GTX Titan (final specs and bench)?

Status
Not open for further replies.
Actually the it would be very unlikely for the 780 to "Crush" the 690. Pretty well no previous generation of card has been able to dominate its predecessor in SLI. That would require a nearly 2x increase, and that really doesn't happen.

Based on some numbers I've ran, the GTX 780 should come out anywhere from 5%-15% faster than the GTX Titan, but with lower power use (roughly 5%-10% higher than GTX 680) and a smaller price tag of between $500-$600 thanks to 22nm fab process. The numbers I ran are on the assumption that Nvidia wouldn't be stupid enough to release a 780 that's slower than the Titan, which is where my original estimation of the 780 would have fallen (had it been released now, as opposed to the Titan). All they've done with the Titan is put out a "True" 680 as they were supposed to initially (remember the 680 was never intended to be the 680). They used the same die size in the titan as they had in the gtx 580. So the cost for Nvidia goes up. But it allows them to get one last hurrah out of the 28nm process.

Some numbers you ran up? lol I would like to see where you got those numbers from. For a start the 780 is just a refresh, It's not going to be a 22nm chip. It's going to be built on the 28nm process. So all your numbers are wrong.

Also ithe 780 is not going to be faster than the Titan. Normal refreshes are generally between 15 and 30% faster. Personally I am hoping that both AMD and Nvidia get a little more performance out of this round of refreshes than previously so I am hoping for a 50% increase.

And can't you get it out of your heads, this isn't a normal production card. It isn't a 680.5 or whatever. It was a card designed for use in supercomputers that is now been turned into a gaming card and in very limited quantities because these cards are been made from binned parts that didn't make the cut for the K20x and K20.

The sale of this won't affect the sales of any other cards. The people who will buy this aren't the kind of people who worry about cost and chances are they already have a high end system.
 
I agree for the most part...I think Nvidia's branding of this as 'Titan' maybe is what I'm having more of an issue with...it's not a titan...like you said it's just the true GTX 680 or as I like to call it, the 680.5...a normal video card refresh that due to various reasons got pushed back

Nothing got pushed back. It's not a 680.5, it's not a normal video card refresh, the 7xx series will be that refresh. And don't you know why it's called the Titan? It's because it's the card that's been used in the Oakridge super computer called Titan.
 
lol wut?

The GTX 780 will be on the same 28nm process.
You think a Kepler GK114 "refresh" will yield a 60-90% boost in performance over the GK104 to beat the Titan?

Laughter.



This is funny too.

wow I actually agree with you about something!! :p First time for everything :D
 
Most people are going by the assumption that GTX780 is based on GK114, which is a refresh of GK104.

From what I've read. GK114 is a refined GK104 and the main differences should lie on clocks, efficiency and power draw. Also cheaper because of a more mature 28nm process.

It should also have 8 SMX units as GK104 and are probably the same SMX as Titan (as GK110 is a refresh of GK100 which never came out)

So imo Titan should hold the performance crown for single GPUs, and the 780 will clearly beat the 7970 1ghz.
 
I run 1440p 120Hz. All settings maxed out. 32x CSAA. Between 4x-8x transparency AA. Doesn't help when a game like Farcry 3 has me limited to as low as 80fps~ in some areas while the GPU's never go above 70%.

The only game that's actually maxed out all 4 cards that I can think of is Skyrim maxed out with mods.

Ya, 4x 680's is overkill for a 120 Hz 1440P, especially with a 3770K at only 4.6 GHz. Even running some of my 8-11 mega-pixel 120Hz display monstrocities I couldn't always max out 4x 680's/7970's.

I plan on running two Titan's for my 130 Hz 1440P with a 5 GHz 3770k until Haswell launches in June. That should be a sweet setup all around and should allow me to max everything at 130+ FPS besides uber-sampling modes in some games.

BTW you are right on CPU limiting, there are some settings in games like Crysis 3 that have serious CPU usage issues and even if you had a 7 GHz CPU you wouldn't be able to get 120+ FPS at all times.
 
Also ithe 780 is not going to be faster than the Titan. Normal refreshes are generally between 15 and 30% faster. Personally I am hoping that both AMD and Nvidia get a little more performance out of this round of refreshes than previously so I am hoping for a 50% increase.

And can't you get it out of your heads, this isn't a normal production card. It isn't a 680.5 or whatever. It was a card designed for use in supercomputers that is now been turned into a gaming card and in very limited quantities because these cards are been made from binned parts that didn't make the cut for the K20x and K20.

