[Rumor]GTX 980Ti to launch somewhere between May 16-26th

[RussianSensation]

Another thing to mention is 980's efficiency, it's a more efficient card than the Titan X so it might be more efficient than the 980 Ti as well.

The difference between them is hardly worth mentioning. Titan X has 12GB of GDDR5 that likely contributes to its 9% hit in perf/watt. I bet a GM200 6GB card will scale linearly in efficiency vs. a 980.

Also, don't forget that real efficiency should be considered in the context of the overall system, not just on a GPU vs. GPU basis since you can't play games without a CPU/mobo, etc.

If we consider perf/watt from a total system power usage, a Core i7 rig with the Titan X is more efficient than one with a 980. That's why perf/watt metrics on a GPU vs. GPU basis are highly misleading when calculating efficiency.

Proof:

Titan X rig with an i7 draws 14% more power than a similar rig with a 980.

But it's 49% faster in the game against the 980!

Therefore, the Titan X is FAR more efficient than a 980 for a gaming rig. It's about time that sites start considering perf/watt in the context of the overall system power usage because measuring perf/watt on a card basis is not providing the reader with the true story of the Gaming Rig's overall efficiency. That's why even if GM200 6GB matches 980 in efficiency, the overall efficiency of the gaming rig will be superior with a GM200 6GB card.

 

[mojothehut]

Just give me the damn 980ti's already. I'll spend the money, throw up a little, get over it, game on.

Here here!
I've saved up cash for it for the last 6 months. Soon as an Asus/MSI model comes out, BAM. Good bye trusty 7950, hello 980ti. (unless Red camp seduces me again with the 390..)

 

[96Firebird]

Therefore, the Titan X is FAR more efficient than a 980 for a gaming rig.

Depends on the review, and probably the game... Anandtech found them to be quite similar using total system power:

392/301 = 30% more power for the Titan X, while:

http://www.anandtech.com/show/9059/the-nvidia-geforce-gtx-titan-x-review/6

Performance is 33.3% faster for the Titan X. This is the average of all results, since AT does not specify which was used for the power consumption number. Worst case, the Titan X is 35% faster.

 

[Aikouka]

16nm/14nm GPUs are atleast 12 - 15 months away. Apple and Qualcomm will clean out any 16/14nm wafer allocation at TSMC/Samsung in 2015.

Do mobile chips really matter here? Mobile chips are typically on a low power variant of a process where GPUs use a high performance variant. I don't exactly profess to be an expert on lithography, but I don't believe these are created on the same line. I guess my point is that Apple and Qualcomm fighting over the LP semi work doesn't really affect companies that need high-speed products like GPUs.

If this rumor is actually true, is NVIDIA not creating a situation where no one wants to buy their flagship Titan cards anymore? It seems like an odd strategy to sell one card as your "extreme high end flagship" for $1000+ and then release a faster card a few months later for less...

Well, in NVIDIA's presentation for the Titan X, they pretty much focused on computation usage. The idea of the 12GB of on-board memory is to satisfy large data sets, and frankly, that benefit will not disappear even if the consumer-oriented card is just as fast. Now, if someone is fine with a 6GB memory buffer, then they might go for the cheaper card.

Therefore, the Titan X is FAR more efficient than a 980 for a gaming rig. It's about time that sites start considering perf/watt in the context of the overall system power usage because measuring perf/watt on a card basis is not providing the reader with the true story of the Gaming Rig's overall efficiency. That's why even if GM200 6GB matches 980 in efficiency, the overall efficiency of the gaming rig will be superior with a GM200 6GB card.

...and we should see even higher efficiency for the rumored 980Ti given less memory chips that will need power (12GB -> 6GB). Albeit, I'm not sure how much juice these chips take, but that would be an interesting number to know.

 

[A.t]

Therefore, the Titan X is FAR more efficient than a 980 for a gaming rig.

I think you're mixing up the words "effectiveness" and "efficiency", because the 980 is a more power efficient card than the Titan X, however, Titan X is the more effective card because it's faster. Why should I look at performance per watt? I don't spend most my time gaming. It's not a more efficient card, it's a less efficient card as it pulls more power.

 

[RussianSensation]

Depends on the review, and probably the game... Anandtech found them to be quite similar using total system power:

392/301 = 30% more power for the Titan X, while:

http://www.anandtech.com/show/9059/the-nvidia-geforce-gtx-titan-x-review/6

Performance is 33.3% faster for the Titan X. This is the average of all results, since AT does not specify which was used for the power consumption number. Worst case, the Titan X is 35% faster.

