Galax RTX 5070 Ti HoF (Hall of Fame) review – To the top with a monster board?

1 - Introduction, overview and technical specifications 2 - Test system and equipment 3 - Teardown: PCB and components 4 - Teardown: Cooling system 5 - Teardown: Material analysis and TIM 6 - Benchmarks: gaming performance 7 - Power consumption, transients, PSU recommendation 8 - Clock rates and overclocking 9 - Temperatures and thermal imaging 10 - Fan curves and noise with audio samples 11 - Summary and conclusion

Total power consumption

The measured values of the Galax RTX 5070 Ti Hall of Fame show a very differentiated picture, which changes significantly depending on the load scenario and allows a clear classification of the efficiency of this card. In idle mode, the power consumption is pleasantly low at 15.1 watts at 60 Hz Full HD and rises to 20.6 watts at 120 Hz in UHD. These values are inconspicuous for a card in this performance class and demonstrate a functioning energy management in idle mode, whereby the higher refresh rate on the monitor draws slightly more power, as expected.

Under light load, here using a browser scenario as an example, the card draws around 164 watts. This value is already in the medium range and shows that the GPU clocks up at an early stage in order to also cover everyday applications smoothly. Around 210 watts are required for 3D CAD, which marks the transition from light to medium load. In everyday gaming, the power consumption depends heavily on the resolution. In Full HD, the average value is 248 watts, which is within the range of what is typical for a card with this power target. In WQHD, the power consumption rises to 279 watts, while an average of 296 watts is achieved in UHD. This shows good scaling, which also makes it clear that the card comes closer to its limit in high resolutions.

The peak values are interesting: In the Torture test, i.e. under synthetic full load, consumption rises to 309 watts, while a maximum of 314 watts is reached in particularly demanding UHD games. The card thus utilizes its power limit of 300 watts up to the upper tolerance limit and shows that the power supply is designed for such loads. It is noticeable that the card requires a relatively high amount of energy in everyday use, even in the medium load range, which is probably also due to the high boost strategy. On the other hand, stable and predictable load values are available, which offer security for overclocking and high continuous loads.

The PCI-SIG specification officially allows a maximum of 5.5 amps for the 12-volt path of the mainboard slot. Multiplied by the voltage, this results in a permissible power consumption of 66 watts, with a certain amount of leeway for tolerances. The card therefore remains clearly within the specifications with less than one ampere and shows that the primary power supply is consistently provided via the external 12V connections. The classification of the well-known Urban Legend that a PEG slot can allegedly supply “75 watts” also fits into the picture here. This information has been circulating in forums for years and is often quoted, but is incorrect in this form. In fact, the PCI-SIG specifies exactly 5.5 amps at 12 volts, i.e. around 66 watts, supplemented by up to 3 amps at 3.3 volts, which are practically no longer used by modern graphics cards. Only if you add both paths together do you get close to 75 watts mathematically, but the 5.5 ampere limit alone is decisive for the GPU supply via 12 volts.

Load peaks during gaming

The series of measurements clearly illustrate how the power consumption of a Galax RTX 5070 Ti Hall of Fame varies depending on the resolution and sampling rate. In the long-term measurement with 20 ms intervals, the average value during a gaming session of Cyberpunk 2077 in UHD is around 300 watts, but the curve constantly fluctuates between just over 200 and at times well over 400 watts. These fluctuations are the normal expression of load peaks that occur when GPU shaders, ray tracing units and memory controllers process strongly varying partial loads. The current diagram shows a very lively picture to match. With an average value of around 25 amps on the 12-volt rail, peaks of over 35 amps occur. Here, too, the distribution appears chaotic, but in fact it follows the changing load of individual functional units of the chip. The fact that the PCIe slot (PEG) is only slightly involved can be seen from the fact that the green line for the slot current is barely visible. The load is therefore concentrated almost entirely on the 12V2x6 connection.

Things get interesting when you refine the time resolution and go into the 10 microsecond range. This reveals that the supposedly gentle fluctuations of the 20 ms resolution are actually a result of abrupt load jumps. The power consumption can rise or fall by more than 100 watts within a few microseconds. The peaks reach values of over 450 watts, although the average remains at around 330 watts. The current curve underlines this, as individual transients briefly push the current up to almost 40 amps. At the same time, the voltage measurements show a drop in the 12-volt rail to around 18 to 19 volts in the most intense moments before the power supply unit counteracts this with its regulation. This behavior poses a particular challenge for power supply units. It is not the continuous load of 300 watts that is problematic, but the extremely short load jumps that have to be absorbed at very high speed. Here, the quality of the voltage regulation and the dimensioning of capacitors determine whether such transients are absorbed cleanly or lead to instabilities. Modern ATX 3.0 power supplies with a corresponding 12V HPWR connection are designed for precisely these scenarios, while older models without a generous transient reserve could be overwhelmed by such jumps.

Power consumption	Currents

Load peaks during the stress test

The stress test shows a completely different picture than the gaming loads. While games with their changing scenes and render paths generate a strongly fluctuating power consumption with partly high transient peaks, the synthetic torture test leads to a comparatively more even continuous load. In the 2-minute monitoring, the card settles at an average of around 310 watts. The regular short dips downwards are noticeable, which are caused less by uncontrolled load peaks than by the test software itself. The intervals between these dips are even, which shows that these are artificially generated load changes and not random fluctuations. Overall, the load curve is much smoother than under real gaming conditions.

In the high-resolution zoom view with microsecond resolution, you can clearly see the short dips in the current flow, which are visible as regular, deep valleys in the power curve. These dips push the load almost to zero for a few microseconds before the card immediately returns to the stable power plateau. For the power supply unit, this means that it has to contend less with unpredictable upward load jumps and more with recurring, clearly defined pulses that have no critical effect due to their short duration. In terms of current, the card is constant at around 26 to 28 amps, which is typical for the GPU. Here too, the dips appear as symmetrical dips that occur in the rhythm of the Torture Tool. The voltage on the 12-volt rail remains largely stable, which shows that the power supply can handle these load patterns without any problems.

Power consumption	Currents

Summary of the load peaks and a power supply recommendation

For the Galax RTX 5070 Ti Hall of Fame, it is better to plan the power supply unit not according to the average power consumption, but according to the measured peaks. The card averages around 300 watts in gaming, but the peak analysis shows up to just under 470 watts for periods of less than one millisecond, around 428 watts for 1 to 5 milliseconds, around 406 watts for 5 to 10 milliseconds and around 371 watts for 10 to 20 milliseconds. This is where the protection logic becomes relevant, as the usual supervisor chips often only react reliably from around 5 to 10 milliseconds. The power supply unit must be able to cope with anything less than this with sufficient buffering and fast regulation without triggering OCP or OPP unnecessarily.

I therefore recommend a current ATX 3.1 power supply with a native 12V 2×6 line. This standard explicitly requires reserves for very high short-term peaks, which improves both operational reliability and component durability. The nominal wattage on the box is less important than the clean transient resistance of the 12-volt rail and a robust primary and secondary capacitor assembly. For typical systems with a processor in the 120 to 150 watt class and standard peripherals, the 5070 Ti HoF results in a realistic gaming peak of around 450 to 550 watts. This means that a good 750-watt ATX 3.1 model positions the load in the efficient range of around 50 to 70 percent and offers sufficient reserves for short load jumps. If you are using a very hungry processor, are planning a lot of drives and fans or are flirting with manual overclocking, you will be more comfortable with 850 watts. In workstations with permanently high CPU loads or encoding applications, a 1000 watt model can also be useful in order to place the continuous load in the sweet spot of efficiency.