NVIDIA GeForce GTX TITAN Video Card Review @ [H]

So perhaps it has been asked already and I missed it, but are monitor overclocking and boost 2.0 hardware/titan exclusive features or will we see them rolled out in drivers/utilities?
 
Careful,instead of recognizing common sense,someone may accuse you of being to poor to afford one. :rolleyes:
Common sense dictates that performance alone cannot be the only potential desirable attribute of a graphics card to all prospective buyers.
 
If your spending $1000 to use this as compute your cheap because if your making money off this card you would have already gotten a Tesla.

I agree with everything else, but freelance and small studio designers to my knowledge, aren't running out buying $3000 cards.
They're probably still running GTX 480/580 cards. Titan on the other hand brings Telsa at a more affordable price point.

Making money off the card doesn't mean you have the need for that "level" of card.
 
Some of you people trying to claim that the 680 was really supposed to be the 660 and the Titan should have been the 680 because of bus width are really, really grasping at straws.
 
Comparing this card to a Ferrari or Lamborghini is a lame argument. The only card I would call a Ferrari (or hell, a corvette really..) is the Ares 2. This is a consumer card, albeit a damnredicualousretardedprice consumer card. Price/performance ALWAYS matters, Always..

Nvidia is in the business of making money, and they obviously think this price point is going to fly or they would not release a card at this price point. I don't think it makes much sense, this is not a limited edition card.. I may be wrong, but I don't think there are that many fanboys with cash to burn on cards that the only thing going for them is that they are brand new.. They already spent their money, last year.. on 690's.. and this, months later, isn't an upgrade. Period.

Almost every review says "This card is amazing! ... BUT ... doesn't make much sense price wise".

Mark my words, The 780 on release will be $800, get the lube..
 
This is what kills me and this gets lost in pages of talking

580 - 384 bit - $500
680 - 256 bit - $500
Titan - 384 bit - $1000

Titan was suppose to be the 680, 680 was suppose to be the 660 ti

Premium building materials and 3gigs of RAM do NOT equal $500

This is a gouge, the smart people agree.

The people that have vested interests do not agree

The people that don't know the facts above have no valid opinion.

If your spending $1000 to use this as compute your cheap because if your making money off this card you would have already gotten a Tesla.

It kills me that people are pussyfooting around the facts, I 100% agree this is just some move by Nvidia to gouge the customer especially when they call it the "TITAN" and not a numbered SKU

Well said, you captured the gist of it.

If you say this they 'whine' that you are too poor. Some of us can easily buy something for $1000 but are not so easily fooled. You can relabel something but if you fall for the marketing you are either unaware, extremely well off, or a fool.

You can't avoid the fact, it's just the GTX 580 successor. They are just taking advantage of the conditioning the market to pay premium pricing and are trying to upsell everything. They call it titan(ic) to pretend it's premium and then they admit it's not a limited run so it's merely the mass produced GTX 580 which they are turning into the pro market (tesla) and are testing what they can get away with.
 
1000$ titan today so where does that put a 780 in a few months? even if its midway between 680-titan speeds, if its 5-600 bucks id be an angry titan owner lol
 
1000$ titan today so where does that put a 780 in a few months? even if its midway between 680-titan speeds, if its 5-600 bucks id be an angry titan owner lol
They will likely continue to diverge the compute and gaming cards further while positioning the Titan as a middle-ground product. I doubt the 780 will compete with the Titan on GPGPU and compute performance.
 
First of all Thanks [H] for the nice review! I cant wait for the tri sli madness!

But the card fails imho, i expect a whole lot more power for that price. They could have gone completely nuts and make a kepler that uses hundreds of watts(300+) and keeps a healthy input output ratio. Thats worth a grant, this aint.

For a lunatic price point, i want something lunatic.
 
Some of you people trying to claim that the 680 was really supposed to be the 660 and the Titan should have been the 680 because of bus width are really, really grasping at straws.

After what Kyle said int he review, I've slowly backed off that argument.
Kyle and Brent has tons more insight and direct connections than I'll ever have, so I'll take their word for it.

But, it sure does seem like flagship parts are now being raised into a new category.
 
Going all the way back to 8800 Ultra days, I decided to cherry pick this quote-

Considering R600 is not looking to be an “8800-killer,” it would seem that NVIDIA has miscalculated the 8800 Ultra product and only ended up marketing against themselves and looking somewhat foolish and arrogant with this overpriced product that seems to be little more than a “PR part.” Sorry NVIDIA, your Ultra is just not worth the money.

