Rumor: Titan V in April, GTX 1080 in June

At least for the last few Titan's, yes, I'd agree.

Maybe not for this one.
 
In my opinion, any single-GPU gaming video card with a $1,000 price tag is ridiculously priced.

A single, top-end card that could best 980Ti SLI performance would not be even remotely overpriced at $1000. It would be a bargain in that a 980Ti SLI combo easily runs ~$1250 right now... and you have all the extra headaches to contend with that SLI can introduce.

Granted, if it does hit and only provides 20% more over what a single 980Ti is capable of delivering, than ya, I'd agree with you.

Would be thrilled with a single card that could beat 980Ti SLI performance. But I'm thinking that the first iteration of Pascal won't bring quite that much brawn with it. Perhaps the "Ti" version of Pascal will...
 
Last edited:
A single, top-end card that could best 980Ti SLI performance would not be even remotely overpriced at $1000. It would be a bargain in that a 980Ti SLI combo easily runs ~$1250 right now... and you have all the extra headaches to contend with that SLI can introduce.

Granted, if it does hit and only provides 20% more over what a single 980Ti is capable of delivering, than ya, I'd agree with you.

Would be thrilled with a single card that could beat 980Ti SLI performance. But I'm thinking that the first iteration of Pascal won't bring quite that much brawn with it. Perhaps the "Ti" version of Pascal will...
That may make sense if there were a direct correlation between price and performance, but you also have to take into account the advancing technology. The GTX 970 offered the same level of performance of a GTX 780 when new, but was $300 cheaper at release. Similarly, a single GTX 980 Ti offers the performance of GTX 780 SLI, but at half the price.
 
A single, top-end card that could best 980Ti SLI performance would not be even remotely overpriced at $1000. It would be a bargain in that a 980Ti SLI combo easily runs ~$1250 right now... and you have all the extra headaches to contend with that SLI can introduce.

Granted, if it does hit and only provides 20% more over what a single 980Ti is capable of delivering, than ya, I'd agree with you.

Would be thrilled with a single card that could beat 980Ti SLI performance. But I'm thinking that the first iteration of Pascal won't bring quite that much brawn with it. Perhaps the "Ti" version of Pascal will...

Totally agreed. If a single GPU comes out that does beat 980TI SLI then it is an easy decision for me as I am committed to never going crossfire/sli again.

Mostly I am looking forward to seeing the temp and noise from the new cards. I'm building an NCase M1 build and a great single GPU with less heat would be amazing.
 
Totally agreed. If a single GPU comes out that does beat 980TI SLI then it is an easy decision for me as I am committed to never going crossfire/sli again.

Mostly I am looking forward to seeing the temp and noise from the new cards. I'm building an NCase M1 build and a great single GPU with less heat would be amazing.

Pascal will be amazing for mini builds like that, that's a given. Performance per watt is going to be improved enormously. What's not a given is raw performance versus the 980 Ti.

I am concerned that rumors about the performance of Pascal may have been corrupted - it's possible that whispers of "50% to 80% improvement in performance per watt" got altered by the rumor mill into "50% to 80% improvement in performance", leading to people thinking the first Pascal card will outperform 980 Ti SLI, a legendary level of disruption we have not seen since the 8800 GTX. If it sounds to good to be true...
 
Pascal will be amazing for mini builds like that, that's a given. Performance per watt is going to be improved enormously. What's not a given is raw performance versus the 980 Ti.

I am concerned that rumors about the performance of Pascal may have been corrupted - it's possible that whispers of "50% to 80% improvement in performance per watt" got altered by the rumor mill into "50% to 80% improvement in performance", leading to people thinking the first Pascal card will outperform 980 Ti SLI, a legendary level of disruption we have not seen since the 8800 GTX. If it sounds to good to be true...

It's very unlikely imho.

Die size is going to be limited by relatively new process so for a smaller die that needs to also have DP units it will be hard to beat 600mm^2 of GM200 fully dedicated to gaming.

