Test of the Real Nvidia Gtx Titan, single-sli-tri Sli, 4 WAY SLI
-
Jotole the input is the power consumed at the outlet and the output is the power delivered to the equipment by the PSU.La the difference is the power lost as heat inside the PSU.I still think that this software does not give correct measurements.
Pepillo I can't tell you about my ASIC right now because I'm not at home, I'm out.About the ASIC we were discussing it at work last year and well the guy in charge of the GPU topic informed us that the ASIC is more or less the quality of the doping depending on how balanced, stable and distributed it is over the core.In the core there are a series of components that must be mixed in percentages per nanometer and this must be as perfect, balanced and exact as possible (homogeneous in certain areas).Obviously this is very difficult to achieve, I refer to the perfection of the doping.Depending on its imperfection the CPU or GPU has a higher or lower drain current, that is, it uses it better or needs more to achieve a speed let's call it X. But here's where it gets complicated, imperfect dopings can cause better performance at certain temperature values, since temperature is fundamental, completely changing the chemical stability of the doping and therefore the performance or consumption (it can be expressed as you want, better use of x nanoamps or more nanoamps to achieve x speed).According to the manufacturers of cores a bad doping gives better performance at X temperature than a good doping.Of course the manufacturer dopes thinking about a temperature range X, if I vary it I break the manufacturer's rule.I call the deviation from the doping rule good or bad but it's really poorly expressed, it's simply different or deviates from what the customer or manufacturer intends and that sometimes, I repeat, is better than what the manufacturer wants.As a general rule perfect dopings improve their performance in the temperature range that the manufacturer determines that core should work.Obviously Titan is designed to work between 75 and 85 degrees and deviated or imperfect dopings can make Titan work better outside of those temperature ranges.But it's also true that higher temperatures cause more drain due to chemical destabilization and if the doping is good this drain is less.
Of course I asked him, then in the graphs that manufacturers release out of the box and improve the temperatures, in theory we are breaking the rules.Of course I already suspected what he was going to answer, they get more speed out of the box by setting the same final temperature as some with less quality and worse heat dissipation.But I asked him again and eVGA in their hydro that lowers a lot the temperature that it mounts, Good or bad chips? (I repeat that they are neither good nor bad just different) His answer was silence he couldn't give me an explanation.That's why I say that the ASIC must have something to do with it, but only to a certain extent.And finally I doubt that the GPU can read the ASIC well, I know that it is engraved at an internal level in the micro when they take them out of the wafers and check them and this is checked with thermography.As far as I know Gpuz reads rom and other values, I think it doesn't have access to certain microcodes of the core, but this is just my assumption -
majo, por curiosidad, ¿Has mirado que ASIC tienen las diferentes tarjetas que has probado, tu que has tenido bastantes? Parece que en las Titan si que se nota un mayor ASIC.
Jotele, pasamos el Unigine así, porque es el preset "Extreme" del benchmark, así es más fácil asegurar que lo que comparamos es lo mismo.
Saludos
Jotole el input es la potencia consumida en el enchufe y el output es la potencia entregada al equipo por la PSU.La diferencia es la potencia perdida en calor dentro de la PSU.Sigo pensando que ese soft no da mediciones correctas.
