• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Gigabyte GeForce GTX 560 Ti Graphics Card Pictured

GTX 560 TI won't compete with 6970 or 6950 2GB
Only will compete with 6950 1GB, and won't beat it, at the best case they will trade blows.
GTX 560 Ti can't even touch 6970 because it will then cannibalize GTX 570.

Oh, it will surely perform on par with HD6950 2GB and once overclocked it will soundly beat stock GTX570 and HD6970. Performance difference between stock GTX580, 570 and 560 is not going to be much different than the one between HD6970, 6950 and 6870. Difference in both cases is around 10% in every step and that's how it's going to be this "generation" because there's a wall (40nm node) that prevents faster premium chips, but there's no such wall for performance/mainstream cards, so they have catched up. The high-end parts still have huge benefits for many enthusiast, like significantly better tesselation and GPGPU performance, also 64 bit support.

Compare these charts to look at the difference (~10%):

perfrel_1920.gif
perfrel_1920.gif


And here is a chart that I made that demostrates the performance that the GTX560 will have based on it's characteristics:

First the resume: performance icrease on Fermi cards is almost linear (+/- 2%) as their GFlops rise, as can be seen in the small chart below (performance numbers taken from W1zzard's HD6950 chart above):

Card | Actual performance | Relative Gflops
GTS450 | 46 | 47.74
GTX465 | 72 | 67.86
GTX460 | 73 | 72.03
GTX470 | 89 | 86.36
GTX560 | ? | 100.00
GTX570 | 109 | 111.59
GTX480 | 108 | 106.86
GTX580 | 125 | 125.53

Here's the complete chart, just in case you want to check the math or if you want to check how, effectively, Fermi cards' performace is pretty much only based on relative GFlops (Gflops = 2*SP number*shader clock):

card | Clock | SP | Gflops | Relative Gflops | ROP | Gpixel/clock | Relative fillrate | memory clock | bus width | memory bandwidth | relative memory bandwidth | texture units | texture fillrate | relative texture fillrate
GTS450 | 783 | 192 | 601.34 | 47.74 | 16 | 12.53 | 47.74 | 3608 | 128 | 57.73 | 45.10 | 32 | 25.06 | 47.74
GTX465 | 607 | 352 | 854.66 | 67.86 | 32 | 19.42 | 74.02 | 3206 | 256 | 102.59 | 80.15 | 44 | 26.71 | 50.89
GTX460 1GB | 675 | 336 | 907.20 | 72.03 | 32 | 21.60 | 82.32 | 3600 | 256 | 115.20 | 90.00 | 56 | 37.80 | 72.03
GTX470 | 607 | 448 | 1,087.74 | 86.36 | 40 | 24.28 | 92.53 | 3348 | 320 | 133.92 | 104.63 | 56 | 33.99 | 64.77
GTX560 | 820 | 384 | 1,259.52 | 100.00 | 32 | 26.24 | 100.00 | 4000 | 256 | 128.00 | 100.00 | 64 | 52.48 | 100.00
GTX570 | 732 | 480 | 1,405.44 | 111.59 | 40 | 29.28 | 111.59 | 3800 | 320 | 152.00 | 118.75 | 60 | 43.92 | 83.69
GTX480 | 701 | 480 | 1,345.92 | 106.86 | 48 | 33.65 | 128.23 | 3696 | 384 | 177.41 | 138.60 | 60 | 42.06 | 80.14
GTX580 | 772 | 512 | 1,581.06 | 125.53 | 48 | 37.06 | 141.22 | 4008 | 384 | 192.38 | 150.30 | 64 | 49.41 | 94.15
 
I like the dual dual-link DVI :)

