***Official Reviews Thread*** Nvidia Geforce GTX Titan - Launched Feb. 21, 2013

[notty22]

The 7970 Matrix is sexy as hell!

You should get out more. J/K

 

[2is]

I believe I addressed everything you said, though I could be wrong. Can you point out exactly what I missed?

There is no proof you can provide that would show a reference board losing to a aftermarket board given the same chip and same cooling.

But by all means, believe the hype.

A better board on Titan would do jack crap. The only thing that would really affect Titan is voltage table and power limit increases.

You didn't, and I don't feel like quiting myself, Everything is there for you to read over again if you wish. You provided guess work skewed to support your own opinions on the subject, in pretty much all of your posts, weather they were directed towards me or someone else.

As far as proof, I have as much as you do, which is to say none at all. But what I do have that you don't (unfortunately for me as it is) is a 680 that appears to be clearly limited by its power delivery. Dud or not.

 

[BallaTheFeared]

You didn't, and I don't feel like quiting myself, Everything is there for you to read over again if you wish. You provided guess work skewed to support your own opinions on the subject, in pretty much all of your posts, weather they were directed towards me or someone else.

As far as proof, I have as much as you do, which is to say none at all. But what I do have that you don't (unfortunately for me as it is) is a 680 that appears to be clearly limited by its power delivery. Dud or not.

Well you've said a lot, how can I know what exactly it is you feel makes your case that I haven't addressed?

I didn't provide any guess work, we already know what chips Titan uses and we can view their specifications online... In fact someone in this thread has already done that and figured out roughly how much power they can actually handle while staying within specifications :hmm:

I have tons of proof, someone above your post added some of their own, and years of buying and overclocking video cards of proof. There has never been anything that lended proof of fact that a better power delivery to a card that already has enough improves anything.

I've addressed the major fact of what the current silicon is actually capable of on air/water, which is roughly between 1200-1300, anything after that would be the domain of much better cooling to allow far less leakage and enable far higher clock speeds as well as warrant such a board as the lightning and matrix which is exactly what they were design for. ()

 

[2is]

Where is your data points that your 'beefier' board components = higher o/c's.
Here is data from techpowerup on o/c's including the stock card at launch.
They are all close to each other.
Including a lightning.

http://www.techpowerup.com/reviewdb/Graphics-Cards/NVIDIA/GTX-680/

My "data points" was my own particular card, which I was very clear about in multiple posts. But I'll use yours too showing the reference PCB right at the bottom of the list. I also didn't say beefier components are alwasy going to = higher OC's, I said they're always welcomed.

 

[BallaTheFeared]

My "data points" was my own particular card, which I was very clear about in multiple posts. But I'll use yours too showing the reference PCB right at the bottom of the list. I also didn't say beefier components are alwasy going to = higher OC's, I said they're always welcomed.

It's actually higher than the AMP! on core, and better on the memory than many of their non reference boards including the lightning and kfa2... So what now?

And yet still none of them are really getting around the silicon hard limit of around 1200-1300...

 

[2is]

Well you've said a lot, how can I know what exactly it is you feel makes your case that I haven't addressed?

I didn't provide any guess work, we already know what chips Titan uses and we can view their specifications online... In fact someone in this thread has already done that and figured out roughly how much power they can actually handle while staying within specifications :hmm:

I have tons of proof, someone above your post added some of their own, and years of buying and overclocking video cards of proof. There has never been anything that lended proof of fact that a better power delivery to a card that already has enough improves anything.

I've addressed the major fact of what the current silicon is actually capable of on air/water, which is roughly between 1200-1300, anything after that would be the domain of much better cooling to allow far less leakage and enable far higher clock speeds as well as warrant such a board as the lightning and matrix which is exactly what they were design for. ()

We will just have to agree to disagree... Recycling the same argument with a couple different words thrown here and there certainly isn't going to convince me that something is "proof" when to me it's far from it. Likewise, nothing I say is going to change someones mind who thinks they've already proven their argument.

 

[BallaTheFeared]

Titan looks to be coming in around 1150Mhz on the OC for most people, meaning even with voltage tweaking and super duper PCBs there is probably only another 50-100MHz left in it's tank before we start becoming limited by the process node.

 

[2is]

It's actually higher than the AMP! on core, and better on the memory than many of their non reference boards including the lightning and kfa2... So what now?

And yet still none of them are really getting around the silicon hard limit of around 1200-1300...

What now? I'm glad you asked....

