NVIDIA GeForce GTX 1080 Founders Edition Review @ [H]

The assumption was that the "Founder's Edition" cards were those sold directly by NVIDIA, branded as NVIDIA cards. Those sold by AIBs should be $599, even if they are reference design. This would have been perfectly acceptable. PNY, Zotac, Gigabyte and even ASUS are not on the same level as EVGA for example when it comes to customer service and RMA. Probably EVGA, followed by ASUS will have the best custom designs. Hoping that they won't cut too many corners.

Like many have said, it's an "early adopter tax".

Don't forget the nVidia page charges state tax too. You literally get to pay another tax. :) bahahaha
 
called it.

5-10% better than an OC 980Ti model.

bring on the OC SLi 980Ti for me.
Why do people compare an older model overclocked card to a newer model @ stock clocks? Makes no sense.... Overclock the 1080 and the gap will widen again. Now, if the 1080 proves to be a terrible overclocker then you have a point, but I do not believe this will be the case.
 
Look at 7:52 here:



I understand that, but if we quietly play along then we'll continue to feed NVIDIA's greed. Of course, if AMD had a solid product, they would have done the same. Just look how they priced the Radeon Pro Duo at $1499... of course, no one is buying it. You could buy two Nano's for $1000 and still have money left over for a nice long weekend vacation... or a decent CPU + motherboard combo.


3D Mark firestrike score doesn't mean too much, but you did give a link.
 
I'm pretty sure that we're saying the same thing here. And no, no one at HardOCP said that FE cards sold by AIBs will cost less than $699. My entire point was that they should cost less, and only NVIDIA should sell them for $699 via their online store and partners such as BestBuy.

If people are willing to pay the premium, let them pay it. No reason to get upset about it. Nvidia is doing what they believe is best for themselves. That's what a for profit company is supposed to do. It's a free market, if consumers don't like the price they don't have to pay for it. They can wait or buy something else.

Correct you are sir, I moved from CF 7970 to 980ti.

Question I must ask is, for someone with an overclocked 980Ti, is this worth an upgrade??

Even from a perf/watt standpoint it doesnt seem so.

Pretty MEH.

EDIT: Even from a 980Ti?

probably not.

For 980 ti owners? Not at all. Though it's not targeted for those of use with 980 tis or Titan Xs. For people with 980-level cards or lower this could be a very nice upgrade, if they want to toss down the money for it. I almost don't want to know what the 1080 ti or next Titan will cost whenever that gets released. Hopefully the 1080 will see a price drop before then so it isn't too outrageous.
 
Also, while its an early test, 980Ti with a 1452-1477 boost clock hit 17420 in Firestrike (15656 stock).

1080 hits 19370 stock and 20468 with 1950-2065 boost clock. 17% increase and while that is just Firestrike, it should give you an idea of whats to come after some more testing. Not only that, driver refinement will help as well. Will be interesting to see what more power and better cooling accomplishes too on these cards.

(guru3d OC benchmarks quoted)
 
IIRC, it was 11" max for a daughterboard. As PC adoption hit its second wave due to the 'net's popularity, you had a lot of requests for a SFF case. That plus overall component reduction led to much smaller cards and case manufacturer's started encroaching on the reserved space, usually with drive bays. It's kind of funny to see them grow again. At this rate we'll need the guides on the front of the mainboard again.

The medium/long term trends are for smaller cases/boards again. HBM will chop a few inches off of card lengths, as it trickles down from halo to mainstream cards we'll return to having a few inches of mobo exposed in front of the GPU again. Independent of that, compact layout cases without a stack of 5.25/3.5" bays up front are becoming a thing with optical media moribund and SSDs that can be either mounted on the mobo of the back of the mobo tray meaning that many gamers won't need spinning rust in their PCs any longer. A few years from now a full ATX case that's only about 16x12 will be a reasonable mainstream option. Bigger cases will almost certainly still be around for people who want triple rads up top or who have a giant media collection and don't want a separate NAS; but even ignoring people switching to smaller mobo form factors we should see baseline gaming PCs shrink a good bit by 2020.
 
Also, while its an early test, 980Ti with a 1452-1477 boost clock hit 17420 in Firestrike (15656 stock).

