Is a reference gtx 880 likely to be faster than a stock 780ti kingpin?

[GhostRyder]

Do the math, the extra power between the 650Ti/750Ti is mainly from the core clock and a few minor IPC improvements, the 650Ti Boost with more ROPS outperforms the 750Ti however, despite a lower clock speed.

928 core clock on a 128bit bus with 768 cuda cores on reference is the 650ti vs 1020 with 1085 boost core clock with 640 cuda cores and a 128bit bus gtx 750ti is the comparison. Not a whole lot different considering (guess you could throw the reference 1gb vs 2gb as well) and the fact it completely takes it down a notch in every game benchmark I've seen shows that the pic improvements are pretty nominal. Even accounting for the high clock rate should not give something with less cores that much of a performance gap.

The GTX 650ti boost has a 192bit bus and a core clock of 980 with boost to 1033 and with the extra memory and memory speed/bandwidth allows it to perform more on par with a 750ti however it still depends on the game as BF4 is better on 750ti while Metro for instance is better on the 650ti boost.

Still in raw performance the 750ti is slightly ahead. Either way though if we include the way things are and the looks the GTX 880 even if it matches the core of a GTx 780ti will perform better.

Still more of a wait and see game, but I still believe the 880 will be the next top dog.

 

[64K]

Probably The GTX 880 will top the GTX 780Ti but the wild card is the Maxwell engineering and not necessarily the 28nm process. The GTX 680 (GK104) 28nm outperformed the Fermi refresh GTX 580 (GF110) 40nm. That's where it gets cloudy. The engineering. We will see soon.

 

[newconroer]

Probably The GTX 880 will top the GTX 780Ti but the wild card is the Maxwell engineering and not necessarily the 28nm process. The GTX 680 (GK104) 28nm outperformed the Fermi refresh GTX 580 (GF110) 40nm. That's where it gets cloudy. The engineering. We will see soon.

I don't see anything particularly great on the architecture lineup for Maxwell that would make it particularly better than Kepler to any significant degree.

UMA and NVLink do not exist until Pascal and Volta. It might be sometime before Nvidia only users need to upgrade.

 

[64K]

I was disappointed when Nvidia sent the Volta design with stacked DRAM and possibly 1 TB memory bandwidth to limbo. A 16nm Pascal flagship will probably be able to drive a 4k monitor on games at max settings but that depends on how 4k monitors catch on. If they don't then Pascal will be more about efficiency than grunt power.

 

[newconroer]

I was disappointed when Nvidia sent the Volta design with stacked DRAM and possibly 1 TB memory bandwidth to limbo. A 16nm Pascal flagship will probably be able to drive a 4k monitor on games at max settings but that depends on how 4k monitors catch on. If they don't then Pascal will be more about efficiency than grunt power.

Do you mean that you were hoping those features would be available on Pascal as well, because they are.

Let's hope Maxwell's success helps Nvidia to continue down this new and exciting path.

 

[RejZoR]

In general, they make new gen mid end as fast as old gen high end. Meaning GTX 870 will most likely be as fast as GTX 780 Ti. GTX 880 will be faster, otherwise there is hardly any point in buying a new high end card if you already bought one last year...

 

[64K]

In general, they make new gen mid end as fast as old gen high end. Meaning GTX 870 will most likely be as fast as GTX 780 Ti. GTX 880 will be faster, otherwise there is hardly any point in buying a new high end card if you already bought one last year...

But you're not taking into account for the die shrink that's coming probably in 2015. The Nvidia release schedule is skewed at this point. The 28nm Kepler mid range GK104 did outperform the 40nm Fermi GK110. Probably the GTX 880 will smoke the GTX 780 and Maybe top the 780Ti. The price though?

 

[newconroer]

But you're not taking into account for the die shrink that's coming probably in 2015. The Nvidia release schedule is skewed at this point. The 28nm Kepler mid range GK104 did outperform the 40nm Fermi GK110. Probably the GTX 880 will smoke the GTX 780 and Maybe top the 780Ti. The price though?

If it's better and more power efficient, then they'll put a premium on it. But how much better than a 780 can it be ?

 

[RejZoR]

Die shrink doesn't mean performance improvement by itself.

 

[64K]

Die shrink doesn't mean performance improvement by itself.

Can you point to a GPU die shrink than didn't have improvement? It's taken for granted that a die shrink will be more efficient and therefore capable of more performance per watt.

 

[bubbleawsome]

There is no way nvidia would release a top end card that didn't beat last gen. No way. The PR issue with their fanboys would be enough.

 

[EarthDog]

My spidey sense is tingling from fanboy hopes in this thread.

SMH at this bait...


The 880 is rumored NOT to be the high end card. So I highly doubt it will beat the 780ti/290x. As was mentioned earlier, it needs officially benchmarked to tell, but judging from the leaked specs, assuming they are true, I would not believe it to beat out 780ti/290x.

Can you point to a GPU die shrink than didn't have improvement? It's taken for granted that a die shrink will be more efficient and therefore capable of more performance per watt.

While this is true, that doesn't mean performance will go up... Think about what you just posted...........

Capable, certainly. In and of itself though... nope.

There is no way nvidia would release a top end card that didn't beat last gen. No way. The PR issue with their fanboys would be enough.

Its rumored not to be top end is the point...

 

[64K]

It's still speculation at this point. The GTX 680 (GK104) did outperform the Fermi refresh flagship GTX 580 (GF110) but the GTX 680 was a die shrink from 40nm to 28nm. The GTX 880 is a new and more efficient Maxwell design but it won't be a die shrink so it's hard to tell how it will perform and no this will not be the high end Maxwell. That will come next year most likely.

