GTX 980 Ti Potentially Launching Mid-Late May

[ADZ_123_!"£]

I'd say GTA V proves otherwise... two Titan Xs in my rig get hammered for 8GB+ of Vram with everything cranked at 3440x1440

GTA V will have DX 12 when it launches won't it.

[spartaman64]

GTA V will have DX 12 when it launches won't it.

gta v already launched and no dx12

[Prastupok]

gta v already launched and no dx12

 

Keep in mind DX12 isn't released yet...

 

-snip-

 

Scenario #2: How does it shaft them? It should perform worse than a Titan X. No one knows how the 3xx cards will do.

 

Scenario #3: It does exist, beats the 3XX but not the Titan X, no one gets shafted.

 

And the most likely scenario: #4) It doesn't exist!

[spartaman64]

I'd say GTA V proves otherwise... two Titan Xs in my rig get hammered for 8GB+ of Vram with everything cranked at 3440x1440

2 980 still beats the titan x in gta v so im not sure if it really uses 8GB

[theelitegamer]

so what is the point of the titan x?

This happened with the 780 ti and the titan.

[ADZ_123_!"£]

No in GTA V it already doesn't show combined vram amounts so when DX12 launches won't it be supported in GTA V.

[d3sl91]

INB4 "What About Tita....

 

 

 

I give up. 

[ChineseChef]

With a 384-bit (6x64-bit) memory controller, you need to put RAM on in multiples of 6. In this case I'm guessing 12x 4Gbit chips or perhaps they'll go back to 24x 2Gbit chips again.

 

Ahh, so it is a technical/mathematical scaling reason.  Something I should have known but never really connected bus size with vram amount before.  Thanks

[DaRk0]

Why do people need more vram because quite a lot of gamers use sli cards and if DX12 does double up vram then any one with 970's 980's or 290's in sli or crossfire will have 8gb which would be enough for games so it's not worth buying it just for more vram.

 

Well, most people don't need that much VRAM. I have a GTX780 3GB and I'ts not bottlenecking me at 1080p resolution. It's worth buying it if you're working with 3D software.

 

from my experience professional use that needs a lot of vram also needs a lot of double precision but i guess there are applications that is not true for. but still 6GB of vram is not worth 300 dollars. also to me what a nvidia titan card is a gaming card that also has a lot of double precision for developers and for that reason the titan x shouldnt have the titan branding

 

You don't need double precision to saturate 12GB of VRAM while working in viewport.

[spartaman64]

This happened with the 780 ti and the titan.

The titan has double precision so it still made sense

[Glenwing]

I'm very new here I only got built my first gaming pc this Christmas.

 

Welcome then

 

FYI you can compare GPUs directly by frequency or by any other spec, but ONLY if the GPU design is the same. You can't compare a GTX 980M (1536 cores) to a GTX 880M (1536 cores) by core frequency alone because the cores aren't actually the same. The GTX 980M uses Maxwell-design cores which are about 1.35x the performance of a Kepler core which the GTX 880M uses. However, if you have two GPUs which both use Maxwell cores then you can compare frequencies directly. Since the core design is the same, the performance-per-cycle is the same, so the only difference in performance comes from the core frequency, or in other words how frequently they cycle.

[spartaman64]

I think nvidia made a bad decision dropping the double precision because there is like no way to make the 980 ti not compete with the titan x

[Briggsy]

WCCF = Wild Computer Centric Fluff

[Glenwing]

Ahh, so it is a technical/mathematical scaling reason.  Something I should have known but never really connected bus size with vram amount before.  Thanks

 

Yeah, while it's not a true restriction and you don't have to follow that rule (see GTX 660 Ti, with 192-bit bus but came in 2GB and 3GB versions) things get preeeeetty weird when you have unbalanced memory configurations which is why you almost never see it. In terms of engineering it's just a lot simpler to work with symmetric configurations across all the memory controllers.

[Gibbons]

There's going to be some salty people out there if this comes out $200+ cheaper and higher clocked so soon after the Titan X. The 6GB extra vram isn't useful for the gamers the TitanX was marketed to.

[BiG StroOnZ]

Err no, I was being sarcastic. The 780 Ti was cheaper and faster than the Titan Black.