I see the non-reference PCB some 70MHz+ over the reference ones.

AMP uses a reference PCB with a BETTER cooler.

So again, what now? It certainly seems according to this table, better cooling isn't doing anything for overclocking but the better PCB's are.

Hope you have ice for that champagne

 

[BallaTheFeared]

Silicon quality and cooling was probably 100% of the difference, not board PCB :\

 

[2is]

Well the AMP certainly put a dent in your cooling guesswork.

 

[BallaTheFeared]

From one sample?

Please, I would never base my opinion on such a limited data point.

 

[2is]

From one sample?

Please, I would never base my opinion on such a limited data point.

You were VERY quick to ring the bells of victory when you thought this ONE sample supported what you were saying.

Funny part is, I'm not even trying to say a better PCB is always going to help, only that it can't possibly hurt and that my particular card appears to be limited by it, but you were to obtuse to even consider that as a possibility, and well... Here we are.

I will fully acknowledge that a better PCB alone may not always translate to better overclocking, but you refuse to even consider that as a possibility. My flexibility and your rigidness on the subject makes it very difficult for you to "prove" me wrong.

 

[BallaTheFeared]

You were VERY quick to ring the bells of victory when you thought this ONE sample supported what you were saying.

Funny part is, I'm not even trying to say a better PCB is always going to help, only that it can't possibly hurt and that my particular card appears to be limited by it, but you were to obtuse to even consider that as a possibility, and well... Here we are.

I will fully acknowledge that a better PCB alone may not always translate to better overclocking, but you refuse to even consider that as a possibility. My flexibility and your rigidness on the subject makes it very difficult for you to "prove" me wrong.

That one data point supports what I said, the difference between each card reference to non is trivial and consists mostly of silicon and cooling differences without question.

I never said it would hurt, only that in most cases on air and even water it will not matter. You were trying to play off the fact that your 680 might be a dud as a problem of the PCB, which is where you're wrong. There are many samples of reference cards which have done much better than yours. This was the point I was trying to convey to you, when you kept attempting to play off your one single sample as the definitive example of how a reference PCB would limit a card.

I already said a long time ago that you might get a few more MHz out of a better PCB design, but I've never seen anything to back that up. I was willing to offer that a page back, but you did not want to accept a little bit you wanted to difference between a capable reference power delivery system vs a Matrix or Lightning to be huge, and it's simply not.

I'm trying, perhaps in vain, to help you better understand silicon and power systems on cards. I'm not trying to "win" anything, it doesn't matter to me at all if you continue on believing something incorrectly, however I am trying to help you out here you're just unwilling to accept that there are much bigger factors at play with GPU's than power delivery.


You have not shown that Titan has insufficient power delivery to sustain silicon capable clock speeds this generation. The fact of the matter is that it does as far as voltage and cooling currently allows. Titan is hitting around 1150MHz from what I've seen around the web with end users. 28nm does not offer much more MHz headroom, so Nvidia has delivered a decent product with good perf/w and decent overclocking capabilities while limiting the more zealous of users who are willing to accept 100w more draw for 5% more performance. Given it's limited TDP of 265w, assuming you could increase that to say 300w, there is still nothing to show that Titan in reference form could not handle such power draw, and again nothing placed forth by you has shown that Titan with the same limits would overclock higher with a beefier power delivery system.

It's like RAM... Having too much doesn't hurt, but doesn't really help anything either..., having too little causes major problems (stutter, or in this case, blown vrms).

 

[2is]

There are a few facts about that card

1) It's a reference PCB
2) It has better cooling (one of two main factors you say contributes to overclocking)
3) It is not a good overclocker

It having inferior silicon is something you cannot prove. The facts (what we KNOW) about that card go completely against what you're saying, and the unknown is well, an unknown, and you're just running with it.

Guesswork.

Yeah, it's one sample, and I wasn't even going to use it against you for that reason until you attempted to use it in your favor, ignorant to what the card actually is.

And I fully understand...

http://forums.anandtech.com/showpost.php?p=34683789&postcount=962

As far as Titan, I haven't said ANYTHING about titan, but the same principles apply there anyway.

 

[3DVagabond]

The idea that somehow a beefier PCB makes a card faster, it doesn't. It might allow higher overclocks, but that still based on silicon, doesn't matter if you have the best PCB in the world if you have a crap chip it's just going to woof down the power and do almost nothing with it.

