1080 vs 1440 graphics performance - SLI 470 vs SLI 480
-
And what conclusion have you come to? I mean, for example, does this performance difference also occur for resolutions lower than 1080p compared to 1080p? If not, we could conclude that it is a memory problem due to the resolution used? Because we must remember (if my memory doesn't fail me;D) that the amount of memory available in an SLI for textures is that of a single card.
PD: A table with the comparative results would be very good
-
Praimus on high resolution monitors, when you put a 1080 it loses some performance if you compare it with a native 1080, it has to represent the game and the desktop, but well I don't think it's much difference and both configurations would be equal, especially if both were passed on the same monitor and winmode.@krampak:
¿ And what conclusion have you reached? I mean, for example, at resolutions lower than 1080p this performance difference also occurs compared to 1080p? If not, we could conclude that it is a memory problem due to the resolution used? since we must remember (if my memory doesn't fail me ;D) that the amount of memory available in an SLI for textures is that of a single card.
I think the conclusion is that at 1440p the 470 are weak due to vram, but in addition the 480 are more powerful due to the amount of SP and the larger data bus that should also make them hold up better at high resolution
regards
-
If you compare it to a native 1080, it has to render the game and the desktop, but well I don't think it's much of a difference
In windowed mode I don't think it will make a difference.
I think the conclusion is that at 1440p the 470 struggles with vram, but also the 480 are more powerful due to the amount of SP and the larger data bus that should also make them hold up better at higher resolution
But it's logical to think that this happens, right?
-
¿ Y a qué conclusión has llegado? Me refiero a que, por ejemplo, a resoluciones inferiores a 1080p también ocurre esta diferencia de rendimiento respecto a 1080p? Si no es así podríamos llegar a la conclusión de que se trata de un problema de memoria debido a la resolución utilizada? ya que hay que recordar (si no me falla la memoria ;D) que la cantidad de memoria disponible en un SLI para las texturas es la de una sola tarjeta.
La conclusión ha sido a que a mayor resolución se necesita mas potencia grafica, ¿cuanto? pues hay estan las pruebas para mas o menos hacerse una idea.
Antes de quedarte sin Vram te quedas sin potencia, salvo casos excepcionales 1.5g de Vram es mas que suficiente.
@javete:
Praimus en monitores de esos de alta resolucion,al poner un 1080 se pierde algo de rendimiento si lo comparas con un 1080 nativo,tiene que representar el juego y el escritorio,pero bueno no creo que sea mucha diferencia y las dos configuraciones estarian en igualdad,sobretodo si se pasaron las dos en el mismo monitor y modo winmode.
saludos
Hola compañero, las pruebas a 1080 están pasadas con mi monitor P2770HD por supuesto un 1080 nativo y las pruebas pasadas a 1440 con mi Dell 1440 un 1440 nativo.
Un saludo.
En modo ventana no creo que influya.
Pero es lógico pensar que eso ocurra no?
El modo ventana influye al rendimiento en cuanto a ventana completa.
Un saludo.
-
In windowed mode I don't think it has an impact.
In some things it does have an impact, in 16/10 monitors I think it ends up scaling to 1200p and especially in Unigine we already tested that using native monitor 1080 performs a bit better 4 fps or 5 I think.
although it was a 2560x1600 monitor and there is some difference between putting 1080p on that monitor or putting a native 1080p one.
maybe ELP3 still saves the captures.
regards
-
@javete:
Praimus on high resolution monitors, when you set a 1080, you lose some performance if you compare it to a native 1080, it has to render the game and the desktop, but well I don't think it's much of a difference and both configurations would be equal, especially if both were passed on the same monitor and winmode.
I think the conclusion is that at 1440p the 470 lag due to vram, but also the 480 are more powerful due to the amount of SP and the larger data bus that should also make them hold up better at high resolution
regards
I base this on the tests I did back in the day, and in those tests the 470 didn't run out of Vram, it's true that nowadays I've come across a couple of games that I have to lower to 4x because the 3waySLI of 480 runs out of vram, but the difference in graphic performance of a sli of 470 vs trio of 480 is very large.
As Elp3 once said, you shouldn't give importance to the Vram issue if it's not a case of multimonitor and 4 cards.
-
I base myself on the tests I did back in the day, and in those tests the 470s didn't run out of Vram, it's true that nowadays I've come across a couple of games that I have to lower to 4x because the 3waySLI of 480 runs out of vram, but the difference in graphic performance of a sli of 470 vs trio of 480 is very big.
As Elp3 once said, you shouldn't give importance to the Vram issue if it's not a case of multimonitor and 4 cards.
If I don't give importance, but it's clear that lately the games that put something more of SSAA consume more Vram, like sleeping dogs, Hitman, although they would be in configurations of 3 cards, with one or two it will be very difficult for vram to run out because before it runs out of power, but if they continue to apply SSAA to the games it increases the consumption of Vram, I also put in MOHW that one also consumes a lot of vram, although perhaps it's a game bug.
That said, I don't worry about vram to play at 1920x1080, it will be noticed a little at higher resolutions and certain new games.
I also told you back in the day that the 480s would be noticed at your resolution, not just for vram, they are more powerful for those resolutions.
