NVIDIA GeForce GTX 1080 Ti Specifications Leaked, Inbound for Holiday 2016?

[GreiverBlade]

It does seem a bit ridiculous, however they maybe assuming since it has the bigger bus its not necessary and that's how they keep the cost down. I would find it ridiculous as well, but this would not completely surprise me if its true.

That being said, way to early to judge.

If this does turn out to be true, well it will still be a decent card just not as great as people had hoped.

looking at your syspecs i wonder about clock speed ... you write 2025 but it's totally not that it's like for me to write 2100 while the core is actually 1700, i wish people would drop max boost clock and use the regular clock

other than that, enjoying your TXP? (i hope so ... it's still "THE beast" of the actual lineup )

"GTX 1080 Ti" is scheduled for Q4. It will probably perform ~30% over GTX 1080 or better.

aherm ... ~30% over a 1080 is already what a TXP give .... soooo ... sure?

since the TXP is not "way" ahead a 1080 ...

although the 780Ti and 980Ti did also made Titan and Titan X owner sad because they where so close, and even above in some case involving a better OC'ability, for a "fraction" of the price (not for me ... most Ti were priced like a Titan counterpart ... at my retailer/etailer nonetheless )

 

[efikkan]

aherm ... ~30% over a 1080 is already what a TXP give .... soooo ... sure?

since the TXP is not "way" ahead a 1080 ...

Titan X (Pascal) performs 35% over GTX 1080 when placed in a closed case and pre-warming benchmarks, which most benchmarkers don't do.

 

[GhostRyder]

looking at your syspecs i wonder about clock speed ... you write 2025 but it's totally not that it's like for me to write 2100 while the core is actually 1700, i wish people would drop max boost clock and use the regular clock

other than that, enjoying your TXP? (i hope so ... it's still "THE beast" of the actual lineup )


aherm ... ~30% over a 1080 is already what a TXP give .... soooo ... sure?

since the TXP is not "way" ahead a 1080 ...

although the 780Ti and 980Ti did also made Titan and Titan X owner sad because they where so close, and even above in some case involving a better OC'ability, for a "fraction" of the price (not for me ... most Ti were priced like a Titan counterpart ... at my retailer/etailer nonetheless )

Well, thats what the boost is telling me it does when overclocked . Normally though even around there it hovers around 1950mhz, I am more displaying its max boost it shows on the MSI graph while playing.

Yea I am enjoying it, great card with very little issues to speak of with it. Will be water cooling it soon enough but I am waiting until I first get the G-Sync monitor and then decide if I want a second one since I don't want to rebuild the GPU part of the loop twice.

Its going to be probably around 20-25% better than the GTX 1080 based on the leaks if we believe them completely.

 

[qubit]

This was posted one month ago on reddit then -> wccf -> oc3d -> techpowerup.

Really? Got a link for it?

 

[64K]

History is pretty much repeating itself here.

Maxwell Titan X was 24% faster than a 980 at 4K and 23% faster than a 980 at 1440p

Spoiler: Maxwell

Pascal Titan X is 24% faster overall than a 1080 at 4K and 21% faster overall than a 1080 at 1440p

Spoiler: Pascal

There is room for a 1080 Ti

I expect a non reference factory overclocked 1080 Ti to be a little slower than a reference Pascal Titan X but faster than a non reference 1080. Pretty much the same story as the Maxwell Titan X , 980 Ti, 980.

Whether it's worth the $$$ over the 1080 is unknown yet.

 

[Slizzo]

It does seem a bit ridiculous, however they maybe assuming since it has the bigger bus its not necessary and that's how they keep the cost down. I would find it ridiculous as well, but this would not completely surprise me if its true.

That being said, way to early to judge.

If this does turn out to be true, well it will still be a decent card just not as great as people had hoped.

It makes no real sense. Major sticking point for GDDR5X existing, is that it doesn't cost a great deal more than traditional GDDR5, while allowing greatly increased bandwidth. I'm not even sure that GP102 has a GDDR5 memory controller anyway, I thought it only has a GDDR5X memory controller.

 

[Ascalaphus]

Yeah. Odd that they would be using GDDR5 instead of GDDR5X. Could it be cost??