Titan is capable of 265w based on Nvidia's limits, most likey guys picking it up will expand that quickly if you're right than we'll see Titans blowing vrms, if you're wrong than we'll see Titan still limited by voltage even with additional TDP headroom. Most likely the board is more than capable of delivering over 300w of power based on the specs of the chips on the board.

"Quality" mosfets and chokes is mostly pure marketing bs.

What am I supposed to be right or wrong about? I've not claimed anything. LOL

What are you on about? I'm just showing what PCB's that are designed for O/C'ing are built like. People who buy top of the line GPU's want to have state of the art PCB designs and components too. Some people are excited that there might be Titan's that are designed, and maybe even allowed to be over volted, so they can be maxed out. Relax, nobody's making fun of Titan. You don't have to come running to it's defense.

 

[BallaTheFeared]

There are a few facts about that card

1) It's a reference PCB
2) It has better cooling (one of two main factors you say contributes to overclocking)
3) It is not a good overclocker

It having inferior silicon is something you cannot prove. The facts (what we KNOW) about that card go completely against what you're saying, and the unknown is well, and unknown, and you're just running with it.

Guesswork.

Yeah, it's one sample, and I wasn't even going to use it against you for that reason until you attempted to use it in your favor, ignorant to what the card actually is.

:thumbsup:

Well alright then buddy.

What am I supposed to be right or wrong about? I've not claimed anything. LOL

What are you on about? I'm just showing what PCB's that are designed for O/C'ing are built like. People who buy top of the line GPU's want to have state of the art PCB designs and components too. Some people are excited that there might be Titan's that are designed, and maybe even allowed to be over volted, so they can be maxed out. Relax, nobody's making fun of Titan. You don't have to come running to it's defense.

Then probably nothing, other than your assertion that Titan could be "improved" with better power delivery.

Those PCB's are design for LN2... Big difference. Reference is almost always fine for typical overclocking would be my point. People who want more for LN2 could just do what Kingpin did.

I'm not defending Titan so much as I am defending reference board overclocking...

 

[wand3r3r]

So now it's time to defend crappy cut down and very minimalistic reference boards? Wow.

For a $1k board there is no excuse to be cutting corners and yet some marketers sit in an enthusiast forum and try justify weak pcb's as if they shouldn't be better.

For such a "premium" board the only thing premium is the cooler and ultra premium raping price.

 

[Lonbjerg]

So now it's time to defend crappy cut down and very minimalistic reference boards? Wow.

For a $1k board there is no excuse to be cutting corners and yet some marketers sit in an enthusiast forum and try justify weak pcb's as if they shouldn't be better.

For such a "premium" board the only thing premium is the cooler and ultra premium raping price.

Try locating the target group...it not you...but the demographics are often talked about in reviews...try looking it up.

 

[3DVagabond]

:thumbsup:

Well alright then buddy.



Then probably nothing, other than your assertion that Titan could be "improved" with better power delivery.

Those PCB's are design for LN2... Big difference. Reference is almost always fine for typical overclocking would be my point. People who want more for LN2 could just do what Kingpin did.

I'm not defending Titan so much as I am defending reference board overclocking...

Let me put it another way. First we put up with your position of buy used 2 generation old cards, put them under expensive custom water cooling and O/C the piss out of them is the ultimate perf/$ setup. Then you've gone on for months and months about midrange cards masquerading as high-end selling for high-end prices. Now you are defending PCB's that are merely good enough for these so called midrange cards being used on a true high-end $1000 card.

I'm gonna put you on ignore (for the second time). Just because you are impossible to have a discussion with and I don't need your agro.

 

[wand3r3r]

Try locating the target group...it not you...but the demographics are often talked about in reviews...try looking it up.

“You’re going to see some people who just say, I want maximum frame rate,” says Nvidia’s Tom Petersen. “And if you want maximum frame rate, GTX 690 is you. If you want the best experience, if you want the best acoustics, then Titan is for you.”

According to that and everything about the card being it's a normal production card (not a limited edition even) speaks volumes. The only thing premium is the ?, well you can guess. I'll leave that as a user exercise.

 

[Lonbjerg]

According to that and everything about the card being it's a normal production card (not a limited edition even) speaks volumes. The only thing premium is the ?, well you can guess. I'll leave that as a user exercise.

Could you cherrypick more?
http://www.anandtech.com/show/6774/nvidias-geforce-gtx-titan-part-2-titans-performance-unveiled

Meanwhile with the launch of Titan NVIDIA has repositioned their traditional video card lineup to change who the ultimate video card will be chasing. With a price of $999 Titan is decidedly out of the price/performance race; Titan will be a luxury product, geared towards a mix of low-end compute customers and ultra-enthusiasts who can justify buying a luxury product to get their hands on a GK110 video card. So in many ways this is a different kind of launch than any other high performance consumer card that has come before it.