1080 hits 19370 stock and 20468 with 1950-2065 boost clock. 17% increase and while that is just Firestrike, it should give you an idea of whats to come after some more testing. Not only that, driver refinement will help as well. Will be interesting to see what more power and better cooling accomplishes too on these cards.

(guru3d OC benchmarks quoted)

Who cares about firestrike? Go read computerbase's review, they did OC vs OC in games.
 
Yep, the PCB is designed to handle up to a 6+2 power phase but they went for 5+1. Being an optimist, this could actually bode well for the Ti and or Titan version.

I suspect the reason's even simpler. The PCB layout was either finalized before they'd locked down the power level for the GPU (IOW they might've had to go +20W and more power delivery); or they're throwing a bone to the AIB companies by letting them bump power on semi-custom cards without having to do a custom board layout first. Could be both.

I suspect the Ti/Titans will probably be build around GP100 (GP 200??) chips when availability exceeds what they need for their super computing/enterprise markets. Both of which can pay several grand per card if the performance justifies it; and both of which have largely been left out in the cold because the 9xx series gutting FP64 performance in order to get enough of an FP32 boost to be an upgrade for gaming.
 
Repeating what a company tells you verbatim as truth and swearing that these people who you've known for two decades couldn't possibly be lying, is not reporting, it's acting as a propaganda agent for the company. Asking them very hard questions and hammering on it when they are caught outright lying is reporting. Though doing actual reporting may endanger your chance to get a $2000 paid vacation to that company's next event. The fact you are friends with these people should clearly be stated in any review of their product. Journalism requires ethics.
The problem is the overclocking tools are not fully there yet for consumers.
TPU manage to achieve a 2114MHz boost and 1450MHz memory OC; this meant BF3 at 1440p in their test went from 137fps to 155fps.
However one also has to understand how Boost3 behaves to see the OC working moderately well, meaning to OC one has to know the profile behaviour-threshold requirement of fan speed/voltage/frequency in a more dynamic way than before.
To quote them:
These are very good results, but the GTX 1080 seems to be limited by the board's power limit and temperatures going above 83°C, which will both result in lower clocks due to NVIDIA Boost. Still, the real-life performance gains are there and are significant
.
While some may complain about temperature/OC, this will be crude until one gets to grip with the behaviour and variables of Boost3 and importantly how they set those parameters including fan and maintain that fine-tuning - TPU did this the old fashioned way where going forward one needs something like the beta EVGA OC utility that was used at the NVIDIA announcement.
But lets be honest, true OC comes from the AIB designs with their greater spec/components, cost comparison is a different discussion to this point.

Importantly to assist with this point;KitGuru did a bit using the EVGA beta tool which you can read to get a feel how the variables mean this is more complex and dynamic than in the past: Nvidia GTX 1080 Graphics Card Review
Cheers
 
Last edited:
Who cares about firestrike? Go read computerbase's review, they did OC vs OC in games.
So they used a Gigabyte 980Ti gaming on their review which has a 19% overclock over a base 980Ti (max boost). They OC the 1080 to a max overclock of 17% but stated that it could not sustain due to power/temps and dropped to 8-11% (which will most likely be resolved with AIB cards) and still the 1080 was about 15% faster...... Seems to me that it looks pretty promising with AIB cards w/ more power and cooling.
 
Thanks for the review. It's a very nice card but I don't really see that it's worth upgrading my 980ti when I game at 1440p. Sure it's 30% faster on average but I'm maxing out graphics settings now in every game I play with no noticable issues.
 
No, not upset at all... I've been arguing that the extra cost of the FE cooler is justified. I was just trying to put the idea into terms simpletons who were bitching about the FE costing the same from AIB partners as Nvidia could understand. I think Nvidia spent a shit ton of money designing a badass fucking cooler for the OG Titan that performs extremely well... And the last two generations they have offered that cooler on their higher end reference cards... now they want to get paid for their efforts. Absolutely nothing wrong with that.

Maybe I've been up too long and I'm becoming incoherent. We keep butting heads when we are actually agreeing, lol.

I just think the way Nvidia released the info was mishandled. They could have just said something like:

"We're offering the 1080 with a Titan cooler for $699. To give our customers other options we are offering a PCB only (no cooler) option to our AIB partners with a suggested retail price of $599. It will then be up to our board partners to come up with a cooling solution and price the product accordingly."

