About: ROPs and FLOPS and pixels and textels

[GoGeTa006]

Ok, this is a well, boiled enough version of what a ROP is:

My question to you, that article is very old (talking about the 6600 GT (posted 2006))

and I have come to a climaxing decision in my next build:

What I have:

- Fractal design R4 case
- Led Strip lightning

What I am SURE to get:

- 120 GB SSD
- 16 GB RAM
- intel i5 Ivy bridge 3570K
- at least 600 W PSU (depends on sale price)


What I am still deciding:

- Video Card:

Here is where the forever debate continues,

I have read multiple articles and forum posts comparing

eVGA Geforce GTX 660 Ti FTW (Overclocked to ~1 Ghz)
whichever is the cheapest Geforce GTX 670 at the time of purchase (Newegg's presidents day sale)
Radeon 7950 Overclocked (forums suggest OCing to 1.2 Ghz)

------------ Now here is my question to you BakaBT members,

From those numbers. in that webpage which according to the internetz is one of the most reliable sources of data (techpowerup)

Nvidia have something called SMX while the Radeon have something called computing units, fuck me if I know what that is. . .anyways

my question is about the ROPs:

The GTX 670 has 32 ROPs just like the 7950, but the 660 Ti FTW has higher GTextel and GPixel rates than both the 670 and 7950
now according tothis ROPs were "decreasing" because the age of ROPs was comming to an end with shaders, but now we see this increase in rops with increase in model (670 > 660 Ti )

So,  if TL;DR

What are ROPs used for nowadays, are they crucial to Anti-Aliasing? (which AFAIK is the new eye candy that everyone wants to maximize)

I assume more ROPs is better, but how much are we talking about here?? (I mean from 24 to 32, is that a HUGE gap?)
cause I am boiling down my decision based on price range between the 660 Ti FTW and the 7950 OC

(click to show/hide)

I know at some point it MIGHT boil down to what Im going to do with it, so here it is:

- Photoshop
- Adobe Premiere (very rarely, hobby)
- Flash (rarely, hobby)
- Solidworks (3D-CAD) (constant, for school)

Gaming (I know some games are "optimized" for physX and shit like that)

- DOTA 2 (always)
- CS:GO (always)
- Amnesia (rarely, I have it tho)
- COD:MW ?? (newest) (I might buy it)
- BF3 (I might buy it)
- Blade & Soul (maybe) <- Seems highly AA
- Black Dessert Online (maybe)

I dont buy too many games because TBH I only play 2 or 3 (diff genres) so I might change my FPS from CS:GO to something else if it tickles my fancy

Please elaborate on your posts with links if possible

Thank you
much love

[kitamesume]

i'll get back here after finishing eating, but for now go with benchmarks and reviews, they do tell whats their difference clear enough.
also comparing two different architectures is quite hard as well, they lack the "IPC" in CPU terms to be comparable.

ok i'm back - ROPs, TMUs and pixel shaders(they all go by ratios, note from GPU charts) definitions from wiki to hasten up the explanation.
http://en.wikipedia.org/wiki/Render_Output_unit
http://en.wikipedia.org/wiki/Texture_mapping_unit
http://en.wikipedia.org/wiki/Pixel_shader#Pixel_shaders // http://en.wikipedia.org/wiki/Unified_shader

(click to show/hide)

The Render Output Unit, often abbreviated as "ROP", and sometimes called (perhaps more properly) Raster Operations Pipeline, is one of the final steps in the rendering process of modern 3D accelerator boards. The pixel pipelines take pixel and texel information and process it, via specific matrix and vector operations, into a final pixel or depth value. The ROPs perform the transactions between the relevant buffers in the local memory — this includes writing or reading values, as well as blending them together.

Historically the number of ROPs, TMUs, and pixel shaders have been equal. However, as of 2004, several GPUs have decoupled these areas to allow optimum transistor allocation for application workload and available memory performance. As the trend continues, it is expected that graphics processors will continue to decouple the various parts of their architectures to enhance their adaptability to future graphics applications. This design also allows chip makers to build a modular line-up, where the top-end GPU are essentially using the same logic as the low-end products.

