Review AMD RX 480 By Javisoft

[Javisoft]

This post is being processed/translated. The original version will be shown:

Overclocking

**
Por Aire**

**
Asic**

**
Consumo**

**
Graficos De Consumo Globlal Con Medias Independientes**

**
Datos Complementarios**

Bueno he estado viendo el tema de pasarla por agua, resulta que tengo aqui un EK thermosphere, pero viene con anclajes para nvidia :(, asi que tendria que pedir los anclajes para AMD e ingeniarmelas para hacer un disipador para los mosfets ya que el bloque full cover no esta disiponible aun :eoh:

Para mas inri, viendo la calidad de los componentes termicos de la tarjeta y con ansias de subir mas el nivel de oc, he cambiado los pads termicos por unos fugipoly normales tanto a la vram como a los mosfets y aparte he cambiado la pasta por AS5.

¿ Resultado ? Dos grados menos en full pero identico overclock … No sube ni 5 mhz mas ni en core ni en vram y para ser honesto, no creo que rasque mucho con agua visto lo visto.

**
Computacion Asincrona ( DX11 Vs DX12 OFF Vs DX12 ON ) Full OC**

Buenas noticias !! con el ultimo parche para Rise Of The Tomb Raider Crystal Dinamics ha dotado a este juego con la tan polemica computacion asincrona (Asynchronous Compute) y como no, he decidido probarla.

El juego por defecto ya trae la computacion asincrona ON cuando seteamos DX12, si queremos desactivarla a placer hay que entrar en el registro de windows y cambiar el valor 1 ( activada ) por 0 ( desactivada ) segun queramos:

HKEY_CURRENT_USER\SOFTWARE\Crystal Dynamics\Rise of the Tomb Raider\Graphics\AsyncCompute ( cambiamos el valor hexadecimal a 0 o 1 segun se quiera )

Los settings son los mismos que en el resto de la review, el oc es 1375 mhz/ 2250 mhz.

DX11 ( Computacion Asincrona OFF )

DX12 ( Computacion Asincrona OFF )

DX12 ( Computacion Asincrona ON )

Bueno, estoy un poco decepcionado, ya sabiamos que la computacion asincrona no es una varita magica, que para ver sus mejoras se necesita una programación muy especifica para ello y sinceramente pocos estudios creo que se quieran meter en tal follón y mas viendo resultados asi.

He pasado el test interno del ROTT por que en un gameplay las diferencias serian practicamente inapreciables, si es un test y casi no llegamos a 1 fps, imagina si estas jugando ;D.

En favor de crystal dinamics podemos decir que el juego no está integramente diseñado bajo DX12 y con computacion asincrona en mente, todo ha venido despues mediante parches.

Pero en honor a la verdad en juegos diseñados bajo DX12 al 100% como hitman o ashes of singularity han demostrado tambien muy poca efectividad/impacto y mucha dificultad segun sus desarrolladores, no ha habido ese salto que tanto anunciaba a bomba y platillo AMD, o por lo menos no por ahora:

En el futuro con la entrada masiva de titulos bajo esta api podremos hablar con mas claridad y determinación.

Conclusiones de los datos mostrados ( Tercera Parte )

Bueno esto llega a su fin, habia que ponerle la guinda al pastel con el overclock a la pequeña ;D, no podia dejar pasar la oportunidad de apretarle al maximo con ese pedazo de asic y ya de paso probar la computacion asincrona y los shaders asincronos de polaris, sinceramente tenia muchas ganas de verlo con mis propios ojos .

Para meternos en faena vamos a comentar el consumo, como bien habreis visto en otras review, el consumo de este ejemplar en concreto es buenisimo, muy por encima de la media, el asic dice mucho en estos modelos ciertamente, tenemos un consumo medio de 151w, lo que esta justamente dentro de los margenes que AMD dió como TDP en su presentación.

