I’ve been running an outdated pc for a while and want to start with VR and multiple monitor setup.
I’m putting together a list of components for my custom PC with the following specs:
Intel Core i9-13900KS
4090 x 2 which version is the best?
128 gb ddr5 ram
980 Pro NBMe x 2 (cache and storage)
liquid cooling.
z790 mobo if it can fit 2 x video cards
1600 watt power supply
eATX case
Are there any benefits to installing 2 x 4090s? Can the game utilize both graphic cards or are there no benefits?
Also wondering if there is anything specific I should be looking for building this computer. I want to have the best system that will last me another 10 years.
I will also use this computer for media production.
If I ever have free time, I wouldn’t mind building a sim or the least a motion simulation chair. What are your thoughts and feedback on this setup? Anyone have experience with dual video cards?
Just winging it here. Probably the game can use more than one card in multi monitor setups. However, now it needs to copy game geometry data twice. So it would run slower, not faster.
No more than any other game if it were optimised to use two cards at once. I think the issue really is a single 4090 is more than enough for MSFS, and it being a predominantly CPU bound program, the second 4090 wouldn’t get to do much, even if MSFS were capable of using it.
If you play any games out there that do take advantage of dual GPU’s I expect they would run amazingly well, but they might anyway with a single 4090 as I doubt there is much out there that could tax it, and if they can they are probably un-optimised ports from another platform.
I used to run SLI in the past, even going back as far as having dual 3Dfx Voodoo 2’s. The benefit there was I could play Quake at 1024x768 instead of 800x600. In later incarnations, with Nvidia cards, you would typically get a 50-60% boost in frame rates, but it was never even close to 100%.
I wouldn’t go as far as to say it’s dead, if you count NVLink as is successor, but its hardly prominent. Some more modern titles support it, such as Borderlands 3, GTAV, Dying Light etc. but the list is getting more and more bare as time goes on.
In the past, what some people like myself did was to have you main GPU for rendering games, and then you had a smaller, cheaper GPU, and you assigned PhysX to it, taking the load off the main card.
It might be possible for Asobo to allow more complex physics computations to be offloaded to a GPU, or some other custom ASIC, rather than the CPU, and have that as an option for multiple GPU systems. That might also be a lot of work for a tiny fraction of the user base, smaller even that VR.
I agree with others. A second 4090 would be a waste of money for gaming. As for creative apps, take for example DaVinci Resolve, which only saw a 7% increase in performance with dual unbridged A6000 GPU’s. For compute apps and AI, yes.
Also, 128GB RAM is way overkill, unless you’re creating large RAM drives. And 4 x 32 GB DDR might run slower than 2 x 32 DDR.
If there is a technology that would allow the cards to use each other’s memory at top speeds, then you are correct. Otherwise, it’s all about the hardware, it will run slower. Chips are real, they need data. If you have them double, you need to copy twice.
Older methods involved plugging in a cable, or small bridge, across an intrerface exposed at the top of the card. NVLink did away with that, I believe, and it could use the PCIe interface itself, but at a reduced speed I expect.
For gamers, creators, and workstation users who want a multi-GPU setup with the latest RTX 4000 series, Nvidia is switching from NVLink as a bridge to the PCIe Gen 5 standard.
“Because Ada [Lovelace] is based on PCIe Gen 5, we now have the ability to do peer-to-peer cross Gen 5 that’s sufficiently fast, and that’s a better trade off,” Huang added.
Current, and last gen Intel boards can already do PCIe 5.
But MSFS most possibly won’t use the second GPU. Even if you have multiple displays… i think the rendering workload is always spread on a single GPU for msfs…
If MSFS can show its content on a monitor connected to the second GPU, it is using that second GPU.
It’s all abstracted by the OS en DirectX.
Efficiency is another matter, but technically I see no problem.
Well, I am not a graphics programmer so cannot validate or invalidate your statement, but even if what you are saying is true, i think the extra overhead caused by having to distribute and ( perhaps synchronize, not sure if that is needed) the data will endup causing very little to no benefit of having the second GPU. But unelless someone tests and proves or disproves this, i dont know.
Windows uses RAM as a cache for the file system. So just looking at what an application like MSFS consumes memory wise, does not give you the whole picture. As long as there is RAM, windows will continue to use it for file system caching ( maybe there is a setting for the amount, I have no idea, but that doesn’t really invalidate the point I am making ), so more RAM, up to a point, is never a bad idea. I have 32GB, I wish I had 64GB.
