GTX 1080s, are people kidding themselves?

[Dackzy]

companies will always like to oversell their product and a lot of people will believe them. People also just look at clock speeds, but they don't scale very well, when it comes to the 1080.

[App4that]

6 minutes ago, xAcid9 said:

#~2.1GHZgate  

 

Fun to see when people justifying their extra $100(in Msia) purchase on 1070 Strix with "better memory overclock" compare to 1070 Jetstream.

Similar thermal, similar performance, similar warranty period.

Now AIB releasing "higher" factory overclock 1070/1080 with higher price just to milk the shit of their customer.

And it's any different with Maxwell? Or Hawaii? Owned both. Paid more for a OC edition Nitro that would OC for shit. It was only 30 bucks but still... 

 

The difference here is jealousy, pure and simple. Once Pascal is sitting on shelves few people will question the different models. 

[xAcid9]

4 hours ago, App4that said:

And it's any different with Maxwell? Or Hawaii? Owned both. Paid more for a OC edition Nitro that would OC for shit. It was only 30 bucks but still... 

 

The difference here is jealousy, pure and simple. Once Pascal is sitting on shelves few people will question the different models. 

Yes, the term "silicon lottery" or "binned/cherry picked" doesn't apply to Pascal. Just because your last gen cards can't OC shit doesn't mean other people that bought the same card can't overclock higher than yours.

 

It's good and bad at the same time but i cringed everytime people recommend more expensive 1070/1080 with reasons like "they overclock better because it use better binned chip"

[Hunter-97-G]

I'm reaching 2063-2076 Mhz without touching voltage, so if I do I'm hoping to get to 2.1. It won't be easy, though.

 

Still, remember 2000-2100 is about 400 to 500 Mhz overclock from the base clock of the card, which is GREAT overclocking. People have to remember that GPU Boost is already automatic overclocking. But sure, 400 Mhz is 25% of 1600 when it was 40% of 980 Ti's 1000 Mhz base clock. It is awesome to hit 2 Ghz, IMO.

[Kinda Bottlenecked]

21 minutes ago, Hunter-97-G said:

I'm reaching 2063-2076 Mhz without touching voltage, so if I do I'm hoping to get to 2.1. It won't be easy, though.

 

Still, remember 2000-2100 is about 400 to 500 Mhz overclock from the base clock of the card, which is GREAT overclocking. People have to remember that GPU Boost is already automatic overclocking. But sure, 400 Mhz is 25% of 1600 when it was 40% of 980 Ti's 1000 Mhz base clock. It is awesome to hit 2 Ghz, IMO.

 

Can you really count the base clock? By default, these cards are already boosting to a boost clock of 1900mhz on average because of GPU boost. Base clock by now kinda means nothing.

 

[Godlygamer23]

6 hours ago, ivan134 said:

Yes, but making rational arguments gets you labeled a fanboy. I've never said the 1080 and 1070 were bad cards. They're actually amazing cards. Node shrink allowed Nvidia to fit 601mm2 performance into 314mm2. I've only said they don't cost what Nvidia is charging for them. 314mm2 chips do not cost what Nvidia is charging for them.

Companies never charge what the cost actually is if they intend to make a profit - that's just an absurd expectation.

[Lays]

3 minutes ago, Godlygamer23 said:

Companies never charge what the cost actually is if they intend to make a profit - that's just an absurd expectation.

I think he's forgetting marketing costs, research and development costs etc.

 

Just because something costs $50 to make doesn't mean they should sell it for $75.   There's so many other things at stake that they have to fund.

 

During they keynote they said they spent billions developing Pascal, they have to make that money back somehow.

[Godlygamer23]

2 minutes ago, Lays said:

I think he's forgetting marketing costs, research and development costs etc.

 

Just because something costs $50 to make doesn't mean they should sell it for $75.   There's so many other things at stake that they have to fund.

Exactly. If Apple charged what it cost them to build all of their products, they wouldn't be here because their profit margin would remain at $0. They would have no way of paying employees, electricity, research and development, marketing, etc.

[DAnewguy]

7 hours ago, The Spark said:

The GTX 980 ti was very early spotted as an excellent overclocker with proper cooling, some of the aftermarket cards overclocked extremely well when provided with the right care.

