bandwidth or rop heavy stresses it the most. so deferred rendering with msaa, anything with lots of alphas or transparency, any type of ssaa or trssaa, particle heavy scenes, scenes with lots of overdraw etc (shanghai in bf4 doesnt fit any of these scenarios other than the msaa but even then, its a great case for msaa due to the very simple geometric nature of the map)

edit - id be willing to bet the big quick drops hardocp shows are when your close up with a vehicle explosion and smoke/fire fills your screen. thats a huge problem area in general due to the poor way in which particles and alphas are handled in fb3
 
14636166464UR8bKmzfk_1_5_l.gif


Did you read the article I linked? Average clock during BF4 test was 2061mhz.

Am I missing something, or does the card overclock just fine?



14636166464UR8bKmzfk_1_6_l.gif


The fan is at 100% but the temperatures are also ~20 C below the thermal "throttling" limit at its highest. Would it not follow, then, that the fan curve could be scaled back?

It's incorrect to state the card CANNOT overclock properly. It can. The fan is designed to work up to and including 100% of its possible speed.

I've been trying to figure out whether I should pick one of these up and the misinformation being thrown around in this thread about the overclocking capabilities of the card has made it amply difficult to understand what exactly the fuck is going on. Glad HardOCP actually did some tests. Card seems fine, if anything the stock cooler just isn't all that great.

People also running into power throttling regardless of Temps, if you do want to run your fan at 100%.
 
power loads vary depending on the game and even the scene. i tested this a lot when i had a reference titan X. hardocp tests bf4 on shanghai. that map does not stress the power target. i was able to sustain clocks in the 1400s on that map. when i turned on TRSSAA and tested on caspian border, clocks would constantly throttle down to the high 1200s and low 1300s. same for the heavy forested maps of evolve. i think it is heavily influenced by how hard bandwidth and rops are being hit. those seem to be heavy hitters for power use. witcher 3 was another game that would power choke my titan x. when i switched to a g1 980ti with a better power system, ive never seen throttling. the sillicon lottery will also play a much bigger role in reference cards. maybe you will get lucky and get a 1080 that has a very high asic and wont be as prone to throttling. then again, maybe you wont. custom boards get rid of this variance.

id suggest you experiment with how much power is used in x game during x scenario by monitoring the power readout in afterburner. i think youll find that anything bandwidth or rop heavy stresses it the most. so deferred rendering with msaa, anything with lots of alphas or transparency, any type of ssaa or trssaa, particle heavy scenes, scenes with lots of overdraw etc

Thanks for the awesome insightful answer. I wish HardOCP had done some additional testing. They are usually thorough, so seems kind of out of character for them to only test a non-demanding map.

I guess we'll have to wait longer for there to be actual trustable OC numbers for the 1080. I'm just trying to decide whether or not to upgrade from a 980 Ti that overclocks to 1500mhz presently.
 
Thanks for the awesome insightful answer. I wish HardOCP had done some additional testing. They are usually thorough, so seems kind of out of character for them to only test a non-demanding map.

I guess we'll have to wait longer for there to be actual trustable OC numbers for the 1080. I'm just trying to decide whether or not to upgrade from a 980 Ti that overclocks to 1500mhz presently.

id suggest you check out the pcgameshardware.de review linked in the OP. and yeah, shanghai is one of the best performing maps in bf4 multiplayer(outside of metro and locker) as long as you avoid that trouble spot in the middle of the map by c with the burning cars and all that poorly optimized smoke.
 
I'm not planning on Overclocking the 1080... Does that means that the FE would still always run at 100% fan or that's just for OC?

Fan speeds are usually controlled automatically. I don't think I've ever had a gpu run its fan at 100% without touching anything. These 100% graphs are probably from the GPU being forced to, which you can do with programs like afterburner.
 
Was thinking to maybe sell my 970 and get a cheap(ish) used 980ti, but might just go for a 1070 for potential VR benefits.

Timing is tricky though. Sell now to not lose too much and be left without a GPU for a month or more, or wait until 1070 is available and my 970 will be worth peanuts

Yeah, same boat here lol. I think used 970s will probably be going for $125-150 once the 10x series rolls out.
 
We'll see about that. A 980ti with a standard overclock is pretty much on par with a stock 1080. Not sure where that leaves the 1070.
Where's this coming from? My 980Ti is what I consider standard overclock (1241 MHz) and that gives it +15% of performance. Even a heavily overclocked 980Ti would be lucky to hit stock 1080 performance which is +30%.
 
I'm not planning on Overclocking the 1080... Does that means that the FE would still always run at 100% fan or that's just for OC?

You won't need to worry about the fan, it'll quietly do its thing well short of 100%. I do the same, treating a card like an appliance. My overclocking days are well in the past.

Just this week I've loaded that EVGA tool just to see what my 980's have been doing these years. Been comparing it to the 1080 review results and all that, confirm that I'll be just fine with the 1080.

So glad I lost interest in tracking down cherry CPU samples for OC, making sure to do the same with GPUs and doing voltage mods and such.. yuck. Was fun back in the day, so I get it..
 
Where's this coming from? My 980Ti is what I consider standard overclock (1241 MHz) and that gives it +15% of performance. Even a heavily overclocked 980Ti would be lucky to hit stock 1080 performance which is +30%.

a 980ti at 1500/8000 is probably roughly on par going by the pcgh special benchmarks
 
Where's this coming from? My 980Ti is what I consider standard overclock (1241 MHz) and that gives it +15% of performance. Even a heavily overclocked 980Ti would be lucky to hit stock 1080 performance which is +30%.