What would be good to check is if it influences in Unigine that the monitor is 16/10 or 16/9, perhaps in a 16/9 you don't notice a difference in putting 1440p or 1080p.
regards
-
@javete:
If I don't give it importance, but it's clear that lately the games that put in a bit more SSAA consume more Vram, like sleeping dogs, Hitman, although they would be in configurations of 3 cards, with one or two it will be very difficult for Vram to run out because before it runs out of power, but if they continue to apply SSAA to the games it increases Vram consumption, I also put in MOHW that it also consumes a lot of Vram, although perhaps it is a game bug.
That said, I don't worry about Vram to play at 1920x1080, it will be noticed a little at higher resolutions and certain new games.
I also told you back in the day that the 480s would be noticed at your resolution, not just for Vram, they are more powerful for those resolutions.
What would be good to check is if it influences in Unigine that the monitor is 16/10 or 16/9, perhaps in a 16/9 you don't notice a difference in putting 1440p or 1080p.
regards
The games you mention are current and I wouldn't know what to tell you, as I say I'm starting to notice a lack of Vram but with three 480s and at 1440, but they are two games, the MOHW and the other you'll forgive me but I can't remember right now jejeje.
Yes, in this test it makes clear what you told me back in the day and for which many people labeled me as crazy and it is nothing more or less than the 480s are more powerful than the 470s at higher resolution, I say this because for a while it was said that the 470 at 800mhz rendered like a series 480 and as can be seen in these tests that is not the case.
Sorry, let me see if I understood what you propose correctly, do you want me to run Unigine on the Dell at 1080 and then run it on the Samsung at native 1080 to see if there are differences? If so, it can be tested and we can clear up doubts.
regards.
-
The games you mention are current and I wouldn't know what to tell you, as I'm already starting to notice a lack of Vram with three 480s and at 1440, but they are two games, MOHW and the other one, I'll forgive you, I can't remember right now jejeje.
Yes, this test makes it clear what you told me back then and for which many people called me crazy and it's nothing more or less than the 480s being more powerful than the 470s at higher resolutions, I say this because for a while it was said that the 470 at 800mhz performed like a stock 480 and as you can see in these tests that's not the case.
Sorry, let me see if I understood what you propose correctly, do you want me to run an Unigine on the Dell at 1080 and then run it on the Samsung at native 1080 to see if there are differences? If so, we can test it and clear up any doubts.
Best regards.
No, what I was referring to was comparing Unigine at 2560x1600, 2560x1440 and 1080p all at 1920x1080, to see if the monitor has something to do with it or if it's a matter of 16/10 that in reality passes it to 1920x1200.
because a while ago ELP3 already did it on his 2560x1600 monitor and when he switched to a native 1080 monitor it gave him something more, he probably has all three resolutions, so we could clear up doubts about whether he loses because of the 16/10 or if it's also noticeable in a 16/9.
I've noticed that sometimes in windowed mode some games seem to perform worse, but your tests are good and they're all on the same monitor so there's equality.
I mention Unigine because it's tessellation and there's not much difference due to OC, if it's run on the same equipment and same frequencies it shouldn't vary much.
regards
-
Hello and thanks for the Praimus test.
I think we need to demystify a bit the issue of Vram. Firstly, because in Win7 it is impossible to know exactly how much is being consumed. Programs like Afterburner measure more the "allowed" memory located than the used one, and in addition, it adds the one that the system itself is eating which in Win7 is variable, therefore as I say we can make approximations but we can never say this game spends X Vram because it is impossible.
The normal thing is that a graphics card runs out of power much earlier than out of Vram. Even in multi-gpu configurations. There are very exceptional cases, like the dual AMD 5970 that only had 1GB of real Vram, which if they have extra power to move more, but they fall short of Vram in some specific cases, but it is the exception that confirms the rule.
It is also true that recently very demanding Vram games have come out, like BF3, the new MOFH, or games like Hitman Ansolution or Sniper Elite, which due to the somewhat indiscriminate and "marketiniano" use of MSAA have a very high cost, especially the last two that I mention.
But we are talking that for example to exhaust 1.5GB of Vram, it would take at least 3 graphics in TRI SLI like the GTX 480 of Praimus, a minimum resolution of 1440p, and those games mentioned at maximum or others with mods to exhaust Vram before the available power. Speaking always of course, of a search for frames good.No to the case made to exhaust the Vram and that the playability is disastrous...
But everything that has been said before I think it is extrapolable to that a single graphics card is not going to run out of Vram before running out of power, and many times people confuse lack of optimization of the game, drivers etc.. with lack of real Vram. A total lack of Vram produces a total collapse of the game leading it to 0 fps and being absolutely unplayable, until the buffer of the graphics card is filled again and you can play normally again... although again the collapse will occur again soon and we will enter a loop.
Best regards.
P.D.
On the subject of resolutions, indeed a 16:10, as it is not a multiple of 16:9 or 1080p some benchmarks are not totally reliable since they make a mess when passing resolutions 16:9 in 16.10. To make 1080p in a 16:10 you need to put black bands at the top and bottom, the Unigine on the contrary does not do it and fills the screen by elongating it, this translates into that although you select 1080p in the menu, the screen really tells you that it is passing to 1200p, and the results are a little worse than a 16:9, because obviously it passes it to more resolution. I recently did tests with the Unigine on my Dell U3011 of 16:10 and a native 1080p screen with a 660Ti, and indeed it gave 3 or 4 more frames on the 1080p screen with the same options and resolution due to the previously mentioned. I do not know if this is solved with some Unigine update or patch.. but what javete comments is true..
