• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Nvidia GeForce GTX 1080 reviews and benchmarks

Seems like the inherent limitations of moving to 16nm, that everyone should have seen coming, but everyone chose to believe instead that having a ton of power pins and phases would allow EXTREME INSANEXXX overclocks, even though the highly qualified engineers at nVidia determined 8 pins was sufficient for the limitations of the silicon.

So overclocking is going to become smaller and smaller as we get more node shrinks? So when we eventually get to ~8nm cards will pretty much be stuck at stock speeds?
 
The 980 last time was a pointless product with the existence of the 970, the difference between the 1070 and 1080 this time seems to be bigger, even if the former is still a much better value proposition. Who knows how much the ti would be. The popularity of the 1080 FEs isn't exactly going to deter Nvidia from trying something like this again.

Hmm okay, I'm thinking of going for the 1080, just hoping it's not completely eclipsed in value proposition just by waiting 6 more months. I don't really need to upgrade from my 970 and would keep my 970 for a year or more from now if I didn't jump into VR, I can wait, I'm a patient man. I just want something better for my Vive since it doesn't quite cut it for a smooth experience in everything. Hard to choose between 1080 and 1070, or whatever AMD is bringing at the higher price range since the 480 seems pretty cool.
 
So overclocking is going to become smaller and smaller as we get more node shrinks? So when we eventually get to ~8nm cards will pretty much be stuck at stock speeds?

I doubt that's going to necessarily be the case. It seems as though the 1080 shouldn't be clocked as high as it is, but since it lacks certain features compared to something like the 980ti (such as number of ROPs), they're forced to make up the difference with higher base clocks.
 
So overclocking is going to become smaller and smaller as we get more node shrinks? So when we eventually get to ~8nm cards will pretty much be stuck at stock speeds?
No one can fully predict this, but just look at the development with Intel CPUs over the past 8 years or so.

The combination of physical limitations with hardware that gets better and better at performing up to its own limits out of the box certainly doesn't seem likely to increase the relative overclocking potential.

We won't see a Celeron 300A again, that's for sure.

I doubt that's going to necessarily be the case. It seems as though the 1080 shouldn't be clocked as high as it is, but since it lacks certain features compared to something like the 980ti (such as number of ROPs), they're forced to make up the difference with higher base clocks.
I don't really see any basis for this claim. It's clearly extremely power efficient at the clocks it uses out of the box, so why shouldn't it run at those?
 
No one can fully predict this, but just look at the development with Intel CPUs over the past 8 years or so.

The combination of physical limitations with hardware that gets better and better at performing up to its own limits out of the box certainly doesn't seem likely to increase the relative overclocking potential.

We won't see a Celeron 300A again, that's for sure.

I don't really see any basis for this claim. It's clearly extremely power efficient at the clocks it uses out of the box, so why shouldn't it run at those?
Would a bigger chip make room for better overclocking? Let's say a 1080Ti at 500-isch mm^2.
 
Person at home ignored the doorbell, UPS ignored my note to leave the package, nobody's home tomorrow.

:|
I know the feeling.. and it sucks.

That actually happened to me when I was even home one time. Ironically I was playing a game waiting for my new graphics card to arrive and didn't hear lol
 
Would a bigger chip make room for better overclocking? Let's say a 1080Ti at 500-isch mm^2.
Not inherently.

I mean, you could probably get higher clocks by making a larger chip and then only using parts of it (like e.g. Intel does for hideously expensive specialized Xeons), but that would obviously defeat the purpose in something which scales as well with parallelization as GPU workloads.
 
I don't really see any basis for this claim. It's clearly extremely power efficient at the clocks it uses out of the box, so why shouldn't it run at those?

Well, switching to a smaller manufacturing process does lend to better efficiency. When I say it "shouldn't" run at those clocks, I mean that traditionally a GPU's stock clocks aren't pushed as close to the limit of what's possible.
 
I wonder how many of these cards are being bought by scumbag scalpers. Between them, others placing orders on multiple cards just to see if they can get one instead of what they really want, and then legit buyers, this has been an incredibly frustrating experience.

Not as rage inducing as trying to get a Focus RS, but it's up there. lol.
 
Would a 1080 be worth getting, it feels like I never heard mention of 980s once the 980ti came out so is that expected to happen again this time? How much more is the ti version expected to cost based on previous gens?

This is impossible to know. A 780 offered better bang for the buck than the 780Ti but the 980 was just annihilated by the 980Ti on overall performance and value.

As someone with a highly overclocked 980Ti the 1080 is not appealing because I only see marginal improvements. If the 1080Ti blows away there 1080 I will probably upgrade. Additionally, when the 1080Ti comes out the price of the 1080 might drop.
 
Most likely. There may also be something to be said about NVidia's reference cooler being better than usual. It may have allowed a higher stock clock than they would have set on older reference designs. And since we view overclocking potential as % above stock, it could be coloring perception a bit.