Render Output Pipelines (ROPs)
The main articles for this category are Raster graphics and Raster Operations.

The Render Output Pipeline is an inherited term, and more often referred to as the Render Output Unit. Its job is to control the sampling of pixels (each pixel is a dimensionless point), so it controls antialiasing, where more than one sample is merged into one pixel. All data rendered has to travel through the ROP in order to be written to the framebuffer, from there it can be transmitted to the display.

Therefore, the ROP is where the GPU's output is assembled into a bitmapped image ready for display.

A texture mapping unit (TMU) is a component in modern graphics processing units (GPUs), historically it is a separate physical processor. A TMU is able to rotate and resize a bitmap to be placed onto an arbitrary plane of a given 3D object as a texture. In modern graphics cards it is implemented as a discrete stage in a graphics pipeline,[1] whereas when first introduced it was implemented as a separate processor, e.g. as seen on the Voodoo2 graphics card.

Texture Mapping Units (TMUs)
The main article for this category is Texture mapping.

Textures need to be addressed and filtered. This job is done by TMUs that work in conjunction with pixel and vertex shader units. It is the TMU's job to apply texture operations to pixels. The number of texture units in a graphics processor is used when comparing two different cards for texturing performance. It is reasonable to assume that the card with more TMUs will be faster at processing texture information.

In the field of computer graphics, a shader is a computer program that runs on the graphics processing unit and is used to do shading - the production of appropriate levels of light and darkness within an image - or, in the modern era, also to produce special effects or do postprocessing.

Shaders calculate rendering effects on graphics hardware with a high degree of flexibility. Shading languages are used to program the graphics processing unit (GPU) programmable rendering pipeline, which has mostly superseded the fixed-function pipeline that allowed only common geometry transformation and pixel-shading functions; with shaders, customized effects can be used. The position, hue, saturation, brightness, and contrast of all pixels, vertices, or textures used to construct a final image can be altered on the fly, using algorithms defined in the shader, and can be modified by external variables or textures introduced by the program calling the shader.

Types of shaders

There are three types of shaders in common use. While older graphics cards utilize separate processing units for each shader type, newer cards feature unified shaders which are capable of executing any type of shader. This allows graphics cards to make more efficient use of processing power.

(click to show/hide)

Vertex shaders

Vertex shaders are run once for each vertex given to the graphics processor. The purpose is to transform each vertex's 3D position in virtual space to the 2D coordinate at which it appears on the screen (as well as a depth value for the Z-buffer). Vertex shaders can manipulate properties such as position, color, and texture coordinate, but cannot create new vertices. The output of the vertex shader goes to the next stage in the pipeline, which is either a geometry shader if present, or the pixel shader and rasterizer otherwise. Vertex shaders can enable powerful control over the details of position, movement, lighting, and color in any scene involving 3D models.


Geometry shaders

Geometry shaders are a relatively new type of shader, introduced in Direct3D 10 and OpenGL 3.2; formerly available in OpenGL 2.0+ with the use of extensions.[2] This type of shader can generate new graphics primitives, such as points, lines, and triangles, from those primitives that were sent to the beginning of the graphics pipeline.[3]

Geometry shader programs are executed after vertex shaders. They take as input a whole primitive, possibly with adjacency information. For example, when operating on triangles, the three vertices are the geometry shader's input. The shader can then emit zero or more primitives, which are rasterized and their fragments ultimately passed to a pixel shader.

Typical uses of a geometry shader include point sprite generation, geometry tessellation, shadow volume extrusion, and single pass rendering to a cube map. A typical real world example of the benefits of geometry shaders would be automatic mesh complexity modification. A series of line strips representing control points for a curve are passed to the geometry shader and depending on the complexity required the shader can automatically generate extra lines each of which provides a better approximation of a curve.