Says to me NVIDIA has a Winner and the red team keep whining about everything else...than performance ^^

 

[skipsneeky2]

Kinda lost for words with how much drama this card has brought,maybe i am what is called a newbie as i have only pc gamed since 2006 but wow i never seen so much negativity about a gpu .

Gtx280 does sorta come close,i know its a gpu i regret purchasing and i got it at $649 but frys was nice enough to refund me $150 cash when the price drop happened and i still regret it after that.:awe:

 

[blackened23]

Some interesting developments with GPU Boost 2.0 - apparently it throttles more than GK104 throttle, has anyone else experienced this? From callsignvega (He's an enthusiast on the level of AdamK) over at hardocp/ocn:

Well, I was full expecting my benchmark numbers to be complete tonight but I have run across some interesting aspects with Titan. I think I may have an inkling why some of these performance numbers and charts are low and all over the place. Apparently, even with all "limiters" set to max: (106% power target, 94 deg C, +38mv for an even 1.2v), the cards are throttling pretty bad for some reason.

The fan's max out at 85% even though you can set up to 100%. (they are super quiet btw which is amazing), and with the GPU's maxed out at 85% fan I am staying below 60 C. (I also removed the rear brackets which tend to slow the air down and create noise). Power % virtually never goes over 100% so it's not hitting the 106% limiter, they aren't even on the same planet as the temperature limiter, so not quite sure what is going on here and why they are throttling down.

As an example, my cards in SLI are limited to about a 1150 MHz core. Running some benchmarks like Valley 1.0, the GPU usage will be maxed out yet the core frequencies drop down to the ~1050 MHz range with a power % draw of ~90's well under the 106% limit and temps under 60 C. It makes no sense (not a CPU limit as the GPU utilization stays at 99%). It's like the cards aren't properly reacting to the demand. Not only that, the voltage drops down from 1.2v to as low as 1.1v under full load when the clocks drop down and then the voltages bounce around wildly. The only thing I can think of us that the launch drivers are just crap or that EVGA Precision 4.0 isn't reading stuff correctly.

I'll have to play around with the cards more but I'm not too impressed so far. All of these cards will max out around the same clocks and there is no reason to water cool them unless you need to lessen the already fairly low noise.

Yes, the voltage and clocks stay set perfect in 2D desktop but revert back to crazy all-over the place in-games. Makes zero sense.

Sounds like over-voltage is once again, gimped on the Titan as it was on the GTX 680.

 

[BallaTheFeared]

Let me put it another way. First we put up with your position of buy used 2 generation old cards, put them under expensive custom water cooling and O/C the piss out of them is the ultimate perf/$ setup. Then you've gone on for months and months about midrange cards masquerading as high-end selling for high-end prices. Now you are defending PCB's that are merely good enough for these so called midrange cards being used on a true high-end $1000 card.

I'm gonna put you on ignore (for the second time). Just because you are impossible to have a discussion with and I don't need your agro.

:hmm:

"Put up with", lol too funny.

I never once told anyone to buy two 470s for $120ea, two blocks for $40ea, get a rasa kit for $120, and blow the doors off a $600 7970. Not once, you're welcome to try and find it but you won't. Your information is wrong, which is not surprising.

More like a year, I went on for a year about mid-range performance at the $500+ price point. First I said it with the 7970 release, then I said the same thing about the 680 (GK104). I said things looked bad for consumers, I said GK110 (high end) would come to desktop and crush both cards. :awe:

And here you are yet again talking about things you don't know about. GK110's PCB is more robust than the GTX 480s, it is however limited to less power consumption than the GTX 480. Not only was the reference GTX 480s design more than adequate for stock and FURMARK, but it was more than capable once overclocked, or put on blocks and taken well beyond 300w draw. Titan isn't mid-range, it's the high end card I've been expecting since the 7970 came out.

The problem here is you have no clue what you're talking about, and you can't even remember things that happened only a year ago. You call Titan mid-range as if I did, yet I never did and you attempt to smear my name with misinformation, either through lack of mental ability or plain old ill will.

 

[wand3r3r]

Thankfully you guys have cleared it up now so that we can't talk about and consider anything else other than the performance of titan. (puts on nvidiot blinders, go nv)