Pepillo mi ASIC no te lo puedo decir ahora porque no estoy en casa, estoy fuera.Sobre el ASIC lo estuvimos comentando en el trabajo el año pasado y bueno el que lleva el tema de GPU nos informó que el ASIC es mas o menos la calidad del dopaje en función de lo equilibrado, estable y repartido que este sobre el núcleo.En el núcleo van una serie de componentes que deben estar mezclados en tantos por cientos por nanómetro y esto debe ser lo más perfecto, equililibrado y exacto posible( homogeneo en determinadas zonas).Lógicamente esto es muy dificil de conseguir, me refiero a la perfección del dopaje.Dependiendo de su imperfección la cpu o gpu tiene mayor o menor corriente de drenaje,es decir la aprovecha mejor o necesita mas para conseguir una velocidad llamemosla X. Pero claro aqui es donde la mata, dopajes imperfectos pueden ocasionar mejores rendimientos en determinados valores de temperatura, ya que la temperatura es fundamental, variando completamente la estabilidad química del dopaje y por ende el rendimiento o el consumo ( se puede expresar como se quiera, mejor aprovechamiento de x nanoamperios o mas nanoamperios para conseguir x velocidad).Según los fabricantes de núcleos un mal dopaje da mejor rendimiento a X temperatura que un buen dopaje.Claro el fabricante dopa pensando en una margen de temperatura X, si yo lo varío rompo la regla del fabricante.Yo a la desviación de la regla del dopaje lo llamo bueno o malo pero realmente esta mal expresado, simplemente es diferente o se desvia de lo que pretende el cliente o fabricante y que en ocasiones repito es mejor que lo que el propio fabricante quiere.Por regla general dopajes perfectos mejoran su rendimiento en el rango de temperaturas que el fabricante determina que debe funcionar ese núcleo.Lógicamente Titan esta pensada para trabajar entre 75 y 85 grados y dopajes desviados o imperfectos pueden hacer que Titan trabaje mejor fuera de esos rangos de temperatura.pero tambien es cierto que a mayor temperatura mayor drenaje por desestabilización química y si el dopaje es bueno este drenaje es menor.
Claro yo le pregunté, entonces en las gráficas que los fabricantes sacan oceadas de serie y mejoran las temperaturas, en teoría estamos rompiendo las reglas.Claro ya lo suponía lo que me iva a responder, sacan mas velocidad de serie estableciendo la misma temperatura final que unas con menos calidad y peor disipación.Pero yo le volvi a preguntar y eVGA en sus hidro que baja mucho su temperatura que monta, Chips buenos o malos? ( repito que ni son buenos ni malos simplemente diferentes) Su respuesta fue la callada no supo darme explicación.Por eso digo que el ASIC si tiene que ver, pero hasta cierto punto.Y por último dudo que el GPU puede leer bien el ASIC, yo sé que queda grabado a nivel interno en el micro cuando los sacan de las obleas y los comprueban y esto lo comprueban con termografia.Gpuz que yo sepa lee rom y otro valores, pienso que no tiene accseo a cieros microcodigos del núcleo, pero esto es simplemente suposición míaA mí tambien me extrañan esos consumos, pero ya son 2 "medidores", los que me dan una cifra parecida, intentaré contrastar con otro metodo..
Este es el cuadro que decía antes que vi, se puede sacar directamente con el gpu-z.
-
¡Esta publicación está eliminada! -
Majo, piece of explanation. I've got the concept, but I lack technical wisdom...
I haven't had time to flash it yet, I have it pending. What I will also do, if it's not too much trouble for you, is to send you my BIOS so that you can leave it at 1.16 fixed.
Auer I got home late, today I'm working, but, even though I got home late I ran a test on it. The unigine crashed at 1110 and 1.15, after 20 minutes. However at 1097 it didn't. This with 1.15. It insists on going down to 1.15. If the consumption is not very high and the GPU is cool it gets stuck at 1.16. There are games that even though the GPU usage is at 99% the power target is not at the limit. I imagine it's because of the different use that the game makes of the graphics.
The biggest hog, by far, is Crysis 3. With it, if I put more than 1032 of the core it goes into throttling. At 1032 it stays stuck.
About the ASIC, well according to what Majo says, well what I interpret is that you may have an ASIC that's not very good and that it has a good overclock when the temperature is contained.
Jotole has a pretty good ASIC, and, even so, he doesn't have it at full capacity out of the box.
Mine is from a 66.9. A bit of a chestnut.
Jotole, I think they come capped because if not nobody would get the 690. That's why e, maybe, for some other reason...