Another unknown factor is frequency of those new GPUs
 
pretty sure there are no more dedicated DP units...... the ALU's are all capable of mixed precision for Pascal....
 
A single, top-end card that could best 980Ti SLI performance would not be even remotely overpriced at $1000. It would be a bargain in that a 980Ti SLI combo easily runs ~$1250 right now... and you have all the extra headaches to contend with that SLI can introduce.

Granted, if it does hit and only provides 20% more over what a single 980Ti is capable of delivering, than ya, I'd agree with you.

Would be thrilled with a single card that could beat 980Ti SLI performance. But I'm thinking that the first iteration of Pascal won't bring quite that much brawn with it. Perhaps the "Ti" version of Pascal will...

That may make sense if there were a direct correlation between price and performance, but you also have to take into account the advancing technology. The GTX 970 offered the same level of performance of a GTX 780 when new, but was $300 cheaper at release. Similarly, a single GTX 980 Ti offers the performance of GTX 780 SLI, but at half the price.

Exactly. You don't compare new tech to old tech and scream "see that's good value!"

Pascal will be amazing for mini builds like that, that's a given. Performance per watt is going to be improved enormously. What's not a given is raw performance versus the 980 Ti.

I am concerned that rumors about the performance of Pascal may have been corrupted - it's possible that whispers of "50% to 80% improvement in performance per watt" got altered by the rumor mill into "50% to 80% improvement in performance", leading to people thinking the first Pascal card will outperform 980 Ti SLI, a legendary level of disruption we have not seen since the 8800 GTX. If it sounds to good to be true...

GP100 might rival 980 Ti SLI assuming average (~70%) scaling. But GP104 rivaling 980 Ti SLI? Not a snowball's chance in hell.
 
Pascal will be amazing for mini builds like that, that's a given. Performance per watt is going to be improved enormously. What's not a given is raw performance versus the 980 Ti.

I am concerned that rumors about the performance of Pascal may have been corrupted - it's possible that whispers of "50% to 80% improvement in performance per watt" got altered by the rumor mill into "50% to 80% improvement in performance", leading to people thinking the first Pascal card will outperform 980 Ti SLI, a legendary level of disruption we have not seen since the 8800 GTX. If it sounds to good to be true...

If Nvidia indeed had that kind of improvement, they'd most likely milk all they can from it by releasing increasingly faster cards over the next two years. Depending on how AMD's Polaris performs.
 
Exactly. You don't compare new tech to old tech and scream "see that's good value!"

What?! If the old tech in SLI significantly outperforms the new tech at a much lower price point, then hell yes, I would say it is a better value. Likewise, just like I said, if pascal comes in faster and cheaper, then it is the better value.

I simply was comparing what kind of performance you get for the $ NOW, regardless of tech/age of their top end cards. Odds are the 780Ti/Titan X cards won't plummet in price when pascal hits unless it is significantly faster and priced aggressively to match... Typically the newer tech does get faster/cheaper, but the bottom line for me is what is highest level of PERFORMANCE I can get from their top tier product, either last gen or this gen for the $ spent... And I'm willing to go SLI to single card in the performance to value comparisons.
 
Well let's think about Maxwell for a sec. The 980 launched at $550, and was like what, barely 10-15% faster than 780 Ti? But despite that the 780 Ti lost 50% value overnight, and you could get a 780 Ti SLI setup for not too much more than a single 980. Or if you don't mind buying used it gets even cheaper. Of course the 970 at $330 didn't help things either, even though it was about 10% behind a 780 Ti at launch.

So the 780 Ti suddenly became good value overnight, but only because new tech forced it to become very cheap. That's what I was getting at, you can't use the NOW (pre-release of new tech) price vs the new tech's release price to quantify value, because unless the new tech had ridiculously bad perf/$, the old tech will have a massive drop in price, and become good value.
 
Alright, I grant you that - we simply don't know what the drop in price for the 980Ti cards are going to be when pascal hits. Nor how much pascal will cost... nor how much better exactly it will perform.

