[WCCF/Videocardz] Nvidia is prepping a new GK110 GPU - GTX Titan Ultra coming soon

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
I have no problems dropping $600-1k for a gaming card, makes no difference to me. As it should not make any difference to someone who can only afford at $300 card.

You must be a huge baller than if to you it makes no difference if a GPU is priced at $600 or $1,000.

Gigabyte Windforce 3x 770 4GB = $450
MSI TwinFrozr 7970 1.05Ghz version with 3 games = $269

67% more expensive, can't really outperform it.
http://www.computerbase.de/artikel/grafikkarten/2013/nvidia-geforce-gtx-770-im-test/4/

I think I would rather go for a nice dinner than pay $130-150 more for 3% more performance the 770 offers over 7970Ghz. Problem is even that statement isn't correct since 770 oc cannot beat 7970 oc in games. This is exactly why NV prices their cards so high because their loyal customs keep paying $100s of dollars more for little to no performance gain. When NV finally drops the prices on 770, I wonder what 770 owners will say.

7970Ghz vs. 770 is very similar to $299 4870 vs. $399 GTX260.

----

If NV releases a 2880 cuda core Atlas card, it implies there are a lot of fully yielding GK110s. In that case, it would make some sense to sell some of them as consumer cards as it's doubtful NV will be able to sell every single one of those 2880 chips to professionals while they could in theory raise the price of 2880 Titan to $1,200 and make even more $. Releasing a DP-crippled Titan as 785 at $649 would be a lot nicer for us gamers, putting 2880 Titan II at $1000.
 
Last edited:

Crap Daddy

Senior member
May 6, 2011
610
0
0
If NV releases a 2880 cuda core Atlas card, it implies there are a lot of fully yielding GK110s. In that case, it would make some sense to sell some of them as consumer cards as it's doubtful NV will be able to sell every single one of those 2880 chips to professionals while they could in theory raise the price of 2880 Titan to $1,200 and make even more $. Releasing a DP-crippled Titan as 785 at $649 would be a lot nicer for us gamers, putting 2880 Titan II at $1000.

I don't think for the moment Nvidia needs an Atlas for consumer market. For all we could gather from bits and leaks, the 290X will trade blows with Titan while being priced to 780 levels or slightly higher. NV needs a GK110 based card between the 780 and the 770 while dropping the price of the 770 close to 280X. I don't think Titan is selling now. It's 8 months old and whoever wanted to get already did, they could lower the price and try to push more but I don't think it's important for NV. They have to make the 780 competitive to the new shape of the market after the 290X launch and offer a card for to compete with the 290 (about which, oddly enough, we don't know a thing)
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
It seems the Titan Ultra rumours comes from the Colorful iGame Titan Ultra Edition with watercooling. And mixed with the K40. I think everyone can put the last math together.

Just a shame it goes back to june.

Colorful-iGame-GTX-TITAN-Ultra-1200x691.jpg
 
Last edited:

Slomo4shO

Senior member
Nov 17, 2008
586
0
71
Great name eh?, I thought Titan was cool, but Atlas...NV sure have a cool PR dept

Considering that Atlas was a Titan... Olympian would have been a better name as the Olympians were the the successors of the Titans.
 

SolMiester

Diamond Member
Dec 19, 2004
5,331
17
76
I thought Titans were monsters, like the Kracken...whereas Atlas was a demi-God...LOL
 

blackened23

Diamond Member
Jul 26, 2011
8,548
2
0
I don't think for the moment Nvidia needs an Atlas for consumer market. For all we could gather from bits and leaks, the 290X will trade blows with Titan while being priced to 780 levels or slightly higher. NV needs a GK110 based card between the 780 and the 770 while dropping the price of the 770 close to 280X. I don't think Titan is selling now. It's 8 months old and whoever wanted to get already did, they could lower the price and try to push more but I don't think it's important for NV. They have to make the 780 competitive to the new shape of the market after the 290X launch and offer a card for to compete with the 290 (about which, oddly enough, we don't know a thing)

This mirrors my thoughts pretty much exactly. I don't think the GK180 will be bound for the consumer market because NV doesn't need it - if the 290X costs 600$, I don't believe another 1000$+ GPU halo product would sway decisions at all, except for those who want an excessive amount of VRAM. In fact, i'm nearly 100% sure that another 1000$ halo product will make little difference, because the GTX 780 OC already made the Titan irrelevant for most buyers. If anything, nvidia should place a product between the 770 and 780 and do some slight price cuts among the 700 series and that's it. If the 290X is 600$ and offers greater than 780 performance while trading blows with the Titan (let's say %35-65 with the Titan, and the Titan wins more - while the 290X beats the GTX 780), then a 50$ pricecut for the 780 would be fine. As well, the 780OC performs exceptionally well so that's another factor as to why another Titan II makes little sense, unless someone wants more VRAM.

It seems that NV is indeed doing this. There are rumors of 770 and 780 pricecuts, as well as a 770ti product to fit the bill between the 400$ 770 pricetag and the 650$ 780 pricetag.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
Because they'll get a lot more for the same cards badged as Quadro's.

What you're saying here applies if there aren't enough full function dies to go around. All GK110's come from the same wafer. It's down to yields. Just as an example, what if all GK110's are able to be certified fully functional then they actually have to artificially cripple them to make Titans and 780's. That appears to have been the case with Cayman when 6950's could simply have their extra shaders unlocked and be fully functional 6970's. Then they are actually making more money leaving some (as much as the market can hold) fully functional and selling them for more money. Contrary to what some believe, groups who use Tesla cards for scientific research, for example, aren't buying Titans instead. They need the features and support they only get from Tesla cards, and are willing to pay for it. Also, they don't pay retail. ;)

Anyway, this far along in the GK110's life yields are probably fine to support a fully functioning GK110 Geforce card. Although, it only shows a K40 in the slides. So, where this Ultra Titan is coming from, I don't know. Wishful thinking, maybe? Besides, how much faster is it likely to be? 3%?

Marketing wise it would probably be more effective if the actual performance was left unknown and people could use the, "nVidia could make a faster card if they wanted to. They just make too much money selling Quadro/Tesla cards to be bothered", spiel. Like the reason they didn't want the console business.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
There you have it guys. The value of a better brand.
Look how enormous the effect is on the profit for a company.
In NV and AMD case, the difference accounts for aprox 25% resale value. Thats more or less equal to 25% revenue difference. In a market with perhaps 10% profit. That difference can be between losing and having a very, very profitable company.
Ofcourse there is cost to maintain the brand - and eg. good oem relations and good shopfloor placement and sales training, but i hardly guess its more than 5-10% of revenue. Still at least 15% advantage.
It clearly shows that the absolutely best effect for selling GFX with a profit is better marketing and branding and not more engineering hours.

You need to save that for the shareholders. As a consumer the company making more money from us isn't a desirable thing. If that's what we wanted we could all buy 7970's and send AMD a check for $100. Then they'd get all of it. They'd make heaps more money. The only additional cost would be having to hire people to count it. ;)