Pero viendo los modelos de otras review hay casos llegando a los 200w con full oc, lo que me parece una autentica burrada, si ya con mi ejemplar pata negra nos vamos a unos peaks en slot pcie y conector de 6 pines altos, con 200w simplemente son brutales.

En mi tarjeta con full oc nos vamos a 33w en el slot pcie y 118w en el conector de 6 pines de media en cada uno, peaks fuera, está un poco descompensado, pero lo prefiero sinceramente, la fuente aguanta mucho mas y mejor la carga por conector que por placa base.

En los graficos se ve perfectamente que el heaven no va nada bien en esta tarjeta, hay varios picos de caidas tanto de consumo, como de fps, timeframe y carga.

Son problemas que espero con versiones superiores de los drivers se corrijan, siendo un test no hay que darle demasiada importancia. Lo que si hay que tener en cuenta es la variacion de voltaje y el vdroop que se produce cuando baja la carga.

En el grafico correspondiente vemos que la carga de voltaje oscila entre 1.125v y 1.118v y en cuanto la carga de trabajo se relaja se dispara hasta 1.13v y 1.14v, es apenas un 1% de diferencia en peak, pero denota que los mosfets y las fases estan sufriendo.

Moviendonos hacia las frecuencias, como podemos ver, son una roca, una vez aplicas la frecuencia que quieras tienes que llegar a los 90º o +50% de Power target para que entre en throttling, cosa que evidentemente en este ejemplar no pasa ( recuerdo que el ventilador está a tope, a stock si habria throttling logicamente ).

Para terminar esta conclusión y dar cerrada la review, comentar el ultimo grafico, muy curioso la carga del controlador de la vram, una media del 40% de carga con peak del 80%, no es de extrañar que intente sacar el maximo provecho de la memoria este test, se ganan bastantes fps si consigues llevarla alta, pero es curiosa la demanda del chip siendo gama media.

Sobre la temperatura, bueno, pasado 1 min aproximadamente ya entramos en los 80º, llegando a 84º como peak, con el ventilador en full me parece bastante temperatura, pero viendo la calidad del mismo no sorprende que pasen estas cosas, entra dentro de la logica :ugly:.

Por ende el nivel de ruido es bastante insoportable, si eres muy delicado con los db está claro que el modelo reference no es el tuyo, mejor esperar a los customs donde parece ser que empresas como asus o powercolor van a sacar modelos muy muy buenos.

Por ultimo, este modelo NO tiene ningun tipo de coilwhine, ni de stock ni con oc, AMD ha hecho un gran trabajo en este sentido, muy muy bueno, 0 patatero ;D.

Y con esto me despido, ha sido un gran esfuerzo, contando que curro mis 8 horas de rigor y despues me siento a hacer todo lo que acabas de leer ( 81 horas de curro en total para la review )

Gracias por leer la review y por soportarme, espero que te haya sido de interés y agrado, espero que todas las dudas acerca de esta pequeña hayan sido disipadas :happygirl.

Si te ha gustado mi trabajo y quieres aportar para que me tome una cerveza o para que pueda seguir haciendo reviews de nuevas gpus u otro hardware o simplemente por que crees que lo merezco XD, aqui tienes mi paypal:

PayPal.Me

¡¡¡ No dudes en preguntar o comentar lo que necesites !!!

¡¡¡ Un saludo y un abrazo !!!

[krampak]

It seems like a good option for 1080p/1200p but it scales quite well to 2k. What are we comparing it to in terms of price range? Maybe the GTX970? Although I imagine the logical comparison would be to wait for the release of the 1060.

How is it in terms of temperature and noise? I don't see any data, and then to finish, you'll tell us if it goes up in revs or not; D

[Fassou]

What a great review

I can't wait to see the versions of the assemblers, and if the GTX1060 can compete with this, or if it will be a slower and more expensive mess.

Cheers!

[Javisoft]

@krampak:

It seems like a good option for 1080p/1200p but it scales quite well to 2k. What are we comparing it to in terms of price range? Maybe the GTX970? Although I imagine the logical comparison would be to wait for the 1060 to come out.