And, stick with this today. :p
 
If your spending $1000 to use this as compute your cheap because if your making money off this card you would have already gotten a Tesla.
You have absolutely no idea what you're talking about. The GPGPU applications I use do not need double precision or ECC and work with regular gaming cards. It would be a complete waste to spend 3K on a single K20X. A 1K Titan is much more reasonable. Just because I use the hardware to make money doesn't mean that I have an unlimited budget. You have to go with the best price/performance/feature ratio for your needs. I also have to contend with software licensing costs that run into the thousands which means that I need to pack as much compute power in a single workstation as possible. A triple Titan setup with 6GB for 3K would allow me to work on larger scenes (more geometry and textures) with less noise and power usage than a triple 690 3K. It would also cost 8K less than a triple K20X for 9K. That's 8K that can go towards other more important expenses. That's not called being cheap; that's called being smart with your money.
 
Last edited:
The fact is that nobody outside Nvidia knows why things happened the way they did or what was originally intended.

I think that it's equally plausible - if not moreso - to follow the rumor mill which in late 2011/early 2012 had the GTX 680 slated as being GK110 at ~7970+40% performance, scrapped due to poor yields and replaced by the midrange GK104 chip, and now finally released at a crazy markup.
 
You have absolutely no idea what you're talking about. The GPGPU applications I use do not need double precision or ECC and work with regular gaming cards. It would be a complete waste to spend 3K on a single K20X. A 1K Titan is much more reasonable. Just because I use the hardware to make money doesn't mean that I have an unlimited budget. You have to go with the best price/performance/feature ratio for your needs. I also have to contend with software licensing costs that run into the thousands which means that I need to pack as much compute power in a single workstation as possible. A triple Titan setup with 6GB for 3K would allow me to work on larger scenes (more geometry and textures) with less noise and power usage than a triple 690 3K. It would also cost 8K less than a triple K20X for 9K. That's 8K that can go towards other more important expenses.



Titan is not artificially handicapped like the 580/480's were, All their compute performance is unlocked. See Anand's article for more details, moreover this card is also an entry level K20 compute card. You can actually go into the drivers and switch it to this mode. Only trade off is you loose clocks/boost clocks when you unlock the Tesla K20X like performance but you gain all the DP performance. (Stuff that is missing are just features only good for compute clusters such as the OpenMPI stuff, ECC, etc..)
 
The biggest thing that I hate about the $1000 price, other than the fact that it's clearly a price-gouge? The fact that Radeon 7970's and GTX-680's will not drop in price one iota. Hell - at $500 and at 30% less performance, they now look like STEALS compared to Titan's pricing.

Now for a Tesla on the cheap? Sure - Titan's actually a good buy. But for most gamers who will likely never use this feature? Get out of my face.
 
1000$ titan today so where does that put a 780 in a few months? even if its midway between 680-titan speeds, if its 5-600 bucks id be an angry titan owner lol

Why? I'd think that the people who are willing to pay a premium for top performance will still be happy because they'll still have the fastest thing possible. Maxwell is due for what, 2014, for all we know there might not be a faster card for a long time.
 
Are [H] going to address the fact that they state they used "latest" caps with the13.2 beta drivers?

(12.11 CAP2 are the latest caps and ar NOT meant to be used with the current 13.2 beta drivers. They actually reduce performance in many of the games tested because the overwrite the later optimised profiles built into the drivers.

I honestly can't believe that mistake was made tbh.
 
If you consider this release as getting next gen performace early, but for a price, then it puts things into perspective a little.
 
Titan is not artificially handicapped like the 580/480's were, All their compute performance is unlocked. See Anand's article for more details, moreover this card is also an entry level K20 compute card. You can actually go into the drivers and switch it to this mode. Only trade off is you loose clocks/boost clocks when you unlock the Tesla K20X like performance but you gain all the DP performance.
I was trying to illustrate that not all 'professional' compute users need DP as mt2e suggested that professionals should have bought the K20X months ago if they weren't being cheap.

I'll just re-post this request to Brent and Kyle just in case it got lost in the shuffle:
@Kyle and Brent - A favor to ask: Would it be possible for you to gauge CUDA performance by installing the free Octane v1.1 Demo here:

http://render.otoy.com/downloads.php

The benchmark scene 'Octane_Benchmark.ocs' can be found here:

http://render.otoy.com/downloads/OctaneRender_1_0_DemoSuite.zip

To start the bench, right click on the node called 'RenderTarget PT' and select 'Render'.

