GTX680 or GTX590

[GSalden 4,655]

Currently having a GTX580 I am having the opportunity to buy either a GTX680 or a GTX590 ( both second handed ) for 400 euro.

 

Having read the topic about the amount of shaders being related to fps in clouds I already was thinking about upgrading.

At the moment I have a GTX580.

 

 

GTX580 512 shaders 2x clockspeed = 1024

 

GTX680 1536 shaders at clockspeed = 1536

 

GTX590 512 shaders x 2 clockspeed x 2 = 2024

 

 

As FSX does not use SLI, I was wondering if the extra shader power from the GTX590 could be an advantage over the GTX690 regarding the drops in framerate in clouds.

 

What do you think ?

[GSalden 4,655]

I forgot to mention that I use a TH2Go for one Wide outside view in 5040 x 1050.

[dazz 69]

Keep in mind the 590 doesn't really run at 2x the 680's shader speed since the base clock is like 600 vs 1000, so it's 1200 vs 1000.

Besides, if you overclock, the 680 easily beats the 590

GTX680, all the way

[GSalden 4,655]

My framerate drop is with spotplane view.

Then clouds are spread over 3 monitors .

 

The SLI tests ( see sticky ) show that with external view SLI gives a enormeous boost

 

SLI OFF 1920x1200 --> 48

SLI OFF 2560x1600 --> 32

 

SLI ON 1920x1200 --> 26

SLI ON 2560x1600 --> 17

 

However this was done with just 1 monitor.

 

 

 

Iceman2 went from a GTX590 to a GTX680.

Quote " I have noticed after going from a GTX590 to a GTX680 4g, the impact on FPS for multi screens is greater. This is assuming we consider a single 590 card as SLI.

 

There was no increase in overall FPS with the 680 on a single 30in screen which I was expecting from other posts about FSX & SLI not playing nice together.

 

The 680 is certainly a nicer card with quality and color but now I'm thinking that the SLI was helping with higher resolutions in demanding scenery. "

[Ben Cap 89]

GTX680... no question

[GSalden 4,655]

GTX680... no question

 

Because ?

 

Keep in mind that I have a muli-monitor outside view setup with a high resolution.

 

 

 

@Dazz

You are right . My calculations where off.

 

I remarked that the calculation for the Shader power in the GPU Benchmark tests where done with the Core clock speed and not with the Shader clock speed.

 

If the outcome of the calculations where correct then the Shader speed has nothing to do higher FPS, but the amount of Shaders icw the Core speed.

 

 

Otherwise the calculations had to be with th efollowing cards:

 

GTX580 --> Gigabyte GTX580 OC

GTX590 --> EVGA GTX590 Classified

GTX680 --> Asus GTX680

 

 

Shader power :

 

GTX580 512 shaders at 2x 1.590 Ghz Shader clock = 1628

 

GTX680 1536 shaders at 1.006 Ghz Shader clock= 1536

 

GTX590 512 shaders at 2x 1.544 Ghz Shader clock x2 (sli) = 3162

 

 

But if higher FPS depends on the amount of Shader processors and the Core Speed my first calcultions should be like :

 

 

GTX580 512 shaders 2x clockspeed ( 0.855 Ghz ) = 876

GTX680 1536 shaders at clockspeed ( 1.006 Ghz ) = 1546

GTX590 512 shaders x 2 clockspeed ( 607 ) x 2 = 1243

 

 

If the last calculations are true than the GTX680 is preferred over the GTX590.

 

However in the SLI topic it shows that with external view SLI is being used.

Then it seems that the extra Shader power comes to use.

 

When not doing things in the cockpit I like spotplane view and use that a lot.

Even when flying a STAR over 3 monitors it is very impressive.

Only many cloudlayer can make the FPS drop till 14-15.

 

Getting that to 20 or more would be very welcome.

[GSalden 4,655]

Here are 2 screenshots to get the idea.

 

The zoom and eypoint are not set yet. So to much fisheye effect.

