[TPU] 1080 TI specs leaks

[tviceman]

https://www.techpowerup.com/225849/...pecifications-leaked-inbound-for-holiday-2016

No pricing, but I'm guessing $799/899 MSRP for AIB and FE cards. Looks like a calculated move by Nvidia to ensure the 1080 TI can't match Titan XP, even with OC, and to also leave wiggle room in case they want to refresh/rebadge Pascal in the event Vega overperforms.

 

[DamZe]

Lol if true, they are downgrading the memory to regular GDDR5 @8005Mhz, pure nVIDIA value for money right there.

 

[Krteq]

Isn't this a same, month old, speculation (probably a fake) from a HWbattle server?

http://www.hwbattle.com/bbs/board.php?bo_table=news&wr_id=40409&ckattempt=1

 

[RussianSensation]

If true, this is shaping up to be one of the worst overpriced, if not the worst, GPU generations in NV's history. The Titan XP was already nothing more than a $349 GTX570 successor card with double the VRAM. Cutting that further down as a 1080Ti is a close successor to the $289 GTX560Ti 448 Core:
http://www.anandtech.com/show/5153/nvidias-geforce-gtx-560-ti-w448-cores-gtx570-on-a-budget

With upper-mid-range 1080 purposely positioned as a fake intermediary marketing flagship and selling for $629+ market prices (bifurcating a generation strategy), and the heavily cut-down 1080Ti likely positioned at $799+, the GPU market has never been this grossly overpriced. Meh, makes it very easy to keep skipping generations and picking up flagship cards at their real historical prices just 15 months later (aka $699 EVGA Classy 980Ti ---> $360 just 15 months later).

This will only get worse as the dGPU market keeps shrinking, ensuring that both AMD and NV will continue to raise prices for every tier. While the GPU market has shrunk 2X in less than a decade, GPU prices roughly doubled:
http://www.anandtech.com/show/10613...-grabs-market-share-but-nvidia-remains-on-top

The naysayers keep denying these facts though despite the 3rd consecutive GPU generation since 2012 Kepler/GCN clearly showing it's exactly what's happening. The bad part is that in an oligopoly when NV raised prices so high in the Pascal generation, even if AMD beats them in price/performance in 2017, it hardly matters since AMD's cards will also be overpriced. The severely cut-down 1080Ti is more or less just a GTX780/R9 290 tier but no way AMD prices 2nd tier Vega for $399 if 1080Ti is $799+.

The more expensive they are, the harder they fall!

September 2014 $550 980
September 2016 $168 1060:
http://slickdeals.net/f/9092647-gig...-s-h-new-customers-only?src=SiteSearchV2Algo1

 

[Arachnotronic]

If true, this is shaping up to be one of the worst overpriced, if not the worst, GPU generations in NV's history. The Titan XP was already nothing more than a $349 GTX570 successor card with double the VRAM. Cutting that further down as a 1080Ti is a close successor to the $289 GTX560Ti 448 Core:
http://www.anandtech.com/show/5153/nvidias-geforce-gtx-560-ti-w448-cores-gtx570-on-a-budget

With upper-mid-range 1080 purposely positioned as a fake intermediary marketing flagship and selling for $629+ market prices (bifurcating a generation strategy), and the heavily cut-down 1080Ti likely positioned at $799+, the GPU market has never been this grossly overpriced. Meh, makes it very easy to keep skipping generations and picking up flagship cards at their real historical prices just 15 months later (aka $699 EVGA Classy 980Ti ---> $360 just 15 months later).

This will only get worse as the dGPU market keeps shrinking, ensuring that both AMD and NV will continue to raise prices for every tier. While the GPU market has shrunk 2X in less than a decade, GPU prices roughly doubled:
http://www.anandtech.com/show/10613...-grabs-market-share-but-nvidia-remains-on-top

The naysayers keep denying these facts though despite the 3rd consecutive GPU generation since 2012 Kepler/GCN clearly showing it's exactly what's happening. The bad part is that in an oligopoly when NV raised prices so high in the Pascal generation, even if AMD beats them in price/performance in 2017, it hardly matters since AMD's cards will also be overpriced. The severely cut-down 1080Ti is more or less just a GTX780/R9 290 tier but no way AMD prices 2nd tier Vega for $399 if 1080Ti is $799+.

The more expensive they are, the harder they fall!

September 2014 $550 980
September 2016 $168 1060:
http://slickdeals.net/f/9092647-gig...-s-h-new-customers-only?src=SiteSearchV2Algo1

Thou doth protest too much.