Underneath the image, on the left-hand corner should be a figure X.XX Ms/sec. For reference a GTX 680 averages 3.01 Ms/sec

Thanks!
 
Last edited:
Maxwell is shaping up to be a beast looking at Nv GPU map.
If we see another, let's say GM104 (GTX 880) part for $500, and a GM110 part (Titan 2) for $1000, this Titan thing could become the norm.

Hopefully there will be competition vs. the GM110 parts, so we'll see lower prices on the high end parts. Looking forward to see what Maxwell will do though. :)
 
The fact is that nobody outside Nvidia knows why things happened the way they did or what was originally intended.

I'd wager than a decent amount of people on the fab side of things know a bit too (or at least can put 2+2 together).
 
I'd wager than a decent amount of people on the fab side of things know a bit too (or at least can put 2+2 together).

Well yeah but you know what I meant... guys writing reviews at tech sites and people following things and posting on tech sites don't really know what's up. All we can do, and all the reviewers here can do, is go for a best guess approach.
 
Thanks for the great review, [H]. Way to go, nVidia, as you just set the single-GPU bar 10 miles higher than what it was. I can hardly wait to see some performance goodness with the 700 series when it's time for me to upgrade.


This is what kills me and this gets lost in pages of talking

580 - 384 bit - $500
680 - 256 bit - $500
Titan - 384 bit - $1000

Titan was suppose to be the 680, 680 was suppose to be the 660 ti

If nVidia had released the Titan as their flagship 680 on the 600 series launch day, then what, pray tell, was AMD to answer that with? That would place their flagship single-GPU 7970 as a competitor to what you refer to as "the real" 660ti. Which, in turn, would place AMD in a very bad position. AMD fans and/or supporters of ongoing advancement and healthy market competition should pat nVidia on the back for NOT doing that. Aside from that, it's hard to deny the performance that the Kepler architecture can provide with 2GB of VRAM and a 256 bit bus in 680 fashion is damned impressive...and now we can see what that architecture does with triple the memory and 50% more bus bandwidth. Great things to come for the next generation of GPU's, for certain!
 
theirs many of us who never have issues with sli.

I wish you were around yesterday and said THIS

rinaldo00 Limp Gawd, 8.0 Years Status:

Quote:
Originally Posted by Zarathustra[H]
That is a good question.

All the comparisons to multi-GPU solutions are just silly. Granted Nvidia's SLI is better than AMD's CFX, but they still come off as being public beta tests, unstable, problems, and never as fluid as a good single GPU solution.

I have run SLI for over 5 years without any problems, what am I doing wrong?



then heard this

cannondale06 [H]ardForum Junkie, 5.2 Years Status:

Quote:
Originally Posted by Skakruk
Well the first thing you're 'doing wrong' is being immune to microstutter. You're very lucky.

and claiming no problems is simply not true. there is zero chance that he can run sli for 5 years and not have issues at some point. its like the people that say they have never had a driver issue ever. it is just nonsense and selective memory.
 
Hopefully there will be competition vs. the GM110 parts, so we'll see lower prices on the high end parts. Looking forward to see what Maxwell will do though. :)

AMD doesn't build big die chips. I doubt we'll see anything. AMD hopes to kill it with crossfire.
 
I honestly can't believe that mistake was made tbh.

Taken directly from the review.

"For the AMD Radeon HD 7970 GHz Edition we are using the newly released Catalyst 13.2 Beta 6 and latest CAP. "

http://www.hardocp.com/article/2013/02/21/nvidia_geforce_gtx_titan_video_card_review/2#.USaXTldi5I0

Now look at the graphic shown on the test setup page of the review. It clearly states that Catalyst 13.2 Beta 6 and Cap 12.11 CAP 2 are used for the AMD HD 7970 GHz card.

So yes, until [H] can confirm otherwise they did make that mistake. Using these redundant CAPs will overwrite newer more optimised profiles for Far Cry 3, Crysis 3 and potentially other games tested. So some of the AMD results in this review are suspect. If we stick to the scientific method this mistake renders the entire AMD test results invalid by virtue of the fact it is configured incorrectly. I would be saying the same thing if [H] used drivers from 3 months ago for the Nvidia cards.
 
Last edited:
People need to realize that scale matters also. They aren't producing a ton of these cards, and that affects pricing. If this was their next line of cards to mass produce, it would be gouging. For a halo card, produced in limited numbers, with no other competition, the price makes sense.
 