Didnt Gamespot do some benchmarks where an OC'd 980Ti was virtually the same performance wise as a stock 1080?(the 1080 was maybe ~3% faster)
 
Where's this coming from? My 980Ti is what I consider standard overclock (1241 MHz) and that gives it +15% of performance. Even a heavily overclocked 980Ti would be lucky to hit stock 1080 performance which is +30%.

That sounds like the base clock + overclock but NOT figuring in boost, which would jump to > 1400.

I think I'm at ~1350, and I'm boosting to 1480.

I'd say a standard overclock + boost for the 980TI would put you at 1450. 1500 is on the high end for air cooling.
 
That sounds like the base clock + overclock but NOT figuring in boost, which would jump to > 1400.

I think I'm at ~1350, and I'm boosting to 1480.

I'd say a standard overclock + boost for the 980TI would put you at 1450. 1500 is on the high end for air cooling.

I'd go a bit lower than 1450 as a standard boost oc. And if you aren't pushing power and voltage sliders your boost clock likely isn't all that stable for sustained sessions at load. The temp and power throttling isn't new. The reference 980 ti had the same problem. Doesn't mean performance is terrible or anything. Just that a bios modded aftermarket 980 ti is going to sustain higher boost clocks much better. The 980 ti g1 overclock . net bios modding forum is a pretty good source of info. Also for g1 970 and 980 models. The zoson thread on the evgas is also a good one.

The problem is just put under a microscope because Nvidia is attempting to charge $699 for a ~300mm reference card while simultaneously stating it has superior cooling and build quality which is a load of garbage, and then the added confusion of saying the msrp is $599, without detailing what a $599 card will look like (an even worse blower cooler and build quality from aibs, as seen in the galax clown example).

People in this thread are repeatedly making the assumption that all the aftermarket cards will be $599, which is silly. If an aftermarket card has a better pcb and cooling, why would it be cheaper? Shady nvidia marketing is clearly working.

Once everyone has sold their old card thinking they can get a new card at $379 or $599, and then realize they can't, they'll just grudgingly shell out $450+ or $699+. It's an extremely deceptive way to drive new card purchases and increase margins.
 
well even your 980ti at 1241 is going to be boosting pretty damn close to 1500 in actual games

Not even close, 1380 at max, 1367 sustained. 1241 is G1's rated boost clock. 1500 is a rather high overclock which only some 980Tis are able to reach without some advanced modding. Saying that it's "standard" is just misleading, comparing that to reference 1080 on stock clocks is straight up pointless.

1080 FE should be able to hit 2GHz OC. The drops noticed are related to the new boost 3.0 thingie and the obvious deficiencies the reference has in power supply and cooling. I'm expecting some rather spectacular OC numbers from Strix/Xtreme/etc cards, probably bigger ones than which was possible with 980Ti (which isn't really surprising considering the difference in TDP). The only thing which may be limiting them is the upcoming GP102 cards which might get a lot less interesting if 1080OC will hit 2+ GHz in a factory.
 
Not even close, 1380 at max, 1367 sustained. 1241 is G1's rated boost clock. 1500 is a rather high overclock which only some 980Tis are able to reach without some advanced modding. Saying that it's "standard" is just misleading, comparing that to reference 1080 on stock clocks is straight up pointless.

1080 FE should be able to hit 2GHz OC. The drops noticed are related to the new boost 3.0 thingie and the obvious deficiencies the reference has in power supply and cooling. I'm expecting some rather spectacular OC numbers from Strix/Xtreme/etc cards, probably bigger ones than which was possible with 980Ti (which isn't really surprising considering the difference in TDP). The only thing which may be limiting them is the upcoming GP102 cards which might get a lot less interesting if 1080OC will hit 2+ GHz in a factory.

It's been fairly well covered the FE will not sustain a 2ghz boost clock for any meaningful amount of time unless you have an extremely aggressive fan curve, which a lot of people won't be willing to do because of noise.

Doesn't mean the FE is a bad card (it kind of is at $699, but ignoring price), just that an OCD 980 ti will perform closer to an FE than some of the reviews suggest. In this thread the difference was drawn out to only be a 3% advantage for the FE. I'm sure someone can dig up the posts if necessary. It does change the value proposition significantly in upgrading from the 980 ti to an FE (imo do not upgrade from a 980 ti to to an FE 1080).

With sustained 2ghz boost clocks w/ an aftermarket card, it's a better value, but I'm still not sure it's worth upgrading if you have a good clocking 980 ti, given the cost.

If you want to get the most out of your G1, I suggest bios modding it. You may be able to squeeze 1450mhz out of it. I have the same card and it does 1500/8000 with a bios mod. It isn't hard to do, and the G1 is built for it with its dual bios set up.
 
Not even close, 1380 at max, 1367 sustained. 1241 is G1's rated boost clock. 1500 is a rather high overclock which only some 980Tis are able to reach without some advanced modding. Saying that it's "standard" is just misleading, comparing that to reference 1080 on stock clocks is straight up pointless.

The benchmarks I've seen with cards overclocked, where there's barely a difference between 980ti OC and 1080 stock, were not running the 980ti at 1500Mhz. It was 1400, something that many 980ti owners can easily achieve. And I'm sure at 1300 the difference will still be small.
 
Top Bottom