Test of the Real Nvidia Gtx Titan, single-sli-tri Sli, 4 WAY SLI
-
Updated with the 4WAY…
That thing that doesn't work well, we're going to have to replan it..eh?..jeje
Salu2.
Joe, what a setup, but of course with 4 way sli, you don't have to play do you??, they must do everything the graphics, you just sit back and watch and the graphics save the world for you, hehe. It must be impressive to see the 4 there with their green lights, hehe.
I had 4 way sli of the 680 shortly after it came out, they were reference, but I didn't get the optimal performance out of them, really with two I already worked as I wanted, obviously I didn't push them as hard as you do in resolutions and stuff, hehe, I have a small 24" LG 3D monitor at 1080 that I rarely put in 3d, now I have an amd 8350 mounted on it and the truth is, I'm afraid to try the titans there in case the pc gets disassembled, I don't know if it will be able to assimilate all the power of the graphics or at least as optimized as intel.
Have you observed the temperature thing I was talking about?? or doesn't that happen to you?
-
Joe, what a setup, but of course with 4 way sli, you don't even have to play, right??, they should do everything the graphics, you just sit back and watch and the graphics save the world for you, hehe. It must be impressive to see the 4 there with their little green lights, hehe.
I had 4 way sli of the 680 shortly after it came out, they were reference ones, but I didn't get the optimal performance out of them, really with two I already worked as I wanted, obviously I didn't push them as hard as you do in resolutions and stuff, hehe, I have a small 24" LG 3D monitor at 1080 that I rarely put in 3d, now I have an amd 8350 on it and the truth is, I'm scared to try the titans there in case the pc comes apart, I don't know if it will be able to assimilate all the power of the graphics or at least as optimized as intel.
Have you noticed the temperature thing I was talking about?? or does that not happen to you?
Yes, indeed the graphics card even if you set the target or the highest temp, doesn't listen. I think it might be a problem with the software for doing OC, that's why it doesn't apply voltages or maintain clocks at high ocs.
As for kinping, there's nothing left to do but drool over his stuff..nothing is stock and everything has a volt-mod, graphics and software specially modified for him by Evga. That's not for mere mortals..;)
Updated with the sleeping..not much improvement…certainly, the 4 Way is going terribly...:facepalm:
-
Yes, indeed the graph, even if you set the target or the highest temp, doesn't listen. I think it may be a problem with the software for OC, that's why it doesn't apply voltages or maintain clocks at high OCs.
As for Kinping, there's nothing left to do but drool over his stuff.. nothing is stock and everything has volt-mod, graphics and software specially modified for him by Evga. That's not for mere mortals..;)
Updated with the sleeping.. hardly any improvement… certainly, 4 Way is going terribly...:facepalm:
Well you see, no matter how much I read around, I see that in the end everyone comes to the same conclusion, just as Nvidia has set up these titans, they are beastly cards in performance, with good overclocking, but they are a "coffee for everyone" since they go up equally because what limits them is not the chip itself no matter how good it is, but the temperature it reaches, there they have cured themselves in health, this will avoid a few rma's, hehe, but the card, once this "traffic jam" is cleared and it will be, either by software or by bios and they can start to go up to see who can stop a beast like this even with the reference cooler that does a good job and not to mention with water…... why put to play crysis with ln2 I don't think even Kingpin would do it, hehe,
... as for the 4way there must be so much power in there that the pc must go crazy, hehe
-
When you run them through water, they don't have to have so many frequency drops, at least with the 680s they don't go higher, but the frequencies, if the consumption doesn't skyrocket, stay as solid as a rock.
-
@josele.:
When you run it through water, it doesn't have to have so many frequency drops, at least with the 680 they don't go up any more but the frequencies if the consumption doesn't skyrocket it stays like a rock.
But it's strange that by putting the fan at high revs it doesn't maintain frequencies, because the cards are usually at 55º, so they also consume less because they are cool.
The normal thing would be that if the GPU is cool it maintains the frequencies better and in these you have to leave the fan on auto and let it get hotter so that the frequency is maintained.
The normal thing is that it will lower the frequency when it's hotter and maintain it when it's cooler.
regards
-
But it is strange that by putting the fan at high revs it does not maintain frequencies, because the cards are usually at 55º, so they also consume less because they are cool.