Regardless, extra power pins don't seem to serve any purpose unless we see a Custom BIOS that allows people to reach 2200+.

I'm sure BIOS editors will be out sooner or later, and people will unlock the voltage up to around the maximum hardlocked 1.25v. However, I think going higher than the stock 1.080v will start to significantly decrease the life of the silicon for Pascal. Electromigration and other low level effects are a real concern on 16nm chips.

The maximum I'd be willing to try on my card is ~1.15V, and it's watercooled.

So overclocking is going to become smaller and smaller as we get more node shrinks? So when we eventually get to ~8nm cards will pretty much be stuck at stock speeds?

Yah. The smaller you go, the lower the threshold is for voltage. And damage often increases exponentially with linearly increasing voltage. If anything, it's a miracle that 1.25V is allowed at all with hard modding, as the recommended voltage for 16nm is 900mV.

I think what we're seeing is nVidia optimizing clocks closer to the maximum threshold for voltage.
 
Not to sound like a broken record but... Europe? Any news around here? :(
UYSjmBe.png


There's not a single partner card in stock anywhere here in Germany (apparently some Palit card was according to a GAFfer, but that one is actually more expensive than the FTW lol).
The three shops I contacted told me all stock dates are just rough estimates.
 
UYSjmBe.png


There's not a single partner card in stock anywhere here in Germany (apparently some Palit card was according to a GAFfer, but that one is actually more expensive than the FTW lol).
The three shops I contacted told me all stock dates are just rough estimates.

Welp. I'm holding out on my apparent pre-order of the G1 on Amazon UK. No chance in hell I'm paying €850 for a FE card...
 
Finally got one of the EVGA SC's earlier on newegg in what felt like an epic of battle of refreshes and cancelled orders. Twice I actually got the card ordered and got a confirmation only to get an "Order voided" email a few minutes after. I managed to catch them in stock a third time and the order went through.
 
Finally got one of the EVGA SC's earlier on newegg in what felt like an epic of battle of refreshes and cancelled orders. Twice I actually got the card ordered and got a confirmation only to get an "Order voided" email a few minutes after. I managed to catch them in stock a third time and the order went through.

Congrats. I'm glad I ended up with this one as well. I had three ordered at one point and then cancelled down to the SC. Watching JaytwoCents' video reinforces my happiness.

Man, the wait for my FTW is excruciating. Ship it already EVGA! My new PC build is waiting.

At least you know it's going out soon.
 
Contacted Gigabyte on facebook and asked about Amazon orders
"Unfortunately, I don't have an exact date for you but I'm expect to see it release towards the end of the week.
Early next week at the LATEST."
 
So overclocking is going to become smaller and smaller as we get more node shrinks? So when we eventually get to ~8nm cards will pretty much be stuck at stock speeds?

What? no

pascal is clocking WAY higher than anyone could have possibly expected before it was shown
TSMC's process has offered huge clockspeed increases over 28nm planar

The only difference is that nvidia seems to have clocked their cards a lot higher out of the box this time around, while with maxwell their reference clocks were pretty low

They could have clocked the 980ti or 980 at 1.2ghz reference instead if 1000mhz /1126 mhz but didn't for whatever reason



It's funny to see how expectations alter people's perception:

offer huge clockspeeds that didn't seem possible a month ago? dissapointment based on hope of even higher speeds on custom boards

pay 700 dollars for a 300mm² card? That's fine because you're used to being gouged anyhow

It's like opposite world in here sometimes
 
I see I missed the MSI card, that price is outside what I'm looking to pay but I'm getting super desperate for a custom 1080 D:
 
is that not a power limit issue due to just the 8 pin?

if not...da fucvk

No, it's not a power limit issue. Turns out, nVidia knew just what they were doing when they put a single 8-pin on the FE.

8+6 cards have been overclocking the same or below. Until someone BIOS mods some cards to get closer to the 1.25V limit, we won't see anything above ~2100mhz unless the chips are extremely golden.
 
Anyone else feel like these board makers just started producing these cards at the beginning of the month. It feels like they are so damn disorganized on actually having these things in a warehouse and ready to actually sell to the masses.
 
Anyone else feel like these board makers just started producing these cards at the beginning of the month. It feels like they are so damn disorganized on actually having these things in a warehouse and ready to actually sell to the masses.

Yeah for sure, can't think which card it was in a YouTube video where they said the design wasn't final, and that was a week or so ago. The the EVGA hybrid hasn't gone into manufacturing yet, I think gamers nexus had a clip where they said EVGA wernt settled on a final design.
 
I managed to get one of the EVGA ACX 3.0 while I was at work. I didn't feel it was worth the extra money for the SC since I doubt there will be much difference when the cards are OC'd as much as possible. Funny thing is that I was only refreshing because I missed the chance for the MSI card earlier. Would have paid a hundred dollars more for that. I guess things worked out.
 
Top Bottom