The Titan X is faster by more than 33% against the 980 when you use GPU limited games/resolutions. At 4K the Titan X is nearly 40% faster than the 980.

Per the AT chart you linked, AT specified Crysis 3. The Titan system uses 30% more power but it's 35% faster in Crysis 3. Also, the ONLY 980 card that can exhibit that power usage per AT is the reference blower card. Not everyone buys those which basically means the Titan X rig should end up efficient than just about any GTX980 setup unless one plays a lot of CPU limited titles.

I think you're mixing up the words "effectiveness" and "efficiency", because the 980 is a more power efficient card than the Titan X, however, Titan X is the more effective card because it's faster. Why should I look at performance per watt? I don't spend most my time gaming. It's not a more efficient card, it's a less efficient card as it pulls more power.

I never said you should care about perf/watt. I am simply stating perf/watt should be measured on a Total System basis. Stating that a reference 980 is more power efficient than a reference Titan X ignores the actual efficiency a PC gamer "pays for" / experiences when he is gaming.

Do people drive a car's engine or the entire car? That's why the car's overall fuel economy efficiency is dictated by a combination of factors: Engine's efficiency, coefficient of drag, rolling resistance of the car's tires, etc.

In a similar fashion, comparing the efficiency of a GPU on a card basis is something an engineer might care about. As a gamer/user, what we want is the total system efficiency because our entire rig is used to generate IQ/FPS. That makes the Titan X rig both more effective and more power efficient.

Well, in NVIDIA's presentation for the Titan X, they pretty much focused on computation usage. The idea of the 12GB of on-board memory is to satisfy large data sets, and frankly, that benefit will not disappear even if the consumer-oriented card is just as fast. Now, if someone is fine with a 6GB memory buffer, then they might go for the cheaper card.

What I meant in the gaming scene, GM200 6GB (MSI Lightning/EVGA Classified treatments, etc.) or R9 390X will make the Titan irrelevant. For those rendering and doing semi-professional work, sure the Titan X will have its place.

...and we should see even higher efficiency for the rumored 980Ti given less memory chips that will need power (12GB -> 6GB). Albeit, I'm not sure how much juice these chips take, but that would be an interesting number to know.

Agreed. That's why once the Titan X launched, I right away called it that a 1200mhz+ GM200 6GB with after-market coolers is really what we want. The Titan X is a compromise on 2 fronts: (1) shoved arguably useless amounts of VRAM for gaming at the expense of power usage and GPU clocks (2) limited air cooling options to the reference blowers which resulted in compromised noise levels in overclocked states. GM200 6GB should address both of those points.

Just months after the OG Titan's launch, AIBs released factory pre-overclocked 780 cards that made the Titan obsolete for gaming.

And they managed to hit 30 dBA at load (!), while operating at 76C in max overclocked states (!)

If NV doesn't neuter the consumer GM200 6GB card too much, MSI Lightning and EVGA Classified GM200 6GB cards should beat the Titan X for $200+ less.

 

[Mondozei]

I think you're mixing up the words "effectiveness" and "efficiency", because the 980 is a more power efficient card than the Titan X, however, Titan X is the more effective card because it's faster. Why should I look at performance per watt? I don't spend most my time gaming. It's not a more efficient card, it's a less efficient card as it pulls more power.

Effiency is performance per watt. It's not overall power draw. If you don't know what effiency means, please don't try to lecture others.
(But bonus points for unintended hilarity)

What is the most effective card? That depends on your priorities.

 

[A.t]

I never said you should care about perf/watt. I am simply stating perf/watt should be measured on a Total System basis. You can'd drive a car's engine without a car. That's why the car's overall fuel economy efficiency is dictated by a combination of factors: Engine's efficiency, coefficient of drag, rolling resistance of the car's tires, etc.

In a similar fashion, comparing the efficiency of a GPU on a card basis is something an engineer might care about. As a gamer/user, what we want is the total system efficiency because our entire rig is used to generate IQ/FPS. That makes the Titan X rig both more effective and more power efficient.

Let me put it this way, whether you go by custom or reference 980's, the 980 will, under whatever case, pull less power than a Titan X.

The Titan X is faster, but the 980 is more power efficient, and if you don't believe it, check out the reviews of the cards.

