Hi all
I'll be taking my watercooled system apart over the next couple of weeks to rebuild it with a different airflow configuration. At the same time I want to move the gfx cards about as I have found out that the bottom 8xPCI-E slot holds the master card in a tri-SLI configuration, not the top 16xPCI-E that I expected.
Each card is connected to a 24" monitor for a surround setup.
Now, I have 3 cards that each behave differently.
C1: Ocuk(Palit) gtx680. Consistently runs 26MHz core above the 2 others. Power usage approx +7-9% higher than the others. Temp 4-5c above the others
C2: Zotac gtx680. Runs same speed as C3. Uses 2% more power than C3. runs 1-2c hotter than C3.
C3: MSI gtx680. 48c on load. core 1100Mhz.
So the question now is which card to use as master. Here is what I came up with:
Top PCI-Ex16 : Sync with Master : MSI : PhysX card
Middle PCI-Ex8: Sync with Master : Zotac
Bottom PCI-Ex8: Master : Ocuk
It looks to me like the master card runs one or two percent below what the sync cards does possibly due to the sync cards having to keep in sync with the master. I have tried finding info about this on google but cant really find anything about the so I have not confirmed thats how it works. If it does then putting the Ocuk card in the master slot at the bottom would free it from having to keep in sync, releasing a few % to raise the baseline as it were.
MSI card goes in the x16 slot and gets designated as the physx card. Its the coolest running card so should be better able to cope with physx and also, gets the x16 slot because of that.
Zotac card goes in the middle on the remaining x8 slot.
If the master - sync relationship does not work as I assume above then the positioning could be different. If the master card has the highest workload then the MSI card needs to be in the bottom slot....
Decisions, decisions....
Anyways - hope someone has a few pointers on what, if any, performance differences there are between a master and the sync systems in tri-SLI.
I'll be taking my watercooled system apart over the next couple of weeks to rebuild it with a different airflow configuration. At the same time I want to move the gfx cards about as I have found out that the bottom 8xPCI-E slot holds the master card in a tri-SLI configuration, not the top 16xPCI-E that I expected.
Each card is connected to a 24" monitor for a surround setup.
Now, I have 3 cards that each behave differently.
C1: Ocuk(Palit) gtx680. Consistently runs 26MHz core above the 2 others. Power usage approx +7-9% higher than the others. Temp 4-5c above the others
C2: Zotac gtx680. Runs same speed as C3. Uses 2% more power than C3. runs 1-2c hotter than C3.
C3: MSI gtx680. 48c on load. core 1100Mhz.
So the question now is which card to use as master. Here is what I came up with:
Top PCI-Ex16 : Sync with Master : MSI : PhysX card
Middle PCI-Ex8: Sync with Master : Zotac
Bottom PCI-Ex8: Master : Ocuk
It looks to me like the master card runs one or two percent below what the sync cards does possibly due to the sync cards having to keep in sync with the master. I have tried finding info about this on google but cant really find anything about the so I have not confirmed thats how it works. If it does then putting the Ocuk card in the master slot at the bottom would free it from having to keep in sync, releasing a few % to raise the baseline as it were.
MSI card goes in the x16 slot and gets designated as the physx card. Its the coolest running card so should be better able to cope with physx and also, gets the x16 slot because of that.
Zotac card goes in the middle on the remaining x8 slot.
If the master - sync relationship does not work as I assume above then the positioning could be different. If the master card has the highest workload then the MSI card needs to be in the bottom slot....
Decisions, decisions....
Anyways - hope someone has a few pointers on what, if any, performance differences there are between a master and the sync systems in tri-SLI.