Pixel shaders

Pixel shaders, also known as fragment shaders, compute color and other attributes of each fragment. Pixel shaders range from always outputting the same color, to applying a lighting value, to doing bump mapping, shadows, specular highlights, translucency and other phenomena. They can alter the depth of the fragment (for Z-buffering), or output more than one color if multiple render targets are active. In 3D graphics, a pixel shader alone cannot produce very complex effects, because it operates only on a single fragment, without knowledge of a scene's geometry. However, pixel shaders do have knowledge of the screen coordinate being drawn, and can sample the screen and nearby pixels if the contents of the entire screen are passed as a texture to the shader. This technique can enable a wide variety of two-dimensional postprocessing effects, such as blur, or edge detection/enhancement for cartoon/cel shaders. Pixel shaders may also be applied in intermediate stages to any two-dimensional images in the pipeline, whereas vertex shaders always require a 3D model. For instance, a pixel shader is the only kind of shader that can act as a postprocessor or filter for a video stream after it has been rasterized.

so to point out:
1) ROPs handles the output from GPU to Monitor
2) TMUs handles the assembly of the image to be rendered
3) Shaders handles whats there to be proccessed
---
4) ROPs only needs specifically enough amount to handle the size of the image and the resolution of the screen, in other words its the doorway from the GPU to the monitor hence it doesn't need a huge amount of pipelines to do so.
5) TMUs only need enough amount to not bottleneck the shaders, as a rule of thumb they need to fill and map the meshes at the rate of (screen refresh rate times pixel count) and should not go under it.
6) shaders are the graphics proccessors themselves.

PS: before you start purchasing anything you should pay attention to the market, AMD (sea islands) HD8000 is near it's launch date
http://en.wikipedia.org/wiki/Sea_Islands_%28GPU_family%29
xbitlabs HD8000

edit: ok heres a god-awfully-done-chart of the HD8000 with a resized HD7000 for comparisons


to note:
HD8870 would be a little bit faster than HD7950 in single-precision computations while being nearly half it's price, the cons? double-precision plummets.
expect HD8930/HD8950 to be a little bit cheaper while a whole level faster than HD7950, estimated prices at $350-HD8930, $420-HD8950, $500-HD8970.

[GoGeTa006]

*havent read everything i g2g but

My budget is in the 300 range, + tax (330ish)

for that price I can either get that 660 Ti FTW or the HD 7950 or, as I've seen people comment I believe the Zotac 670 has dropped at some point to 314 or something along those lines.

Ok, read it. . .and saw that the 8950 still has less Gtextels and Gpixels than the 660 Ti FTW
It does however have a lot more GFLOPs
And the wiki table doesnt say how many ROPs does it have (or didnt see it in the table)


Which is more important Textels/Pixels or ROPs (Because even the 8950 which will probably be too expensive has less Gtextels and Gpixels than the FTW)
It does however have a lot more FLOPS (even the 7950)
Then again, how does this translate to performance? And why should I care about these numbers

*EDIT*

Found this, looks a lot more powerful, but has different numbers than that wiki table

http://www.techpowerup.com/gpudb/1933/AMD_Radeon_HD_8950.html


The 660 Ti FTW is pretty close to the 760 Ti rumored stock values

[kitamesume]

those numbers roughly estimates how much the GPU can do, pixel fillrate represents how fast it can fill a full picture per frame per second (1920x1080p has 2,073,600 pixels per frame [1GP/s is 1billion pixels per second]), texture fillrate is how fast it can fill meshes per second, and (giga)FLoating-point Operations Per Second represents how much it can crunch instructions per second.

direct comparison with GTX760Ti would be HD8930 @ (shader)2048:(TMU)128:(ROP)32, given varying software scaling, IPC and clock frequency, the deciding winner would be based on price point alone, HD8930 should have an approximate SRP of $320-350(based on last price point of HD7870).