 

[swilli89]

If true, this is shaping up to be one of the worst overpriced, if not the worst, GPU generations in NV's history. The Titan XP was already nothing more than a $349 GTX570 successor card with double the VRAM. Cutting that further down as a 1080Ti is a close successor to the $289 GTX560Ti 448 Core:
http://www.anandtech.com/show/5153/nvidias-geforce-gtx-560-ti-w448-cores-gtx570-on-a-budget

With upper-mid-range 1080 purposely positioned as a fake intermediary marketing flagship and selling for $629+ market prices (bifurcating a generation strategy), and the heavily cut-down 1080Ti likely positioned at $799+, the GPU market has never been this grossly overpriced. Meh, makes it very easy to keep skipping generations and picking up flagship cards at their real historical prices just 15 months later (aka $699 EVGA Classy 980Ti ---> $360 just 15 months later).

This will only get worse as the dGPU market keeps shrinking, ensuring that both AMD and NV will continue to raise prices for every tier. While the GPU market has shrunk 2X in less than a decade, GPU prices roughly doubled:
http://www.anandtech.com/show/10613...-grabs-market-share-but-nvidia-remains-on-top

The naysayers keep denying these facts though despite the 3rd consecutive GPU generation since 2012 Kepler/GCN clearly showing it's exactly what's happening. The bad part is that in an oligopoly when NV raised prices so high in the Pascal generation, even if AMD beats them in price/performance in 2017, it hardly matters since AMD's cards will also be overpriced. The severely cut-down 1080Ti is more or less just a GTX780/R9 290 tier but no way AMD prices 2nd tier Vega for $399 if 1080Ti is $799+.

The more expensive they are, the harder they fall!

September 2014 $550 980
September 2016 $168 1060:
http://slickdeals.net/f/9092647-gig...-s-h-new-customers-only?src=SiteSearchV2Algo1

We as Americans, at least those that deal with American companies (all major Semiconductor companies), should be used to companies maximizing profits but the sting is never softened by this fact of theory. The fact remains that nVidia exists not to deliver value as its 1st tenant of business, but to maximize profits thereby making Jen and his shareholders as rich as possible. In that line of thought, without a competitive high end from AMD this is exactly what you should expect. Does the blame lie with AMD? Well if blame even matters, I would say it lies with both AMD and consumers. Why did the 960 outsell the 380x when it was slower? Lots and lots of examples like this.

So we have even smaller die sizes yet even higher prices. Imagine if Samsung never gained traction with Android, Iphones could easily be over a grand brand new. Luckily there is competition there that helps keep prices down. Look at Intel, 200mm^2 CPUs cost over $1000! Unluckily there is currently no competition there.

The more I read on these forums the less I see AMD vs nVidia and the more I see Corporation vs Consumer cheerleaders. Some wish for AMD to go bankrupt but in that very same breath they are wishing higher prices and slower paces of innovation on themselves! Yesterday's marketshare gain announcement for AMD was a win for all consumers of graphics cards. Yesterday on the radio I heard about a KISS concert where the lead singer was condemning Colin Kapernick for not standing during the anthem with his rebuttal being "its ALWAYS cool to be patriotic". Well let's adapt that to our world:

Its ALWAYS cool to be pro-consumer.

How ANYONE can argue that nVidia is delivering more value this generation is totally beyond me, and calling out members for complaining about lack of consumer value is even more mindblowing. If consumers had a collective consciousness like nVidia (a company has 1 leader and acts in unison) and could act in unison we would purchase less this generation to vote with our collective wallets to show nVidia consumers ARE educated and we DO demand more value for these non-life essential goods.

 

[Headfoot]

can we keep this restricted to actual 1080 ti speculation/spec leaks please?

 

[Arachnotronic]

How ANYONE can argue that nVidia is delivering less value this generation is totally beyond me, and calling out members for complaining about lack of consumer value is even more mindblowing.

Consumer value on these cards is phenomenal if you judge them based on the performance they deliver and the experiences they enable, rather than based on die size. Pretty hilarious that people use the die size argument without considering the substantially increased R&D intensity required per chip as well as the flat-to-rising cost/xtor with newer generation mfg technologies.

It's also funny that making a profit is a crime these days in the eyes of consumers. Companies that deliver great products deserve to earn a healthy profit on them.

Anyway, back on topic, the 1080 ti looks like it will be about as fast as the Titan X out of the box due to the higher clocks. Effective memory bandwidth should be much higher than that on the 980 Ti thanks to the new memory compression tech. And I guess for those who really care about VRAM buffer size, 12GB should be nice too.

 

[swilli89]

can we keep this restricted to actual 1080 ti speculation/spec leaks please?

I'm not trying to derail so please forgive me, but if we are discussing 1080ti specs then I think its fair to discuss the value of said specs and their potential performance/$.

 

[Headfoot]

All being said - I don't believe the rumor that it'll come with regular 8gbps GDDR5. That just seems too little bandwidth over a 1080 and if I was paying in the $700-800 range for a card I would be really turned off by not getting the latest and greatest tech on it.

