First tests of Nvidia's 980 and 970
-
Very interesting those Gigabyte, although their 31cm is not for everyone, having 2 with 8 pin connectors gives me the feeling of having more OC potential than the other customs.
Are they mounted on socket 2011 or 2011-3?
Regards.
They are reference ones Patagonico, I can't mount 4 customized ones.
First contact.
Unigine 1080p everything to the extreme and 8 layers of aa 1080p and an OC close to 1500MHZ with 0,5 v of extra voltage..
Not bad at all the result.
Regards.
uff, I made a mistake, I didn't put ultra in quality…then I do it..;) the lack of custom..jeje
-
Unigine doesn't seem to be its strong point, according to reviews, it runs like a 780Ti or close to it but it's not where they excel, you can see that tessellation no longer increases even though it's the performance chip, which isn't bad either, the big one should definitely stand out clearly.
What I find interesting is the DSR thing, I don't think that can be done with Kepler.
Regards
-
Patagonico is the reference, I can't mount 4 customized ones
Oh I didn't see any Gigabyte reference, even on their pages the only ones published are the custom ones
http://ar.gigabyte.com/products/list.aspx?s=43&p=52,53@53@53,62&v=1,8@1@2,29
Regards.
-
Ostia I didn't see any reference to Gigabyte, even on their pages the only ones published are the customs
http://ar.gigabyte.com/products/list.aspx?s=43&p=52,53@53@53,62&v=1,8@1@2,29
Saludos.
This is although maybe not even gigabyte has it on their website, they will have several websites, they will have it on the main one
Gigabyte GeForce GTX 980 4GB GDDR5 GV-N980D5-4GD-B Graphics Card
For SLI that cooler is very good.
Saludos
-
This is although perhaps not even a gigabyte on its website, it will have several websites, it will have it on the main one
Gigabyte GeForce GTX 980 4GB GDDR5 GV-N980D5-4GD-B Graphics Card
For SLI this cooler is very good.
Regards
Nice it reminds me of the Titan, I suppose the GTX 970 will be the same pointed design to keep it in mind.
Nvidia GeForce GTX 980 SLI / 3-way SLI / 4-way SLI review
http://uk.hardware.info/reviews/5623/nvidia-geforce-gtx-980-sli–3-way-sli--4-way-sli-review
Regards.
-
Bonita me recuerda a la Titan, supongo que la GTX 970 será mismo diseño apuntada para tenerla en cuenta.
Nvidia GeForce GTX 980 SLI / 3-way SLI / 4-way SLI review
http://uk.hardware.info/reviews/5623/nvidia-geforce-gtx-980-sli–3-way-sli--4-way-sli-review
Saludos.
Quiza con la 970 sea mucho mas difícil de ver ese disipador, la 770 no vi ninguna con el, solo al principio alguna reviews, vi alguna de referencia de MSI y es malo parecido al de 680, no se si veremos alguno, debe ser caro y no se lo ponen.
Las que veo aquí 970 ninguna tiene ese, la palit tiene uno de referencia pero malo.
http://www.aussar.es/tarjetas-graficas/saludos
-
Well, let's see, I have already passed it under conditions.
I preferred to use the MSI, since I am more familiar with it and you can increase the final boost frequency more, it has been around 1506MHZ.
I think the result is excellent.
Why? Because I used this test on purpose, since I know it is not their strong point, and even though it has been ages since I benchmarked this, and it is possible that there has been an improvement in drivers, I would be almost sure that a Titan or 780Ti should go above 1375MHZ to match/surpass this score. With the difference that those, especially the TITAN, would need to be done with modified bios, voltage up to the hilt and RLS. While here you take a reference card, you give it 100% to the fan, which does not pass in the 65º temperature test, you barely put voltage (I will explain the voltage issue later, more voltage is worse) and you don't change the bios, nor RLS nor anything.
I do not justify a change for anything, not even from 780 for this, but it has to be recognized that nvidia has done a great job with this thing... because it performs amazingly and consumes, already in these overcloks, like 150W less than the TITAN to give a similar score.
Best regards.
-
I already have them around here.