PS: GP/s and GT/s of HD8930 and HD8950 is still unknown, but do note that HD8870 has 35GP/s : 123GT/s : 3940GFLOPs, so expect the others to be much higher.
to note, GTX760Ti only has 29GP/s : 118GT/s : 2,841GFLOPs

[Tatsujin]

*havent read everything i g2g but

My budget is in the 300 range, + tax (330ish)

for that price I can either get that 660 Ti FTW or the HD 7950 or, as I've seen people comment I believe the Zotac 670 has dropped at some point to 314 or something along those lines.

Ok, read it. . .and saw that the 8950 still has less Gtextels and Gpixels than the 660 Ti FTW
It does however have a lot more GFLOPs
And the wiki table doesnt say how many ROPs does it have (or didnt see it in the table)


Which is more important Textels/Pixels or ROPs (Because even the 8950 which will probably be too expensive has less Gtextels and Gpixels than the FTW)
It does however have a lot more FLOPS (even the 7950)
Then again, how does this translate to performance? And why should I care about these numbers

*EDIT*

Found this, looks a lot more powerful, but has different numbers than that wiki table

http://www.techpowerup.com/gpudb/1933/AMD_Radeon_HD_8950.html


The 660 Ti FTW is pretty close to the 760 Ti rumored stock values

Do more reviews on the 660 Ti FTW before you purchase it. You need to unlock the fan speed and set it at 50% or higher (minimum), otherwise you'll have black screens and restarts. I am not sure if the updated drivers would fix this, but the default option for the fan speeds is the worst thing you can get for the FTW edition.

those numbers roughly estimates how much the GPU can do, pixel fillrate represents how fast it can fill a full picture per frame per second (1920x1080p has 2,073,600 pixels per frame), texture fillrate is how fast it can fill meshes per second, and (giga)FLoating-point Operations Per Second represents how much it can crunch instructions per second.

direct comparison with GTX760Ti would be HD8930 @ (shader)2048:(TMU)128:(ROP)32, given varying software scaling, IPC and clock frequency, the deciding winner would be based on price point alone, HD8930 should have an approximate SRP of $320-350(based on last price point of HD7870).

PS: GP/s and GT/s of HD8930 and HD8950 is still unknown, but do note that HD8870 has 35GP/s : 123GT/s : 3940GFLOPs, so expect the others to be much higher.
to note, GTX760Ti only has 29GP/s : 118GT/s : 2,841GFLOPs

Can I see the sources to that?

[kitamesume]

^wiki and gogeta's link.

eitherway this'll be a long discussion if she still asks.

also: price points as to why i got that estimations - [ price : name : core-config ]

$109:HD7750(512:32:16) - $120:HD8750(768:48:16) - ?? perf increase (in terms of GFLOPs)
$159:HD7770(640:40:16) - $150:HD8770(896:56:16) - ?? perf increase (in terms of GFLOPs)
$249:HD7850(1024:64:32) - $199:HD8850(1536:96:32) - 70% perf increase (in terms of GFLOPs)
$349:HD7870(1280:80:32) - $279:HD8870(1792:112:32) - 54% perf increase (in terms of GFLOPs)
$359:HD7870XT(1536:96:32) - $320-349:HD8930(2048:128:32) - ?? perf increase (in terms of GFLOPs)
$450:HD7950(1792:112:32) - $420-450:HD8950(2304:144:48) - ?? perf increase (in terms of GFLOPs)
$550:HD7970(2048:128:32) - $500-550:HD8970(2560:160:48) - ?? perf increase (in terms of GFLOPs)

notice: core-config alone should tell you how the performance increase reached over 50%
PS: if these numbers does tell us whats coming then GK110 should indeed be coming as GTX780 with a wooping monstrous performance increase over GTX680, otherwise to hell with Nvidia (no less than 40% perf increase or Nvidia would fall behind AMD this time around).

[GoGeTa006]

If by she you meant me, Im a he, just FYI.

Anyways, so the 8950 seems out of reach if its going to be around the 400 mark

The 8870 couldve been if it was already out. . .

So from what will be available as of mid february, I am debating between the 660 Ti FTW, the 7950 or if price drops enough on a zotac 670.