 

The GTX 1080 is overclocked almost to the max already when shipped by manufacturers, the average overclock will max at around 2050 mhz and the lucky ones may get it just beyond 2100 mhz, the highest I personally have seen is my neighbours FE edition with 2110 mhz and he is happy to have won the lottery. Mine reaches 2050 and wont go any further in spite of heat not being an issue.

 

I am seeing some very fancy solutions coming out for the GTX 1080 which includes awesome water cooling but that has to for the noise right? I have seen a lot of testing and reviews of this card and they all stop at 2000-2100 mhz unless they have been modded heavily (home made), with an unlocked bios there may be an extra 100 mhz in storage. Overall however the 1080 unlike the 980ti just doesn´t show the same improvement when overclocked.

 

My MSI 1080 gaming X is extremely silent even when under heavy load, it just wont get much better than that and due to the locked bios the extra 6 pin power connector is just there for the kicks, I guess a hacked bios might change things if I am willing to scrap the warranty on the card but with my card stopping already at 2050 I will never take that bet.

 

Are people cheating themselves when paying absurd amounts for cards which will deliver nothing more than the cards available today offer?

 

With the 980 ti I could really understand water cooling and overall improved additions to the card, as for 1080 ... it just seems rather pointless.

 

 

Wait for 1080ti we will see

[ARod]

Point blank, they did most of the work for us. I for one can appreciate purchasing a card that is already maxed out in terms of potential. But, we're talking about it at a point where the market has yet to be flooded with these cards. Folks need to sit tight until last week of July or wait till August when prices will become more reasonable though.

 

To keep things in perspective, the Zotac GTX 1080 AMP has an MSRP of $659.99. I can pick one up at distribution(the channel) for $588.00. Distributors in the channel only mark things up about 5% or so, I imagine their cost to acquire direct from Zotac in mass is about $550.00 per card. So, in the end overall from beginning to end it's a 20% markup. That's about the norm.

[DXMember]

it doesn't have as many shaders, that's why overclocks don't help that much in performance scaling

[Yamoto42]

7 hours ago, Hunter-97-G said:

I'm reaching 2063-2076 Mhz without touching voltage, so if I do I'm hoping to get to 2.1. It won't be easy, though.

 

Still, remember 2000-2100 is about 400 to 500 Mhz overclock from the base clock of the card, which is GREAT overclocking. People have to remember that GPU Boost is already automatic overclocking. But sure, 400 Mhz is 25% of 1600 when it was 40% of 980 Ti's 1000 Mhz base clock. It is awesome to hit 2 Ghz, IMO.

I remember the good old days of the MHz wars...

 

Overclocking is very much a percentile effect, which means it's diminishing returns are exponential.  So from that regard, the 40% overclock of the 980Ti is far superior to the 1080's 25%.

 

Another factor of performance gains in overclocking is core count.  Unless someone REALLY wants to argue that 2560 > 2816...this is again a clear, albeit potentially marginal victory for the 980Ti.

 

Finally there's IPC.  Now this doesn't have a clear cut answer...Pascal should be some IPC gain, but there have been sources saying there was actually an IPC decrease...so who knows...I don't care to validate that.  Honestly what we do know is that there's little change.  If you want to find specific numbers, that's fine.  I have neither the time nor care.

Now I'm not math expert...but I'd place my money on 1.4*2816 being greater than 1.25*2560.

Now, IPC doesn't really factor into the 'quality' of an overclock.

 

If someone wants some fun homework though, I would be interested to see a 95% confidence interval (using the above formulas to remove speed and core count factors) on the percentile IPC difference going from the 980Ti to the 1080...I don't care enough to do it myself, but i would be interested to see it.

 

Either way, I'm perfectly happy with my 980Ti until/if SMP becomes widely used.

 

6 hours ago, Pohernori said:

 

Can you really count the base clock? By default, these cards are already boosting to a boost clock of 1900mhz on average because of GPU boost. Base clock by now kinda means nothing.

 

Also this time 9001.

[Vellinious]

12 hours ago, The Spark said:

The GTX 980 ti was very early spotted as an excellent overclocker with proper cooling, some of the aftermarket cards overclocked extremely well when provided with the right care.

 

The GTX 1080 is overclocked almost to the max already when shipped by manufacturers, the average overclock will max at around 2050 mhz and the lucky ones may get it just beyond 2100 mhz, the highest I personally have seen is my neighbours FE edition with 2110 mhz and he is happy to have won the lottery. Mine reaches 2050 and wont go any further in spite of heat not being an issue.

