MSI GeForce RTX 5070 Ti Ventus MSRP card review – Interesting chip for gamers and the end of the GPU shortage?

1 - Introduction and details of the Blackwell GB203-300-A1 GPU 2 - Test system and equipment 3 - Teardown: PCB, components and cooler 4 - Material analysis and heat conducting materials 5 - Gaming: Full-HD 1920x1080 Pixels (Rasterization Only) 6 - Gaming: WQHD 2560x1440 Pixels (Rasterization Only) 7 - Gaming: Ultra-HD 3840x2160 Pixels (Rasterization Only) 8 - Gaming: WQHD 2560x1440 Pixels, Supersampling, RT & FG 9 - Gaming: Ultra-HD 3840x2160 Pixels, Supersampling, RT & FG 10 - DLSS4 and MFG: Cyberpunk 2077 in detail 11 - DLSS4 and MFG: Alan Wake 2 in detail 12 - PCIe 5 problems, power consumption in practice 13 - Load peaks native vs. DLSS4, PSU recommendation 14 - Cooler, temperatures, thermography, noise 15 - Summary and conclusion

Teardown

During the teardown of a graphics card, the card is systematically dismantled, starting with the documentation of the original condition. After removing the backplate and cooling components, the board is exposed and examined in detail, focusing on the layout, power supply and soldering quality. Individual components such as GPU, memory chips and voltage regulators are analyzed, as well as the cooling design and the heat conducting materials used. This approach enables a sound evaluation of the card’s design, performance and efficiency and provides valuable insights for end users and professionals.

Circuit board and components

As usual, the board is almost tiny and, like NVIDIA’s reference design, relies on three large voltage rails and several smaller ones. The voltage converters for the NVVDD as the actual GPU core voltage are nothing new. However, similar to Intel and AMD, NVIDIA now uses separate voltages for the GDDR7 memory and the frame buffer. While the voltage rails for the GPU cores and the memory are familiar, the frame buffer is new to NVIDIA in this form.

The frame buffer in a graphics card is a special area of the memory that is responsible for storing the pixel information of the displayed image. It contains data such as color depth, transparency and image resolution and is continuously updated by the GPU to enable image output on the monitor. The frame buffer is directly connected to the graphics memory, which operates under the voltage MSVDD. This memory provides the physical resource that the frame buffer accesses. In short: FBVDD ensures the stability and accuracy of data transfers between GPU and memory, especially at high clock rates.

The MSVDD voltage, on the other hand, regulates the operation of the memory chips themselves. This voltage directly influences the speed and stability of the memory, as it meets the electrical requirements of the memory cells and the memory controller logic. MSVDD and FBVDD work closely together, as the memory logic and the frame buffer must communicate with each other in order to exchange image data efficiently between GPU and memory. The separate regulation of MSVDD and FBVDD enables precise adjustment of the voltages to the respective requirements. The board is no witchcraft and I see 17 control loops, i.e. 10 x NVVDD (core, 0.8 to 1.1 V) 4x MSVDD (memory, 0.8 to 1.1 V) 3 FBVDD (frame buffer, 0.9 t0 1.24 V) and various other voltages.

The power supply of modern graphics cards is based on the precise coordination of several control and power components. The MP29816 (this time on the back) is a highly efficient and complex PWM controller that regulates the GPU core voltage (NVVDD), the memory voltage (MSVDD) and the frame buffer voltage (FBVDD). These components are crucial for the stability and efficiency of the graphics card, as they supply the main load of the GPU and memory. The MP29816 controls a multiphase system (IntelliPhase) that distributes the electrical and thermal loads while enabling precise voltage regulation.

The actual power regulation is carried out in all control circuits with higher loads (NVVDD, MSVDD) by the MP87993 DrMOS devices from Monolith, which convert the signals controlled by the PWM controllers into the corresponding output voltages. Three inexpensive uP9646 modules from UPI are used for the frame buffer (FBVDD), which are completely sufficient with a continuous current of up to 50A. These modules integrate the high-side and low-side MOSFETs as well as the gate drivers in a compact housing and are capable of efficiently processing the high currents of the NVVDD, FBVDD and MSVDD rails. Their design not only minimizes switching losses, but also saves space on the circuit board. The DrMOS components also ensure that the graphics card is protected by integrated safety mechanisms such as temperature and short-circuit protection.

There is also nothing sensational to be found on the rear. However, MSI, like NVIDIA, has only installed MLCC under the socket and no polymer caps. I already discussed the reasons for this years ago. In addition to the fuse and the shunt for the PEG, we also see the large PWM controller for NVVDD and MSVDD as well as a smaller one for FBVDD and the obligatory supervisor chip for power monitoring.

The NCP45492 can also be found here. This is a monolithic high-performance IC that can be used to simultaneously monitor bus voltages and currents on up to four high-voltage power supplies. Key features of the NCP45492 include the ability to translate and scale the shunt and bus voltages, as well as enabling the monitoring of up to four power supplies with a single device. Each channel is individually programmable through the selection of external resistors, allowing flexible customization for specific applications. In addition, the device offers a fast settling time and a real-time display of the validity of all bus voltages. This makes the chip ideal as a supervisor for the 12V lines (12V2X6 and PEG) of the power supply.

Here is a high-resolution microscopy view of all the important components:

VHX_000000

VHX_000001

VHX_000002

VHX_000003

VHX_000005

VHX_000006

VHX_000007

VHX_000008

VHX_000009

VHX_000010

VHX_000011

VHX_000012

VHX_000013

VHX_000014

VHX_000015

The cooler

The rear backplate is made of aluminum and does not additionally cool the circuit board by means of an attached heat conducting pad, which is a pity. The backplate only contributes to mechanical stability and does not improve cooling. Together with the central cooling block as a supporting element, the structural integrity of the card is additionally increased, which guarantees stable operation under high loads.

The rather lightweight cooler of the MSI GeForce RTX 5070 Ti Ventus, weighing only 553 grams, is a compromise between cool operation and cost efficiency, which is achieved through a combination of innovative technologies and a very high air flow rate. Cooling is made possible by a (small) integrated vapor chamber, which serves as the primary heat dissipation element.

It transports the heat directly from the GPU and the VRAM to the four huge, so-called core pipes with a diameter of 8 mm. These square-shaped heat pipes behind the chamber optimize thermal contact with the chamber and ensure even heat distribution. The heat is then dissipated through a network of precision-manufactured fins, which are designed for a very high throughput in order to be able to handle the slightly more than 300 watts of waste heat.

The fans of the cooler are each equipped with nine fan blades optimized for throughput, which we already know from the older Ventus models. In addition, the Zero-Frozr function offers the option of stopping the fans completely at low loads to enable silent operation. Another design element is the thermal pads, which I will discuss later, which provide additional heat dissipation from critical components such as the voltage converters.

The cooling system represents an acceptable compromise between size, weight and performance, but should also be somewhat more audible than the monster coolers of the more expensive models. Let us be surprised!