Advantages of Crossfire vs Non-Crossfire

[skippytheturtle]

Quick question

Why run in crossfire if you have to give up the vram on the second card? Wouldn't it be better to run them in non-crossfire to get all of your vram available especially at large resolutions.

You can run two graphics cards in one system without crossfire, another advantage of doing this would be not being as limited on how many outputs you can use.

 

Please give answers for both high resolution gaming 

and high resolution NOT gaming at all.

 

Thanks

[Guest]

EDIT - oops, I actually don't know about the VRAM issue

[FlashingKool_Z]

Why would you run 2 cards in non-crossfire/sli? You would be literally putting it there for looks 

[DIV1D3]

You can't have two GPUs work together, without CF, so I think your outlook is flawed lol. 

Now I'm curious to know if having 3 displays hooked up to a GPU each would work. 

 

Edit:

With just one monitor, and without CF enabled (along with the CF bridge with most AMD GPUs) only the GPU that has the monitor connected would be used, the second would just stay at idle, doing nothing. 

[Izaya Orihara]

Mate you are horribly missinformed. 

[skippytheturtle]

You can't have two GPUs work together, without CF, so I think your outlook is flawed lol. 

Now I'm curious to know if having 3 displays hooked up to a GPU each would work. 

Yes you can use each card independently 3 (sometimes more, sometimes less) per card used for large multimonitor setup (like 12 monitors for instance)

So for gaming 2 graphics cards in crossfire and for not gaming I would assume not to use crossfire to get the extra vram

I was originally going to do an 8k setup (4x 4k monitors), bit it did not go so well

 

Thanks 

[DIV1D3]

Yes you can use each card independently 3 (sometimes more, sometimes less) per card used for large multimonitor setup (like 12 monitors for instance)

So for gaming 2 graphics cards in crossfire and for not gaming I would assume not to use crossfire to get the extra vram

I was originally going to do an 8k setup (4x 4k monitors), bit it did not go so well

 

Thanks 

You don't seem to understand how CF works, the reason why it exists is because then two or more AMD GPUs can work together, without it, well, they would work together. 

I actually have no idea as to what would happen if you don't enable CF and use multiple GPUs for multiple monitors. 

[skippytheturtle]

You can't have two GPUs work together, without CF, so I think your outlook is flawed lol. 

Now I'm curious to know if having 3 displays hooked up to a GPU each would work. 

 

Edit:

With just one monitor, and without CF enabled (along with the CF bridge with most AMD GPUs) only the GPU that has the monitor connected would be used, the second would just stay at idle, doing nothing. 

 

 

I was meaning multi monitor with monitors on both cards, thanks 

[skippytheturtle]

You don't seem to understand how CF works, the reason why it exists is because then two or more AMD GPUs can work together, without it, well, they would work together. 

I actually have no idea as to what would happen if you don't enable CF and use multiple GPUs for multiple monitors. 

IF frame rate does not matter, just displaying semi static information like graphs that update in 1 min intervals, the two GPUs don't really need to work together. 

Although in this case the extra vram would be useless anyway

[FlashingKool_Z]

I was meaning multi monitor with monitors on both cards, thanks 

Then I don't see a reason why you should even buy the same card, just buy like a gt610 for powering the monitors? 

[DIV1D3]

I was meaning multi monitor with monitors on both cards, thanks 

Yes, I'm aware of that. But I'm pointing out that with one monitor, only the GPU that is connected to the monitor would be used. 

I'm certain if you don't enable CF and try to run multiple monitors off of more than one GPU only the primary GPU would have a display, the others would again, not do anything. 

[skippytheturtle]

Then I don't see a reason why you should even buy the same card, just buy like a gt610 for powering the monitors? 

It was just a coincidence they are the same card 

[skippytheturtle]

Yes, I'm aware of that. But I'm pointing out that with one monitor, only the GPU that is connected to the monitor would be used. 

I'm certain if you don't enable CF and try to run multiple monitors off of more than one GPU only the primary GPU would have a display, the others would again, not do anything. 

You can use all cards in a system to use "extend desktop" in windows. They dont need to have a crossfire bridge. You can even use onboard video and a graphics card to add even more monitors. 

[RatedBlam]

You can use multiple cards not in CF/SLI but you use like gt 610 for that.

The OS or the software or something in that chain might just use the card No 1 and the other one would just be not taken use of. If not ran in CF.

[Sklo]

Even without CF you can still use all display outputs, but only the "main" card does stuff.