Again, it all comes down to this question:

Which are more important, FLOPs, ROPs, Gtextels or Gpixels performancewise?.

 on a 1080p monitor single (+ a TV for watching anime) but I wont be extending a single fullscreen application over 2 monitors.

And in a hierarchy because at the end of the day its pretty much comparing the 7950 with the 660 Ti FTW

if (FLOPS/ROPs >> Pixels/Textels)
     then I get 7950
Else
     660 Ti FTW



*EDIT*

as I keep reading reviews, as Tatsu said the FTW seems to have some issues with the cooling and many report stability problems, something thats probably fixed (some posts seemed relatively old)


ANYWAYS I am starting to lean towards the Gigabyte HD 7950 OC (900 Mhz) and probably OC it to 1Ghz, I am still seeing that stuff, although I have a question:

I CANT FIND FUCKING BACKPLATES!!
however, I do have access to a 3D printer, so I could potentially print a backplate myself but I am not sure if ABS plastic is a good material for backplates. . .has anyone owned a backplate? are they plastic or metal?

The thing is, this new Rig im building I am highly going for aesthetics and I am trying to keep everything in a black&white color scheme to go with my Fractal design R4, I am getting white light instead of colors, and I am getting an ASRock motherboard just cause they're black, and I am getting  BLACK RAM too, and the fans are black n white and shit. so I really want a back plate cause that blue PCB will just completely ruin my rig

[kitamesume]

Which are more important, FLOPs, ROPs, Gtextels or Gpixels performancewise?.

 on a 1080p monitor single (+ a TV for watching anime) but I wont be extending a single fullscreen application over 2 monitors.

And in a hierarchy because at the end of the day its pretty much comparing the 7950 with the 660 Ti FTW

if (FLOPS/ROPs >> Pixels/Textels)
     then I get 7950
Else
     660 Ti FTW

in terms of raw computation its always FLOPS > anything else, but overall performance has a balance between pixel:texture:flops, go out of ratio and one bottlenecks the other. but theres a catch, each program has a unique ideal ratio, meaning you can't have the perfect ratio e.g. one prefers more flops or the other prefers more textures.
this is why you go strictly for benchmarks on the given application that you want to see it run on, compare and voala.
but to point out, HD7950 has more headroom than GTX660Ti, its approximately in between GTX660Ti and GTX670 in terms of overall performance.

and yeah, if you're after going for a "buy by feb" then you're out of luck, neither Nvidia or AMD will be releasing earlier than june.

most of the backplates i see are made of metal, or some kind of hard plastic you see that doesn't go soft at 110-125c.


save up some money for a couple more weeks and stack more dollars, try to keep stretching your schedule and aim for those nice newegg sales(if you're getting one from newegg), worst thing that can happen is you getting impatient and pulling the trigger

[Saras]

If by she you meant me, Im a he, just FYI.

Anyways, so the 8950 seems out of reach if its going to be around the 400 mark

The 8870 couldve been if it was already out. . .

So from what will be available as of mid february, I am debating between the 660 Ti FTW, the 7950 or if price drops enough on a zotac 670.

Again, it all comes down to this question:

Which are more important, FLOPs, ROPs, Gtextels or Gpixels performancewise?.

 on a 1080p monitor single (+ a TV for watching anime) but I wont be extending a single fullscreen application over 2 monitors.

And in a hierarchy because at the end of the day its pretty much comparing the 7950 with the 660 Ti FTW

if (FLOPS/ROPs >> Pixels/Textels)
     then I get 7950
Else
     660 Ti FTW



*EDIT*

as I keep reading reviews, as Tatsu said the FTW seems to have some issues with the cooling and many report stability problems, something thats probably fixed (some posts seemed relatively old)


ANYWAYS I am starting to lean towards the Gigabyte HD 7950 OC (900 Mhz) and probably OC it to 1Ghz, I am still seeing that stuff, although I have a question:

I CANT FIND FUCKING BACKPLATES!!
however, I do have access to a 3D printer, so I could potentially print a backplate myself but I am not sure if ABS plastic is a good material for backplates. . .has anyone owned a backplate? are they plastic or metal?