The TITAN are monsters, but it must be acknowledged that they haven't gone very far... I'm having problems with them and I haven't put voltaje.De on one for ages. In fact, I have no choice but to do an RMA on one already, because it's more broken than working.
That's one of the main reasons.Si. If all 4 were working at their full potential, I wouldn't even bother trying them out... but well, it's already impossible to find TITANs and besides, the price would be absurd.
I'll upload something later.
bye.
Well, it was already strange that you weren't interested in taking a look at graphics cards like these given that they seem to have come out so curiously, even though you had been a bit disconnected for a while. :troll:
I think you'll like them quite a lot even though at first glance you might see them as "weak" graphics cards against the more seemingly robust Titan (more VRAM, more bandwidth, more more... ), the general impression I've been left with from so many reviews and cameos with the little maxwell is that nvidia had found a way to greatly improve the original "mojo" of the kepler, to make them better where they were weakest and improve a little in everything else.
I already told you that the numbers aren't good companions to judge them, let's see how these ones go, but I think they'll surprise you more than disappoint, even though you're already used to similar levels of performance.
-
If you set that as the reference, the custom ones should go higher for having better phases, which seems like GPU-Z is wrong, it says 32 Rops and they are 64 I think, I guess it's a mistake.
They don't look bad because with 4 of those you won't have a power supply problem, nor will you need RL, I think we'll see a lot of SLI of these because you shouldn't need much power supply nor does it seem like the 256bits will be much of a bottleneck.
Also now you have to take into account that it still gets hot, now you can't watercool like in winter and even less with stock cooling, now the phases and the GPU get hotter, but these cards aren't bad, I'm going to try a 970 and retire the 480s that have already served their time, I'm not switching to a 780, I'll keep both.
Regards
-
Well, let's see, I have already passed it under conditions.
I preferred to use the MSI, since I am more familiar with it and you can increase the final boost frequency more, it has been around 1506MHZ.
I think the result is excellent.
Why? Because I used this test on purpose, since I know it is not their strong point, and even though it has been ages since I benchmarked this, and it is possible that there has been an improvement in drivers, I would be almost sure that a Titan or 780Ti should go above 1375MHZ to match/surpass this score. With the difference that those, especially the TITAN, would need to be done with modified bios, voltage up to the hilt and RLS. While here you take a reference card, you give it 100% to the fan, which does not pass the 65º temperature test, you barely add voltage (I will explain the voltage issue later, more voltage is worse) and you don't change the bios, nor RLS nor anything.
I do not justify a change for anything, not even from 780 for this, but it must be recognized that NVIDIA has done a great job with this thing... because it performs amazingly and consumes, already in these overcloks, like 150W less than the TITAN to give a similar score.
Best regards.
ELP3 leaving the fans below 75% what temperature do the graphics reach?
Will you try some game benchmarks to see the SLI 3-4 way scaling?
Best regards.
-
ELP3 leaving the fans below 75% at what temperature do the graphics reach?
Will you try some game benches to see the 3-4 way SLI scaling?
Best regards.
It's just an OC test. So I don't know..
The profile is superconservative, and they are quieter than any previous GTX that already eran.Es possible that below 75% it doesn't increase more than 4 or 5 º…
I'm going to set up the 4 Way. But first I'll see which one is the best to make master..
Best regards
-
It's just an OC test.So I don't know..
The profile is super conservative,and they are quieter than any previous GTX that already eran.Es possible that below 75% it doesn't increase more than 4 or 5 º…
I'm already going to set up the 4 Way.But first I'm going to see which one is the best to make master..
Regards
Don't forget to remove the mini-lid from the backplate of the ones that form a "sandwich".... it's supposed to make them breathe better in tight SLIs.
-
Don't forget to remove the mini-lid from the backplate of the ones that form a "sandwich"…. it's supposed to make them breathe better in tight SLIs.
I don't know if I should leave it..
That's a backplate, but it's true that it accumulates a lot of calor.Si if I see that they are going bad, I remove it.
Best regards.
-
It's just an OC test.So I don't know..
The profile is super conservative,and they are quieter than any previous GTX that already eran.Es possible that below 75% it doesn't increase more than 4 or 5 º…
I'm already going to set up the 4 Way.But first I'm going to see which one is the best to make master..