You sound awfully sure of yourself. If the 780 was only 15%-30% faster than than the 680, that would leave it behind the performance of the 8000 series AMD cards.
 
lol wut?

The GTX 780 will be on the same 28nm process.
You think a Kepler GK114 "refresh" will yield a 60-90% boost in performance over the GK104 to beat the Titan?

Laughter.

Can you show where it's confirmed that the 780 will be the same 28nm process? And a 60%-90% boost after a year and a half isn't too out of place with GPU power increases in the past and moore's law regarding the matter.
 
Can you show where it's confirmed that the 780 will be the same 28nm process?
Nvidia's own roadmap shows that Maxwell will be the first Nvidia card to be manufactured on the 20nm process.
You sound awfully sure of yourself. If the 780 was only 15%-30% faster than than the 680, that would leave it behind the performance of the 8000 series AMD cards.
I wasn't aware that the 8000 series has been tested and benchmarked.
 
Isn't the 8000 series simply a rebrand sold only to OEMs?

For what they have out right now at least, same with nvidia and the 300 series. They will most likely use the 9000's.

Considering how hard a time nvidia had with the titan gpu I don't think they will attempt another big consumer core. So saying a 30% increase lends a reasonable assumption.
 
Nvidia's own roadmap shows that Maxwell will be the first Nvidia card to be manufactured on the 20nm process.

I wasn't aware that the 8000 series has been tested and benchmarked.


AMD said it would be about 45% faster than the 7970. (note: this was before they said the 8000 series was just a rebrand of 7000 series for oem. So for the purpose of our discussion, I suppose we're talking about what is likely to be named the 9000 series) It's just hearsay at this point and the release is delayed until the very end of this year, but it does follow the same trend as the 580->680 and 6970->7970 in terms of generation over generation performance increase.

Looking back...I actually can't remember a time when one series to another had a performance increase of only 15%. If the 780 is going to be a Q2 launch, then a performance increase of 25%-33% is realistic. But I can't foresee any performance increase lower than that. And even being that low, as I mentioned, would have to be a Q2 release. If it ends up being a Q3 launch it would inevitably have to go along with a further delay to the release of Maxwell. So essentially Nvidia's game plan has to be one of the following:

1) Early release of the 780 based on 28nm process using a 380-420mm die to take a clear lead over the 7970 which has been doing better lately thanks to the new drivers and its attractive lower price point, followed by the release of Maxwell shortly after AMD's next generation card is out.

2) Delay release of the 780 until Q4, using 28nm process with a 400mm-450mm die, putting its performance on par and potentially slightly ahead of the next AMD card with a 40%-60% improvement over the 680, and save the Maxwell cards for later in 2014. (though this would seem less likely, as Maxwell would be cheaper to get out with better performance on a smaller die, if it happens to be ready at that point)

Remember Nvidia's own actions show us that they don't like to dominate their competition. They like to put out something just slightly better than their competition to save their higher up stuff for whenever it's actually needed. Makes it cheaper for them. And means they always have a couple cards up their sleeve if needed. (pun intended). And with AMD announcing a delay in the release of their next gen card, Nvidia will have no need to launch Maxwell. Which means they can stretch out their release schedule as well.

So the conclusion is that the 780's performance will be largely determined by when it's going to be launched. It's so hard to get info on 780 rumours right now because the majority of rumours around it were really about the GTX Titan. So...it's going to be a little while before we hear anything, I think.
 
Last edited:
You sound awfully sure of yourself. If the 780 was only 15%-30% faster than than the 680, that would leave it behind the performance of the 8000 series AMD cards.

I am not sure of anything, only making logical estimates. I don't know what the performance of the 8000 series will be, I have seen several articles on various sites saying the performance of the 8970(or whatever) will be anything from 15 to 50% better than the 7970.

Again just looking at the 7970 the biggest complaints have been power use compared to the 680. I think that they will try to reduce that in the 8 series. Concentrate on getting better GPGPU performance per watt.

Whereas nvidia have more room to play with. And as I said in my last post, I am hoping that they will bring a 50% increase and think they can.
 
Can you show where it's confirmed that the 780 will be the same 28nm process? And a 60%-90% boost after a year and a half isn't too out of place with GPU power increases in the past and moore's law regarding the matter.