The normal thing would be that if the GPU is cool it maintains the frequencies better and in these you have to leave the fan on auto and let it rise in temperature so that the frequency is maintained.
The normal thing is that it will lower the frequency when it is hotter and maintain it when it is cooler.
regards
It probably won't be the VRM? That you can assume that even if you ventilate more strongly, with more load on the VRMs for raising the frequencies...
-
3D Mark 11 xtreme…
After Kingpin...the guy..by air,without volt-mod and without chosen graphics..
Kpablo..you're going to have to consider that the 4 Way scales poorly...I'm stuck..
Best regards.
-
3D Mark 11 xtreme…
After Kingpin...the guy..by air,without volt-mod and without chosen graphics..
Kpablo..you are going to have to consider that the 4 Way scales badly...I am anodized..
Regards.
:ugly::ugly::ugly::ugly: CONGRATULATIONS!!!!!!
-
Probably not the VRM? Because you can assume that even if it ventilates harder, there is more load on the VRMs when increasing the frequencies…
I think it's a fan issue, the VRM thing would happen when the GPU is cool, it would happen more when it's hot. The fan on auto when it's almost at 80º is when it holds the frequency better, maybe it detects the fan at max and associates it with being hot and goes into throttling, it could be a fan control error.
Maybe the driver or that Nvidia has capped it so it doesn't go up and stays close to 690.
It could also be some state, this Titan has states similar to Intel's, like C1E, C6, maybe it goes into some state and doesn't maintain the frequency, plus low voltage, it lowers both frequency and voltage at the same time, it could be a state.
The VRM thing doesn't seem right to me, because a hot GPU can consume more than when it's cool and puts more strain on the VRMs.
Regards
-
Mother of god with 4 WAY SLI!!!
If with these drivers they already get that difference to the 680, you could get up to a +70% extra performance, right?
Regards -
With two balls,…..
and green,...with a push of drivers and stability in the OC-Frequency,...watch out if it doesn't hover around 80%
regards
-
Mother of God with 4 WAY SLI!!!
If with these drivers they already get that difference to the 680, you could reach a +70% extra performance, right?Greetings
I don't think so, they can get big differences in demanding games and high resolution, seeing how it pulls Sleeping Dogs at 2560x1600, it's clear that these graphics at high resolution go very well, but in less demanding games and at 1080p they won't get much out of the 680, the drivers must improve, what could really get a lot out of it is the 4way to a 4way of 680, it could get a good margin, especially with multimonitor and high resolution in the most demanding games.
Greetings
-
I don't think so, they can get big differences in demanding games and high resolution, seeing how sleeping dogs runs at 2560x1600, it's clear that these graphics at high resolution work very well, but in less demanding games and at 1080p they won't get much out of the 680, the drivers need to improve, what could really get a lot out of it is the 4way to a 4way of 680, it could get a good margin, especially with multimonitor and high resolution in the most demanding games.
Regards
The architecture is the same as that of the GTX 680, so people should forget about making themselves think of 70-80% extra. In the same way that on other sides it is said that only 20% is obtained compared to other monogpu, let's not fall into exaggerating towards the other side.
The biggest improvements will come from applications very limited by shaders or by bandwidth, and sleeping dogs seems a pretty good case when it is used in it, in addition to all these GPGPU techniques written to the letter for GCN that AMD has so much taste for them (see beta drivers 314.14 from nvidia and their impressive improvement in Sniper V2, proven that it is for this), it uses Supersampling so it is a direct case where bandwidth becomes very important.
But anyway, some extra adjustment may appear, but it will be to maintain at most a 50% difference with the GK104, and thank you. Mind you, average is not the same as performance in a specific test, then what happens happens… :facepalm:
-
3D Mark 11 xtreme…
After Kingpin...the guy..by air,without volt-mod and without chosen graphics..
Kpablo..you're going to have to consider that the 4 Way scales poorly...I'm stuck..
Regards.
That's why I commented "according to the article" because I don't have them to test it myself… :troll:
-
3D Mark 11 xtreme…
After Kingpin...the bastard..by air,without volt-mod and without chosen graphics..
Kpablo..you're going to have to consider that the 4 Way scales poorly...I'm stuck..
Best regards.
