Asus Strix R9-390X DCIII - quick test

[FlanK3r]

Now the results of the card with stock settings, first PC rig

Intel core i7-6700K 4GHz, cooled by Gelid Tranquillo
2x 4 GB RAM Corsair Vengeance Performance XMP 3200 MHz
Asus Maximus VIII Hero
Asus Radeon Strix R9-390X DCIII OC
zdroj Corsair AX 1200
SSD HyperX 120 GB
HDD Seagate 3TB

software:
Windows 10 Home Edition 64-bit
AMD Catalyst driver 15.8 beta

Info CPUZ and GPUZ


In terms of quality, it looks like a good compromise that could be good both in the air and on the extreme (although according to the general evaluation ASIC is the rather less suitable for LN2, but the reality may be different). 78.8% is in theory such a light above average.



Benchmarks and games:
3D Mark11 Performance
3D Mark11 Extreme
3D Mark Sky Diver
3D Mark Fire Strike
3D Mark Fire Strike Extreme
Catzilla 720P
Catzilla 1080P
Catzilla 1440P
Unigine Valley - preset
Unigine Valley Extreme HD
Resident Evil 6 (max settings at FHD)
Heavensword:Final Fantasy XIV (max DX11)
AVP test (max DX11 FHD)
Sleeping Dogs
Tomb Raider
GTA V
Sniper Elite 2
Total war Shogun 2

I'm not pure game player and therefore my game selection is limited to what I have in my house.As the measurements I used the integrated test or Fraps log. For a rough outline that will suffice goal is to show the general performance of the card and overlcocking, which is much closer to me than the analysis of individual gaming sites and FCAT frames (although in practice they are essential for players)

3D Mark11 - tests performance and Extreme



3D Mark Sky Diver, Firestrike, Firestrike Extreme




Unigine Valley and Extreme HD




Catzilla 720P, 1080P and 1440P




Heavensword FF XIV benchmark
-settings of the game test


Who would want a clearer display settings, including the results of this test, utilizes a preview of the image, and beneath it is already resulting point score:



Resident Evil 6 - settings and result bellow



Sniper Elite 2 - settings and result




Tomb Raider - settings and result




Alien vs Predator -settings+result




Sleeping Dogs - settings+result



GTAV - part of the flyby fighter and ride in the city




Total War Shogun 2, two settings and results, one custom settings from screenshots and second results is at Extreme DX11



extreme DX11 result:


Later, I will summarize the results in a clearer table with OC results for each tests. Now it remains available on temperature. The temperature in the room were around 22 C. Temperatures in idle were 38 C at the card. Fans at graphics card I have left on automatic and must say that the behavior and the load was very quiet. Indeed, at the fan speed to search for yourself. The same temperatures. It is seen that higher ASIC is a little warmer, but the temperature at a given noise good card has provision for overclock (but it needs a good PSU which is AX1200). Temperatures in the load ranged up to 84 C at peak times, or about 80 C most time. The acoustic speech card was bee-friendly. At rest, avoid turning the fan and the card to work in passive mode. On such a monster like the core of Fiji and its later derivative is commendable. If the card was running at a lighter load in the 3D mode, it was clear to hear the gentle whistling coils, but when the card switched to higher load capacity utilization, sound speech spools ceased, and there was only a slight murmur of fans.


The power consumption of all PC (without LCD) was 76-78W in idle and 206W in CPU test, and 392W in combined test.

Next time we will look at the tooth of overlcocking card. Look also at the PCB itself, and then replace the original paste for Gelida Extreme = lower temperatures and compare your before and after . Finally, I will summarize the results in the table.

 

[Loosenut]

Nice, thanks. I can't wait for the TIM change and see the difference.

 

[FlanK3r]

I hope for 4C in load or so... We will see.

 

[Moofachuka]

plz add unigine 4.0 thx

 

[FlanK3r]

U mean Unigine Heaven?

 

[Moofachuka]

U mean Unigine Heaven?

yeh

 

[FlanK3r]

Unigine Heaven Will be added later

Overclocking with default voltage

Finally, we have the first results of the OC. The first task was to achieve such a daily routine overclocking, with the default voltage and tolerable noise and power consumption. I have to stress, therefore, not touch and were only two things. The first is increase of power limit up to maximum (for sure), and then scroll to set the right clock for the GPU core and memory clocks. And I also wanted to let fans still patches of about 50%. If we let the automatics, sometimes causing the artifacts in the load.
I would say that this is a slightly above average piece 390x cards. I managed to get the core to 1150 MHz (1160 MHz on some tests have been artifacts) and the memory at 6600 MHz effective.


How is this reflected in the performance? I'll start again with synthetic tests and gradually moving on to a few games. Up completely at the end of the entire testing will bring comprehensive and transparent measurement table. All settings are identical, as I said in the previous "chapter", where they can be viewed.

3D Marks






Unigine Valley Extreme HD



Catzilla tests (720P, 1080P a 1440P)




Resident Evil 6


Final Fantasy XIV HeavenSword


AVP


Sniper Elite 2


Tomb Raider


GTA V (the bridget, airfighter and the city ride)


Sleeping Dogs


Total War Shogun (custom settings+DX11 preset)



Performance of the card obviously increased, but it has risen slightly and noise, yet it is acceptable to my ears (eg. fan on the my CPU FX-9590 got noisier). Thanks to the high speed fans, or perhaps of baking paste dramatically changed the temperature of the card. They now load reached only about 68 C! Consumption was also unchanged voltage changed much.