AMD doesn't build big die chips. I doubt we'll see anything. AMD hopes to kill it with crossfire.

Killing it with crossfire is futile. Considering that Titan is the worst performance/dollar GPU ever, even Nvidia's own SLI options outperform it for less $. Titan is more suited for those that want the best performing single GPU card, or highest performing multi-GPU option regardless of price/performance.

The Titan is not the fastest card out there, but the fastest single GPU card and thats where AMD should compete with it. The Titan already have multi-gpu alternatives as competition, even from Nvidia themselves. Lets hope they build bigger die chips to give the Titan some real competition. :)

Titan is in a segment of its own at the moment.
 
I wish you were around yesterday and said THIS

rinaldo00 Limp Gawd, 8.0 Years Status:

Quote:
Originally Posted by Zarathustra[H]
That is a good question.

All the comparisons to multi-GPU solutions are just silly. Granted Nvidia's SLI is better than AMD's CFX, but they still come off as being public beta tests, unstable, problems, and never as fluid as a good single GPU solution.

I have run SLI for over 5 years without any problems, what am I doing wrong?



then heard this

cannondale06 [H]ardForum Junkie, 5.2 Years Status:

Quote:
Originally Posted by Skakruk
Well the first thing you're 'doing wrong' is being immune to microstutter. You're very lucky.

and claiming no problems is simply not true. there is zero chance that he can run sli for 5 years and not have issues at some point. its like the people that say they have never had a driver issue ever. it is just nonsense and selective memory.

ive had many differnet versions of SLI starting with the my BFG 6800gtx series till my 680 classifeies. and have never experienced micro stuttering on any of my personal rigs.

Ive seen the video's and can see and understand what others have had issues with it. But i personally have never seeen it on any of my builds ive done personally or for friends/family
 
I knew HardOCP would give this a gold award. They always seem to have a boner for expensive and impractical hardware.
 
I knew HardOCP would give this a gold award. They always seem to have a boner for expensive and impractical hardware.

Thats kind of the point of being a hardware enthusiast site...and thats why we come here right?
 
There shouldn't be $1000 cards in the first place :rolleyes:

That's like saying you should be allowed to post because we have differing opinions. :rolleyes: I would suggest that NVIDIA will sell every Titan they intend to build. That alone renders your argument invalid.

Who else thinks this Titan release makes way for a 780 on release with a price tag of $800?

I mean.. If they can get $1000+ for this card whats stopping them?

If we see prices like that I would suggest that AMD is the "problem," not a halo card built in small quantities.

Everyone praises the Titan because it is great enough that we can finally do away with SLI problems.....then in the next sentence they say "I WANT 2 TITANS IN SLI"

Is that sane?

Yeah, that seems sort of odd the way it is laid out. This is basically referring to the real market for this card, and that is boutique system integrators. SIs would rather stay away from an enthusiast technology like SLI just because of support issues. SLI for enthusiasts is a no-brainer, but not for everyone else.

Question, what happens to the review cards?

Those are destroyed immediately upon completion of the review.

Titan was suppose to be the 680, 680 was suppose to be the 660 ti

It kills me that people are pussyfooting around the facts, I 100% agree this is just some move by Nvidia to gouge the customer especially when they call it the "TITAN" and not a numbered SKU

Facts....yeah. Just because something gets repeated thousands of times does not make it a fact.

So perhaps it has been asked already and I missed it, but are monitor overclocking and boost 2.0 hardware/titan exclusive features or will we see them rolled out in drivers/utilities?

Yes, these are technologies that you WILL see again out of NVIDIA.

Well yeah but you know what I meant... guys writing reviews at tech sites and people following things and posting on tech sites don't really know what's up. All we can do, and all the reviewers here can do, is go for a best guess approach.

I do not very often share "opinions" of mine in review content unless I have an extremely good understanding of the subject matter addressed. I have been doing this for 15 years and generally know who to talk in the industry to either confirm or disprove my thoughts. If I thought my shared opinions were not valid, I would keep them to myself in the context of a review.
 
I swear I've seen [H] reviews trash products for being too expensive, though. I just don't see the value in this, outside of MAYBE SFF...not really even then TBH.
 
I swear I've seen [H] reviews trash products for being too expensive, though. I just don't see the value in this, outside of MAYBE SFF...not really even then TBH.

Yes we have, and yes we will continue to do so. Halo product for a niche of a niche market to be produced in very small quantities. As just mentioned above, we very much talked about value more than once.
 
Back
Top