Nvidia Quadro Questions

[StarlightDash]

Hello! So recently Im building a cheap editing workstation off of old cheap hardware that I know will work (Please no judging) and recently I was rewatching a linustechtips moving vlog and videos in general as they help me learn more about pc components and how things work. In one of the videos it looked like his editor had a GTX Titan AND an nvidia being used in one system. I was really curious as to what uses there are in this and how does that work as well as uses for spare gpus or cheap gpus in general besides SLI. I'd really like to learn about this and curious if anyone could shed light on this.

 

[k1114]

Sli and cf are gaming features. Other uses will mess up with it on and require you to disable it or get better performance with it disabled. For content creation that is gpu accelerated and can use multiple gpus, they use anything in the system no matter if they are different. There are somethings that have to be the same like cuda version, or nvidia with nvidia because how they implement opencl vs amd. Some will take nvidia with amd. But beyond that, you can take a tesla, quadro and geforce all at the same time, all working together. Especially with 3d rendering as some renderers compartmentalize.

Quadros tend to do much better at the preview window in video editors. But you just can't get the raw performance plus cheaper price that geforce offers which does better during final output. If you have the money to just throw at the pc, sure why not have the best of both worlds.

If your uses don't use multiple gpus, spare or cheap gpus are useless. Even if they do, the cheap ones maybe too weak or just use more power vs the performance increase that it's still a waste. Others can use multiple cheap gpus to have more monitors like traders.

 

[StarlightDash]

Sli and cf are gaming features. Other uses will mess up with it on and require you to disable it or get better performance with it disabled. For content creation that is gpu accelerated and can use multiple gpus, they use anything in the system no matter if they are different. There are somethings that have to be the same like cuda version, or nvidia with nvidia because how they implement opencl vs amd. Some will take nvidia with amd. But beyond that, you can take a tesla, quadro and geforce all at the same time, all working together. Especially with 3d rendering as some renderers compartmentalize.

Quadros tend to do much better at the preview window in video editors. But you just can't get the raw performance plus cheaper price that geforce offers which does better during final output. If you have the money to just throw at the pc, sure why not have the best of both worlds.

If your uses don't use multiple gpus, spare or cheap gpus are useless. Even if they do, the cheap ones maybe too weak or just use more power vs the performance increase that it's still a waste. Others can use multiple cheap gpus to have more monitors like traders.

Wow thats really interesting. GPU acceleration is almost like using the vram on another gpu to speed up specific processes? Sorta like intel optane to harddrives or SSDs except for certain things. I guess in a gaming application in a sense if your bottleneck is VRAM you can just pop an old GPU in some cases and that might solve your problem? Sorry to be bothersome this is really fascinating to me atm. Please correct me if Im wrong. Im trying to understand on my sorta thinking.

 

[dontlistentome]

Things can get complicated depending on your specific use and the specific software you use. Generally the videocard ram does not add up. If you run two 2gb cards, most use cases 'see' only the 2gb ram per card,not 4gb of pooled ram. What can be combined is the processing power. Although don't expect double the processing, since other factors in most cases limit the performance gains so you wind up with less than double the power.

This why for most people it's better to use ONE good card rather than two lesser cards. If you need a certain amount of videocard ram, get one card that has that amount. If you need twice the power of a certain card, get one card that has twice the power.

Quadros can provide higher image quality than gaming cards. In some applications this can be very important, such as in CAD. However, to take full advantage of this you'll want a monitor that's capable of displaying that level of quality image. It defeats the purpose to use a Quadro with a cheap gaming monitor for example.

 

[StarlightDash]

Things can get complicated depending on your specific use and the specific software you use. Generally the videocard ram does not add up. If you run two 2gb cards, most use cases 'see' only the 2gb ram per card,not 4gb of pooled ram. What can be combined is the processing power. Although don't expect double the processing, since other factors in most cases limit the performance gains so you wind up with less than double the power.

This why for most people it's better to use ONE good card rather than two lesser cards. If you need a certain amount of videocard ram, get one card that has that amount. If you need twice the power of a certain card, get one card that has twice the power.

Quadros can provide higher image quality than gaming cards. In some applications this can be very important, such as in CAD. However, to take full advantage of this you'll want a monitor that's capable of displaying that level of quality image. It defeats the purpose to use a Quadro with a cheap gaming monitor for example.