The refresh from both companies is going to be on the 28nm process. Actually I read somewhere that maxwell might be on the 28nm process as well. It's been a very costly changeover for both companies. Becuase it's been such a difficult transistion, I think the performance increase that we expected to come from changing to 28nm will only come with the refresh. So the 780 will be that 60-90% faster than the 580.
 
AMD said it would be about 45% faster than the 7970. (note: this was before they said the 8000 series was just a rebrand of 7000 series for oem. So for the purpose of our discussion, I suppose we're talking about what is likely to be named the 9000 series) It's just hearsay at this point and the release is delayed until the very end of this year, but it does follow the same trend as the 580->680 and 6970->7970 in terms of generation over generation performance increase.

Looking back...I actually can't remember a time when one series to another had a performance increase of only 15%. If the 780 is going to be a Q2 launch, then a performance increase of 25%-33% is realistic. But I can't foresee any performance increase lower than that. And even being that low, as I mentioned, would have to be a Q2 release. If it ends up being a Q3 launch it would inevitably have to go along with a further delay to the release of Maxwell. So essentially Nvidia's game plan has to be one of the following:

1) Early release of the 780 based on 28nm process using a 380-420mm die to take a clear lead over the 7970 which has been doing better lately thanks to the new drivers and its attractive lower price point, followed by the release of Maxwell shortly after AMD's next generation card is out.

2) Delay release of the 780 until Q4, using 28nm process with a 400mm-450mm die, putting its performance on par and potentially slightly ahead of the next AMD card with a 40%-60% improvement over the 680, and save the Maxwell cards for later in 2014. (though this would seem less likely, as Maxwell would be cheaper to get out with better performance on a smaller die, if it happens to be ready at that point)

Remember Nvidia's own actions show us that they don't like to dominate their competition. They like to put out something just slightly better than their competition to save their higher up stuff for whenever it's actually needed. Makes it cheaper for them. And means they always have a couple cards up their sleeve if needed. (pun intended). And with AMD announcing a delay in the release of their next gen card, Nvidia will have no need to launch Maxwell. Which means they can stretch out their release schedule as well.

So the conclusion is that the 780's performance will be largely determined by when it's going to be launched. It's so hard to get info on 780 rumours right now because the majority of rumours around it were really about the GTX Titan. So...it's going to be a little while before we hear anything, I think.

I can go along with some of what you say here. But there are a couple of things I don't agree with.

First, I don't think either company is ready for the transistion to 20nm. Look at all the problems they had moving to 28nm. So it might be very wrong to say that Maxwell would be cheaper to get out.

And history has shown that Nvidia does like to dominate the oppostion. Look at the 8800 GT, the 8800 GTX and the 8800GTX ultra as the most obvious example. ATI had nothing to compete with the 8800GT, but they still released the 8800GTX ultra!!

You could well be right about release date deciding how powerfull the 780 will be. It's all speculation at this stage!!
 
The refresh from both companies is going to be on the 28nm process. Actually I read somewhere that maxwell might be on the 28nm process as well. It's been a very costly changeover for both companies. Becuase it's been such a difficult transistion, I think the performance increase that we expected to come from changing to 28nm will only come with the refresh. So the 780 will be that 60-90% faster than the 580.
so you think just a few months later they will release a 780 that will be as fast as the $900 Titan? to be clear a gtx690 is 95% faster than a gtx580 and Titan is supposed to be about 10% slower than that.
 
Whereas nvidia have more room to play with. And as I said in my last post, I am hoping that they will bring a 50% increase and think they can.

My apologies. I must have misunderstood your original post. I thought you were saying there would be a substantially smaller performance increase.
 
so you think just a few months later they will release a 780 that will be as fast as the $900 Titan? to be clear a gtx690 is 95% faster than a gtx580 and Titan is supposed to be about 10% slower than that.

Depends what you call a "Few Months." With the Titan launching in February, and the 780 possibly coming as late as Q4, that's potentially a good 3/4th of a year between releases. And with the Titan being a limited edition VERY SMALL QUANTITY run for enthusiasts only who want to get top technology many months ahead of others, it's not that unreasonable.
 
And history has shown that Nvidia does like to dominate the oppostion. Look at the 8800 GT, the 8800 GTX and the 8800GTX ultra as the most obvious example. ATI had nothing to compete with the 8800GT, but they still released the 8800GTX ultra!!

