Nvidia GeForce GTX 1080 rumor mill just blew up.

defaultluser said:
It would help if you could EXPLAIN YOURSELF, which you obviously cannot if all you can do is call me names Tell us in exact words what was so broken with the GTX 980 Ti, or expect us to ignore your ass.

And don't just bitch about the GTX 980 Ti - you have to justify why this problem on the 980 Ti has any potential to appear on the GTX 1080.
Click to expand...

the reason reference 980tis hit the tdp wall so easily was, in part, due to the power delivery not being up to the task, the output mosfets in GPU VRM are rated for a certain current depending on temperature, as they get hotter this current is reduced. this is why custom PCB are the best
 
Alkaladur said:
the reason reference 980tis hit the tdp wall so easily was, in part, due to the power delivery not being up to the task, the output mosfets in GPU VRM are rated for a certain current depending on temperature, as they get hotter this current is reduced. this is why custom PCB are the best
Click to expand...

Thank you for communicating. I've hit a fucking wall of silence with Someguy, who expects me to mind meld with him :D.

Your post backs-up my own - if the cooler or VRMS aren't up to ask, just wait for the custom PCBs. It's annoying I know, but you can wait an extra month of two for the real deal.

And just because it happened with the 980 Ti doesn't mean it will happen with the GTX 1080. But I can see why you are concerned now, at least for the viability of the launch cards.

And crappy VRMs DOES NOT INVALIDATE MY EARLIER STATEMENT. If the total power handling of the card is 225w, the target TDP will be significantly below that. That's the whole fucking statement I saw and I jumped into his thread to quash, when Someguy was dreamily believing that 225w could be the 1080's TDP, and not it's peak overclock headroom.

He was comparing the 225w power handling of the GTX 1080 to the 250w TDP of the 980 Ti (actual theoretical power handling = 300w). And then he got derailed bitching about how crappy the 980 Ti VRMS were. WHAT THE FUCK?
 
Last edited:
Armenius said:
I could run my Titan X at 425W all day if I had the cooling for it.
Click to expand...

I know someone running dual 980tis on water... At 1640mhz.

I'll poop in my hat and eat it if an overclocked gp104 card is significantly faster than that, a single card that is.

Remind me.

Speaking of which, if the consumer lineup of the pascal cards *does* half double fp16 throughput, to what extent to we think it will be used in games, and how long will it take for it to potentially takeoff?

That is something I've never see mentioned in all the pascal and gcn 1.3 discussions

I'm asking by the way, I have no idea where and how fp16 can be used in games
 
Last edited:
people seems to forget that since kepler Nvidia boost clock create false expectation that the cards have a low overclock headroom which the fact is, the card will be always running well above the advertised base clock and even more specially for highly customized AIB cards.

Reference TDP mean really shit unless one is specifically interested in run reference board, non-reference cards always run at max TDP as possible so they will provide the higher boost clock out of the box as possible which in some cases are way higher than even the max achievable overclock in reference cards, Again that produce a false sensation that the GPU have a low overclock headroom, specially to those who have little knowledge on overclocking and how the GPU behave, some people have to remember that the advertised TDP also mean that TDP is at the base clock of reference boards. which just nobody run at since kepler, that's the whole point of GPU boost, auto-overclock as max as possible out of the box between the advertised parameters.
 
Araxie said:
people seems to forget that since kepler Nvidia boost clock create false expectation that the cards have a low overclock headroom which the fact is, the card will be always running well above the advertised base clock and even more specially for highly customized AIB cards.

Reference TDP mean really shit unless one is specifically interested in run reference board, non-reference cards always run at max TDP as possible so they will provide the higher boost clock out of the box as possible which in some cases are way higher than even the max achievable overclock in reference cards, Again that produce a false sensation that the GPU have a low overclock headroom, specially to those who have little knowledge on overclocking and how the GPU behave, some people have to remember that the advertised TDP also mean that TDP is at the base clock of reference boards. which just nobody run at since kepler, that's the whole point of GPU boost, auto-overclock as max as possible out of the box between the advertised parameters.
Click to expand...

the boost clock is fixed though, even compared to the boost clock there was tons of headroom. Tons.