If a pair of 980Ti's in SLI runs roughly the same cost (after they drop in price) as a single pascal card (and a pascal card doesn't handily beat them when it comes to overall performance), then for me at least, the performance/value nod would still go to the 980Ti's. That's the value comparison I simply was trying to make.
 
I hope they call it the Titan V and it comes out in April. I bet it will be a huge upgrade from my GTX 780 Ti lol.
 
I read that Pascal Titan is coming around April, GTX 1080/1070 during the summer, and 1080Ti June 2017....
JUNE 2017!!!

Think I'll just buy a second 980 Ti. Probably someone will be selling their cheap once Pascal Titan is near release.

as I saied in other thread Titan series are only for enthusiasts that has more money than good sense.

the Ti series is the hot spot for me, a pair of 980 Ti is faster than a single Titan X and gives better performance for quite the same money.

I'm really interested in the Ti series of Pascal, hope to see it soon.
 
Nvidia is super good at marketing. With the Titan line, they effectively made a 650$ video card look like a "deal" and somehow 1250$ worth of video card is not outrageous and uncommon since its only 250$ more than a Titan for twice the performance.

I built my i7 920 system for the same amount it would cost me ONLY in video cards today to get the same relative performance (thats without the fact that us canadians have to pay 900$ for a 980 ti lol).

I am super excited for Pascal but I really hope the pricing wont see a huge hike again.
 
Nvidia is super good at marketing. With the Titan line, they effectively made a 650$ video card look like a "deal" and somehow 1250$ worth of video card is not outrageous and uncommon since its only 250$ more than a Titan for twice the performance.

I built my i7 920 system for the same amount it would cost me ONLY in video cards today to get the same relative performance (thats without the fact that us canadians have to pay 900$ for a 980 ti lol).

I am super excited for Pascal but I really hope the pricing wont see a huge hike again.

If this was true, why are they only making 10% profit and AMD can't undercut them and take market share back? AMD is bleeding money, that shouldn't be happening if you were correct. Well you are correct they have good marketing!

Costs increase as we go lower in node size. Gets tough when you have billions of gates, that are supposed to flip billions of times a second, and are only a few atoms wide... Perf/$ has been going down, but since the Titan V is rumored to have twice the transistors and they aren't half the price there's a good chance the prices will move a bit.
 
Last edited:
as I saied in other thread Titan series are only for enthusiasts that has more money than good sense.

the Ti series is the hot spot for me, a pair of 980 Ti is faster than a single Titan X and gives better performance for quite the same money.

I'm really interested in the Ti series of Pascal, hope to see it soon.

Normally, I'd agree, but I made a post which was laughed at which made in the context of the information that Titan V will debut a whole year before Ti series, which was the main reason for me to expect the Titan V rather than Ti version (a year is a LONG time to wait for a $350 savings).

Titan X, I'd fully agree with you. I was not remotely impressed with the performance on that thing compared to the price tag, I held it in lower regard than 980, which I initially thought was a total rip off compared to 970, even with the whole VRAM fiasco.

But now that Titan V has been 'corrected' to Q1/17 release (a much more reasonable release date, albeit far too inconvenient), I guess I will once again look forward to 1080ti. Just not too sure if I should wait for 1080ti and put up with my 970 SLI for a year longer or get 980ti SLI and forget about pascal (I don't sell GPU's, not where I live anyhow).
 
Tesla cards have much higher margins than GeForce parts, yields for 16nm FF+ likely to be questionable initially, makes much more sense to harvest every working die for a Tesla rather than a GeForce part.
 
Look at the launch prices for Tesla cards. Why would you sell Titans for a quarter of the price before the demand for Tesla cards fell below the supply?