You could well be right about release date deciding how powerfull the 780 will be. It's all speculation at this stage!!


I think that changed with the release of the GTX 680. It was supposed to be the 670 or even 660, considering how much smaller of a die they used for it compared to the gtx 580. But if they had released that as the 660, and released the actual 680 as the titan, that would have been a card literally 2x faster than the 7970 at the time. And the technology was there to do it. But the opportunity for more profit was there, and they took it. I think AMD is destroying technology. Look at both Intel and Nvidia. Because of AMD's poor performance, neither of the 2 have felt the need to push ahead too far. Just put out something a bit better than the competition, and you're done. No need to go all out.

Basically the way I'm looking at it is from a business standpoint. Nvidia has the technology today, to put out a card faster than what AMD will put out a year from now. That would be the Titan. But if people just buy the best card on the market, and their card is already the best one out there, they don't feel much of a need to compete against their own product until one of the following happens:

1) AMD comes out with a better card, and Nvidia is forced to top them.
2) They feel the GTX 680 has been out for long enough, and they feel they have something with enough of a performance advantage to make people who bought a 6-series card do a hardware upgrade (similar to Apple's business model with the iPhone). That would require a significant enough performance increase to work.

Anything outside of those 2 scenario's would be a bad business move, and if we go off of Nvidia's reaction to the 294mm die GTX 680's performance, then they're quite aware of this.
 
ON the rumour side of things, I just read on xbitlabs that there is only 10000 Titans available worldwide.

http://www.xbitlabs.com/news/graphi...d_Up_Launch_of_GeForce_GTX_Titan_Reports.html

Based on the recent rumours, the GeForce GTX Titan will not beat the dual-chip GeForce GTX 690 or Radeon HD 7990 in benchmarks, which means that gamers looking for maximum performance should not care about the new product unless they have problems with multi-GPU configuration of the model 690.

lol, this seems kind of backwards, unless they meant to say "single slot gamers".
 
Depends what you call a "Few Months." With the Titan launching in February, and the 780 possibly coming as late as Q4, that's potentially a good 3/4th of a year between releases. And with the Titan being a limited edition VERY SMALL QUANTITY run for enthusiasts only who want to get top technology many months ahead of others, it's not that unreasonable.

10,000 cards isn't THAT small. That ASUS 7990 is only 1,000 units.
 
The people in here comparing prices of this and that are missing the fact that this company doesn't care what passed prices were, or are. This company can make the price what ever the fuck they want. And that is that. If they want it at this price, so be it. Whining about it only shows how poor you are. Just deal with your current graphics and dry your tears, life will still be fine. There will be better prices on this gear later, and the drivers will get better, and the games will be better optimized. It gets better. Just calm down.
 
Anyone who thinks people who complain about price do so because they are poor has little experience in this thing we call life. There are varying personalities among the human race, some of the wealthiest people I know are also some of the most frugal.

I have a fair amount of buying power and I think I'll complain, because I can and it makes me feel better. Just like nVidia can dictate price because it makes them feel better.

And the word you're looking for is "past" not "passed"
 
Anyone who thinks people who complain about price do so because they are poor has little experience in this thing we call life. There are varying personalities among the human race, some of the wealthiest people I know are also some of the most frugal.

I have a fair amount of buying power and I think I'll complain, because I can and it makes me feel better. Just like nVidia can dictate price because it makes them feel better.

And the word you're looking for is "past" not "passed"

IMO, those who buy this might not even have the money to afford it. They're not called enthusiast for nothing, and that's what the Titan is marketed towards.

People can complain all they want. But it won't change the fact (speculation) of the performance of the Titan and that they will all sell.

Will I buy one? Nope. I don't have much reason to. Can I afford one? Yes...I can buy three without much problem (tax refund) or worry about losing my house/car/etc. I'm just not that [H]ard, and already have a GTX 680 that i'm waiting to SLI later on this year when the used market has cheap prices.
 
I'm curious if anyone has considered that Nvidia might not include an SLI connector on these cards..The reason I mention this is that several people have stated "Well you can get a single Titan now and then in 2-3 years, get another one and SLI them" in order to stay ahead of Maxwell etc..

Well, considering they are only making 10K of these cards, that might make them decide to save a few pennies and not include an SLI connector..And no, I am not trolling..
 