An average non-reference 970 will clock to 1500 core, that's a good 25% increase over reference boost, Many higher quality/higher binned non-reference 970s clock to 1550-1600. That's a lot of headroom man. Boost clocks could easily have been 1350~ for 970/980/980ti/titanx with decent cooling. maybe not the reference blower

NVIDIA GeForce GTX 980 Ti 6 GB Review
 
the GTX 970 is a whole thing different as that card doesn't have a reference model, every one of those cards is customized, and at the worse case possible scenario some use the GTX 670 PCB, "the reference" TDP of that card was intended to 145W, however a card like the EVGA 970 FTW is closer to 170-180W because of the way higher boost clock those cards can achieve out of the box, most tend to go above 1400mhz out of the box and this is what cause confusion in people, as they think the overclock headroom is what they can achieve above out of the box clock, which isn't true, the true margin of overclock have to be taken from the base clock as you said Alkaladur just compare your link of the reference 980TI review to this one Gigabyte GTX 980 Ti G1 Gaming 6 GB Review

Reference card
211W Average load
238W Peak Load
277W Maximum Load

G1 Gaming
259W Average Load
293W Peak Load
359W Maximum Load.

its a whole point to consider when overclocking.


Also as talking about VRM, some people also tend to forget the role of the customized PCB, some AIB offer more efficient power circuitry which can offer higher clocks at lower power which is the case for example of MSI cards, but we have others like Gigabyte which are focused on the highest possible overclock capacity but also tend to be way more inefficient in the power consumption.
 
Araxie said:
the GTX 970 is a whole thing different as that card doesn't have a reference model, every one of those cards is customized, and at the worse case possible scenario some use the GTX 670 PCB, "the reference" TDP of that card was intended to 145W, however a card like the EVGA 970 FTW is closer to 170-180W because of the way higher boost clock those cards can achieve out of the box, most tend to go above 1400mhz out of the box and this is what cause confusion in people, as they think the overclock headroom is what they can achieve above out of the box clock, which isn't true, the true margin of overclock have to be taken from the base clock as you said Alkaladur just compare your link of the reference 980TI review to this one Gigabyte GTX 980 Ti G1 Gaming 6 GB Review

Reference card
211W Average load
238W Peak Load
277W Maximum Load

G1 Gaming
259W Average Load
293W Peak Load
359W Maximum Load.

its a whole point to consider when overclocking.


Also as talking about VRM, some people also tend to forget the role of the customized PCB, some AIB offer more efficient power circuitry which can offer higher clocks at lower power which is the case for example of MSI cards, but we have others like Gigabyte which are focused on the highest possible overclock capacity but also tend to be way more inefficient in the power consumption.
Click to expand...

Maximum load is furmark, peak and average are what really concern most people, and the tdp is 250w, which is very much in line with the peak on the reference card
clock_vs_voltage.jpg


Considering this blower runs at 84c load, I can see why it tops out at 1200mhz :p

G1 Gaming:
clock_vs_voltage.jpg


MSI is actually not more efficient i don't know where you get that
MSI GeForce GTX 980 Ti Gaming 6 GB Review

The GPU will draw X power at whatever clock and voltage you have set, the hotter it is, the more power it draws for a set clock and voltage. The cards that draw less at a given setting are likely better cooled
 
Last edited:
I was talking in general, not only about the 980TI Gaming, and not only about GTX 980TI, but about more generally and historically speaking (im one of the old heavy OC'ers =) ) just to point more things into the mix look at the MSI 980TI lightning MSI GeForce GTX 980 Ti Lightning 6GB Review the much better and optimized power delivery circuitry allow this card to reach higher boost with higher voltage at a lower power consumption than the Regular MSI GTX 980TI Gaming.