Regards
Do you test the OC well because for games there are times when the GTX 980s are mounted on the source? as for the reduction of consumption do you calculate that they should be around in the whole equipment?
Regards.-
-
What a powned you've got, right?, after saying against all odds that they were slower than the 780/Ti… :ffu::ffu:
Zap! That's the sound of that. 7.5% faster than the Ti, at 4K. The same as it beats at 1080p. And cheaper, by the way.
And with OC it gains more performance, out of the factory tight, nothing:
Given that it gains 14.6% with OC (a good OC anywhere in the world of GPUs), that multiplies the previous average result of 107.5% of a Ti, this is:
123% of the performance of a GTX 780 Ti (yes, and I refuse to explain that percentages do not add up, who does not know why should review their math).
Where are those cards that were going to be slower? Ayayayyy….
Sorry to tell you that from stock it is faster yes, has opted for the 2048 ccs to maintain consumption and raise clocks like crazy to be able to pass it, the problem you find when you see this:
http://i1145.photobucket.com/albums/o504/Gibbo205/HeavenExtreme.jpg
Same clock, 980 slower, voltage of the 780 ti 1.062v for the 1200 mhz, mine of course, they have to go up a lot the 980 in difference with the 780 ti to compensate for the power, so no kid, it is not more powerful in general terms, they have wanted to justify the 550 pavos giving as novelty its reduced consumption and speed.
And do not come with the excuse of different architectures, not to compare the mhz, in high end you look for the best performance, you squeeze everything you can/let, and seeing the average of oc in the 980 according to all the reviews I have seen I place them at 1400 mhz, the average in 780ti is at 1300 more or less.
The issue of 4k you should focus it from the point of view of the color compression that they have put and the anticipated suppression of the hidden pixels of Z (early z-culling), if not so possibly at more resolution it would come down due to the bus, since they increase the effective bandwidth and improve the general rendering performance.
For me a scam removing the consumption, we have not gained anything in performance.
Regards.
-
Well, let's see, I have already gone through it under conditions.
I preferred to use the MSI, since I am more familiar with it and you can increase the final boost frequency more, it has been around 1506MHZ.
I think the result is excellent.
Why? Because I used this test on purpose, since I know it is not their strong point, and even though it has been ages since I benchmarked this, and it is possible that there has been an improvement in drivers, I would be almost sure that a Titan or 780Ti should go above 1375MHZ to match/surpass this score. With the difference that those, especially the TITAN, would need to be done with modified bios, voltage up to the hilt and RLS. While here you take a reference card, you give it 100% to the fan, which does not pass the 65º temperature test, you barely add voltage (I will explain the voltage issue later, more voltage worse) and you don't change the bios, nor RLS nor anything.
I do not justify a change for anything, not even from 780 for this, but it must be recognized that nvidia has shone with this thing... because it performs amazingly and consumes, already in these overcloks, like 150W less than the TITAN to give a similar score.
Best regards.
That is a good point, to reach, click, increase voltage and perform very very well with low temperature and low consumption.
Totally agree with you on the rest XD.
-
I don't know if I should leave it as it is...
That's a backplate, but it's true that it accumulates a lot of heat.If I see that they are going bad, I'll remove it.
Best regards.
Mmm… I don't know if I've been understood, just in case, I'm referring to this small part of the backplate:
Which is removed by taking out the screw there and pushing the piece outwards:
Nvidia says that with this minimal change, they can keep a backplate without hardly noticing the extra obstruction it might make in case of very close SLIs, since that area must give the margin for there to already be enough airflow towards the blower of the card that is right above the card from which that piece of the backplate is removed.
It's a pretty curious detail, possibly no one thought that at Nvidia they would be giving so much thought to the problem of using backplates with "close" SLIs, lol :ugly:
-
Mmm… I'm not sure if I was understood, just in case, I'm referring to this small part of the backplate:
Which is removed by taking out the screw there and pushing the piece outwards:
Nvidia says that with this minimal change they can maintain a backplate without hardly noticing the extra obstruction it could make in case of very close SLIs, since that area must give the margin for there to already be enough air current towards the blower of the card that is right above the card from which that piece of the backplate is removed.