But I dont like the 450W :(



I don't think it is such a big deal

New features of OpenGL 4.1 include:
Full compatibility with OpenGL ES 2.0 APIs for easier porting between mobile and desktop platforms >> Not a feature or performance benefit - just simpler coding/development for the mobile market
The ability to query and load a binary for shader program objects to save re-compilation time >> Potentially useful, but I that doesnt help performance in any existing code/games
The capability to bind programs individually to programmable stages for programming flexibility >> I think this is more of a CUDA enhancement and making a OpenGL 4.1 compliance board also capable of more efficient OpenCL libraries. Doesnt affect any existing code/games
64-bit floating-point component vertex shader inputs for higher geometric precision >> OUCH. That will be a performance killer and probably more important in the CAD/medical market
Multiple viewports for a rendering surface for increased rendering flexibility >> Interesting feature basically allowing hardware 3D rather than having to do it via software/driver enhancement

While OpenGL 4.x is a continuous improvement, and I like that, is seems that the feature set isnt going to significantly change what a PC enthusiast/gamer can expect. It seems quite targeted at specialised markets. Perhaps the 3D is interesting for us though! (multiple viewports).
oh wait. it turns out that the card is opengl 4. the pic above is a typo
t2capaxfxaxxxxxxxx35039.jpg


ps even the gtx 460 is opengl 4.0
 
What exactly does Ti stand for in this context? The last card I remember bearing "Ti" in the name was a Geforce 4 Ti 4200....and from what I remember it was a bit of a let-down.
 
What exactly does Ti stand for in this context? The last card I remember bearing "Ti" in the name was a Geforce 4 Ti 4200....and from what I remember it was a bit of a let-down.

The GeForce4 models are exactly what I thought of when I saw this. I don't remember seeing "Ti" used since.
 
I assume as some of the things that were fixed between GF100 and GF110 were included on the 460 this will be around an overclocked 460, i convinced a friend to get a 460 due to one model being within his price range and he managed to push it up to 950/975mhz (i can't remember which was the stable max for him), i would expect the 560 to hardly beat a 460 at such a higher clock so anyone with a 460 should have no reason to upgrade but anyone with an older card would find it to be a very nice mid range upgrade.

Really both Nvidia's 5xx cards and AMD's 6xxx cards are a nice improvement but nothing amazing due to both being 40nm still, i might get a 2GB card as I'm running a triple screen setup but with 28nm chip set to launch this year i think myself and most others should try to avoid temptation :laugh:
 
Oh, it will surely perform on par with HD6950 2GB and once overclocked it will soundly beat stock GTX570 and HD6970. Performance difference between stock GTX580, 570 and 560 is not going to be much different than the one between HD6970, 6950 and 6870. Difference in both cases is around 10% in every step and that's how it's going to be this "generation" because there's a wall (40nm node) that prevents faster premium chips, but there's no such wall for performance/mainstream cards, so they have catched up. The high-end parts still have huge benefits for many enthusiast, like significantly better tesselation and GPGPU performance, also 64 bit support.

Compare these charts to look at the difference (~10%):

http://tpucdn.com/reviews/HIS/Radeon_HD_6950/images/perfrel_1920.gifhttp://tpucdn.com/reviews/HIS/Radeon_HD_6970/images/perfrel_1920.gif

And here is a chart that I made that demostrates the performance that the GTX560 will have based on it's characteristics:

First the resume: performance icrease on Fermi cards is almost linear (+/- 2%) as their GFlops rise, as can be seen in the small chart below (performance numbers taken from W1zzard's HD6950 chart above):

Card | Actual performance | Relative Gflops
GTS450 | 46 | 47.74
GTX465 | 72 | 67.86
GTX460 | 73 | 72.03
GTX470 | 89 | 86.36
GTX560 | ? | 100.00
GTX570 | 109 | 111.59
GTX480 | 108 | 106.86
GTX580 | 125 | 125.53

