Oh! I made a mistake, I have a 4500 (and not a 5000), the TDP is 200W and the length is also 267mm.
However, beware of the Pro version, the power plug is sometimes located at the back of the card.
I got tricked, the dimensions on paper fit in the case, but I had to cut the front panel to pass the cable.
As Jeremy indicates, The PCIe port are backward compatible. However you must well distinguish the transfer speed between your RAM and the VRAM and the work of the GPU with its VRAM. My RTX 4500 PCIe 5 is mount on my old PC with a PCIe 4 port.
The data that are transferred between the RAM and the VRAM cannot go faster than the PCI 4 port allows it (even if the card would accept a faster transfer speed). Once the data in in the VRAM, the GPU is absolutely not limited by the PCIe port. It runs like on a PCIe 5 card. the PCIe was the standard of the physical connector on the motherboard, the limitation only applies to the RAM/VRAM transfer speed. In the case of Rhino which sends a lot of update to the VRAM, I exploit the RTX 4500 as fast as I exploited the GTX 1080Ti. But for a games for example, once load, the speed will be that of an unbridled card.
I confirm what Jim say, even if it is daring to say that a card is THE best of all, the RTX 1080Ti remains a histirical anomaly. over-dimentioned for its time, a bus size of 352 bit, even the 5080, cannot read/write as much VRAM data at the same time by CUDA cores. It’s only because I had very specific need that I had to change it.
About the 250 watts, you have to know three things, this consumption is not the one when you turn on the PC, at rest or to browse the internet and even to display Rhino model, the card does not consume that much. If on the other hand you do a Cycle render, that you play a games or do AI on your machine, there yes, the card can ask up to 250 watts to its power supply.
The second thing (but I confirm nothing, because I never did it myself) is that overclocable card have not for obligation to be overclocked upwards. If you have access to the frequency of the card you can also reduce it, and it seems to me having read that some card also allow to throttle the consumption (the Watts).
In case, I prefer to clarify, if a power supply is deliver to deliver 100W, one could be led to think that even if an equiment asks for 150W, the power supply not being able to deliver more and that the hardware will be throttle. I therefore invite you to be wary of this kind of talk. Some power supply have safeties, apple tends to put good hardware but I wouldn’t count on it despite everything.