It's a pretty curious detail, possibly no one thought that at Nvidia they would be giving so much thought to the problem of using backplates with "close" SLIs, lol :ugly:
It's a plus for Nvidia, although we would have to compare with and without the backplate XD. Elp3 the work is piling up for you!!! ;D
I take this opportunity to leave this branch of overclockers where many models with their main features appear, if not all:
http://forums.overclockers.co.uk/showthread.php?t=18624822
Wouldn't it be necessary to open an official Maxwell branch? Let's see if we can all work on it together and leave the info clearer.
Best regards.
-
This post is being processed/translated. The original version will be shown:
Siento decirte que de stock es mas rapida si, ha optado por los 2048 ccs para mantener el consumo y subir clocks como cabrones para poder pasarla, el problema te lo encuentras cuando ves esto:
http://i1145.photobucket.com/albums/o504/Gibbo205/HeavenExtreme.jpg
Mismo clock, 980 mas lenta, voltaje de la 780 ti 1.062v para los 1200 mhz, las mias claro, mucho tienen que subir las 980 en diferencia con las 780 ti para compensar la potencia, asi que no chico, no es mas potente en terminos generales, han querido justificar los 550 pavos dando como novedad su reducido consumo y velocidad.
Y no me vengas con la escusa de diferentes arquitecturas, para no comparar los mhz, en high end buscas el mejor rendimiento, las exprimes todo lo que puedas/dejen, y viendo la media de oc en las 980 segun todas las review que he visto las situo en los 1400 mhz, la media en 780ti está en 1300 mas menos.
El tema del 4k deberias enfocarlo desde el punto de vista de la compresión del color que han metido y la supresión anticipada de los píxeles ocultos de Z (early z-culling) , de no ser asi posiblemente a mas resolucion se vendria abajo debido al bus, ya que aumentan el ancho de banda efectivo y mejoran el rendimiento general de renderizado.
**Para mi un timo quitando el consumo, no hemos ganado nada en rendimiento.
**
Saludos.Hay que tenerlos bien grandes para llamar a esto "timo", cuando se están ofreciendo niveles de rendimiento superior por bastante menos pasta que antes, y con una serie de mejoras extras. Que me timen así todos los días, por favor.
¿Cómo que es una eXcusa lo de hablar de diferentes arquitecturas? ¿crees que son iguales las interioridades de cada cpu? ¿hacemos el absurdo de comparar voltajes y frecuencias entre Fermi y Kepler para valorar una en función de la otra, sí (qué bien para kepler, forzar a las Fermi a bajar frecuencias hasta que sus CC funcionen a la misma frecuencia que kepler, victoria doble)?
Algunos contáis calorías pero no conocéis ni el concepto que hay detrás (es bonito el usar ciertos conceptos simples pero engañosos, uno de éstos es contar "unidades de ejecución" que ni siquiera son tales al ser parte de una única unidad SIMD), pero no sabéis qué es lo que importa realmente en este tipo de arquitecturas, la unidad MINIMA de ejecución de un shader NO son los CUDA core, son los SMX/SMM, los CUDA core son tan, tan independientes, que son parte de una batería de unidades de ejecución vectorial (32x), mira tú qué "átomo" valoras tú tanto en arquitecturas nvidia, ejecutan exactamente la misma instrucción en cada ciclo sobre datos contiguos de una serie de vectores (normalmente datos de píxeles contiguos). Pero nada, a contar CCs que es lo que vale…. :wall:
nvidia ha aligerado el hard de cada SMM para facilitar el meter más hard (unidades auténticamente básicas, los SMX, no los CC) y más funcional por un área. Ha hecho tantos cambios importantes que es absurdo ir con comparaciones de números de CCs a estas alturas (cachés, no sólo la L2 tan grande, sino las L1 de cada "procesador" o "core" de la gpu, los SMX, el tamaño del archivo de registos disponibles, etc, todo esto afecta a la capacidad de ejecución positivamente, evitando muchos stalls).