-
I just bought the backplate for the Titan from eVGA, this one:
I'll tell you how it goes, I had it on the GTX 295, and on the Titan that has the memory exposed from the back, I thought it was a good addition. At the very least it looks much nicer
Regards
-
Majo, I've tried the bios you sent me.
The first one, the most aggressive, raises the voltage without understanding why up to 1.21 and 1201 boost (the modded one from the SC). It's a disaster.
Then I tried Naennon's ones, specifically the 1150 and 1.16. In theory it's at 1.16, but it raises the voltage up to 1.21 and 1150. It's a disaster.
Then I modded one myself, setting the voltage to 1.16 and 115% power target. Here's the curious thing: I'm not sure if it's my power supply: the graph stays stuck at 1.16, but the clocks oscillate, even when I set the power target to the maximum. The clocks go down, but the voltage stays at 1.16.
I'm going crazy.
-
Majo, I've tested the bios you sent me.
The first one, the most aggressive, raises the voltage without understanding why up to 1.21 and 1201 boost (the modded SC). It's a disaster.
Then I tested Naennon's ones, specifically the 1150 and 1.16. In theory it's at 1.16, but it raises the voltage up to 1.21 and 1150. It's a disaster.
Then I modded one myself, fixing the voltage to 1.16 and a 115% power target. Here's the curious thing: I'm not sure if it's my power supply: the graph stays stuck at 1.16, but the clocks oscillate, even when I set the power target to the maximum. The clocks go down, but the voltage stays at 1.16.
I'm going crazy.
The clocks may be going down either because of temperature, or because that power target is too low for that frequency.
What temperatures are you getting on the card, cooling?, power supply?.
Regards…
-
The clocks may be lowering either because of temperature, or because that power target is too low for that frequency.
What temperatures do you have on the card, cooling?, power supply?.
Salu2…
Hello Jotole. Let's see, I flashed with Naennon's, the one with 1150 and 1.16. I started unigine and it went to 1202 and 1.21.
The one from Majo, same, it does weird things to me. I start it and it goes up to 1202, then it goes down.
The temperatures on the card are normal, 80-82, stock cooling (case open), and the power supply is a thermaltake thoughpower of 1000
I've tried several bios and nothing. I don't know why in Crysis 3 with the one I had modified it stayed at 1.16, which is what I was looking for, but it kept going down to 1071…1058...1006 and so on. The consumption was seen to be skyrocketing up to 95-98 when the Power target was at 115%, this last both in my cutrebios and in the ones from Naennon and Majo.
-
Hello Jotole. Let's see, I've flashed with Naennon's, the one with 1150 and 1.16. I start the unigine and it goes to 1202 and 1.21.
Majo's one, same thing, it does weird things to me. I start it and it goes up to 1202, then it goes down.
The temperatures on the card are normal, 80-82, stock cooling (case open), and the power supply is a thermaltake thoughpower of 1000
I've tried several bios and nothing. I don't know why in Crysis 3 with the one I had modified it stayed at 1.16, which is what I was looking for, but it kept going down to 1071…1058...1006 and so on. The consumption was seen to be skyrocketing up to 95-98 when the Power target was at 115%, this last thing both in my cutrebios and in Naennon and Majo's ones.
When you start the unigine, you monitor it with AB, or the Precision?, if so, I don't know why those tools don't pick up the modified bios well.
Reset to default values with those programs, and try to run unigine with them reset, or uninstalled and reinstalled. They did weird things to me at the beginning too.
-
When I restart the computer after the modified bios, the EVGA Preccision is in Default. I leave it as is.
I start unigine and it goes to 1201 and 1.21 volts. I don't understand why when, in theory, it is the bios of 1150 and 1.16. Also, even with this bios, I enter Crysis 3 and it goes into throttling, but only the frequencies, not the voltage.
Should I uninstall the Precission? What strange things did it do to you?
-
When I first started, with the modified bios, it did that to me, it went to the maximum, even though the EVGA was on Default. You have to manually reset the values, so that it lowers the clock to the "normal" ones so that they don't go up to 1200...