That will not happen, they would get a mountain of shit if they didn't include the connector
 
I'm curious if anyone has considered that Nvidia might not include an SLI connector on these cards..The reason I mention this is that several people have stated "Well you can get a single Titan now and then in 2-3 years, get another one and SLI them" in order to stay ahead of Maxwell etc..

Well, considering they are only making 10K of these cards, that might make them decide to save a few pennies and not include an SLI connector..And no, I am not trolling..

I'd say that's about as likely as an asteroid crashing into the moon and sending it orbiting around Mars, becoming the catalyst for life on that planet... But sure, it could happen
 
10,000 cards isn't THAT small. That ASUS 7990 is only 1,000 units.

I'd think that the number of people looking for dual-gpu cards is probably a LOT smaller than the number of people that would pony up for a really fast single gpu.
 
I'd think that the number of people looking for dual-gpu cards is probably a LOT smaller than the number of people that would pony up for a really fast single gpu.

But someone could get nearly 3 GTX 670's (on sale) for the same price as a single Titan. I think the Titan is targeted towards a different crowd. The crowd that will spend whatever money without worry. Especially knowing that it's a limited production. If the performance numbers are real, I'm looking at getting either 2 or 3 of them myself. I'm sure there are quite a few others out there. The Titan would be considered a Crippled card without SLI because it'd be amazing as a single GPU, but weaker than a 690 or 680 sli with no option to increase performance.
 
yup the crowd who goes for bling bling...

I don't think that's fair. My reason for doing it is that I'm finding some compatibility issues with Quad-SLI, even though I enjoy the performance I get from it. I had to disable and unplug one of my cards, running in Tri-SLI now to resolve my problem. 2 or 3 Titan's would be a great solution for me. And considering I do a shit ton of video encoding, these cards would be golden for me for at least 2 years.
 
Hasn't GPU encoding shown to provide considerably worse results (quality) than CPU?
 
Hasn't GPU encoding shown to provide considerably worse results (quality) than CPU?

Not that I've heard....and especially not for lossless codecs. Nor with higher end modern gpu's and opencl. Unless you have info that I'm missing? Besides even if there were some loss in quality, it would be nothing compared to the quality loss from YouTube's final compression.
 
Not that I've heard....and especially not for lossless codecs. Nor with higher end modern gpu's and opencl. Unless you have info that I'm missing? Besides even if there were some loss in quality, it would be nothing compared to the quality loss from YouTube's final compression.


A quick google search with the following results, there's a lot more. In addition to worse quality, it appears that the file size is also larger.

http://www.extremetech.com/computing/128681-the-wretched-state-of-gpu-transcoding

http://techreport.com/review/23324/a-look-at-hardware-video-transcoding-on-the-pc/6

Putting it on youtube is a moot point, the better the source, the better it's going to look on youtube regardless of their compression.
 
A quick google search with the following results, there's a lot more. In addition to worse quality, it appears that the file size is also larger.

http://www.extremetech.com/computing/128681-the-wretched-state-of-gpu-transcoding

http://techreport.com/review/23324/a-look-at-hardware-video-transcoding-on-the-pc/6

Putting it on youtube is a moot point, the better the source, the better it's going to look on youtube regardless of their compression.

From the link you listed, the GTX 580 encode looks much better than the cpu encode. But upon further review you seem to be right. Looks like the majority of my encoding has been purely CPU based. There is still one thing I'm looking forward to with the Titan, however. And that's Smooth Video Project finally being able to run 120fps 1440p content. There's no SLI support and a single 680 falls slightly short. :(
 
The editors of the articles certainly don't think so. Quality was just one area of concern with GPU encoding. Maybe its better today, but I haven't seen anything suggesting that it is.
 
The editors of the articles certainly don't think so. Quality was just one area of concern with GPU encoding. Maybe its better today, but I haven't seen anything suggesting that it is.

You're certainly correct. I know the kepler series actually has h264 encoder built in to the hardware outside of generic gpu acceleration. I wonder if that helps at all. Or how Tesla or Titan compare to the 580 they tested in the article. Sounds like something I may need to do a review on? =D
 
Oh man I get so giddy when new graphics hardware is around the corner. All this talk about pricing and such is lame
 
what makes this card hilarious is not the price, but the fact that it will most likely be voltage locked limiting overclocks. one thing I can say AMD does right is allows relatively unlimited overclocking options (i know there are recent PCB's from brands like sapphire that lock certain voltages, but most 7970's are not voltage locked)
 
Status
Not open for further replies.
Back
Top