Status
Not open for further replies.

killerskulls

Commendable
Jan 31, 2017
2
0
1,510
So i have a 1080 classified as my main card, and was going to use my 980 classified as a dedicated physx card. However, ive seen people say its not worth doing so anymore, but they dont say why. I would have the 1080 in a 16 slot, and the 980 in an 8 slot. ive seen a few posts saying that could reduce performance in total? im kinda just looking for some clarification on this process.
 
Solution
It looks like the BEST CASE is about 15% (Batman AK) on average, but I found other evidence that showed over 25% during heavy PhysX scenarios. It's still hard to justify keeping the card in the case IMO.

In THEORY you could reduce performance but only with a weaker dedicated PhysX card.

You're basically switching from the GTX1080 to the GTX980 to do the PhysX, physics processing. If you get a huge explosion or some big physics even that managed to use 20% of the GTX1080, but do that on the GTX980 instead then you FREE UP that 20% of the GTX1080 to handle the rest of the game.

So 100/80 = 1.25, means in theory you might see up to 25% or so gains during heavy events.

(I'd still rather TWEAK the game settings slightly to maintain a...

Gregg Eshelman

Reputable
Mar 25, 2014
46
0
4,540
First off you either have to use old drivers or hack the newer ones. Second, it will only work with specific games that can use a card separate from the main GPU for PhysX. So if the PhysX capability integrated into your main GPU is better than in the secondary card you want to use for PhysX, there's no point to doing it.

The only reason you'd want to do this is if your main GPU doesn't have PhysX, in which case it has to be a pretty old nVidia card and would be worse than your 980.

Here's a video https://www.youtube.com/watch?v=H9nZWEekm9c
 
1) You absolutely CAN gain some performance in SOME games. (and in some games you can choose different PhysX levels)

2) You would need to TEST this for yourself by using an in-game benchmark or repeatable scenario (that involves a PhysX event or events) by comparing:
a) GPU PhysX on GTX1080
vs
b) GPU PhysX on GTX980

(change in the NVidia drivers and record FPS for both scenarios)

3) Worth it?
That depends what results you create, but you are adding NOISE, HEAT, and you may not have many games using PhysX that justify this.

OTHER:
Hack the newer drivers?
Are you sure about that? I doubt it.
 
It looks like the BEST CASE is about 15% (Batman AK) on average, but I found other evidence that showed over 25% during heavy PhysX scenarios. It's still hard to justify keeping the card in the case IMO.

In THEORY you could reduce performance but only with a weaker dedicated PhysX card.

You're basically switching from the GTX1080 to the GTX980 to do the PhysX, physics processing. If you get a huge explosion or some big physics even that managed to use 20% of the GTX1080, but do that on the GTX980 instead then you FREE UP that 20% of the GTX1080 to handle the rest of the game.

So 100/80 = 1.25, means in theory you might see up to 25% or so gains during heavy events.

(I'd still rather TWEAK the game settings slightly to maintain a solid 60FPS or whatever my goal was then stick in the GTX980, but you can at least EXPERIMENT for fun)
 
Solution

Gregg Eshelman

Reputable
Mar 25, 2014
46
0
4,540
Yup, hack the drivers. Watch the video. Using a separate GPU for PhysX is disabled, I assume mainly to keep people from using those old and now pitifully wimpy standalone PhysX cards, which will slow things down.
 

killerskulls

Commendable
Jan 31, 2017
2
0
1,510


Yea, im kinda against it after some tests. Ive noticed in Borderlands 2, a very phsyx heavy game (if you have the setting on) and the game doesnt run smooth. After watching that one video Gregg provided it makes sense. so i think im going to take it out. Thanks for the help!
 
Status
Not open for further replies.