What a bestiality of results
-
The architecture is the same as the GTX 680, so people forget about making fools of themselves with 70-80% extra. In the same way that others say that only 20% is obtained compared to other single GPUs, let's not fall into the trap of exaggerating the other way around.
The biggest improvements will come from applications that are very limited by shaders or bandwidth, and sleeping dogs seems a pretty good case when used in it, in addition to all these GPGPU techniques that are written to the letter for GCN that AMD enjoys so much (see beta drivers 314.14 from nvidia and their impressive improvement in Sniper V2, proven that it is for this), uses Supersampling so it is a direct case where bandwidth becomes very important.
But come on, some extra adjustment may appear, but it will be to maintain at most a 50% difference with the GK104, and thank you. Mind you, average is not the same as performance in a specific test, then what happens happens… :facepalm:
Of course it is the same generation, both the 680 and the Titan, you can't ask for miracles, although as you rightly say in specific games that use GPGPU techniques and at high resolution the Titan can get a good margin over the 680.
Then Nvidia is accused a lot of sponsoring games so that they don't run on AMD and now I don't see those complaints, with games putting SS to the max or GPGPU techniques, in addition to the hair thing in Tom Raider, which seems to take quite a bit of performance even on AMD.
Personally, I think it's great that they put in physics to make games more realistic, although I would like it to be something for everyone, but for now the physics I see through DirectCompute don't seem as good as PhysX, even the hair effects in Alice Madness, I think they are pretty good and don't ask for too many resources, besides that it's not just the hair, that game has good physics.
regards
-
Claro es que es la misma generacion,tanto la 680 como la Titan,no se pueden pedir milagros,aunque como bien dices en juegos concretos que usan tecnicas GPGPU y a alta resolucion la Titan si pueden sacar un buen margen a la 680.
Luego se acuso mucho a Nvidia de patrocinar juegos para que no tiren en AMD y ahora no veo esas quejas,con juegos metiendo SS a saco o tecnicas GPGPU,ademas lo del pelo en Tom Raider,que parece quita bastante rendimiento incluso en AMD.
A mi personalmente me parece bien que se metan fisicas para hacer mas realistas los juegos,aunque me gustaria que fuera algo para todos,pero de momento las fisicas que veo por directcompute no me parecen tan buenas como las de PhysX,mismamente los efectos del pelo en Alice Madnes,me parecen bastante buenos y no pide demasiados recursos,ademas que no es solo el pelo ese juego tiene buenas fisicas.
saludos
Es que lo de Tomb ni siquiera es un motor de física, que veo a mucha gente confundida, es simplemente una simulación de pelo y nada más. Es como una pieza de un motor de coche, el radiador por ejemplo. Evidentemente el radiador por sí sólo no es un motor de coche, aunque sea una de las partes necesarias para tener uno.
El caso de Tomb está trayendo cola, por cosas como ésta:
http://www.hardocp.com/news/2013/03/05/nvidia_gpu_performance_issues_in_new_tomb_raider
De lo que me quedo con estos extractos:
Unfortunately, NVIDIA didn't receive final code until this past weekend which substantially decreased stability, image quality and performance over a build we were previously provided.
…
Hi Brent, this isn't solely a driver issue. The developer needs to make some changes on their end to fix their issues on GeForce GPUs. This will require code changes to the game.
...
I myself am experiencing slow performance, (R.E. 20's FPS) with GTX 680 SLI at maxed out in-game settings, with FXAA. There is definitely a performance issue that needs resolving. The only thing you can do right now, if you are experiencing slow performance, is turn TressFX off, back to Normal, and try lowering in-game setting, starting with SSAO, to try and improve performance.
O sea, el juego funcionaba MEJOR claramente con una compilación anterior a la usada en el gold, compilación final que nvidia por supuesto no ha visto hasta el lanzamiento básicamente (gaming evolved en acción). Normalmente cuando haces recompilaciones es para mejorar el rendimiento y arreglar problemas, no para hundir el rendimiento (en el rival de tu patrocinador ¬¬).
Es necesario que el fabricante haga cambios en el código para parchear los problemas (o sea, que hay un punto de ineficiencia fuerte en sea como sea que se gestionan ciertos recursos, apuesto a que temas de GPGPU vista la increible mejora de los 314.14 beta de nvidia con SniperV2, algo a lo que le dedicaré un tiempo en breve).