 

[swilli89]

Consumer value on these cards is phenomenal if you judge them based on the performance they deliver and the experiences they enable, rather than based on die size. Pretty hilarious that people use the die size argument without considering the substantially increased R&D intensity required per chip as well as the flat-to-rising cost/xtor with newer generation mfg technologies.

It's also funny that making a profit is a crime these days in the eyes of consumers. Companies that deliver great products deserve to earn a healthy profit on them.

You're using generalizations in your argument. You are easily intelligent enough to know "making a profit" and "highest profits ever" are QUITE different things. Apples to oranges, my friend.

In the AMD marketshare thread you cheered nvidia's "record breaking profits" despite their marketshare loss. So please reconcile these two facts:

• Substantially increased R&D intensity required per chip as well as the flat-to-rising cost/xtor with newer generation mfg technologies.
• Record breaking profits

Either costs are rising and they are increasing their prices to maintain profit margins, or as we know with their higher than ever margins they are increasing prices consumers pay more than the cost increases they are seeing. I'm not saying this is evil, I'm not even bringing morality into this, I'm simply talking economics and trying to educate consumers to vote with their wallets just like Capitalism begs us to do.

Anyways, done with my market imbalance argument, hopefully the 1080ti knocks 1080 down and occupies the $650 slot just like its two predecessors.

 

[Arachnotronic]

All being said - I don't believe the rumor that it'll come with regular 8gbps GDDR5. That just seems too little bandwidth over a 1080 and if I was paying in the $700-800 range for a card I would be really turned off by not getting the latest and greatest tech on it.

You get 12GB of VRAM and more CUDA cores. Using GDDR5X would remove a key point of differentiation relative to the Titan X. This time, NVIDIA seems to be trying not to make the Titan buyers regret their purchases.

 

[nathanddrews]

The 1080Ti is a given, always has been. Not sure they'll launch it over $800 though... that seems a bit high... but I didn't expect the Titan XP to be $1200 either.

Let's not forget that AMD has taken an ~11% chunk of Nvidia's share of a shrinking dGPU market in the last couple quarters. The pressure is being applied, just not yet at the high end. If AMD can release a competitive high-end product (Vega?), Nvidia will have no choice but to reduce prices of their high-end stack. Of course, there's nothing to stop AMD from using the existing market values (set entirely by Nvidia) of high-end GPUs to maximize their own profits.

God forbid we just wait and see...

 

[nurturedhate]

Consumer value on these cards is phenomenal if you judge them based on the performance they deliver and the experiences they enable, rather than based on die size. Pretty hilarious that people use the die size argument without considering the substantially increased R&D intensity required per chip as well as the flat-to-rising cost/xtor with newer generation mfg technologies.

Performance in a vacuum is a poor way to judge any product as it allows any price to be set. What RS is doing is comparing previous generations versus this one and that is indeed the correct way to look at it. Every new generation will cost more than the previous generation did based on a flat value simply due to time and inflation. A 2000sq ft house costs more today than in 1990, 1990 costs more than 1970. This is how things work. What must be looked at is what the actual end product is in relation to where it would have stood in the past versus what we have today then add in inflation and time.

My 1070 would have been a 560 in 2011 and MSRP was $199 based on where it sits in Nvidia's own product stack. The 1070 released in 2016 at $449. Has prices on everything else around us more than doubled in the past 5 years. I can buy a 6700k for roughly what I could buy a 2600k for. So Intel can but Nvidia isn't able to? I highly doubt that. They simply do not want to. They want their high prices. That is their right, their responsibility to their shareholders. But it is also our job to state such things and call out the discrepancies when they are present in a market. If a 2018 Ford F-150 started at $65k would that be ok? It obviously cost more to produce than any F-150 before it? If your gallon of milk was $12 tomorrow would that be ok? It will obviously cost more to produce that milk tomorrow than it did yesterday.

The question must be is the increase in pricing today versus what it was yesterday justified. A 225% price increase over the past 5 years is a little much. Going back 5 years before 2011 we have the 7600 GS which had an MSRP of $149. Do I need to do the math for you. Why on earth should we all accept a fleecing from a company and then go on forums defending it like we should all be thankful Nvidia even exists to provide us with products. The simple truth is we provide them with a market to sell to. To blindly go along with something is idiotic.

 

[Arachnotronic]

Performance in a vacuum is a poor way to judge any product as it allows any price to be set. What RS is doing is comparing previous generations versus this one and that is indeed the correct way to look at it. Every new generation will cost more than the previous generation did based on a flat value simply due to time and inflation. A 2000sq ft house costs more today than in 1990, 1990 costs more than 1970. This is how things work. What must be looked at is what the actual end product is in relation to where it would have stood in the past versus what we have today then add in inflation and time.