 

I am seeing some very fancy solutions coming out for the GTX 1080 which includes awesome water cooling but that has to for the noise right? I have seen a lot of testing and reviews of this card and they all stop at 2000-2100 mhz unless they have been modded heavily (home made), with an unlocked bios there may be an extra 100 mhz in storage. Overall however the 1080 unlike the 980ti just doesn´t show the same improvement when overclocked.

 

My MSI 1080 gaming X is extremely silent even when under heavy load, it just wont get much better than that and due to the locked bios the extra 6 pin power connector is just there for the kicks, I guess a hacked bios might change things if I am willing to scrap the warranty on the card but with my card stopping already at 2050 I will never take that bet.

 

Are people cheating themselves when paying absurd amounts for cards which will deliver nothing more than the cards available today offer?

 

With the 980 ti I could really understand water cooling and overall improved additions to the card, as for 1080 ... it just seems rather pointless.

 

 

Thermal levels, just like with Maxwell, are more what decide core clocks, as opposed to voltage.  Seems to be even more prevalent with Pascal.  The lower the temps, the higher it'll overclock.  Obviously there are more factors that will play into that, but....it's looking more and more like, if you want a better overclock than the "average", you'll need some really good cooling.  Think in the line of, below 30c max core temp.  The lower the better.

[SteveGrabowski0]

10 hours ago, App4that said:

The HOF boosts to over 2GHz on its own lol. Pascal overclocks like crazy. My buddy with a 1080 FE runs his over 2GHz. Still makes me giggle just typing it lol. 

 

The 1080ti will be to 1440 as the 980ti was to 1080, total overkill. Why I'm getting a 1080 HOF or AMP extreme

You're upgrading from a 980 Ti Hybrid to a 1080? Man if I had that kind of disposable income. 

[App4that]

19 minutes ago, SteveGrabowski0 said:

You're upgrading from a 980 Ti Hybrid to a 1080? Man if I had that kind of disposable income. 

Need a card for a second rig, community rig for me and the roommate hooked to a TV. So might as well replace the 980ti and put it in the second rig. But it's also a sizeable upgrade in most of the games I play. Looking at the FTW hybrid, AMP Extreme, and HOF. Both the HOF and Extreme fit my build theme, but I'm guessing the Hybrid might OC better. We'll see.

 

I'd wait for the 1080ti but I need to build the HTPC by Sept and the 1080ti looks to be overkill for 1440.

[Tech Inquisition]

20 hours ago, App4that said:

And it's any different with Maxwell? Or Hawaii? Owned both. Paid more for a OC edition Nitro that would OC for shit. It was only 30 bucks but still... 

 

The difference here is jealousy, pure and simple. Once Pascal is sitting on shelves few people will question the different models. 

Preaching to the choir... I have an EVGA GTX 980 Classified... it's ok as from the factory... only able to push 55 on the core stable for gaming...

 

[winningsince1337]

23 hours ago, The Spark said:

snip

 

 

This, and Jayztwocents said the only cards he had that consistently hit 2100 were his founders cards

[done12many2]

23 hours ago, The Spark said:

The GTX 980 ti was very early spotted as an excellent overclocker with proper cooling, some of the aftermarket cards overclocked extremely well when provided with the right care.

 

The GTX 1080 is overclocked almost to the max already when shipped by manufacturers, the average overclock will max at around 2050 mhz and the lucky ones may get it just beyond 2100 mhz, the highest I personally have seen is my neighbours FE edition with 2110 mhz and he is happy to have won the lottery. Mine reaches 2050 and wont go any further in spite of heat not being an issue.

 

I am seeing some very fancy solutions coming out for the GTX 1080 which includes awesome water cooling but that has to for the noise right? I have seen a lot of testing and reviews of this card and they all stop at 2000-2100 mhz unless they have been modded heavily (home made), with an unlocked bios there may be an extra 100 mhz in storage. Overall however the 1080 unlike the 980ti just doesn´t show the same improvement when overclocked.

 

My MSI 1080 gaming X is extremely silent even when under heavy load, it just wont get much better than that and due to the locked bios the extra 6 pin power connector is just there for the kicks, I guess a hacked bios might change things if I am willing to scrap the warranty on the card but with my card stopping already at 2050 I will never take that bet.

 

Are people cheating themselves when paying absurd amounts for cards which will deliver nothing more than the cards available today offer?