Y por último resulta que hay que bajar muchos parámetros, no sólo lo de TressFX. Se huele el uso de las "optimizaciones" para PC para reventar el rendimiento ajeno. Esto es lo que AMD entiende por "promocionar las tecnlogías abiertas", la de usarlas de tal forma que sean un boicot efectivo a todo menso a su propio hard.
Dentro de un par de meses todo esto estará arreglado y tal, pero mientras a obtener los réditos del bench integrado en el juego (made for gaming evolved!!!) obteniendo puntuaciones hinchadas en cualquier review donde se les ocurra usar Tomb. Esto ya lleva pasando en varios títulos del 2012, así que no debería sorprender, pero a partir de ahora empezaré a llamarle directamente a las cosas por su nombre.
Boicot, puro y duro.
-
It's that Tomb isn't even a physics engine, which I see a lot of people confused about; it's simply a hair simulation and nothing more. It's like a part of a car engine, the radiator for example. Obviously the radiator alone isn't a car engine, even though it's one of the necessary parts to have one.
The Tomb case is causing a stir, for things like this:
http://www.hardocp.com/news/2013/03/05/nvidia_gpu_performance_issues_in_new_tomb_raider
From what I gather from these excerpts:
So, the game clearly ran BETTER with a build prior to the one used in the gold, the final build that NVIDIA of course hadn't seen until launch basically (gaming evolved in action).
Normally when you do recompiles it's to improve performance and fix problems, not to tank performance (on your sponsor's rival ¬¬).
It's necessary for the manufacturer to make changes to the code to patch the problems (meaning there's a strong inefficiency point in how certain resources are managed, I bet it's GPGPU issues given the incredible improvement of NVIDIA's 314.14 beta with SniperV2, something I'll dedicate some time to soon).
And finally it turns out that many parameters have to be lowered, not just TressFX. You can smell the use of "optimizations" for PC to break someone else's performance. This is what AMD understands by "promoting open technologies", using them in such a way that they are an effective boycott of everything less than their own hardware.
In a couple of months all this will be fixed and all, but in the meantime they'll get the benefits of the built-in benchmark in the game (made for gaming evolved!!!) getting inflated scores in any review where they think of using Tomb. This has already been happening in several 2012 titles, so it shouldn't be surprising, but from now on I'll start calling things by their real name.
Boycott, pure and simple.
Note that I run Tomb Raider at 60 fps with everything enabled (1200p FXAA) with a simple SLI of 480. Sure, when the camera gets close to the hair, the density goes up and the performance drops to 40 fps. But that also happens on AMD. If I remove the hair and High Precision (which isn't noticeable) I can go up to SSAAx2 while maintaining 60 fps all the time. So it's not that it's a horror of optimization or a sabotage of NVIDIA. The thing is that TressFX is something disproportionate at short distances whether you have NVIDIA or AMD. And maybe the low performance of the 6XX series in DirectCompute also comes into play.
I'm sure ELP3 with their Titans is wallowing in the mud with Tomb Raider at whatever resolution they want.
-
Come on, this is what Nvidia has been doing for years and years. What now affects your favorite company and it's no longer good that they send engineers to "help" optimize games for their hardware, what used to be a vision of the future and great for the user, is now terrible. This has always happened, ever since hardware companies started sponsoring software companies and vice versa, and it's always bad for the end user. -
Note that I run Tomb Raider at 60 fps with everything enabled (1200p FXAA) with a simple SLI of 480. Of course, when the camera gets close to the hair, the density increases and the performance drops to 40 fps. But that also happens on AMD. If I remove the hair and High Precision (which is not noticeable) I can go up to SSAAx2 while maintaining 60 fps all the time. So it's not that it's a horror of optimization or a sabotage of NVIDIA. The thing is that TressFX is something disproportionate at short distances whether you have NVIDIA or AMD. And maybe the low performance of the 6XX series in DirectCompute also comes into play.
I'm sure ELP3 with his Titans is wallowing in the mud with Tomb Raider at whatever resolution he wants.
Basically that's it...I can't put more things because I can't..;)
1600P 4XSSAA everything on ULTRA:
The minimum bench scores are not real, they are the typical ones of AMD framebuffers..
Best regards.