780 Ti was faster than the original Titan not the Titan Black. Titan Black and 780 Ti peform identical since they both have the same amount of Cuda Cores

[ADZ_123_!"£]

There's going to be some salty people out there if this comes out $200+ cheaper and higher clocked so soon after the Titan X. The 6GB extra vram isn't useful for the gamers the TitanX was marketed to.

The thing is the titan x is a completely stupid product as it's not aimed at professionals yet it has the 12gb of vram which is completely useless to gamers.

[sgloux3470]

I'd say GTA V proves otherwise... two Titan Xs in my rig get hammered for 8GB+ of Vram with everything cranked at 3440x1440

 

How?  What are you doing running 8x MSAA on top of 4K?

 

It uses 3.5GB ish for me at 4K with Very High Textures and Shadows.  (I do have some other settings on High though, but they shouldn't push me over 4GB.)

 

Chances are the game is just being smart enough to cache as much as possible in the RAM.  

 

 

There's going to be some salty people out there if this comes out $200+ cheaper and higher clocked so soon after the Titan X. The 6GB extra vram isn't useful for the gamers the TitanX was marketed to.

 

I think they'd be really stupid to release a 980 Ti.  It's going to cannibalize sales for both the 980 and TitanX.  The 780Ti was a unique case because the 780 was always a cut back design while the 980 was fully unlocked.  

 

I think it's more likely we'll see a TitanXY if Nvidia needs a flagship to be considered the highest performance GPU.  

 

And then we can watch it get rekt by a 395x2 as history repeats itself again.

[ChineseChef]

The thing is the titan x is a completely stupid product as it's not aimed at professionals yet it has the 12gb of vram which is completely useless to gamers.

 

You keep saying these things like it matters.  The purpose of the Titan X is not to make sense.  It's purpose is literally to be the most expensive.  That's it.  It is supposed to have the biggest numbers, and cost the most.  That is it's only purpose.  It never was and never will be about it being the "best" card in a performance per dollar or even total performance metric.  Despite anything marketing might try to tell you, it's only point is to be the most expensive. 

[ADZ_123_!"£]

You keep saying these things like it matters.  The purpose of the Titan X is not to make sense.  It's purpose is literally to be the most expensive.  That's it.  It is supposed to have the biggest numbers, and cost the most.  That is it's only purpose.  It never was and never will be about it being the "best" card in a performance per dollar or even total performance metric.  Despite anything marketing might try to tell you, it's only point is to be the most expensive. 

So it's for people who have more money than sense. Who really need the best and need 12gb of vram and as if one isn't enough they then get 2 or more bloody cards.

[Glenwing]

780 Ti was faster than the original Titan not the Titan Black. Titan Black and 780 Ti peform identical since they both have the same amount of Cuda Cores

Yeah I thought the 780 Ti had a higher stock clock but I guess not.

[Toddwjp]

this doesnt make sense, i mean maybe this card will exist if amds new top tier card is amazing,

[Prastupok]

-snip-

 

4x MSAA, everything maxed out otherwise. 

[patrickjp93]

how many people use 12GB of vram and the 980 ti will have at least 6GB so thats already plenty

Generally Titans fall into a niche of CUDA research cards which work well for schools. The large framebuffer is also good for caching tasks on a shared system or experimenting with scaling of data set size and performance for various workloads.

 

And now you can run OpenCL 1.2 on them too in addition to CUDA and OpenACC, so they're great for practicing heterogeneous computing since they have the same frame buffer sizes as Teslas, even if you don't get the DP with the Titan X you would with a Tesla.

[patrickjp93]

You keep saying these things like it matters.  The purpose of the Titan X is not to make sense.  It's purpose is literally to be the most expensive.  That's it.  It is supposed to have the biggest numbers, and cost the most.  That is it's only purpose.  It never was and never will be about it being the "best" card in a performance per dollar or even total performance metric.  Despite anything marketing might try to tell you, it's only point is to be the most expensive.

The Titan line's purpose is to provide schools with cheap heterogeneous computing research cards and rich gamers with stupid solutions which make Nvidia money. Nothing in life is without a purpose.

 

With the same framebuffer size and overall performance profile to a Tesla, they're great for doing proof of concept research due to the much more competitive price.