Pero nada, según tú, hay que comparar a "igualdad de frecuencias", y te marcas una medio verdad diciendo que:
1.- La media de las GTX 780 Ti de OC es de 1300, más falso que un duro de a cuatro. Aquí estamos hablando de OCs usables por aire, ESTABLES al 100% y para jugar, no para pasar benchs, ni tampoco sobrevoltajes locos que vuelvan a tu gráfica en un futuro muerto viviente.
La grandísima mayoría de GTX 780 Ti que hay por ahí no pasan o lo hacen poco de 1200 para un uso real, deja tanto de benchear y mira más OCs reales. Conozco a unos cuantos con gráficas como ésta, y es curioso el que haya algunos que han probado más de una docena de GTX 780 Ti y sólo hayan llegado a 1300 en una unidad aislada, ¿verdad? en fín. Que NO, clara sobre-estimación para intentar "ganar" a costa de tergiversar la realidad.
Que si te salen todas de 1300 con voltajes potables y demás, oyes, de puta madre, bien por ti, pero no es lo normal. Deberías saberlo, muchas firmas, pruebas, capturas subidas de 780 Ti vistas ya, por 1200 o poco más para creerme ahora que lo normal son "una media de 1300" (para arriba y para abajo, más o menos igual, claro).
2.- Te marcas la jugada de decir que "apenas pasan de 1400" las GTX 980. Es parte de una estrategia que empata con lo anterior, por un lado sobreestimas el OC de las GTX 780 Ti, y por otro reduces aún llegando a rozar la mentira (porque lo es, joder, tienes delante de tus narices una captura de ELP3 con una de las suyas a algo más de 1500, he visto como una decena de reviews y en TODAS, por aire y sin sobrevoltaje o mínimo, consiguen superar con mucho los 1400 y alguna pasan de 1500).
TODAS las unidades de TPU! han pasado generosamente de 1400:
http://www.techpowerup.com/reviews/EVGA/GTX_970_SC_ACX_Cooler/28.html
ESTA en concreto pasa de 1500, casi lo hace con el boost inicial, y es fácil comprobar su boost máximo comprobando en la siguiente página que éste sube unos 37 MHz más, o sea, literalmente pasa de 1500.http://www.techpowerup.com/reviews/ASUS/GTX_970_STRIX_OC/28.html
1400 y… mucho. Más cerca de 1500 que de 1400 raspados (otra vez, misma historia, boost máximo fácilmente comprobable mirando página siguiente).
http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/29.html
Otra que tal baila.
Gigabyte GeForce GTX 970 G1 Gaming review - Overclocking The Graphics Card
Una que pasa de 1500, 1516 en concreto.
OTRA:
http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review/22
¿Sigo? Porque son mayoría. Muy por encima de 1400 la mayoría, no es cierto eso que dices que "les cuesta llegar a" ni de coña, una de muchas no podrá, pero más bien pocas. Más si es para benchear como me parece que debes hacer tú más que jugar, que tanto crees que las Ti tienen una frecuencia de uso normal tan elevada.
Se compara a tope una y otra, si quieres comparar con OC, comparas ya churras con merinas porque dependerá de las unidades, así que o comparas dos "pura sangre" o no comparas. Y ni siquiera sabemos cuál es el límite de una "pura sangre" de las GM204, sabemos que está por encima de 1500, sin toquetear apenas nada (nada de desbloqueo de voltajes a lo bestia ni historias), pero NO cuanto.
La ventaja que tienen de serie las GTX 980 a la Ti le cuesta ponerse con algo de OC el recortarla, y lo que le quede en reserva dudo que iguale OCs así de simples como en las reviews de "por aire y con límites de TDP/voltaje", OCs que le hacen obtener una mejora de rendimiento en pruebas reales de un 15-20% más.
Para hacerse una idea de qué rendimiento hablamos:
http://tpucdn.com/reviews/NVIDIA/GeForce_GTX_980/images/perfrel_3840.gif
GTX 980 de referencia, 100 (referencia de todas), GTX 780 Ti de referencia, 93. La GTX 980 le saca 100/93 = Un 7,5% de rendimiento. Aparentemente "poca cosa", pero como tiene su propio OC, miremos cuánto rendiría cogiendo los resultados no de OC siquiera, sino de mejora de rendimiento real con OC:
15% -> 1,15 * 1,075 = 1,2365, esto es, en un caso de estimación muy conservadora (4K, OC de sólo 15% como el de TPU! aún viendo mejores resultados en otros lados), la GTX 980 con OC le sacaría un 23,65% de rendimiento de media a una GTX 780Ti.