 

With the 980 ti I could really understand water cooling and overall improved additions to the card, as for 1080 ... it just seems rather pointless.

 

The 980 Ti will overclock better in extreme overclocking only, which only pertains to an extremely limited amount of people.  Otherwise a 980 Ti that is overclocked on ambient cooling methods is not comparable to a 1080 overclock with ambient cooling methods.  The other difference is when pushed to their max, the 1080 will actually remain game playable.  The same could not be said for the 980 Ti.  Just because you can get it to pass a benchmark doesn't mean that it's stable enough to game with at it high overclock.

 

Water cooling and upgraded power delivery setups aren't necessarily a waste on the 1080, but they aren't going to take you much higher than the 2100 region.  It's just the way it is.  A stock FE card will run the same clock rates as a custom PCB card with 2 x 8 pins provided you adjust the stock fan curve.

 

I run water blocks on my 1080 SLI just because my computer is dead silent and I don't need some GPUs messing that up.  I was glad that the 1080s don't produce the coil whine that my 980 Tis did.  They were horrible for coil noise.

 

The thing that screwed up perception of the 1080 the most with regards to overclocking was how nvidia attempted to mislead everyone into thinking these cards would overclock insanely high.  They showed a 1080 idling at 2100 at the debut and overclockers automatically assumed that they were going to be able to push it a lot further.  Nvidia knew what they were doing and this led to a pretty big let down for a lot of people. 

[done12many2]

21 minutes ago, winningsince1337 said:

This, and Jayztwocents said the only cards he had that consistently hit 2100 were his founders cards

 

This has been my experience when compared to buddies setups.  My FEs are on water and the cards just run at 2100+ all day long if I want.  Friends with some non FE cards are having trouble staying around 2000 or just above.  It might be the fact that I'm on water, but I'm seeing a lot of FE cards holding 2100 on their stock air coolers.  Who knows?

[winningsince1337]

10 minutes ago, done12many2 said:

 

This has been my experience when compared to buddies setups.  My FEs are on water and the cards just run at 2100+ all day long if I want.  Friends with some non FE cards are having trouble staying around 2000 or just above.  It might be the fact that I'm on water, but I'm seeing a lot of FE cards holding 2100 on their stock air coolers.  Who knows?

Nvidia says they are not cherry picking cores, for the sake of this discussion we'll assume they aren't. The only other possibility is that nvidia knows exactly what components produce the most consistent outputs and they then built their board around the core. By picking the board components around the core's design requirements Nvidia's engineers, who inherently know more about the core then AIBs, can build a better board than AIBs. Makes sense if you think about it

[done12many2]

3 minutes ago, winningsince1337 said:

Nvidia says they are not cherry picking cores, for the sake of this discussion we'll assume they aren't. The only other possibility is that nvidia knows exactly what components produce the most consistent outputs and they then built their board around the core. By picking the board components around the core's design requirements Nvidia's engineers, who inherently know more about the core then AIBs, can build a better board than AIBs. Makes sense if you think about it

I wouldn't argue that point at all. 

[SSL]

22 minutes ago, winningsince1337 said:

Nvidia says they are not cherry picking cores, for the sake of this discussion we'll assume they aren't. The only other possibility is that nvidia knows exactly what components produce the most consistent outputs and they then built their board around the core. By picking the board components around the core's design requirements Nvidia's engineers, who inherently know more about the core then AIBs, can build a better board than AIBs. Makes sense if you think about it

 

Occams Razor applies here. This sounds needlessly convoluted; so there's supposedly "components" that allow the chip to perform better? But not additional power delivery or whatever.

[winningsince1337]

1 hour ago, SSL said:

 

Occams Razor applies here. This sounds needlessly convoluted; so there's supposedly "components" that allow the chip to perform better? But not additional power delivery or whatever.

I'm a mechanical engineering students and have taken a number of electrical classes before. TLDR: More phases doesn't equal higher overclocks. Electrical chip stability from the power aspect has more to due with tight voltage control and hitting an ideal impedance than just throwing more mosfets in to the power delivery system. A mosfet is really just a big switch more or less. It doesn't do anything for power delivery than potentially adding unwanted capacitance and/or impedance to the power delivery circuitry unless the mosfets don't have the amperage capacity to provide enough power. However, being that 1080 is such a low power card, the power per phase is really low compared to a 980ti.