The thing is, this new Rig im building I am highly going for aesthetics and I am trying to keep everything in a black&white color scheme to go with my Fractal design R4, I am getting white light instead of colors, and I am getting an ASRock motherboard just cause they're black, and I am getting  BLACK RAM too, and the fans are black n white and shit. so I really want a back plate cause that blue PCB will just completely ruin my rig

1st. Ignore the 8xxx series. AMD won't be introducing them anytime soon, due to them having their hands full producing stuff for the ps4. I wouldn't be expecting it at a price point worth mentioning in 2013 if at 2013 at all. For nVidias 7xx series, the released cards will be way out of your price range.

if you want a cool looking black motherboard, why not go for the Asus Sabertooth http://www.asus.com/Motherboard/SABERTOOTH_Z77/#gallery

The 7950, especially if overclocked, floors the 660 ti. Spec wise, if the game has no bias one way or another, the 7950 will kick the 660s ass. However, in the end, that rare and you'll need to look at the games you play. Borderlands 2 were made to be played with nVidia cards, likewise, Far Cry 3 was made for ATI. Look at the games you want to play and take a look at what they're made for.

PS: if these numbers does tell us whats coming then GK110 should indeed be coming as GTX780 with a wooping monstrous performance increase over GTX680, otherwise to hell with Nvidia (no less than 40% perf increase or Nvidia would fall behind AMD this time around).

No it won't. The GK110 is coming in a month as the GeForce Titan A 1000$ single GPU card. It's a Tesla card, not a GTX one. The 780 will be coming with a die about the same size as the 680. nVidia, like Intel is doing everything it can to increase their profit margins and both of them have been cutting their die sizes a fast as they can.

[kitamesume]

PS: if these numbers does tell us whats coming then GK110 should indeed be coming as GTX780 with a wooping monstrous performance increase over GTX680, otherwise to hell with Nvidia (no less than 40% perf increase or Nvidia would fall behind AMD this time around).

No it won't. The GK110 is coming in a month as the GeForce Titan A 1000$ single GPU card. It's a Tesla card, not a GTX one. The 780 will be coming with a die about the same size as the 680. nVidia, like Intel is doing everything it can to increase their profit margins and both of them have been cutting their die sizes a fast as they can.

source?

proof of estimation is AMD's incoming HD8970 being about 20-30ish percent faster than HD7970 just by looking at the increase in shader count, we know at the moment that a HD7970-GE is right at GTX680's foot in terms of performance, much more so with an overclocked HD7970-GE.


without GK110, i can't imagine a GK204(refreshed GK104 aka GTX660/660Ti/670/680) competing at that vast difference in performance, unless they took a few SMX off GK110 and named it as the GK204 instead. i also can't imagine Nvidia going for 2nd place.

edit: one thing that came to mind is Nvidia going with GTX785 as their main monster and not GTX780, its highly likely too.
the story would go around like "HD8970 way better deal than GTX780... wat!? 'nvidia throws in GK110 GTX785' nuff said"

[Saras]

The 8970 isn't a direct competitor, because it ain't coming out any time soon
http://wccftech.com/amd-confirms-radeon-hd-8000-sea-island-delay-hd-7000-series-remain-primary-focus/

The Geforce Titan
http://wccftech.com/nvidia-reportedly-preparing-geforce-titan-gpu-gk110/
http://wccftech.com/asus-geforce-titan-6-gb-listed-online-1300-launch-imminent/

[kitamesume]

i wasn't talking about titan you know, i'm talking about what the GTX780 will actually be... its the people's hype about titan that made a false rumor anyway.

[Saras]

That's out of my own assumption. There's little else to go by with at the time of writing.

nVidia has been doing everything it can to keep the production cost as low as possible, I don't see them going against what they were doing for the last few years and just supersizing their die. The profit margins are low as they are already, their share holders wouldn't quite like them to sell cards at a loss.