How is it in terms of temperature and noise? I don't see any data, and then to finish up, you'll tell us if it's overclocking or not;D

The current price range would be practically equivalent to the 970 but as you say, the logical comparison will be against the 1060, which by the way seems to be both faster and more expensive.

It's doing poorly in terms of temps and noise, let's not kid ourselves, AMD's cooler is very weak, it gets the job done, but at the cost of being a turbine... when I finish with the games I'll get into all these topics like RL and OC.

On the latter... The GPU promises, 89% ASIC, top-notch, does 1375 mhz by air with 1.125v, now at 85º, it's expected that with water cooling it could go a bit higher at the same voltage, but we'll see if it's true XD.

Best regards!!

[garfield]

It seems that dx12 performs better than dx11. Or am I the one who doesn't understand a damn thing about graphics?

[Javisoft]

@garfield:

It seems that dx12 performs better than dx11. Or am I the one who doesn't understand a damn thing about graphics?

You're reading them wrong, you lose fps but you gain a better timeframe ;).

[garfield]

And beware of oc. I don't like what I see. The reference ones with the 6-pin connector seem to be killing the boards when they do oc:
reddit.com: over 18?

[Obione]

From what I've seen in reviews, it's like a 970 but with 8GB, which will be appreciated later, although with less OC. I think the 1060 will cost the same and will be more successful despite having 6GB, I think it's a good amount of memory to last a few years at 1080p. I suppose AMD will have to lower prices a bit and put more effort into the refrito.

It's rumored that the 1060 will perform 15% better and what's clear is that it will consume much less, if a 1070 is at 170W if I remember correctly, I think the 1060 was at 120. In terms of temperatures, I think Nvidia will also be better. If I had to buy a custom GPU for 300 euros, I would wait for Nvidia.

I was hoping it would be closer to the 980

[Javisoft]

@garfield:

And be careful with oc. I don't like what I see. The reference ones with the 6-pin connector seem to be killing the boards when they do oc:
reddit.com: over 18?

I'm aware, according to amd today Tuesday a new driver comes out with a hotfix for this, even so I haven't been able to replicate the problems of websites like pcperspective about total consumption.

Surely they are some extreme and circumstantial conditions that have happened to several media.

I suppose it's a bad asic joined to a bad pcb and other circumstances.

I haven't been able to go over 160w even with full oc.

When I put a block I will try to take it to the limit ;).

Best regards.

[Javisoft]

@Obione:

From what I've seen in reviews, it's like a 970 but with 8GB that will be appreciated later, even if it has less OC. I think the 1060 will cost the same and will be more successful despite having 6GB, I think that's a good amount of memory to last a few years at 1080p. I guess AMD will have to lower prices a bit and work on the rehash.

There are rumors that the 1060 will perform 15% better and what is clear is that it will consume much less, if a 1070 is at 170W if I remember correctly, I think the 1060 was at 120. In terms of temperatures, I think Nvidia will also be better. If I had to buy a custom GPU for 300 euros, I would wait for Nvidia.

I was hoping it would be closer to the 980

I've read that it's 5% above a 980, what is clear is that in Timeframe NVidia has won the battle for a while now...

100 fps are of little use if the Timeframe dances more than a Brazilian at carnival XD

[Obione]

@Javisoft:

I've read 5% above a 980, what's clear is that in Timeframe NVidia has won the battle for a long time now …

100 fps are of little use if the Timeframe dances more than a Brazilian at carnival XD

xD

I meant 15% above 480.

1080 = cucumber
1070 = 980 Ti
1060 = 980 (that's what I hope for)

[amd125]

Very good work, it's a pleasure to see ;D

Congratulations.

[Javisoft]

Update with Dark Souls 3, Crysis 3, Battlefield 4, DOOM (2016) and Starcraft 2 Legacy Of The Void.

With these I close the FIRST part of the review pending loose ends or general information, tomorrow I'll get on with the second XD.

Goodbye!!!

