NVIDIA GeForce RTX 3090 Ti Production Halt Could Be Due to a Reused FE PCB Design

NVIDIA has dramatically cut production of their RTX series GPUs, with the cause being a reused PCB design. This could affect other manufacturers as well who have used the same design in recent products such as AMD’s Radeon RX Vega 64 Liquid Cooled Edition and GTX 1080 Ti Founders Edition cards.

The “rtx 3090 founders edition” is a graphics card released by NVIDIA on October 17, 2018. This card was the first to be released with support for ray tracing and DLSS.

NVIDIA image

NVIDIA instructed its AIB partners to suspend manufacture of its newest flagship GPU, the GeForce RTX 3090 Ti, in an unexpected turn of events last week. Igor Wallosek of Igor’s Lab, a well-known feedbacker and hardware tester, has offered a plausible reason for the situation. He believes it has something to do with the Founders Edition’s use of a PCB design that may not be up to the demands of the forthcoming card. He starts by saying that it may be due to anything, but he points out that three AIB partners have told him that their models are operating well.

However, I must limit that down; even three key board partners have said that their cards are working OK, that there are no actual issues with them, and that the hold was imposed by NVIDIA without explanation. Who knows whether NVIDIA is having trouble with its Founders Edition and the rest of the suppliers will have to wait for the “MSRP model”… However, it is no secret that the FE’s adaption is difficult (even down to the cooling).

Igor gives some insight into the development process and speculates that the GeForce 3090 Ti might be a “unicorn” card in the sense that nothing else in the current Ampere generation would be quite like it. If this is accurate, he believes it is illogical for NVIDIA to create a new board from the ground up for just a little performance boost over the old GeForce RTX 3090. He does suspect, though, that NVIDIA may have utilized the PCB for the previous card with the current GPU.

Because the GeForce RTX 3090 Ti is more likely to be a unicorn, developing a separate circuit board for a physically virtually similar card (apart from the GPU and RAM) or this new card, then sending it via NVIDIA’s Greenlight program to certify all individual processes makes little sense.

Designs of AIB vs. FE Boards

He goes on to say that board partners are likely to repeat their own ideas as well. The EVGA GeForce RTX 3090 Ti Kingpin is one known example, since it is rumored to have a twin 12-pin configuration. However, it was also stated that MSI is using their 3x 8-pin design seen on the Suprim and Trio models for the Suprim X instead of the new 12-pin that NVIDIA debuted with the Ampere series, and that “the devil is in the details because the complete power supply of such a card is a very fragile and carefully balanced construct,” and that “the devil is in the details because the complete power supply of such a card is a very fragile and carefully balanced construct.”

The GeForce RTX 3090 Ti will have the highest TDP for a consumer-grade graphics card ever introduced by NVIDIA, at 450 watts, as previously reported. Some AIB models are also known to be capable of even larger power outputs; the Kingpin, for example, might potentially approach 1000 watts. As a result of the higher power consumption, numerous components such as capacitors and shunts face additional demand, as well as increased thermals and the normal firmware and driver upgrades.

Igor’s Lab (photo credit: Igor’s Lab)

The firmware for the GeForce RTX 3090 Ti, on the other hand, stores a substantially greater TBP (from about 480 watts), i.e. the value for the complete graphics card, including all existing 12-volt rails. And, if we remember the components from the design above, things will be a lot of fun thermally, particularly around the GPU, GPU voltage converters, and memory.

Igor’s Lab (photo credit: Igor’s Lab)

Igor put a GeForce RTX 3090 FE board through its paces to determine what its performance limits are. He made adjustments to the original shunts after noticing their limitations. Even with those tweaks and liquid cooling, reaching the aforementioned 480-watt TDP was difficult. It was conceivable, but unreliable and unsuitable for continuous operation. Furthermore, it couldn’t go much higher than 500 watts before failing.

“Unfortunately, despite manual OC, I couldn’t reach any higher than 490 watts since the power runs out at that point.” The 500 watts were unfortunately no longer viable.”

It’s still unclear why NVIDIA instructed its partners to suspend manufacturing, but if Igor is correct, he may have struck the nail on the head, or at the very least been on the right road.

Igor’s Lab is the source of this information.