This is a split board - You can return to the Split List for other boards.

What are the major factors that decides which graphic card is the superior one?

#11DarkZV2BetaPosted 11/11/2013 3:13:54 AM
KabtheMentat posted...
"1GB GDDR5 Memory " = 1GB VRAM

You'll want something that's 2GB, or even 3GB if you can afford it. But a good 2GB card should do well at 1080p with a single monitor.

How much are you looking to spend? If you give us that info, then someone on the board can probably point you towards the best card at that price.

Edit: Just to give you an idea of performance. That card you linked has decent clock speeds, but 1GB VRAM. I have a 7850, which has somewhat lower clock speeds, but has 2GB VRAM, thus my 7850 will get superior performance at 1080p vs. the 7790. But if I played at lower resolutions, then the 7790 might do better than my 7850. I can overclock my card to give it better performance at lower resolutions, but no matter how much you overlock the 7790, it won't do as well as the 7850 at 1080p.

And 720p on a PC monitor...EWWWWWWWW. That's fine for console gaming on regular TVs, but not PC gaming IMO.


Vram has nothing to do with why 7850 does better than 7790. 7850 is just a better card. More cores, more ROPs, more TMUs, wider bus...

As for the OP...
The number of cores will determine how well the GPU handles post processing and advanced lighting effects.(as well as the architecture. ex, until recently, AMD cards had a ton of cores, but they were all pretty much ****, so you'd have AMD cards clocked higher and getting worse performance with 3x the number of GPU cores. GCN and Kepler are fairly even.)
ROPs and bus width, as well as memory capacity to a lesser extent, will affect higher resolution performance more than anything, but are important for performance in general. In the case of memory capacity, you won't see a linear performance difference with more/less RAM, but rather, if you don't have enough RAM, performance will start to suddenly tank, or your game will hitch and stutter. Provided you have enough memory, though, more won't make any difference at all.
TMUs handle textures, and more of them reflects in better performance of anisotropic filtering, multilayered textures, transparent texture effects, and transparency supersampling. These tend to also be impacted by bus width.
Finally, for clockspeed is basically a multiplier. If you have a GPU running 500mhz, and another running 1000mhz, the 1000mhz GPU is twice as fast. Same with memory. A 128bit bus with 5ghz effective data rate is as as good as(or better than) a 256bit bus with a 2.5ghz data rate.

Some of the actual performance results of these are quantitised in best-case-scenario numbers, and can give a rough idea of what to expect. Examples of this are "176GB/s" memory bandwidth, which is the result of bus width x effective memory speed. "pixels/sec" is basically rops x clockspeed. "texture fillrate" is TMUs x clockspeed.

So, to use 7850 vs 7790 as an example:
7790: 896 cores(stream processors) x 1000mhz clockspeed = 896000
7850: 1024 cores x 860mhz = 880640

7790: 16 ROPs x 1000mhz = 16000
7850: 32 ROPs x 860mhz = 27520

7790: 56TMUs x 1000mhz = 56000
7850: 64TMUs x 860mhz = 55040

7790: 128 x 6000mhz = 768000
7850: 256 x 4800mhz = 1228800

As you can see, the 7850 wins out hard in memory bandwidth and ROPs, which perfectly explains the performance curve where the two are very similar in lower resolutions, but separate quite a bit at higher resolutions. We can be rest assured that the 2GB vs 1GB isn't the cause of this because there are 1GB 7850s, and they still walk all over 7790.

This can only be used to compare cards with the same architecture, however, and can never be used to compare between vendors because of that, as well as because of how big a role software can play in these cards' performance.
---
Even people have toenails. Of course PCs have toenails. -claytonbuckley
#12DarkZV2BetaPosted 11/11/2013 3:16:16 AM
DeathHope posted...
I am planning to lead towards the AMD cards as well as nVidia cards is a tad more expensive here.

On a side note, how does the I5-2500 processor pair up with these graphic cards? Will the processor hold back all these graphic cards?
and
How do i determine a good motherboard? What are the things to look out for?


Having a lot of power phases is a good sign.(basically, you can count them by the number of tiny black boxes to the left, and sometimes above the CPU.) But there's not really any way to tell easily.
i5 2500 is fine.
---
Even people have toenails. Of course PCs have toenails. -claytonbuckley
#13KabtheMentatPosted 11/11/2013 3:25:41 AM
DarkZV2Beta posted...
KabtheMentat posted...
"1GB GDDR5 Memory " = 1GB VRAM

You'll want something that's 2GB, or even 3GB if you can afford it. But a good 2GB card should do well at 1080p with a single monitor.

How much are you looking to spend? If you give us that info, then someone on the board can probably point you towards the best card at that price.

Edit: Just to give you an idea of performance. That card you linked has decent clock speeds, but 1GB VRAM. I have a 7850, which has somewhat lower clock speeds, but has 2GB VRAM, thus my 7850 will get superior performance at 1080p vs. the 7790. But if I played at lower resolutions, then the 7790 might do better than my 7850. I can overclock my card to give it better performance at lower resolutions, but no matter how much you overlock the 7790, it won't do as well as the 7850 at 1080p.

And 720p on a PC monitor...EWWWWWWWW. That's fine for console gaming on regular TVs, but not PC gaming IMO.


Vram has nothing to do with why 7850 does better than 7790. 7850 is just a better card. More cores, more ROPs, more TMUs, wider bus...

As for the OP...
The number of cores will determine how well the GPU handles post processing and advanced lighting effects.(as well as the architecture. ex, until recently, AMD cards had a ton of cores, but they were all pretty much ****, so you'd have AMD cards clocked higher and getting worse performance with 3x the number of GPU cores. GCN and Kepler are fairly even.)
ROPs and bus width, as well as memory capacity to a lesser extent, will affect higher resolution performance more than anything, but are important for performance in general. In the case of memory capacity, you won't see a linear performance difference with more/less RAM, but rather, if you don't have enough RAM, performance will start to suddenly tank, or your game will hitch and stutter. Provided you have enough memory, though, more won't make any difference at all.
TMUs handle textures, and more of them reflects in better performance of anisotropic filtering, multilayered textures, transparent texture effects, and transparency supersampling. These tend to also be impacted by bus width.
Finally, for clockspeed is basically a multiplier. If you have a GPU running 500mhz, and another running 1000mhz, the 1000mhz GPU is twice as fast. Same with memory. A 128bit bus with 5ghz effective data rate is as as good as(or better than) a 256bit bus with a 2.5ghz data rate.

Some of the actual performance results of these are quantitised in best-case-scenario numbers, and can give a rough idea of what to expect. Examples of this are "176GB/s" memory bandwidth, which is the result of bus width x effective memory speed. "pixels/sec" is basically rops x clockspeed. "texture fillrate" is TMUs x clockspeed.

So, to use 7850 vs 7790 as an example:
7790: 896 cores(stream processors) x 1000mhz clockspeed = 896000
7850: 1024 cores x 860mhz = 880640

7790: 16 ROPs x 1000mhz = 16000
7850: 32 ROPs x 860mhz = 27520

7790: 56TMUs x 1000mhz = 56000
7850: 64TMUs x 860mhz = 55040

7790: 128 x 6000mhz = 768000
7850: 256 x 4800mhz = 1228800


Huh, I thought the 7850 and 7790 were basically the same. Oh well, learn something new.
---
Big Money. Big Women. Big Fun.
Skillz Ferguson