[Javisoft]

Update with total consumption, 6-pin connector and pciex slot.

[garfield]

I see that the 6-pin connector consumption is over 100w, when it's supposed to not pull more than 75, and that the consumption on the pci-e is 75 and several times during two seconds exceeds 82w. I know that the tolerance is up to 86w, and I suppose it's sustained, but the data I see on consumption has me confused. Not because of an error in your measurements, but because AMD has released that graph with those peaks. The VRM are going to go to hell...

[Javisoft]

I don't understand that such a beastly variation, but with a power meter in hand and duplicate power consumption I can't help but think that it is due to the slight throttling it suffers from the temps and that it tries not to collapse the connector since it abuses this much more than the pcie slot.

I will take data again with the fan at full to avoid the frequency dance and above all to lower the temperature ;D.

Goodbye!!!

Edit: I am uploading more data as I process it, so throughout the night I hope to give it life XD.

[Espinetenbolas]

@Javisoft:

I can't help but think that it's due to the slight throttling it suffers from the temps and that it tries not to collapse the connection

The throttling is not because of the temperature… it's because of the consumption.

You do more throttling against higher temperatures because the VRM conversion performance decreases with the increase in temperature... also at higher temperatures the GPU consumes more since it alters the conductivity of the silicon and the metals that make up the chip.

Consequently the TDP skyrockets above 150W and therefore lowers the Mhz.

That is to say it is not due to thermal causes but to the BIOS... if you modify the TDP in some way you should disappear that throttling either with a ModBios or with the MSI AfterBurner increasing the PowerLimit.

The old Hawaiians suffered from this ailment that depending on the application or game is noticed more or less... if you put the OCCT or the Furmark you'll see how it happens the second you run the test even though the graphics card is as cold as a pozero's ass.

When you put water you'll see that you have the same problem and you only solve it by modifying the TDP.

[Javisoft]

@Espinetenbolas:

Throttling is not due to temperature... it's due to consumption.

It makes you throttle more as the temperature rises because the VRM conversion efficiency decreases with increasing temperature... also, at higher temperatures, the GPU consumes more power since it alters the conductivity of silicon and the metals that make up the chip.

As a result, the TDP skyrockets above 150W and therefore lowers the MHz.

That is to say, it is not due to thermal causes but rather the BIOS... if you modify the TDP in some way, that throttling should disappear, either with a ModBios or with MSI AfterBurner by increasing the PowerLimit.

The old Hawaiis suffered from this problem, which, depending on the application or game, is noticed more or less... if you run OCCT or Furmark, you'll see it happen the second you start the test, even if the graphics card is as cold as a zero's ass.

When you put water on it, you'll see that you have the same problem and you only solve it by modifying the TDP.

The consumption peaks at 124w and averages 112w, which fluctuates a lot, yes, due to the voltage that also marks its bachata steps, but encouraged by the temperature that hovers around 90º XD, if you notice, just when the temperature exceeds 87º, the frequencies begin to drop but the voltage oscillates more or less within the same margins.

The curious thing is that what fluctuates the least, not to say that it is very, very stable, are the frequencies, it holds up like a champion.

Anyway, tomorrow I'm going to run the same tests with the fan at full blast and also measure the noise and fpm with the anemometer XD, surely we'll clear up many doubts.

Best regards!!!

[garfield]

Curious. With the fan at full blast, it draws less from the pcie and draws almost the same from the connector. In the end, it's going to be that those values are so disparate because of the poor cooling of the chip. I think the custom ones will be ahead in that matter, as has always happened to AMD with their heatsinks.
Every time I see more tests, it seems like a pretty decent graphics card for what it costs, and the time frame rate is sure to be improved with new drivers (or so I hope).

[Javisoft]

It is very curious, but the heatsink is a bit of a mess, they could have left an entrance from the back to get more air... The custom ones look good, but I fear that with the GPU at full capacity, they will see similar temperatures to what I have put with the fan at FULL (although with less noise of course), the OC will be another story XD. Best regards!!