Here's the complete chart, just in case you want to check the math or if you want to check how, effectively, Fermi cards' performace is pretty much only based on relative GFlops (Gflops = 2*SP number*shader clock):

card | Clock | SP | Gflops | Relative Gflops | ROP | Gpixel/clock | Relative fillrate | memory clock | bus width | memory bandwidth | relative memory bandwidth | texture units | texture fillrate | relative texture fillrate
GTS450 | 783 | 192 | 601.34 | 47.74 | 16 | 12.53 | 47.74 | 3608 | 128 | 57.73 | 45.10 | 32 | 25.06 | 47.74
GTX465 | 607 | 352 | 854.66 | 67.86 | 32 | 19.42 | 74.02 | 3206 | 256 | 102.59 | 80.15 | 44 | 26.71 | 50.89
GTX460 1GB | 675 | 336 | 907.20 | 72.03 | 32 | 21.60 | 82.32 | 3600 | 256 | 115.20 | 90.00 | 56 | 37.80 | 72.03
GTX470 | 607 | 448 | 1,087.74 | 86.36 | 40 | 24.28 | 92.53 | 3348 | 320 | 133.92 | 104.63 | 56 | 33.99 | 64.77
GTX560 | 820 | 384 | 1,259.52 | 100.00 | 32 | 26.24 | 100.00 | 4000 | 256 | 128.00 | 100.00 | 64 | 52.48 | 100.00
GTX570 | 732 | 480 | 1,405.44 | 111.59 | 40 | 29.28 | 111.59 | 3800 | 320 | 152.00 | 118.75 | 60 | 43.92 | 83.69
GTX480 | 701 | 480 | 1,345.92 | 106.86 | 48 | 33.65 | 128.23 | 3696 | 384 | 177.41 | 138.60 | 60 | 42.06 | 80.14
GTX580 | 772 | 512 | 1,581.06 | 125.53 | 48 | 37.06 | 141.22 | 4008 | 384 | 192.38 | 150.30 | 64 | 49.41 | 94.15

You confirm my words, i was talking about stock speeds.
About overcloking it and soundly beating GTX 570 or HD 6970, i highly doubt it, rebember it is only 1GB of VRAM, it doesn't matter how much you can overclock it, at least at high resolutions it won't be able to compete with 570 or 6970, maybe 6950 2GB.
GTX 560 competition is HD6950 1GB
 
Last edited:
Once these come out GTX 560 Ti SLi here I come!
 
Gigabyte OC and SOC factory OCed 560 Ti's running at 900 Mhz and 1000Mhz respectively?

http://www.fudzilla.com/graphics/item/21566-gigabyte-to-have-two-overclocked-gtx-560-ti-cards

It's Fudzilla quoting some chinese etailer so take with a grain of salt. But if true Nvidia may have a real winner there:

stock 560 Ti vs HD6950
OC 560 Ti vs HD6970
SOC 560 ti vs HD6970 OC

edit: ^^ performance wise. manufacturing costs are far chaper for the 560 Ti, reason for which I think Nidia might have a real winner.

The price for the OC and SOC is $290 and $320 according to the link above. That would probably mean ~$250 for stock clocked cards.


I feel this card GTX560 is gonna rock HD6950 like anything, and there s a possibilty GTX560 1GHz SOC might beat GTX570 and HD6970 too, lets hope it does at a lower power consumption.

No wonder nVidia call the card GTX560 Ti
 
My guess is GTX 470 performance

Probably about the same performance as a GTX 470 with a lot less power consumption. It seems like Nvidia moves product lines in general 25% performance increments. So figure 460 25% slower than 560 which is 25% slower than 570 which is 25% slower than 580.
 
I would be surprised if performance difference between 1 GB and 2 GB HD 6950 is over 2%.
 
I would be surprised if performance difference between 1 GB and 2 GB HD 6950 is over 2%.

up to 1080p maybe, but at 2560x1538 the diff will definitely be higher.
 
Is there a flaw in the 460s?
 
Is there a flaw in the 460s?

IMO the only flaw with the 460 was the fact some of the cuda cores were disabled on the 1GB model.
 
dont feel too confident

Yea right, nVidia has some bad ass card's coming out, AMD better start spitting out more game (I myself am waiting for the 6990 and the 590 benchies) .. this card seems pretty bad ass ...I wonder the o/c potential of this bad boy. Already set at 900/1000 by factory ... that's insane! ... unprecedented!

gadget2.jpg

wowzer!
 
