GPUs Graphics Reviews

Pascal becomes passive: We build a fanless GeForce GTX 1050 Ti

It's been almost four years since we first tried a GeForce GTX 650 in our two-part Silent Gaming project - More graphics performance including thermal protection circuit (here part 1, which is about the PC itself). Temperatures, clock rates and voltages Our measurements have to be carried out in passive mode over a much longer period of time, because it can take up to an hour for the final heating of all components to be carried out.

Temperatures, clock rates and voltages

 

Our measurements must of course be carried out in passive mode over a much longer period of time, because it can take up to an hour for the final heating of all components of the system to be achieved. The graphic with the GPU temperatures shows us that this long test phase was worth it.

 

After about 57 minutes – i.e. almost the estimated hour – the GPU temperature abruptly drops from the 83°C set as the limit and also reached 83°C to 77°C, and then to continue to run constantly. Even after two hours we measure around 77°C, which baffles us quite amazes.

 

The original card with fan can convince in the passive system with comparatively cool 63°C, while the monster cooler in the huge workstation with the two very slow rotating radiator fans guarantees almost an arctic 54°C.

 

But why does the temperature of the passive card in the completely fanless system suddenly drop by almost six degrees Celsius after a longer run time? To fathom this, we need more measurements.

 

 

Therefore, we now look at the GPU clock, which logically shows the same behavior. But it can wither even more: After the longer running time at the set temperature limit of 83°C (factory setting), the boost with its frantic adjustment attempts suddenly no longer exists – the card clocks stable to 1342 MHz and no MHz higher!

 

 

The same behavior can now also be observed at the voltages, which suddenly fall to a constant continuous 0.812 volts. This value appears – as well as temperatures and the clock rate – as a kind of emergency running program. The power consumption is now even less than 40 watts and also remains largely constant. What's interesting now is how the gaming performance behaves.

 

 

We also see the same behaviour at Furmark, where the GAU is much more likely to be achieved here. However, it is pleasing that the card runs completely passive and full load without burning up, but loses significantly in clock (and thus also performance). After all, at least at this moment, you can be sure that nothing will break. However, you should not really exploit this safety solution, because no one knows when the oven is completely off.

 

Note pad #4
• With only a little light airflow, the cooling is even better than the original
• Completely passive operation is possible, albeit with limits
• Clock and voltages are automatically lowered to protect the card

 

We are looking for the limitations

 

After all, over three games (Metro: Last Light to 1440p, GTA V and The Witcher 3 to 1080p) averaged 21 percent less than the air-cooled original seems a lot at first glance, but the power consumption drops to less than 60 percent. If you also take into account that we have already lost a good four percent of performance in the same benchmark run via the reduced Power Target to 80 percent, then the potential loss is absolutely fine.

 

First, let's consider where the GPU as such limits the benchmark flow. This behavior is perfectly fine and always occurs in demanding games. We see again that from the time of the emergency brake something else has to cover the system strongly:

 

 

It can't be at the power limit either, because we see a continuous intervention, but even this one seems to have disappeared at once from about 53 minutes:

 

 

In the end, only the long-lasting high temperature remains as the cause, even if it has sunk after the emergency braking and is far below the predetermined temperature limit:

 

 

Note pad #5
• Performance decreases more slowly than power consumption
• Up to 21 percent loss of power after the limiter has been set up
• Up to four percent performance loss due to the Power Target of 80 percent

 

Infrared measurements of temperature behaviour

 

Let's look at how our conversion behaves with the thermal camera during the measurements. After all, we have only been able to assess the GPU temperature so far, although the graphics card contains many other components, some of which are very temperature-critical.

 

In the Idle, peace, joy and egg cake reign even in the completely passive system – although the temperatures are not even sufficient to be able to bake the latter at all.

 

 

If you place the card in a housing with the minimum airflow, our passive cooler is already in high form. Here, our conversion solution does not even have to hide from the original map with air icon, because it does not cool down any worse.

 

 

Only when you integrate them into the fanless system and torment them for an hour, the thermal penetration of the board is complete to the last angle. But even now, the storage is still cool enough not to fear thermal damage.

 

 

In summary, the conversion has actually been quite successful from a purely thermal point of view, even if we have a bit of a struggle with the fanless location.

 

 

Conclusion

 

Our small test shows once again that completely passive solutions in the 60- to 75-watt class are still only possible with limitations. Either you use at least one very slow-rotating fan as a guarantee for a minimal airflow in the housing or sooner or later you run into a temperature limit, which activates the self-protection of the card and thus also prevents the risk of damage, but cannot really be the purpose and purpose of a passive solution.

 

We see the limit for completely fanless operation again at around 40 watts of power consumption, which is almost completely converted into waste heat. All that really remains is to manually set the power limit to 50 percent or to wait for the card to hit the brakes at some point and achieve a similar power consumption value by the continuous limit. Then you will have to test hard where the thermal limit of your own system really lies – also and above all depending on the ambient temperature.

 

As a conclusion, we again only have to recommend a minimum air movement as a minimum requirement in this performance class. Because if we are really honest, then the starting position has hardly changed since the GeForce GTX 650. The cards are not really more economical at all, but with the same power consumption they have become much faster. This, too, has its charm and helps to overdo a small, slow-turning fan with a noble gesture of nonchalance.

Danke für die Spende



Du fandest, der Beitrag war interessant und möchtest uns unterstützen? Klasse!

Hier erfährst Du, wie: Hier spenden.

Hier kannst Du per PayPal spenden.

About the author

Igor Wallossek

Editor-in-chief and name-giver of igor'sLAB as the content successor of Tom's Hardware Germany, whose license was returned in June 2019 in order to better meet the qualitative demands of web content and challenges of new media such as YouTube with its own channel.

Computer nerd since 1983, audio freak since 1979 and pretty much open to anything with a plug or battery for over 50 years.

Follow Igor:
YouTube Facebook Instagram Twitter

Werbung

Werbung