MSI 980TI Gaming:

MSIGaming.PNG


clock_vs_voltage.jpg


power_peak.gif



980TI Lightning

MSILightning.PNG


clock_vs_voltage.jpg


power_peak.png




The same principle apply to the Gigabyte GTX 980TI Extreme Gigabyte GTX 980 Ti XtremeGaming 6GB Review a card that reach higher clock at higher voltage but still consuming less power than the Gigabyte GTX 980 TIG1 Gaming,.


Gigabyte GTX 980 TI G1 Gaming

G1Gaming.PNG


clock_vs_voltage.jpg


power_peak.gif


Gigabyte GTX 980TI Extreme

Extreme.PNG


clock_vs_voltage.jpg


power_peak.png




this just prove how important is the customization on the PCB to achieve higher amount of performance and still be able to reach a lower than expected power consumption, so again, reference TDP and reference PCB mean always nothing.
 
Last edited:
Big Pascal should be a beast at 15B transistors. Titan X only has 8B. I am not interested in anything else.

780ti -> Titan X was a 50-60% increase with only increasing the transistor count from 7.1B to 8B.

Should be interesting.
 
Dayaks said:
Big Pascal should be a beast at 15B transistors. Titan X only has 8B. I am not interested in anything else.

780ti -> Titan X was a 50-60% increase with only increasing the transistor count from 7.1B to 8B.

Should be interesting.
Click to expand...

This reminds me exactly when I had GTX 480s in SLI. I now have two GTX 680 2GB in SLI, and waiting to see what performance GTX 1080. I was going to do a single GTX 980, but just didn't have enough performance increase from SLI 680's for what I was playing.

Saving on heat/energy alone might be worth it, just like when i was doing two GTX 480's. That rig, with the x58 overclocked, ran VERY VERY hot!

I really wonder what Big Pascal will do...and the cost!
 
DogChainX said:
This reminds me exactly when I had GTX 480s in SLI. I now have two GTX 680 2GB in SLI, and waiting to see what performance GTX 1080. I was going to do a single GTX 980, but just didn't have enough performance increase from SLI 680's for what I was playing.

Saving on heat/energy alone might be worth it, just like when i was doing two GTX 480's. That rig, with the x58 overclocked, ran VERY VERY hot!

I really wonder what Big Pascal will do...and the cost!
Click to expand...

I'll start saving for Big Pascal, or BP100, right now :)
 
Dayaks said:
Big Pascal should be a beast at 15B transistors. Titan X only has 8B. I am not interested in anything else.

780ti -> Titan X was a 50-60% increase with only increasing the transistor count from 7.1B to 8B.

Should be interesting.
Click to expand...

Titan X dropped a lot of double precision performance to make that happen. Going from Titan X to GP104 is the exact opposite with a lot of the extra transistors are going toward DP. You're probably looking at the same 50% increase in games.
 
MangoSeed said:
Titan X dropped a lot of double precision performance to make that happen. Going from Titan X to GP104 is the exact opposite with a lot of the extra transistors are going toward DP. You're probably looking at the same 50% increase in games.
Click to expand...

Screw that then. I'm might have Maxwell for a good long time. :). I need at least 100% increase to really make it feel like it's worth it.

Maybe AMD won't be dicking around like nVidia.
 
Dayaks said:
Big Pascal should be a beast at 15B transistors. Titan X only has 8B. I am not interested in anything else.

780ti -> Titan X was a 50-60% increase with only increasing the transistor count from 7.1B to 8B.

Should be interesting.
Click to expand...

You aren't getting Big Pascal any time soon. The 1080 will release in a couple of months, then Big Pascal will launch with the Titan product at about christmas time, maybe even later. then, six-to-nine months after that, The Big Pascal 'affordable' version will release.
 