The Asus Strix, for example, pulls even less power than the reference 980. What is more is, you mention of entire power system consumption like having a Titan X reduces it. It does not. The card pulls more power, therefore the system also pulls more power. For the reference 980, TechPowerUp pulled a 156W from GPU while the Titan X pulled 223W. For the Strix, as a non-reference example, they pulled even less at an average of 147W. So how come the Titan X becomes more efficient? Aren't we talking about power efficiency here? Because that's what efficiency refers to in a tech forum. If you're going to say the Titan X is the better performing card, there are other ways of saying so, such as, Titan X performs better. Why am I telling this?

Effiency is performance per watt. It's not overall power draw. If you don't know what effiency means, please don't try to lecture others.
(But bonus points for unintended hilarity)

What is the most effective card? That depends on your priorities.

No, it's not. An efficiency of a PSU, for example, has absolutely nothing to do with its performance. It is the total power draw.

 

[Stuka87]

The question I keep asking myself is if there is actually room between a Titan X and a 980 for another card. In some cases they are not that far apart.

When the 780 came out, the Titan had been out for quite some time, and there was a huge difference between a 680 and a Titan. That big gap doesn't exist here.

 

[iiiankiii]

Let me put it this way, whether you go by custom or reference 980's, the 980 will, under whatever case, pull less power than a Titan X.

The Titan X is faster, but the 980 is more power efficient, and if you don't believe it, check out the reviews of the cards.

The Asus Strix, for example, pulls even less power than the reference 980. What is more is, you mention of entire power system consumption like having a Titan X reduces it. It does not. The card pulls more power, therefore the system also pulls more power. For the reference 980, TechPowerUp pulled a 156W from GPU while the Titan X pulled 223W. For the Strix, as a non-reference example, they pulled even less at an average of 147W. So how come the Titan X becomes more efficient? Aren't we talking about power efficiency here? Because that's what efficiency refers to in a tech forum. If you're going to say the Titan X is the better performing card, there are other ways of saying so, such as, Titan X performs better. Why am I telling this?



No, it's not. An efficiency of a PSU, for example, has absolutely nothing to do with its performance. It is the total power draw.

What are you talking about, bro? Efficiency is usually describe in terms ratio. We're not talking about total power draw. That tells us nothing about efficiency.

Definition of efficient: performing or functioning in the best possible manner with the least waste of time and effort; having and using requisite knowledge, skill, and industry; competent; capable:

If a video card gives you 1 fps and consumes 100 watts, does that make it more efficient than card that gives you 50 fps at 110 watts? Well, according to you, that 100 watt is more efficient even though it is practically useless. No, you're wrong. Efficiency, when we're talking about video cards, is best describe as performance/watt.

 

[RussianSensation]

The Titan X is faster, but the 980 is more power efficient, and if you don't believe it, check out the reviews of the cards.

No one is debating that 980 reference is more power efficient than a reference Titan X.

What is more is, you mention of entire power system consumption like having a Titan X reduces it. It does not. The card pulls more power, therefore the system also pulls more power.

I understand all of that. None of that changes the idea that perf/watt should not only be compared on a GPU vs. GPU basis, but also on a total system power basis. Both are valid metrics depending on what you are trying to measure. Perf/watt on a GPU vs. GPU basis is more effective in measuring which architecture/SKU is more efficient. Perf/watt on a total system basis tells the gamer how efficient his/her overall system is in generating a given level of performance. Since we can't play on a Titan X or the 980 without other PC parts, doesn't it make sense that we also look at the overall system efficiency? When we do, a system with the Titan X is not more inefficient against one with a 980.

 

[x3sphere]

The question I keep asking myself is if there is actually room between a Titan X and a 980 for another card. In some cases they are not that far apart.

When the 780 came out, the Titan had been out for quite some time, and there was a huge difference between a 680 and a Titan. That big gap doesn't exist here.

The gap is smaller but not by much - 4% at 1440p.

Here's a site that averaged the perf gain -

http://www.hardwarecanucks.com/foru...992-nvidia-titan-x-performance-review-17.html
http://www.hardwarecanucks.com/foru...orce-gtx-titan-6gb-performance-review-19.html

It also depends what games you're looking at, the TX is way more power than is needed with some games at 1440p and you run into CPU limitations as several games are still not properly threaded these days.