 

[efikkan]

Yeah. Odd that they would be using GDDR5 instead of GDDR5X. Could it be cost??

If true, it's simply because it doesn't need more bandwidth.

 

[chodaboy19]

It's a fake as others have pointed out.

 

[GreiverBlade]

Its going to be probably around 20-25% better than the GTX 1080 based on the leaks if we believe them completely.

which would place it directly on a TXP performances ... meeeeehhhh bummer.

(sooo my 1070 clock better than that TXP, max boost 2100 sustained boost 2078, although OC not needed at all 1556 is fine 1700 is for benchies only )

Titan X (Pascal) performs 35% over GTX 1080 when placed in a closed case and pre-warming benchmarks, which most benchmarkers don't do.

well ... nope number say otherwise although (and case to case scenario are not optimal for review soooo basically : it's like that )

History is pretty much repeating itself here.

Maxwell Titan X was 24% faster than a 980 at 4K and 23% faster than a 980 at 1440p

Spoiler: Maxwell

Pascal Titan X is 24% faster overall than a 1080 at 4K and 21% faster overall than a 1080 at 1440p

Spoiler: Pascal

There is room for a 1080 Ti

I expect a non reference factory overclocked 1080 Ti to be a little slower than a reference Pascal Titan X but faster than a non reference 1080. Pretty much the same story as the Maxwell Titan X , 980 Ti, 980.

Whether it's worth the $$$ over the 1080 is unknown yet.

room? is 24% and 23% enough to place one Ti in between that? that card will be redundant to a factory OC 1080 or a stock TXP (well ... my 980 was fine ... a 980Ti would not have been a significant upgrade so then i would only have considered the Titan X if the price would've been nicer )
although yep that's what nvidia will do ... "f*ck up" early TXP adopter with a XX80Ti

 

[BiggieShady]

Maxwell Titan X was 24% faster than a 980 at 4K and 23% faster than a 980 at 1440p

If Titan X is 100% and 980 is 76%, it is 100/76=1.315789... which is approximately 31.58 % faster (not 24 % by simple subtraction, these are ratios - subtraction is bad m'kay)

 

[64K]

room? is 24% and 23% enough to place one Ti in between that? that card will be redundant to a factory OC 1080 or a stock TXP (well ... my 980 was fine ... a 980Ti would not have been a significant upgrade so then i would only have considered the Titan X if the price would've been nicer )
although yep that's what nvidia will do ... "f*ck up" early TXP adopter with a XX80Ti

The difference in overall performance between the Maxwell Titan X and the 980 warranted the release of a 980 Ti for a lot less than a Titan X. Note the 980 Ti had 6 GB VRAM over the 980 4 GB. 50% more VRAM.

The difference between the Pascal Titan X and a 1080 in overall performance is nearly the same so why not a 1080 Ti that will be close the the performance of a Titan X for a lot less money? Note the 1080 Ti will have 12 GB VRAM over the 1080 8 GB. The same 50% more VRAM for those that need it..

It's when you look at the small difference in performance between a manufacturer overclocked non reference 1080 Ti compared to a reference Titan X for the much cheaper 1080 Ti that I can see it making some sense. You can still overclock both also so they should remain about the same difference in performance if you do that to both.

 

[64K]

If Titan X is 100% and 980 is 76%, it is 100/76=1.315789... which is approximately 31.58 % faster (not 24 % by simple subtraction, these are ratios - subtraction is bad m'kay)

No, the Titan X is 24% faster than the 1080 in that example. What you are thinking of is how much faster would the 1080 have to be to equal the Titan X. 76% x .3158 +76%= 100% so the 1080 would have to be 31.58% faster than it is to equal the Titan X.