The next step will be to increase voltages for maximum of GPU (I estimate about 1200-1250 MHz core and 6800-7000 MHz effective memory). And then we look at actual exchange paste and a new temperature and drops.

 

[FlanK3r]

I tested also Unigine Heaven Extreme as some people wanted, first there OC1 profile-so clocks from last chapter.


I had to new GTA V stock result and Sleeping Dogs (in SD I forogt at V-sync)





Overclocking with the voltage increase
We're going into the last round and overclocking with voltage increase. And then even max overclocking for a couple of 3D Mark. It must first be noted that the core of Grenada and Hawaii are quite challenging at current consumption, and it pays to have not only a good PSU, but also a well-chilled VRM circuit of card. We are now looking for short as well, but the first actual test ...

The maximum stable clocks of card are finished at 1188 MHz for core and 6752 MHz for memory efficiently [/ b]. I set GPU core voltage to 1325 V, the higher was the VRM overheating in longer test. After rebooting the PC, however, beats the software and did not keep within the units were others (eg 6756 MHz for the memory, etc.). Fans however, I have now at 100%, it is not comfortable, however for the highest maximum it was necessary ...


And now on achievement tests

3D Mark11 Performance+Extreme



3D Mark Sky Diver, Fire Strike a Extreme




Unigine Heaven a Valley



Catzilla 720P, 1080P a 1440P




Resident Evil 6


Final Fantasy XIV Heaven Sword



AVP


Sniper Elite 2


Tomb raider


GTAV (the bridge, airfighter and the city drive)


Sleeping Dogs


Total War 2: Shogun
-custom settings (details in first chapter of practice benchamrking)

DX11 high settings



So complete results are behind us, and you will be interested in power consumption, temperature GPU and VRM temperatures and clear the result table. Before we proceed to that, we will look more at the results Strix radeon card for 3D Marks, much more of it was not possible to get the card.
Maximální 3D Marky
The voltage 1.35V, a maximum of what software GPU Tweak allows II, it climbed up to 1203 MHz on the core and 6780 MHz effectively. I was hoping for 6800 MHz memory, but with this setup, I was rewarded with artifacts and application crash . On the harder tests it was impossible to cool the VRM, cooling of GPU core working very well, but on the VRM's just not much in this higher OC case...


Few of the results:





A bit later (today) I will add tables, information about temperatures and little bonus at the end. I hope you liked the article

 

[FlanK3r]

I promised the last part. It concerns the actual behavior of the card in terms of temperature, fan speed, and tables of performance, repast the old TIM for new one at the GPU core. At the outset, I must say that the card is very, very quiet. Maybe too much and about + 10% performance speed would have added an estimated 2 db and certainly reduce the temperature. Temperatures on auto namely through gradual and relatively slow rotating fans and climbs to 84 C. While this is fine and even with a reserve from what the last card, but the actual MOSFET and the entire complex power gets "harder take" and doesn't manage effectively cool . VRM is on automatic around 90 C at full load. This condition occurs as I traced, during a state of gradual increases in speed when the card responds to the state of the GPU core temperature, not the VRM (slow growing speed, but the card has "a long time" to reach higher temperatures for core and VRM also).
Change occurs when I set up since the beginning of the hand fans for about 40-50% of the speed will drop dramatically core temperature GPU card and VRM temperatures.

Let's see how it all looks in general. You can also check the backplate, which holds the smaller screws. The larger ones are used to unlock the aluminum cooling complex and expose the card to you in all its glory.


We see that the basic paste it into the core more than enough. It is neither bad quality but expensive Gelida Extreme is not equal ...


Graphics core is rectangular and round him are equipped with memory chips from Hynix 512MB


Specifically, the model is H5GC4H24AJR


You will also want to see how it fixed the cooling system for the GPU and memory. Heat-conducting tape sitting on MOSFETS literally brilliant. Complex cooling is then connected to the big aluminum pasiv, joint liabilities. Could only be more attracted to this part of the PCB (which I later did, and it helped a little heat to the VRM). Pity that contact with the VRM is also necessary through the heatpipe, it would be on such a great card solution (also it could increase of costs).


I removed original paste carefully and a replacement to Gelid Extreme



Now let us briefly look at what makes up the tab card. Parts on the card are equipped robot, so they are exactly in the plane, no it is not even half a mm askew. It's one of the features Asus Extreme technology. Detailed view of MOSFETs, inductors and capacitors cascade.


There are already the schema of the card. I marked the boundaries of green memory chips card. These together make eight gigabytes of capacity. Pink I have highlighted the power controllers. The larger one is digitally controlled and takes care of most of the phases. The card has a eight phases in total. Red bordered (6 phases) is used for the GPU core and the other two in a blue rectangle for GPU memory. I have not selected a PLL phase, which is responsible for probably smaller controller. They are however not accessible to software management. The actual voltage cascade is very good, there are drmosfets, high encapsulation chokes with ferrite smallgranuled core and decent capacitors.


With new TIM I gained about 4-5 C lower temperatures on the graphics core. They are definitely not limit itself clocked graphics card. But it was VRM temperatures circuit. This is indeed extremely good quality, oversized, as is customary in the Asus, but if the temperatures are about 105 C, turns on the protection circuit. Tighter pulling screws in the impugned section, I reached for VRM circuitry also lower temperatures by about 4 C load. And it fits . If therecould be copper heatpipes in direct contact, it might permit card clocked at 1200 MHz core of stable.

The games results in Table - Power and temperature measurements

Here you have a chart with the results of the setting to specific tests search in the section "Measurement at stock settings."


And here the table with temperatures and power consumption