[dazz 69]

The SLI tests ( see sticky ) show that with external view SLI gives a enormeous boost

 

Yeah, but only in external view. At least when I tested SLI here it didn't do squat for my frame rate in the VC of the NGX and most other VC's in DX9

 

GTX580 512 shaders 2x clockspeed ( 0.855 Ghz ) = 876

GTX680 1536 shaders at clockspeed ( 1.006 Ghz ) = 1546

GTX590 512 shaders x 2 clockspeed ( 607 ) x 2 = 1243

 

these are the right figures:

 

GTX 580: 1581.056 GFLOPS

GTX 590: 2488.320 GFLOPS

GTX 680: 3090.432 GFLOPS (not sure why would this be, this is taken from GPUreview.com but shouldn't it be 1545.216 GFLOPS?)

[GSalden 4,655]

 

 

Yeah, but only in external view. At least when I tested SLI here it didn't do squat for my frame rate in the VC of the NGX and most other VC's in DX9

 

 

 

these are the right figures:

 

GTX 580: 1581.056 GFLOPS

GTX 590: 2488.320 GFLOPS

GTX 680: 3090.432 GFLOPS (not sure why would this be, this is taken from GPUreview.com but shouldn't it be 1545.216 GFLOPS?)

 

Imho you are right about the GFLOPS from the GTX 680.

 

What about a second Gigabyte GTX 580 OC and put those in SLI ?

I am able to buy a second one new for 250 and can put those in SLI.

 

Perhaps the heat will be enormous...

 

My dreamcard : GTX 690. I can buy an EVGA OC version for 780 euro.....

[SAAB340 28]

Yes, the GTX680 is capable of over 3000 GFLOPS alone. You do two operations per clockcycle per shader.

What the 600 series has less of is ROPs or pixel fillrate. As far as I understand the ROPs (or RasterOPerators) are responsible for converting all the texels from all different textures that are angled all aver the place, in to the flat 2dimentional framebuffer. Hence the higher pixel count your display(s) have, the more ROPs or pixel fill rate you need. 1920x1200 or 2.304Mpixels seemed to be ROP limited to just over 20 FPS at high settings with my old G92 8800GTS with a 10.4 Gpixel/Second fillrate.

The ROPs are also tightly connected to the gfx memory so if we are ROP limited we can expect to see an effect by having higher memory bandwidth as well as far as I'm aware.

This might obviously also have an effect for people using multi monitor.

[dazz 69]

Ok,

 

580 = 512 * 772MHz * 2 = 790528

590 = 512 * 607MHz * 2 * 2 = 1243136

680 = 1536* 1006MHz = 1545216

 

double that for the GFLOPS

[GSalden 4,655]

Summary :

 

- The GTX680 gives better FPS in the VC than the GTX580 and GTX590.

- The GTX590 gives better FPS than the GTX680 in external view, but the GTX680 gives better FPS than the GTX580 in external view.

 

I use spotplane view a lot and would welcome 30%+ more FPS.

In the VC better FPS are always welcome.

 

So my solution is the GTX690..... .

Let's see if my wallet also feels that way.LOL

 

I just place a bet of 750 euro for the Asus GTX690.

He asked 850 euro.

[Arklight1 140]

Skip the GTX 690.

 

If you run a very high multi-monitor resolution, just get a 4GB GTX 680 and call it a day.

[GSalden 4,655]

Skip the GTX 690.

 

If you run a very high multi-monitor resolution, just get a 4GB GTX 680 and call it a day.

 

Any particulairy reason for that ?

 

A GTX680 card 2Gb costs about 480-550 euro.

A GTX680 card 4Gb about 620-680 euro.

 

The GTX690 card 4Gb ( second handed ) costs 750 euro .

New it is 999 euro.

 

 

The the GPU Test thread clearly shows that the amount of shaders icw core speed have a direct infuence on the fps with heavy clouds.

The SLO thread shows that in external view SLI is working.

 

So the GTX690 brings a whopping 3072 shader cores for use in external view and 1538 in all other views.

In the cockpit it behaves like a GTX680 , but in spotplane view the SLI comes in play.

 

And that for 100-120 euro more than a GTX680 4 Gb.

[GSalden 4,655]

Palit GTX680 4Gb review on Guru3d :

 

4 Gb, - Realistcally there was no game we tested that could benefit from the extra 2 Gb of graphcis memory.

Triple monitor setups at 5760x1080 ( which is a 6 Mpixels resolution ) , but even there I doubt if 4Gb is really something you'd need to spend money on.

 

It might make a difference at 16xAA and the most stringent games, or if you game in 3D stereo adn triple monitor gaming--I mean sure --at any point graphics memory can and will run out.