Professionala cards were always 4-5x more expensive than comparable consumer cards /essentialy the same GPU/. The difference has been justificated by more advanced hardware (e.g. ECC RAM, components tested for 24/7 operation), drivers supporting high-end professional software, better support and warranty, and so on. That was doubted by many people and even in some cases consumer cards were reflashed to upgrade them to professional cards. Cynics see the 'professional' range just as milking money from organizations for which the card cost is just a freckle considering total cost of developer and projects they do. It will be interesting to see how NV will differenitate this time between the professional and consumer Pascal cards. High-end consumer cards are priced on the pain-threshold for fans meaning there is sufficient population which will buy despite the pain and above the threshold there is huge drop in the size of this population, that threshold was recently around $1K. Think about Titan X, for about $1K quite a few people were still buying (I bought two :)), for $1.5K that would not be the case. The same is in fact with any high-end product, even the Tesla card would not be bought if the price hovered around $10K.
 
A lot of talk about performance, but we can really just guess that.

Now, what I'd like to know, is if anyone can predict when nVidia is actually going to release some info? I guess leaks are harder to trust (there are none anyway). But is there some computer convention coming soon where we might expect nVidia to release some information?
If Titan V is 50%+ faster than 980Ti SLI at 4k, then I'm selling my 980Ti's right now and getting like a used 960/970 while I wait.
 
It's not going to be 50% faster than 980 Ti SLI I can promise you that much (well assuming 70% scaling that is).
 
yeah its probably going to be around the 50% to 80% increase from a single card.
 
Now, what I'd like to know, is if anyone can predict when nVidia is actually going to release some info? I guess leaks are harder to trust (there are none anyway). But is there some computer convention coming soon where we might expect nVidia to release some information?

Well GDC is in the third week of March. If they don't even mention Pascal there then I'd be pretty convinced that we won't be seeing anything before mid-Summer.
 
Well GDC is in the third week of March. If they don't even mention Pascal there then I'd be pretty convinced that we won't be seeing anything before mid-Summer.
Wrong. There is even more important Nvidia GPU Technology Conference on April 7-10 where the Pascal anouncement would fit even better. Remember that Pascal is primarily positioned for the emerging artificial intelligence market a.k.a. deep machine learning.
 
Wrong. There is even more important Nvidia GPU Technology Conference on April 7-10 where the Pascal anouncement would fit even better. Remember that Pascal is primarily positioned for the emerging artificial intelligence market a.k.a. deep machine learning.
Well, they are trying to get into the datacenter as much as possible, but they make 10x as much money on Geforce than Tesla/Quadro, so I don't think they really be quiet during GDC, only to wait for GTC.
 
I read today that AMD might talk about their new GPU at GDC, we'll see if nVidia will keep quiet or have a presentation of their own.
It's been so quiet on both ends, I really that that, at least give us some rumors damnit.
 
Well, they are trying to get into the datacenter as much as possible, but they make 10x as much money on Geforce than Tesla/Quadro, so I don't think they really be quiet during GDC, only to wait for GTC.

Note that Pascal is optimized for computations in a branch of artificial intelligence known as deep machine learning, this is why it has extensive support for flexible floating point including FP16 which has no other use than deep learning. This is potential market hugely bigger than gaming and HPC, think about Pascal chips in every assisted- and self-driving car. Obviously it may happen that NV will announce consumer Pascal cards at the GDC and professional Pascal cards a the GTC.
 
Titan X wasn't the tenth so how you figure Titan V is the 5th?
Because there were four other Titan cards before it.

I sort of expect the Titan name to be dropped for the Volta based "Titan" and instead it be called "Summit" although Titan is probably a better brand name.
 
Call me a skeptic. First GPU will be a 300mm - 400mm^2 midsized part, then only after a year or so will we see the big die version.
 
If Polaris is really as good as the rumor mill has been hinting, NVIDIA will probably go the "Titan Black" or "Titan X" route and go all out to claim the performance crown (or, at least stop AMD from a clear win). Given that TSMC's issues are what prevented AMD from getting the most out of their architecture, I'm thinking that Polaris will be a lot better than what the Fury X currently is -- 30-50% performance gains over the Fury X in the 1440p to 4K categories is the ballpark range I'm thinking of. If anything, I'm more curious to see what NVIDIA does with Pascal and HBM.
 
Back
Top