My 1070 would have been a 560 in 2011 and MSRP was $199 based on where it sits in Nvidia's own product stack. The 1070 released in 2016 at $449. Has prices on everything else around us more than doubled in the past 5 years. I can buy a 6700k for roughly what I could buy a 2600k for. So Intel can but Nvidia isn't able to? I highly doubt that. They simply do not want to. They want their high prices. That is their right, their responsibility to their shareholders. But it is also our job to state such things and call out the discrepancies when they are present in a market. If a 2018 Ford F-150 started at $65k would that be ok? It obviously cost more to produce than any F-150 before it? If your gallon of milk was $12 tomorrow would that be ok? It will obviously cost more to produce that milk tomorrow than it did yesterday.

The question must be is the increase in pricing today versus what it was yesterday justified. A 225% price increase over the past 5 years is a little much. Going back 5 years before 2011 we have the 7600 GS which had an MSRP of $149. Do I need to do the math for you. Why on earth should we all accept a fleecing from a company and then go on forums defending it like we should all be thankful Nvidia even exists to provide us with products. The simple truth is we provide them with a market to sell to. To blindly go along with something is idiotic.

How much of a performance increase is there between a 2600K and a 6700K and what is the performance delta between the GTX 560 and the GTX 1070?

You can also play today's games just fine with a 2600K, but a 560 is unusable today.

 

[railven]

Ooooh..wonder if it will release within my Step-UP window from EVGA?

I excite.

 

[SpaceBeer]

@Arachnotronic
You know, it's much easer to increase performance of chips used for parallel computing, than of those used for serial tasks Especially when new production technologies allow you to put much more transistor in same die size.

If you really want to go that way, you should at least compare performance of (CUDA) core at the same clocks (GPU and MEM) between all those cards. I'm not saying there were no improvements in GPU industry, just that it is much harder to make improvement when you see you are reaching known limits (in both design and production tech.)

 

[Arachnotronic]

@Arachnotronic
You know, it's much easer to increase performance of chips used for parallel computing, than of those used for serial tasks Especially when new production technologies allow you to put much more transistor in same die size.

If you really want to go that way, you should at least compare performance of (CUDA) core at the same clocks (GPU and MEM) between all those cards. I'm not saying there were no improvements in GPU industry, just that it is much harder to make improvement when you see you are reaching known limits (in both design and production tech.)

My point is about delivered value to the consumer. I am not dissing Intel for not making bigger improvements, but just pointing out that the performance on GPUs has gone up a lot and a $249 GPU today opens up fundamentally new/better experiences over any GPU from 2011.

 

[DarkKnightDude]

Wait, 1080 Ti only on GDDR5 while 1080 is on GDDR5X? huh?

 

[Arachnotronic]

Wait, 1080 Ti only on GDDR5 while 1080 is on GDDR5X? huh?

it's actually pretty brilliant, wish I'd predicted this. This way, NV can advertise 12GB of VRAM, it can advertise more memory bandwidth than on GTX 1080, but the Titan X has a nice point of differentiation and it is cheaper to produce (GDDR5X is likely much more expensive, especially since only Micron is making it atm, while there are tons of GDDR5 suppliers).

 

[Bacon1]

Wait, 1080 Ti only on GDDR5 while 1080 is on GDDR5X? huh?

Just a way to cut costs while charging more than ever to maximize profits and reduce longevity of the card so you'll upgrade to the 1180 ti.

 

[nurturedhate]

My point is about delivered value to the consumer.

The delivered value at time of release can only be related to the product stack available at the time of release. Relative value of the 560/1070 is roughly equivalent when compared against their product stack.

I am not dissing Intel for not making bigger improvements, but just pointing out that the performance on GPUs has gone up a lot and a $249 GPU today opens up fundamentally new/better experiences over any GPU from 2011.

And a 2011 gpu "opens up fundamentally new/better experiences over any GPU from" 2006. Of course things got faster, you really felt the need to state that?

 

[Arachnotronic]

The delivered value at time of release can only be related to the product stack available at the time of release. Relative value of the 560/1070 is roughly equivalent when compared against their product stack.


And a 2011 gpu "opens up fundamentally new/better experiences over any GPU from" 2006. Of course things got faster, you really felt the need to state that?

Not the point, but whatever.

I want an NVIDIA Titan X for $1, but NVIDIA is too greedy to sell it to me at that price

 

[nurturedhate]

Continuing this little G73 to GF114 to GP104 comparison.

G73 - 177m transistors
GF114 - 1.95b
GP104 - 7.2b

GF114 had rougly 11x as many xtors than G73.
GP104 has 3.7x as many transistors.

 

[PhonakV30]

I didn't know It was GDDR5 Not GDDR5X ! wow!