 

[Pariah]

The question I keep asking myself is if there is actually room between a Titan X and a 980 for another card. In some cases they are not that far apart.

When the 780 came out, the Titan had been out for quite some time, and there was a huge difference between a 680 and a Titan. That big gap doesn't exist here.

The 780Ti actually beat the Titan in gaming. The advantage the Titan maintained was in double precision computations and twice the RAM. Titan X no longer has the double precision advantage, so all it will have is twice the RAM. Is 6GB of RAM worth an extra $200? It might be to some people. Not the smart ones, but to those who just want to brag that they have 12GB, even if no one else actually cares.

 

[Elfear]

Let me put it this way, whether you go by custom or reference 980's, the 980 will, under whatever case, pull less power than a Titan X.

The Titan X is faster, but the 980 is more power efficient, and if you don't believe it, check out the reviews of the cards.

A 750Ti will draw less power than a 980. I would just buy that.


Better yet, the IGP on your processor will pull less power than the 750Ti, I'd probably go for that instead. :sneaky:


Back OT - Here's hoping we get SOME kind of real info in the next week or two. The GPU world needs to get exciting again.

 

[96Firebird]

The Titan X is faster by more than 33% against the 980 when you use GPU limited games/resolutions. At 4K the Titan X is nearly 40% faster than the 980.

Per the AT chart you linked, AT specified Crysis 3. The Titan system uses 30% more power but it's 35% faster in Crysis 3. Also, the ONLY 980 card that can exhibit that power usage per AT is the reference blower card. Not everyone buys those which basically means the Titan X rig should end up efficient than just about any GTX980 setup unless one plays a lot of CPU limited titles.

I'm confused, are you saying Metro Last Light @ 1440p is more GPU intensive than Crysis 3 @ 4K? And you can't use a performance summary when talking about power consumption, it depends on the game...

A major flaw in using TechSpot's results is that they used Redux for the performance figures, and Last Light for the power consumption.

TechReport found similar results to Anandtech with respect to system efficiency with the Titan X and 980, again in Crysis 3.

 

[Braznor]

If AMD's 3XX series is really that good, AMD can preempt Nvidia by releasing some benchmark figures.

 

[LTC8K6]

If AMD's 3XX series is really that good, AMD can preempt Nvidia by releasing some benchmark figures.

Seems like they would do that if the benches were really good, because it would likely make people wait to make a purchase.

If the card was impressive for the power and price, most people would wait a month to get it over current cards.

 

[digitaldurandal]

Seems like they would do that if the benches were really good, because it would likely make people wait to make a purchase.

If the card was impressive for the power and price, most people would wait a month to get it over current cards.

Maybe by staying tight lipped they have tricked Nvidia into releasing a card preemptively that cannibalizes their profits on Titan X.

Probably faster than 980 and slower than Titan X IMO.

 

[3DVagabond]

If AMD's 3XX series is really that good, AMD can preempt Nvidia by releasing some benchmark figures.

And give nVidia a target? don't think so.

 

[x3sphere]

Maybe by staying tight lipped they have tricked Nvidia into releasing a card preemptively that cannibalizes their profits on Titan X.

Probably faster than 980 and slower than Titan X IMO.

That's one way of looking at it, but I think if Nvidia releases a new card this month it's going to be because of The Witcher 3. One of the biggest PC titles of the year and they are sponsoring it, makes sense to time a new card launch with it. Lots of people will be upgrading.

 

[LTC8K6]

Does NV have anything with HBM ready to go soon?

 

[digitaldurandal]

Does NV have anything with HBM ready to go soon?

If you believe what we have read so far they are waiting for HBM2.

 

[LTC8K6]

Unless NV has a rabbit up their sleeve, I think AMD is in the driver's seat and they need to take full advantage while it lasts.

 

[raghu78]

Unless NV has a rabbit up their sleeve, I think AMD is in the driver's seat and they need to take full advantage while it lasts.

yeah. If AMD has a R9 390X which beats Titan-X by >10% then AMD will be in a good position to make some good margins. They desperately need it. Nvidia will be forced to adjust the prices to reflect the competitive situation.

 

[Stuka87]

yeah. If AMD has a R9 390X which beats Titan-X by >10% then AMD will be in a good position to make some good margins. They desperately need it. Nvidia will be forced to adjust the prices to reflect the competitive situation.

The price of the Titan won't change. Price of the 980 may drop some, but that typically has not been the case.