I see them going the same way as intel really. Want affordable more than good enough computing? There's 1155 platform. Want something more than that? Pay the premium for LGA2011. Just in nVidias case, it's going to be GTX/Titan.

[kitamesume]

going by that assumption Nvidia would have a possibility of losing the title of fastest this year around.


excluding the fact that theres a high chance that AMD will pospone HD8000 series, along with Nvidia doing so as well.
following this chart as a baseline and the claim of HD8970 being 20-30% faster than HD7970, it would result with HD8970 being 15% faster than GTX680.
now with your claim of GTX780 being a refresh of GTX680, possibly pushing the base clock to 1100mhz and maybe a little refinements on the die itself, at most the final product would be 15% faster.

with this we can say that both HD8970 and GTX780 would be equal, but looking at the clock speeds alone, HD8970 would have the advantage of having higher overclocking headroom since it can go about with a baseclock of 900-1000mhz, with both AMD and Nvidia aiming at the price point of $500 it should point out that AMD would have the edge over this competition, much more so with AMD's sly tactics of either dropping price or releasing a revised HD8970.

[Saras]

The Titan is still a single GPU card by nVidia that AMD has no answer to. Even with a 15-30% bump over the current 680.

nVidia will continue to hold the title of having the fastest card, irregardless of where its 780 will stand.

[kitamesume]

i guess so, but titan has a category of Tesla, a workstation card, not really one to be labeled as regular-consumer product.

but eitherway if its about "worth" or "bang for buck", pricepoint alone should indicate that a mere 50-60% speed advantage over GTX680 won't justify it's price of double, much more so against incoming new models with matching pricedrops on older models.
with this, people who considers pure worth would cling to their former equipment, maybe even buy another for multi-card which would result to a much higher performance.

PS: HD8970 theoretically has 1.25 times more SP count which should lead to an increase of 25% at the same clock(1ghz), bumping the frequency to 1.2Ghz results to a 1.5 times ideal performance increase, which comes close to titan(at half the price).
(HD8970)3063x1.5 vs (titan)5000

[Saras]

i guess so, but titan has a category of Tesla, a workstation card, not really one to be labeled as regular-consumer product.

but eitherway if its about "worth" or "bang for buck", pricepoint alone should indicate that a mere 50-60% speed advantage over GTX680 won't justify it's price of double, much more so against incoming new models with matching pricedrops on older models.
with this, people who considers pure worth would cling to their former equipment, maybe even buy another for multi-card which would result to a much higher performance.

PS: HD8970 theoretically has 1.25 times more SP count which should lead to an increase of 25% at the same clock(1ghz), bumping the frequency to 1.2Ghz results to a 1.5 times ideal performance increase, which comes close to titan(at half the price).
(HD8970)3063x1.5 vs (titan)5000

You're talking as if the 7970 or the 680 were bang for buck cards in the first place. They're not, they're premium cards. Bang for buck starts with the 7750/650 and ends with the 7870/660. You're overpaying if you go to either side of the spectrum from that. And if you're an enthusiast, be prepared to have to clear out your wallet for the upgrade.

[kitamesume]

... you missed the point already, i was saying that in terms of "bang for buck" the cards would've been better than titan, i did not say they were the best bang for buck, get your facts straight.

and take note

with this, people who considers pure worth would cling to their former equipment, maybe even buy another for multi-card which would result to a much higher performance.

this is far from being restricted to either HD7970 and GTX680, "former equipment" can range from the ancient HD4xxx GTX9xxx to today's beasts for all i care.

[Saras]

Right. Pardon my reading comprehension.

[GoGeTa006]

Found a pretty nice article, Im getting the 660 Ti
http://techreport.com/review/23981/radeon-hd-7950-vs-geforce-gtx-660-ti-revisited

Its pretty amazing how the radeon lost even with the boost clock and all those fancy higher numbers