 

[El_Mayo]

I'm not the only one trapped, wondering "do I wait, or do I buy a 7xx card now" it seems!

 

[rtwjunkie]

I'm not the only one trapped, wondering "do I wait, or do I buy a 7xx card now" it seems!

Personally, I think it makes perfect sense to buy a 7xx card now, then upgrade to the 9xx series so you can get Maxwell 2.0, like the 5 series was so much better than the 4 series (except the 460, which was damned good card!). The 7xx are fantastic cards, and it will be awhile before the 770 or 780 are not relevant, especially if you're at 1080P.

 

[El_Mayo]

Personally, I think it makes perfect sense to buy a 7xx card now, then upgrade to the 9xx series so you can get Maxwell 2.0, like the 5 series was so much better than the 4 series (except the 460, which was damned good card!). The 7xx are fantastic cards, and it will be awhile before the 770 or 780 are not relevant, especially if you're at 1080P.

I AM at 1080p! see my mistake was when I read "880 and 870 might launch at the same price as the 780 an 770" I took that to mean the price they're at now... turns out the GTX 770 was once £329. I'd pay that much for an 870 if it matches the 780 in performance though so I'm still stuck!

 

[Shambles1980]

i swear im having this exact debate over at a different forum lol.

any way basically what i said their is..

I think the 880 could be faster, but i think nvidia will artificially slow it down by locking off features and shaders etc. then release it a t a reasonably low price (for nvidia) and will be more like an affordable 780 ti slightly slower with less power consumption.
this will mean 780ti sales will continue for a while and some 880s will sell too.
after a while they will unlock the features they locked out and release the card at its full or near full potential. for "$ Lmao are you serious?"
the ones that bought early may then upgrade (but possibly wont, because if they wanted the raw power they would have bought the 780ti) the 780 ti buyers may also upgrade. and the ones that waited will buy the new release of the card.


this means more sales. and they can charge more for the not gimped card.

 

[rtwjunkie]

I think everyone is looking at this the wrong way. The 780 Ti is an extra, or "specialty" card. Nvidia doesn't usually produce a mainstrem series-flagship that outperforms the previous specialty card. They make one to outperform the previous mainstream flagship (which in this case is the 780). The 780 Ti fills the same "specialty" spot that 590 and 690 did the way I see it. Nvidia is not likely to overwhelm the performance of the 780 Ti, only the 780. That will happen if/when they make a "specialty" 8xx card, or when the 980 comes out.

My opinion of course, but that's the way I see it, and feel is the most logical based on past practice.

 

[El_Mayo]

Any indication GTX 800 cards are gonna use less power than their equivalent 700 card? they're both 28nm but different architecture, yeah? I'm trying to avoid needing a new PSU if I ever want to SLI

 

[rtwjunkie]

Any indication GTX 800 cards are gonna use less power than their equivalent 700 card? they're both 28nm but different architecture, yeah? I'm trying to avoid needing a new PSU if I ever want to SLI

Based on the 750 cards, I'd say definately! Maxwell's strong suit is less power consumption, more efficiency.

 

[64K]

Any indication GTX 800 cards are gonna use less power than their equivalent 700 card? they're both 28nm but different architecture, yeah? I'm trying to avoid needing a new PSU if I ever want to SLI

Unless your monitor is 120 Hz or 144 Hz then you probably won't need to SLI a GTX 780 or 780 Ti and probably not need to SLI a GTX 880 anytime soon.

 

[RejZoR]

Can you point to a GPU die shrink than didn't have improvement? It's taken for granted that a die shrink will be more efficient and therefore capable of more performance per watt.

No. Die shrink only lowers power usage. Other than that, it does exactly nothing for performance unless you use different architecture or higher clocks on top of die shrink. Like i've said, die shrink doesn't improve performance by itself. Headroom it gives to stretch it does however.

 

[erocker]

There is no way nvidia would release a top end card that didn't beat last gen. No way. The PR issue with their fanboys would be enough.

X80 cards are no longer Nvidia's high-end cards. The Ti and Titan versions are.

I mean, in this day and age where 1080p is (very) slowly phasing out and we keep hearing "4K!" from the marketeers, one would thing there would be a beefier vram setup on this card... I guess nothing has been confirmed yet though.

 

[GhostRyder]

X80 cards are no longer Nvidia's high-end cards. The Ti and Titan versions are.

I mean, in this day and age where 1080p is (very) slowly phasing out and we keep hearing "4K!" from the marketeers, one would thing there would be a beefier vram setup on this card... I guess nothing has been confirmed yet though.

But even assuming that there is going to be an 880ti (I personally do not think there is going to be one but thats my opinion that im basing off history) the GTX 880 should at least perform on par or higher than a GTX 780ti. Looking back the GTX 670 was much better than the GTX 580 so even assuming that the GTX 880 is not the top tier card I would still think it would be enough to match or beat the previous top tier card.

I could always be wrong, but I can only base it off history and what has been announced/rumored.

 

[bubbleawsome]

X80 cards are no longer Nvidia's high-end cards. The Ti and Titan versions are.

I mean, in this day and age where 1080p is (very) slowly phasing out and we keep hearing "4K!" from the marketeers, one would thing there would be a beefier vram setup on this card... I guess nothing has been confirmed yet though.

Right, but the 680 was "top-end" They wouldn't let the x80 series ever be beat by a card from the last gen. I mean, if they hit a massive production problem I could see them not launching anything but a 870ti and then paper-launch the 880 until they can produce it but the PR horror of being beat by your own card that uses a 2 year old arch would suck.