 

[GreiverBlade]

so why not a 1080 Ti that will be close the the performance of a Titan X for a lot less money?

actually .... nope ... the Ti will be just 200$ under the TXP which is not by any mean "a lot less" nor "much cheaper" more realistically as i saw more 980Ti priced closer, even above sometime, a TXM, i can take a wild guess that the TI will be 100 to 50$ cheaper at consumer level.

i don't care about RAM and performances on that level, only the pricing make it wrong (and i am glad that i got a custom 1070 for 415chf ... )

overpriced is the qualificatif for the 1080/1080Ti/TXP indeed, only some 1070 are priced almost correctly ... thanks to nvidia and the idiotic FE pricing placement at a premium over MSRP (while they are nothing desirable nor have anything to value over a custom so ... logically the FE price became the MSRP )

No, the Titan X is 24% faster than the 1080 in that example. What you are thinking of is how much faster would the 1080 have to be to equal the Titan X. 76% x .3158 +76%= 100% so the 1080 would have to be 31.58% faster than it is to equal the Titan X.

i think he mean the rule of 3 .... but i am not sure myself ... tho it still give me right on my affirmation where i said the TXP already gave us ~30% more on a 1080 so if the Ti would be 30% faster ... it would also be redundant to a TXP, if lower ... then not really worth it over a 1080 (IF THE MSRP WAS RESPECTED ) ok ... +4gb RAM for the elite that run a tri4K setup ... (but will still stutter like mad (joking))

out.

 

[64K]

actually .... nope ... the Ti will be just 200$ under the TXP which is not by any mean "a lot less" nor "much cheaper" more realistically as i saw more 980Ti priced closer, even above sometime, a TXM, i can take a wild guess that the TI will be 100 to 50$ cheaper at consumer level.

i don't care about RAM and performances on that level, only the pricing make it wrong (and i am glad that i got a custom 1070 for 415chf ... )

overpriced is the qualificatif for the 1080/1080Ti/TXP indeed, only some 1070 are priced almost correctly ... thanks to nvidia and the idiotic FE pricing placement at a premium over MSRP (while they are nothing desirable nor have anything to value over a custom so ... logically the FE price became the MSRP for the AIB )

I don't care about 12 GB VRAM either but some people do for whatever reason they need more than 8 GB.

Price isn't known for sure right now but if it is $850 as the article here speculates, and that seems right, then that is $350 cheaper than a Titan X here. The difference in other countries will probably be different so it may be not be so much. Maybe $350 isn't a lot but it certainly is significant if the performance turns out to be close to the Pascal Titan X as the 980 Ti was close to the Maxwell Titan X.

But, yeah from my point of view the 1080 and Titan X are overpriced and the 1080 Ti will be as well. Competition from AMD is badly needed.

 

[BiggieShady]

No, the Titan X is 24% faster than the 1080 in that example. What you are thinking of is how much faster would the 1080 have to be to equal the Titan X. 76% x .3158 +76%= 100% so the 1080 would have to be 31.58% faster than it is to equal the Titan X.

Really? Didn't you write yourself:

Maxwell Titan X was 24% faster than a 980 at 4K and 23% faster than a 980 at 1440p

Pascal Titan X is 24% faster overall than a 1080 at 4K and 21% faster overall than a 1080 at 1440p

And the graph says:

TitanX: 100, 980: 76 at 4k
TitanXP: 100, 1080: 76 at 4k

and

TitanX: 100, 980: 77 at 1440p
TitanXP: 100, 1080: 79 at 1440p

Your numbers are wrong, they should be in the thirties or so ... like this:

Maxwell Titan X was 32% faster than a 980 at 4K and 30% faster than a 980 at 1440p
Pascal Titan X is 32% faster overall than a 1080 at 4K and 26,6% faster overall than a 1080 at 1440p

If Titan X were 24% faster than 980 in 4k, that would mean 76 * 1.24 would have to be 100, and it's not .. it's 94.24

Also, saying "1080 would have to be 31.58% faster than it is to equal the Titan X" is same as saying "Titan X is 31.58% faster than 1080"

 

[fynxer]

No way i am investing in a further cut down GP102 with only GDDR5. Nvidia is fu**ing us over again. Nvidia knows there is no alternative until VEGA in 2017 so they figured why not do a half baked card and squeeze some quick money out of some unsuspecting morons before x-mas and VEGA in 2017.

Do not buy this card, you will be disappointed, within 6-8 months Nvidia will release a more powerful card to counter VEGA.

If VEGA turns out to be a powerful monster you will either see an GP102+ with all cuda cores at 14nm OR a first introduction of Volta for gaming.