I would not agree with the cooler bit - compared to the AMD competition Nvidia design pales. Just look at the temperatures while the AMD Fury is a 250w+ GPU. Too bad Fiji OC like shit. As for OEM coolers, looking at the Nano which is a 175w card and half the size (well not quite but significantly) looks to even cool better then the crumple up paper cooler design.

Reality check - The MSRP for the 1080 is - - - $699. That is what it is going for. When will that change? When supplies are plentiful will help but more importantly if there is competition. Blame AMD, Nvidia is out to make money so they can make even better cards for making even more money.

I was surprised that Async compute is not considered a worthy or big performance avenue here. Fury X goes up 9% when using it, 1080 3% (that I predict will increase). Developer of AoTs said they only moderately used Async Compute yet it is giving a significant advantage. I also think we need to see other titles using this technique and how it plays out in the end but the potential is actually huge.

If one is a 980 owner and wanting to upgrade and get a 1080 for $599 (if that happens this year is to be seen) why not? I can see waiting on the price now and/or what AMD can bring with Vega too. As for FuryX owners, that 4gb ram is really starting to be limiting. 980Ti TitanX owners I would not recommend upgrading at this time since it would be some what minor.

This review is a first look and a very good one at that exposing the clock speeds and getting a feel for the card. Next one will dig deeper which most other sites will probably never get to. Then the following ones. I hope HardOCP also gets a chance to review Polaris cards as well while at this time I do not think they will be direct competitors to the 1070/1080 but could be better perf/$ cards. Also CFX/SLI could be very interesting.

As for the 1070 review, I hope HardOCP looks at to see if the 8gb is fully accessible at full rated speed on their review.
 
It's getting around that Nvidia handed out a 61 page GTX 1080 handbook for reviewers. Is it true?
 
Seems a bit toasty for a cooler they're charging an additional $100 for.

I skipped Kepler and Maxwell so let's see if I can hold out for Volta. :)

You just don't have a need to upgrade from what you had over the last five years.
Upgrade your monitor to 4K and you'll be standing in line for one of these aswell, along with other parts.
 
As for the 1070 review, I hope HardOCP looks at to see if the 8gb is fully accessible at full rated speed on their review.

I sure hope it isn't.

With the 970 essentially EOL, some guys are really going to miss bringing up that VRAM issue.

Hopefully the 1070 will only be 7.5GB to keep you guys in business. Otherwise, the forums will be a lot more boring places.
 
I sure hope it isn't.

With the 970 essentially EOL, some guys are really going to miss bringing up that VRAM issue.

Hopefully the 1070 will only be 7.5GB to keep you guys in business. Otherwise, the forums will be a lot more boring places.

For what it's worth... Jay from Jayztwocents has stated something about the architecture of the 970 and 1070 being different and there being zero chance of another memory debacle.
 
So they used a Gigabyte 980Ti gaming on their review which has a 19% overclock over a base 980Ti (max boost). They OC the 1080 to a max overclock of 17% but stated that it could not sustain due to power/temps and dropped to 8-11% (which will most likely be resolved with AIB cards) and still the 1080 was about 15% faster...... Seems to me that it looks pretty promising with AIB cards w/ more power and cooling.

14% for a $700 card at max OC? I don't see anything promising w/this version. Yeah AIB will be nice and will push it another 10% probably but that's still not worth the jump. That's why big Pascal and Vega are the way to go for 980 Ti and Titan X owners.
 
Last edited:
  • Like
Reactions: STEM
like this
Well yeah, that's why it sucks for everybody.

Another consequence of the reference card being power-constrained is that many people historically purchased them to use with aftermarket cooling like waterblocks. But since the 1080 is power-constrained to 2.1Ghz, and can sustain 2.1Ghz with the weak reference blower at only 55%, extra cooling actually won't help. At all.

DF-1: Comparing the worst-case stock 1080 versus the best-case overclocked 980ti is not particularly meaningful. You'll overclock the 1080 too, even the reference one overclocks like a beast
Has that been proven in long term gaming or just short term benchmarks?
 
Thanks for the awesome in-depth review Brent and Kyle. Definitely time to upgrade from my 760 to a 1070 OC lol.
 