-
I just bought the backplate for the Titan from eVGA, this one:
I'll let you know how it goes, I had it on the GTX 295, and on the Titan which has the memory exposed at the back, I thought it was a good addition. At the very least it looks much nicer
Best regards
It's beautiful Pepillo, but if you can put some pads on the memory areas, like the ones that come with the backplates of the rl blocks, you'll be appreciated... ;).
Cheers...
-
Nothing Jotole. I've flashed it again, I've uninstalled the EVGA, I've reinstalled it and nothing, it goes to 1150 and 1.21 with throttling.
-
Nothing works. I've flashed it again, uninstalled the EVGA, reinstalled it and nothing, it goes to 1150 and 1.21 with throttling.
Well that's very strange, is that your bios? what target does it have? let's see if the graphics card isn't getting the power it needs, or you're not controlling the temperatures, you have to keep it below 80º.
Let's see if Pepillo comes by, he has it air-cooled and will tell you what profile he's using for the fan. He doesn't have throttling and has a Naennon bios, which is more aggressive than the one majo sent you
Salu2..
-
I bet you that you have the problem in the evga precision.Es it is impossible that any of mine will boost you to that if you leave them like that cual.De I throw the second one is an original SC and if you don't boost it with evga or other soft they will never boost.Open the precision, click default, close it, uninstall the evga precision, start the pc, run a registry cleaner, start the gpu-z and the unigine and tell me if like this they boost you to 1201, I assure you that it's not BIOS es.La the first one is at 1,17V when it goes above 1075, below it should stay at 1,16V. And the SC never goes above 1,16V. So listen to me, look for the problem in the evga.Si even so it won't work, uninstall nvidia drivers, run sweepdriver and reinstall, something catches them and it's not BIOS, I repeat
-
Tell me a joke about a coca cola sign that you have the problem in the evga precision.Es it is impossible that any of mine will boost you to that if you leave them like that cual.De I do the second one is an original SC and if you don't boost it with evga or other software it will never boost.Open the precision, click default, close it, uninstall the evga precision, start the pc, run a registry cleaner, start the gpu-z and the unigine and tell me if it boosts you to 1201, I assure you that it's not the BIOS es.La first one is at 1,17V when it goes above 1075, below it should stay at 1,16V. And the SC never goes above 1,16V. So listen to me, look for the problem in the evga.Si even so it won't work, uninstall nvidia drivers, run sweepdriver and reinstall, something leaves them caught and it's not the BIOS, I repeat
That's what I'm trying to tell you, it happened to me as soon as I started the computer once assembled. That they went alone to 1200.
Of course I haven't told you anything about cleaning the registry or driver sweeper….........:ugly:, my fault, I did it myself....
Even though it's not always necessary I always do it when I change drivers, even if they are updates, and also I do a correct installation, so that it doesn't pick up any profile, lost in the registry.
-
Nothing, I give up. I've done the same thing again, I've uninstalled drivers, I've run driversweeper, reinstalled and it goes to 1202. I've messed with the bios, leaving the values at 1.16 (both), and nothing, it goes to 1.21 voltage.
GPUZ shows the same.
The P00 and the other value must be in a different form?, the P00 must be equal to or greater than the P02?
It's the only thing I changed with the Kepler bios tweaker.
-
Nothing, I give up. I've done the same thing again, I've uninstalled drivers, I've run driversweeper, reinstalled and it goes to 1202. I've messed with the bios, leaving the values at 1.16 (both), and nothing, it goes to 1.21 voltage.
. Desistala all trace of EVGA or AB, whichever you use. Clean with Ccleaner. Start a game or bench and monitor with this openhardwaremonitoring.
Leave it running in the background, to see the frequency you reach.
As a last resort try to put the original bios that came with your graphics card.
Salu2…
-
¡Esta publicación está eliminada! -
Set the evga to low offset and click apply and test to see if the frequency decreases.
Read the rom with the gpux and check that it is the one you want to put in, to check if you have flashed it correctly.