The key here is AMD's VEGA, if AMD screws up VEGA then we are all fu**ed. Nvidia will halt the release of any new cards and delay Volta for gaming to around Q3 2018

 

[64K]

No way i am investing in a further cut down GP102 with only GDDR5. Nvidia is fu**ing us over again. Nvidia knows there is no alternative until VEGA in 2017 so they figured why not do a half baked card and squeeze some quick money out of some unsuspecting morons before x-mas and VEGA in 2017.

Do not buy this card, you will be disappointed, within 6-8 months Nvidia will release a more powerful card to counter VEGA.

If VEGA turns out to be a powerful monster you will either see an GP102+ with all cuda cores at 14nm OR a first introduction of Volta for gaming.

The key here is AMD's VEGA, if AMD screws up VEGA then we are all fu**ed. Nvidia will halt the release of any new cards and delay Volta for gaming to around Q3 2018

I doubt Vega will disappoint next year but then Nvidia releases Volta months later and off we go again chasing expensive Nvidia cards.

 

[ensabrenoir]

.......as we squabble and debate price, necessity, performance.........Nvdia:

&

especially if Amd takes the two "reasonable" cards to beat one approach again.....

 

[64K]

Really? Didn't you write yourself:


And the graph says:

and

Your numbers are wrong, they should be in the thirties or so ... like this:

Maxwell Titan X was 32% faster than a 980 at 4K and 30% faster than a 980 at 1440p
Pascal Titan X is 32% faster overall than a 1080 at 4K and 26,6% faster overall than a 1080 at 1440p

It depends on how the word problem is worded.

If you ask how much faster is the Titan X than the 1080 then the answer is 100% - 76% = 24% so the Titan X is 24% faster than the 1080.

If you ask how much faster would the 1080 have to be to match the performance of the Titan X then the answer is:

Let N = the percent needed to be as fast as the Titan X

(76% x N) +76% = 100%
(76% x N) +76% - 76% = 100% - 76%
(76% x N) = 24%
(76% x N) / 76% = 24% / 76%
N = 31.578 ~ 32%

So the answer to that question is that the 1080 would have to be ~32% faster than it is to equal the Titan X but my statement that the Titan X is 24% faster than the 1080 is correct.

 

[BiggieShady]

If you ask how much faster is the Titan X than the 1080 then the answer is 100% - 76% = 24% so the Titan X is 24% faster than the 1080.

Sorry, it doesn't work that way. It cannot be interpreted as both, your second calculation is the correct one.
The source of all data is number of frames per second. If one card gets 100 fps and another 76 fps ... that data would produce 100% and 76% in percentages (normalized for faster card as in the graphs) ... you cant say one is 24% faster just because difference is 24. As you see 100fps/76fps=1.31578

 

[64K]

Sorry, it doesn't work that way. It cannot be interpreted as both, your second calculation is the correct one.
The source of all data is number of frames per second. If one card gets 100 fps and another 76 fps ... you cant say one is 24% faster just because difference is 24 ... that data would produce 100% and 76% in percentages.

It's not interpreted as both. They are two separate questions.

What you are looking at is "How much faster would the 1080 have to be to equal the Titan X" and what I was saying was looking at it from an entirely different direction. "How much faster is the Titan X than the 1080."

But, this is pretty much off topic so I will just say that you are right and I am wrong and we can put aside the math misunderstanding.

 

[BiggieShady]

But, this is pretty much off topic so I will just say that you are right and I am wrong and we can put aside the math misunderstanding.

Yeah we shouldn't, but it's funny how what you consider two separate questions "How much faster would the 1080 have to be to equal the Titan X" and "How much faster is the Titan X than the 1080.", math considers completely the same thing. As in no difference ...
I'm not usually finicky about stuff, must be some childhood education trauma surfacing

 

[HammerON]

I find the possible lack of GDDR5X memory disappointing if in fact it is true. We will just have to wait until it comes out and is reviewed by W1z to see how well it performs...

 

[the54thvoid]

As an Nvidia owner since the original Titan, if the 1080ti comes in at anywhere above the current 1080 price, I'll buy a fucking PS4.

Even I am getting sick and tired of this. Think I'll move in with @RejZoR. Vega better deliver or the pricing structures will be horrendous. And even then, AMD aren't guaranteed to price Vega at a point we need.

£800 Founder Edition 1080ti? Drop dead.