[DIV1D3]

You can use all cards in a system to use "extend desktop" in windows. They dont need to have a crossfire bridge. You can even use onboard video and a graphics card to add even more monitors. 

Aha, never knew that you were able to use multiple GPUs, to run separate displays. 

I'm sure this would never be able to work in games and programs that span the whole 2/3 displays, but for normal use they'd be fine running there own display/s. 

[skippytheturtle]

Aha, never knew that you were able to use multiple GPUs, to run separate displays. 

I'm sure this would never be able to work in games and programs that span the whole 2/3 displays, but for normal use they'd be fine running there own display/s. 

For games it does not work well at all, but for excel sheets and stuff like that it does not do too bad. 

[DIV1D3]

For games it does not work well at all, but for excel sheets and stuff like that it does not do too bad. 

Yeh, thought so, if you need the extra VRAM for multiple extra ultra high resolutions (higher than 4K, which isn't even for enthusiasts atm) for productivity than not running in CrossFireX is a better decision. 

Lol, just realised what I typed , "extra ultra high". 

[Oshino Shinobu]

It doesn't take much vRAM to run a screen on the desktop/webpage/whatever. If you're not gaming, you might as well get a card with a load of DP and just daisychain monitors. 

[JefferyD90]

Guess I don't usually laugh at newbs because most of them are trying to learn... but this made me lol cause people are actually defending non-crossfire/sli modes.

 

Here is the way to look at it.  A graphics card IS actually a computer all in its own.  It has a processor, ram, a PCB board, Vram, input, and outputs.  Much like how 2 computers cant really work on the same task (even though the user can work on each of them at the same time) 2 video cards can not work on the same task together.

 

This was the case until the GeForce 6xxx series cards came along and NVidia revealed the SLI bridge.  This allowed both of the cards to work on the same task.  But to work on the same exact task, it has to have the same information in the ram on the video card.  The SLI bridge works like a LAN network would for a server cluster.  This allows a bunch of servers to communicate across each other to get a common task done.  (also why ram is so important in servers because they too often have to have much of the same stuff in the memory across all servers).

[Samfisher]

  (also why ram is so important in servers because they too often have to have much of the same stuff in the memory across all servers).

 

Not necessarily.  Only very few cases do people actually cluster servers up like this.  Supercomputers sure, but for small, medium and even large corporations, each server is an independent server with it's own tasks and do not work anything like you described.  Fringe cases sure, but a huge majority aren't.

[JefferyD90]

Not necessarily.  Only very few cases do people actually cluster servers up like this.  Supercomputers sure, but for small, medium and even large corporations, each server is an independent server with it's own tasks and do not work anything like you described.  Fringe cases sure, but a huge majority aren't.

Any cloud service uses this method...  Also most gaming services use this method...  Most websites, use this method...  Scaling is huge in todays business environment. 

[skippytheturtle]

Thanks everyone, now it all makes sense!

[Samfisher]

Any cloud service uses this method...  Also most gaming services use this method...  Most websites, use this method...  Scaling is huge in todays business environment. 

 

Cloud sure.  Gaming services don't.  If you're offering a web service, sure.  If you're a regular brick and mortar company, you don't.

[Sadistic_Potato]

Look, look, you're misunderstanding something, unless you're literally gonna buy two different GPU's to do two different things.

 

The point of CF/SLi is to combine the graphical performance of two GPUs. However, in order to do that, the memory has to be split between the cards so they can alternate rendering the same thing. You might think that the VRAM is pointlessly and meaninglessly sacrificed, and that running two separate GPUs would give you the full VRAM of both cards.

 

But doing so completely eliminates the benefits of getting two powerful GPUs. If you want to game on one monitor, and do casual activities on the other, you don't get another powerful GPU, you get a GT720/R5 240 or even use the integrated graphics on Intel Processors. 

 

IF frame rate does not matter, just displaying semi static information like graphs that update in 1 min intervals, the two GPUs don't really need to work together. 

Although in this case the extra vram would be useless anyway

 

As you said, the extra VRAM is pointless in these kind of situations where you truly need to do two separate things, so if you want to reduce the load on your main GPU, you get a super low budget one to watch videos/do spreadsheets for this does not require much graphical horsepower.

 

And what on earth do you mean by 4x 4k? That's not 8k, that is 16k. No Quad-GPU setup yet can even get close to running games at that kind of a resolution, and that requires linking the GPUs. 

 

All in all, linking GPUs is for extra performance in situations where one GPU is not enough. It does not matter how much VRAM you have if your GPUs aren't linked and working together; in fact that would be a bigger waste for you're just wasting the graphical horsepower.