KazeoHin said:
You aren't getting Big Pascal any time soon. The 1080 will release in a couple of months, then Big Pascal will launch with the Titan product at about christmas time, maybe even later. then, six-to-nine months after that, The Big Pascal 'affordable' version will release.
Click to expand...

that seems exaggerated, i think we'll see big pascal launch first, then the Ti version come to compete with vega
 
defaultluser said:
It would help if you could EXPLAIN YOURSELF, which you obviously cannot if all you can do is call me names Tell us in exact words what was so broken with the GTX 980 Ti, or expect us to ignore your ass.

And don't just bitch about the GTX 980 Ti - you have to justify why this problem on the 980 Ti has any potential to appear on the GTX 1080.
Click to expand...
I did several times. Its not hard to follow. You keep making shit up as you go as if i complained about 980TI and other stuff. I never complained i stated several facts and you go ape shit crazy on a high horse. I got the info i needed so i dont really care. I also have better things to go than respond in 2 seconds to your post. I just got to Seoul from Osaka Japan so i got better things to do but keep making stuff up...claiming i said things i never said. lol
 
Sucks that nVIdia is selling a midrange GPU as a flagship, yet again. I'm waiting for the real Pascal, with its cores on.
 
the gp104 is not a flagship product, its a performance product just like they did last year.
 
The highest performance, most expensive product you sell is your flagship.
 
  • Like
Reactions: DF-1
like this
rabidz7 said:
Sucks that nVIdia is selling a midrange GPU as a flagship, yet again. I'm waiting for the real Pascal, with its cores on.
Click to expand...

I'm with you 100%. I guess you can get people to accept anything. I remember when the 680-780 thing happen, people swore they wouldn't never let themselves be hoodwink liked that again, now they are lining up salivating to be abused again.
 
Alkaladur said:
This is dumb.

So Polaris 10 is a flagship?
Click to expand...
Well you need to consider the complete tier and schedules as not all models are released at once; so yeah logically the most expensive product you sell is your flagship and that will be Vega, which falls under the umbrella of this range (being presented by AMD as Polaris 11/10/Vega)..
Same way the consumer flagship card for Pascal (above the 1080) is still not expected until next year or very end of this year if one is optimistic.
I doubt anyone is really thinking the 1080 is the flagship.
Cheers
 
Last edited:
CSI_PC said:
Well you need to consider the complete tier and schedules as not all models are released at once; so yeah logically the most expensive product you sell is your flagship and that will be Vega, which falls under the umbrella of this range (being presented by AMD as Polaris 11/10/Vega)..
Same way the flagship card for Pascal still has not been released for consumers and probably not until next year or very end of this year if one is optimistic.
Cheers
Click to expand...
it doesn't work like that, you can't have an unreleased flagship product. the Polaris is the flagship until the vega is released.
 
well either way you look at it, an uniformed buyer would think Polaris is the top of the line and informed buyer knows it isn't, just as GP104 is to GP100 or what ever has HBM 2 coming with it.
 
shinotenshi said:
it doesn't work like that, you can't have an unreleased flagship product. the Polaris is the flagship until the vega is released.
Click to expand...
Err.
Come again?
By that logic you would always sell the mid range card that releases first at the same price as the officially presented flagship model until that actual flagship model is released???
Sorry but there is a difference between statement-flagship models and that of lower ones, this applies to ALL sectors and beyond GPUs.
I would agree if there was no plans for a higher model, but AMD and NVIDIA has already set their business and project-development strategy regarding what they are developing-manufacturing-distributing, just the timeframe schedule may change a bit but not the planned tiers and the associated product-specs.

Case in point look at how GPUs are presented by AMD and NVIDIA, you see different wording-hyperbole between their lower cards and say those such as Fury X and Titan X (can include 980ti to some extent but not technically a flagship product).
A flagship model is a no holds barred top spec top performance with less cost constraints (one example would be poor yields is more acceptable in a flagship model compared to a lower tier) in terms of engineering-manufacturing and rrp.