I freaked out a little when I saw that the 1080 was consuming 308W... but then I realized that they were measuring the entire system and not just the card. The peak consumption on these is almost exactly the same as the card I'm using now. I suspected this before, but now it's confirmed.

This card is actually looking like a perfect drop-in replacement for my 670 in terms of power consumption. I've been waiting for a decent upgrade that wouldn't tax my PSU much more than my current one or heat up my room. I've been running that thing for 3 years now, and I'm not sure how much overhead it has left.

The only thing I'm wondering about is what's up with these new 8-pin power connectors... my card currently uses 2 6-pin power connectors. Wonder if I'll need to buy an adapter? I'm not sure that my PSU has an 8-pin power connector. It's been a while since I opened it up.
 
The only thing I'm wondering about is what's up with these new 8-pin power connectors... my card currently uses 2 6-pin power connectors. Wonder if I'll need to buy an adapter? I'm not sure that my PSU has an 8-pin power connector. It's been a while since I opened it up.

8-pin isn't new, my 5 year old Ultra x4 1200W modular PSU has 3 of them but the cable splits off to allow for 6-pin connections as well.
 
8-pin isn't new, my 5 year old Ultra x4 1200W modular PSU has 3 of them but the cable splits off to allow for 6-pin connections as well.

Yeah, I just opened it up and checked... apparently it actually is a 6+2 connector, meaning it has the other two pins off to the side. So my system is ready, that's good to know.
 
I freaked out a little when I saw that the 1080 was consuming 308W... but then I realized that they were measuring the entire system and not just the card. The peak consumption on these is almost exactly the same as the card I'm using now. I suspected this before, but now it's confirmed.

This card is actually looking like a perfect drop-in replacement for my 670 in terms of power consumption. I've been waiting for a decent upgrade that wouldn't tax my PSU much more than my current one or heat up my room. I've been running that thing for 3 years now, and I'm not sure how much overhead it has left.

The only thing I'm wondering about is what's up with these new 8-pin power connectors... my card currently uses 2 6-pin power connectors. Wonder if I'll need to buy an adapter? I'm not sure that my PSU has an 8-pin power connector. It's been a while since I opened it up.

It's pretty impressive to think that it's using only as much power as a 670. Also, this card will be a massive upgrade for you and if you've been running a 670 for three years I'm guessing you'll be able to run the 1080 for a long time.
 
Just a quick question. Does the 1080 GTX still support DSR? I assume it does, but would like to make sure.
 
GTX 1080 has an 70% increased clock speed over the FuryX
GTX 1080 has an 30-35% increased clock speed over the 980ti.
Let the GHZ war begin ont he GPU side aswell!
Anyway, Pascal only brings higher clocks and lower TDP, thats it. They had to increase the clocks, otherwise we would have a 980ti with half the TDP and people would scream about performance then.
AMDs job right now is to find a way to increase their clockspeeds aswell, nothing else, because clock for clock, they are on paar with Nvidia.
 
Great card, can't believe the performance this card has at such a low power delta.

Can't wait to see what the 1070 can do as I primarily play at 1080 and 1440, the GTX1080 is almost playable at 4k w/ high or max'sih settings!
 
Yep, the PCB is designed to handle up to a 6+2 power phase but they went for 5+1. Being an optimist, this could actually bode well for the Ti and or Titan version.
No, that's not what I am talking about. I know there is a spot for another phase at the bottom, I was talking about the lack of the second fet on each of the phases that actually are there.
 
GTX 1080 has an 70% increased clock speed over the FuryX
GTX 1080 has an 30-35% increased clock speed over the 980ti.
Let the GHZ war begin ont he GPU side aswell!
Anyway, Pascal only brings higher clocks and lower TDP, thats it. They had to increase the clocks, otherwise we would have a 980ti with half the TDP and people would scream about performance then.
AMDs job right now is to find a way to increase their clockspeeds aswell, nothing else, because clock for clock, they are on paar with Nvidia.
You say 30-35% as if its never lower. Some reviews have it averaging lower than that and in some games its only 15-20% better. The 980 ti also has more overclocking headroom so at the end of the day it would be a pathetically small upgrade when looking at oced 980 ti to oced 1080. I still dont get all this praise as in my opinion this card is a joke for a next gen product at 700 bucks.
 
Back
Top