Introduction
I want to thank Seasonic for lending me the RTX 4090 for this article.
After hearing lots of debates about whether the sense pins on the 12+4 pin connector actually affect the RTX 4090's performance, I decided to find one of these cards and check what's going on. Thanks to
Seasonic and
MIFCOM, I found a Palit GeForce RTX 4090 GameRock, the standard not the OC model, which is among the most affordable RTX 4090 on the market. I know its VRMs are not as strong as in high-end RTX 4090 models, but for this article, it would be fine. Actually, the three legacy PCIe to a single 12VHPWR connector, rated at 450 W, that this card is shipped with will help me make a direct comparison with a native 600 W 12VHPWR connector.
I first completed a full set of tests with a previous generation Seasonic Prime Titanium 1000 W PSU using the provided 12VHPWR adapter. I didn't have any issues, so NVIDIA's claims that the RTX 4090 doesn't require an ATX v3.0 PSU stand up to testing. Even under overclocked conditions, the 1000 W PSU was fine. For detailed power consumption readings, I am using a Powenetics V2 system, which offers more than 1000 readings per second (>1000 Hz polling rate) on 13 sensors so that it can catch every nasty power spike. A significant asset of
Powenetics is that besides the GPU, it also monitors at the same time the system's total power consumption, including the CPU's, so I have a clear image of what is going on.
Since the Palit RTX 4090 GameRock that I received had a power limit of 450 W, and I needed a higher one to see if a 600 W rated 12VHPWR connector (Sense 0 and 1 grounded) would provide any gains, I decided to flash it with the GameRock OC model's BIOS, which I found in TechPowerUp's database. Using a non-public NVFlash version, things went smoothly, and finally the power limit increased to 500 W, which is appropriate for this card's VRMs.
Test System
Test System Specs |
---|
Mainboard | ASUS Prime Z690-A Bios Version 0702 (10/22/2021) |
---|
CPU | Intel Core i9-12900KF (all cores @ 5 GHz) |
---|
GPU | Palit GeForce RTX 4090 GameRock OC |
---|
SSD | XPG SX8200 Pro 2 TB NVMe |
---|
RAM | Corsair Vengeance LPX DDR5 32GB (2 x 16GB) 5200MHz |
---|
Power Supply | Seasonic Vertex 1200 Gold |
---|
Case | DimasTech Bench |
---|
Ambient Temperature | 25 °C ±1 °C |
---|
Drivers | NVIDIA: 522.25 |
---|
I ran Unigine Heaven to evaluate the performance differences in different testing scenarios.
The card's stock clocks and the overclocking frequencies that I applied with both PCIe cables are shown in the GPU-Z screenshots above.
Whether you use a 450 W or 600 W cable, the results are the same at stock clocks.
The 600 W cable made a difference in overclocking, but I cannot say that the difference is notable, at 2.62%.