Hmm ok. Im just trying to find some interesting, fun and unique ways to make use of GPUs old or new in any way.

 

[k1114]

No gpu acceleration is using the gpu to accelerate the process. You can never share vram, yet. The gpu will only use its vram. It's nothing like hdd caching. It is literally using the gpu for processing instead of using the cpu for processing. The advantage of this is that gpus have hundreds of cores, although simpler than cpu cores and can't do everything they can, it is much faster and more efficient for highly paralleled tasks.

When using multiple gpus for content creation, their vram is not mirrored and each gpu uses its own vram for whatever it's doing. They likely will have some same files depending on the work but they are working independently.

The point about one good card vs 2 lesser ones is largely a gaming issue but depends more on the work. Most content creation and multi gpu is linear performance increase as they can work on separate things and aren't limited by other factors.

Quadros do support 10 bit color depth and higher levels of aa but image quality is quite a broad term and many cases isn't the reason people buy quadros. Many uses in cad don't need higher image quality but they will handle more complex assemblies than geforce even with much lower specs. The main benefit of quadros and the reason they cost more is the driver support. Higher end ones will also have ecc vram.

 

[StarlightDash]

No gpu acceleration is using the gpu to accelerate the process. You can never share vram, yet. The gpu will only use its vram. It's nothing like hdd caching. It is literally using the gpu for processing instead of using the cpu for processing. The advantage of this is that gpus have hundreds of cores, although simpler than cpu cores and can't do everything they can, it is much faster and more efficient for highly paralleled tasks.

When using multiple gpus for content creation, their vram is not mirrored and each gpu uses its own vram for whatever it's doing. They likely will have some same files depending on the work but they are working independently.

The point about one good card vs 2 lesser ones is largely a gaming issue but depends more on the work. Most content creation and multi gpu is linear performance increase as they can work on separate things and aren't limited by other factors.

Quadros do support 10 bit color depth and higher levels of aa but image quality is quite a broad term and many cases isn't the reason people buy quadros. Many uses in cad don't need higher image quality but they will handle more complex assemblies than geforce even with much lower specs. The main benefit of quadros and the reason they cost more is the driver support. Higher end ones will also have ecc vram.

hmm ok.

 

[dontlistentome]

Haha, that's why it's never so simple to say "This card is better than that card". What you are doing, which software you're using to do it, and which other components you are using, will all play a part in determining what is better.

If all you want to do is experiment with leftover old parts, then this doesn't matter. Get a motherboard that supports two cards and hook it all up. Then you have your test system to mess with. If you are hoping for an a genuinely useful way to use old parts then that's where things get doubtful. I have three older videocards just sitting in their boxes. If there was a way to add them together into something useful I'd be using them right now.

 

[StarlightDash]

Haha, that's why it's never so simple to say "This card is better than that card". What you are doing, which software you're using to do it, and which other components you are using, will all play a part in determining what is better.

If all you want to do is experiment with leftover old parts, then this doesn't matter. Get a motherboard that supports two cards and hook it all up. Then you have your test system to mess with. If you are hoping for an a genuinely useful way to use old parts then that's where things get doubtful. I have three older videocards just sitting in their boxes. If there was a way to add them together into something useful I'd be using them right now.

Doesnt mean I cant find a way to somehow make them useful in the future :3

 

[k1114]

The longer it is the less likely. You can't even use 200 series or older with the 10 series because of drivers. Fermi has already lost support back in march so even using the 400 or 500 series requires old drivers. They won't be usable with the 11(or 20) series when they come out soon.

There's also the performance. You are getting even lesser of a fraction of the performance. More power and heat for relatively nothing. It's not even worth putting in my old 560ti because the performance is 1/3 but 150% power/heat of a 1060.

 

[StarlightDash]

The longer it is the less likely. You can't even use 200 series or older with the 10 series because of drivers. Fermi has already lost support back in march so even using the 400 or 500 series requires old drivers. They won't be usable with the 11(or 20) series when they come out soon.

There's also the performance. You are getting even lesser of a fraction of the performance. More power and heat for relatively nothing. It's not even worth putting in my old 560ti because the performance is 1/3 but 150% power/heat of a 1060.

I dont have a lot of money to spend a lot of the time. I got an FX 4800 quadro lately. One of the mane reasons I was asking about all this lately. Even if it doesnt work its another to add to the gpu collection.