Cheers
 
Last edited:
CSI_PC said:
Err.
Come again?
By that logic you would always sell the mid range card that releases first at the same price as the officially presented flagship model until that actual flagship model is released???
Sorry but there is a difference between statement-flagship models and that of lower ones, this applies to ALL sectors and beyond GPUs.
I would agree if there was no plans for a higher model, but that is wrong as AMD and NVIDIA has already set their business and project-development strategy regarding what they are developing-manufacturing-distributing, just the timeframe schedule may change a bit but not the planned tiers and the associated product-specs.
Cheers
Click to expand...

a flagship product has to actually exist, it can't be theory, you have to be able to buy it, whether it's a car, a warship, or a video game card, the flagship is your best car, ship, card, that is actually in use or can be purchased. Until that product is is buyable, until you can drive that car, until that warship can hit the seas in formation, it is not a flagship period. a company's intentions are immaterial. AMD might want the Vega to be the flagship but until I can buy it is not, so until then the Polaris 10 is the flagship.
 
Rubbish.
A flagship loudspeaker can be announced long before it is available to the public.
 
shinotenshi said:
a flagship product has to actually exist, it can't be theory, you have to be able to buy it, whether it's a car, a warship, or a video game card, the flagship is your best car, ship, card, that is actually in use or can be purchased. Until that product is is buyable, until you can drive that car, until that warship can hit the seas in formation, it is not a flagship period. a company's intentions are immaterial. AMD might want the Vega to be the flagship but until I can buy it is not, so until then the Polaris 10 is the flagship.
Click to expand...
It is not theory because the business and project-development plan is set in concrete before it is released, meaning the tiers and their associated spec and which model designated has already been set.
These are not pie in the sky products but ones that are pre-planned for the brand in question, for AMD that is they already have a plan for Polaris 11/10/Vega that is in motion and they say their statement enthusiast product Vega is coming out next year. Not Polaris 10 being their high end statement and enthusiast product.
Cheers
 
Nenu said:
Rubbish.
A flagship loudspeaker can be announced long before it is available to the public.
Click to expand...
incorrect, what they announced is that they would like the be the flagship. the speaker that is actually available does not stop being the current flagship, until the new one is available. so what is announce is that will there next flagship when the product is released. Infact almost all those press releases will say, this will be companies X flagship when model Y is released.
 
You guys are arguing about marketing terms, like high speed vs full speed.
 
shinotenshi said:
incorrect, what they announced is that they would like the be the flagship. the speaker that is actually available does not stop being the current flagship, until the new one is available. so what is announce is that will there next flagship when the product is released. Infact almost all those press releases will say, this will be companies X flagship when model Y is released.
Click to expand...
You are talking crap.
There are place holders for the flagship products.
 
xorbe said:
You guys are arguing about marketing terms, like high speed vs full speed.
Click to expand...
Well also engineering tbh, and the planning-development-R&D cycle.
You have to have a plan otherwise you end up with a mess and no product strategy, which is critical when considering the tiers and market/consumer level/etc a set of products for a brand are to be placed in.
Cheers
 
Actually, the Radeon Pro Duo will probably be AMD's flagship until Vega.

And while you can pay for GP100 now, it comes inside a rack-mounted supercomputer with 8 of 'em. And even if you spring for the $129,000 DGX-1 and pop out one of those P100 cards, you won't be able to use it to play The Division-- it doesn't have any graphics ports. Oh, and you won't get it until Q1 2017 unless you're in the compsci research department at stanford, MIT, or NYU.
 
Last edited:
schizo said:
Actually, the Radeon Pro Duo will probably be AMD's flagship until Vega.

And while you can pay for GP100 now, it comes inside a rack-mounted supercomputer with 8 of 'em. And even if you spring for the $129,000 DGX-1 and pop out one of those P100 cards, you won't be able to use it to play The Division-- it doesn't have any graphics ports. Oh, and you won't get it until Q1 2017 unless you're in the compsci research department at stanford, MIT, or NYU.
Click to expand...

You should be able to buy standalone Tesla by January

Anyway, call Polaris 10 a flagship if you want, the important thing is everyone is aware the real GCN 1.3 flagship is Vega 10
 
You must log in or register to reply here.
Back
Top