¿Y con 20% como en otros resultados?
20% -> 1,2 * 1,075 = 1,29. Un 29% más de rendimiento que una GTX 780 Ti.
Suerte para superar esa mejora total con OC en una GTX 780 Ti. Vas a necesitarla, pero… ¿la mayoría lo lograrían? ja. Y hablo de rendimiento medido, no de OCs muy bonitos pero con escalados nulos o bajos de frecuencia, que ésa es otra... :troll:
Cree lo que quieras, está claro que hay ego en el asunto cuando parece que quieres hacer creer que son "algo más lentas" a base incluso de distorsionar ciertos hechos ya comprobados (frecuencias que atribuyes a unas y otras como "normales", no son las que dices). Y que quieres hacer cuentas de la lechera a base de hablar de OCs para intentar "superar" a la novedad. Más cuando ni siquiera han salido al mercado todos los modelos variopintos y "para OC" previstos con las GTX 980, e intuyo que cuentas con todo tipo de posible modificación de voltaje o tarjeta con desbloqueo de fábrica como "hecho dado" en toda Ti.
No pasa nada hombre, no tienes porqué sentirte amenazado por un poco de rendimiento extra, si quieres "lo mismo" con la nueva tarjeta de nvidia, sé feliz con tus 780 Ti si tal, y espera a que salga cosas claramente mejores para un posible cambio. Pero no nos cuentes que esto de las GTX 900 es un timo o que no rinden como demonios, o más que las Ti, por poco que te parezca.
Un poco de verosimilitud, y hay que reconocerle al César lo que es del César, chapeau por nvidia con estas tarjetas.
-
Yes, I have to remove that piece of the backplate. Although curiously there were two that didn't have it... I'll look in their box to see... or maybe they have already been tested...;)
Well, let's see... I have already tested them all. One by one.
The 4 graphics range from an Asic of 69 on the lowest to 75 on the highest. The low ones have a boost of 1265MHZ, the highest ones of 1278MHZ from the factory.
At first we return to the old ASIC standard that I experienced in the GTX 680 and TITAN, although it wasn't always the case, that the higher the ASIC the easier the OC. This is curious because in the 780Ti it didn't happen to me.De In fact, in the 780Ti, the ones with the least ASICs overclocked better. I suppose because the voltage and phase control system is different from the rest of the GTX.
The first one I tried, of course, is the one with the worst ASIC. That one does benchmarks of 1506.La one that has the best ASIC, 75.5, reaches a benchmark of 1600MHZ. At that speed it is clearly faster than an entire TITAN at 1450MHZ. That's something.
Regarding voltage, well, this is really curious because it doesn't help too much, let me explain. As these graphics have a lot of power margin, when you put for example an OC that reaches 1510 and crashes, when you apply voltage what happens is that the boost increases, and it no longer crashes at 1510, but for example at 1530MHZ, so the crash is almost assured since what you are giving it extra, it eats it up on the other side. It also happens that at very high clocks with voltage, you exceed TPD, in these 125, and in the end it stays longer at the lower frequency instead of the target one. Therefore, after fiddling around a bit, I have seen that except for exceptional cases, with a +36mv they have more than enough. Of course this has to be studied in more depth.
But obviously, seeing some reference graphics, crashing at 1600MHZ I think leaves anyone with their mouth open.
In terms of expense there is no dispute. I needed 2 FAs, an Enermax Platimax 1500w and a Corsair AX1200W to be able to benchmark with the TITANs. Cabinet of RL and own power supply apart, not only for the expense, but for a matter of amperage, they sucked so much that the multi-rail power supplies fell short, now I'm with a single Enermax 1500W and there's enough for all sides.No I don't want to anticipate, but surely in extreme benchmarking from TITAN to these, I could be close to -1000W... although I have to confirm this, My complete team of TITANs at the limit had a consumption of about 2200W.
Well, I'm having dinner and I'll test the sli and other things properly.
Regards.