Added statement from Gigabyte.
 
lol, french? I could translate it if it was, but I'm pretty certian that ain't french.

Those are Traditional Chinese.

Translation:

Selling Points:
1. ...
2.The performance equals or is close to HD6950
3.Mini HDMI and HDMI port could be freely changed (by adapter?)
:nutkick:
 
GTX 560 competition is HD6950 1GB

ok, and the difference between a 1gb and 2gb 6950 will be next to nada, save for a CFX situation at high resoluitions like eyefinity for exammple.

the 6950's competition IS a GTX470, as shown in AMD's own slides, and the GTX560 will take the GTX470's place, and given its specs, will be faster than one stock for stock.

the GTX560 will also probably become availiable in non-reference 2gb models, just like the GTX460 did.

why do you find it so hard to believe the 560 can take on a 6950 properly?
 
ok, and the difference between a 1gb and 2gb 6950 will be next to nada, save for a CFX situation at high resoluitions like eyefinity for exammple.

the 6950's competition IS a GTX470, as shown in AMD's own slides, and the GTX560 will take the GTX470's place, and given its specs, will be faster than one stock for stock.

the GTX560 will also probably become availiable in non-reference 2gb models, just like the GTX460 did.

why do you find it so hard to believe the 560 can take on a 6950 properly?

You're saying the same thing i'm saying, you just add more stuff, like 2GB gtx 560, cfx, sli configs.
I said GTX 560 1GB competition is 6950 1GB, meaning they are on par.

Yes of course gtx 470 is the competition i'm not saying it isn't, but like you say gtx 560 will be faster than gtx 470, so what, 6950 already is.

All i'm saying is they will be on par, 1GB vs 1GB, 2GB vs 2GB. I'm not saying gtx 560 will be slower.

EDIT: And i obviously won't talk about o/c because like the update says, completely false info.
 
What exactly does Ti stand for in this context? The last card I remember bearing "Ti" in the name was a Geforce 4 Ti 4200....and from what I remember it was a bit of a let-down.

"Turbo Intercooler" ?? :laugh:
 
You're saying the same thing i'm saying, you just add more stuff, like 2GB gtx 560, cfx, sli configs.
I said GTX 560 1GB competition is 6950 1GB, meaning they are on par.

Yes of course gtx 470 is the competition i'm not saying it isn't, but like you say gtx 560 will be faster than gtx 470, so what, 6950 already is.

All i'm saying is they will be on par, 1GB vs 1GB, 2GB vs 2GB. I'm not saying gtx 560 will be slower.

EDIT: And i obviously won't talk about o/c because like the update says, completely false info.

What you are saying holds true in many instances but certainly not all, in your earlier post you mentioned that the Higher memory cards would perform better at the very high resolutions and even that does not always hold true, if you look at the attached which is showing results at 1920 x 1200 and 2560 x 1600 and look at the comparison between GTX 580 and HD 6970 your statement is supported, you will see that at 1920 the 580 is 13% faster but at 2560 that goes down a little to 10%.
Now take a look at the comparison between the 570 and the 480 (the 480 of course having more memory), you will see that both cards are exactly neck and neck at 1920 but the lower memory card is actually 1% faster at 2560.
Now 1% or even 3% is not noticeable however you cannot always assume that these performance differences are down to the quantity of memory, there could be a number of explanations, such as memory speeds, bandwidth or simply the GPU's architecture.
 

Attachments

  • perfrel_1920.JPG
    perfrel_1920.JPG
    94.1 KB · Views: 405
Last edited:
What exactly does Ti stand for in this context? The last card I remember bearing "Ti" in the name was a Geforce 4 Ti 4200....and from what I remember it was a bit of a let-down.

The GeForce3 300 & 500 and all the GeForce4 cards were "Ti" models.
 
What exactly does Ti stand for in this context? The last card I remember bearing "Ti" in the name was a Geforce 4 Ti 4200....and from what I remember it was a bit of a let-down.

A let-down? are you serious??
Ti 4200 was a legendary card man.
 
Back
Top