PCI Express Scaling Game Performance Analysis review

lanelor

2015-02-19 10:24

The results are pretty much what was expected comparing x16 throughout the generation. Still there are plenty of x16+x4 SLI/CF gaming motherboards and it should have been fun to see what is the actual loss there.

#5016535

fantaskarsef

2015-02-19 10:34

As always, thanks for picking up such basic topics Hilbert 🙂

#5016543

TheDeeGee

2015-02-19 10:39

Would have loved to see 8x Results aswell. I'm bound to 8x with my current Setup.

#5016549

fantaskarsef

2015-02-19 10:49

Would have loved to see 8x Results aswell. I'm bound to 8x with my current Setup.

Maybe I'm wrong, but wouldn't you just be able to take PCIe 2.0 16x as a reference then? Or am I misunderstanding?

#5016551

iancook221188

2015-02-19 10:50

loved this review and would love to see the corssfire review may be with X8 2.0 aswell because i think alot of people are running this, the results would prob be the same as gen 1.1 X16 im running my 780's with gen 2.0 at x8 im wondering if there getting restricted

#5016559

DiceAir

2015-02-19 11:04

Would like to see a test with crossfire

#5016573

rl66

2015-02-19 11:22

Would have loved to see 8x Results aswell. I'm bound to 8x with my current Setup.

more or less 2.0 x8 is the same as 1.1 x16

#5016579

rl66

2015-02-19 11:31

About the "why use the 5*** i7?" it's to have the limit of the port, with a lower CPU you would have the CPU limitation added (2x8 or 1 x16) this way you get rid of any other thing that can influence the test. don't forget it was to test the port ONLY. now that we have seen SLI i can't wait to see Xfire too. could be interesting to see the technologic difference on the port use.

#5016587

k3vst3r

2015-02-19 11:54

Article needs show three cards an crossfire too. NV cards cope better with less pci-e bandwidth. Basically if you covered three cards an tri-fire 290x the article would be golden.

#5016589

stevevnicks

2015-02-19 12:00

quote "f this article invokes enough interest, then in the future we'll also spend some time testing AMDs latest Radeon cards in Crossfire. With the Crossfire bridge removed all composing data now runs over the PCIe bus. That should place more load on bandwidth usage and that could have an impact on performance. So if there is enough interest, we'll dedicate at one point an article to that combo as well." yea that would be some nice data, be good to see the difference with how AMD cards fair.

#5016592

k3vst3r

2015-02-19 12:06

quote "f this article invokes enough interest, then in the future we'll also spend some time testing AMDs latest Radeon cards in Crossfire. With the Crossfire bridge removed all composing data now runs over the PCIe bus. That should place more load on bandwidth usage and that could have an impact on performance. So if there is enough interest, we'll dedicate at one point an article to that combo as well." yea that would be some nice data, be good to see the difference with how AMD cards fair.

Yeah this what on about, theirs more reliance on pci-e bandwidth on AMD cards with removal of the bridge, since communication is fully done across the pci-e lanes.

#5016594

stevevnicks

2015-02-19 12:10

Yeah this what on about, theirs more reliance on pci-e bandwidth on AMD cards with removal of the bridge, since communication is fully done across the pci-e lanes.

how are you tri firing with the Hero VI ? 1x295 and 1 x 290 ? because you have to disable all the pcie x1 ports to use the 3rd pcie 16 slot port at its max of x4 or it will only work at x1

#5016597

Hilbert Hagedoorn

Administrator

2015-02-19 12:15

Article needs show three cards an crossfire too. NV cards cope better with less pci-e bandwidth. Basically if you covered three cards an tri-fire 290x the article would be golden.

I'll probably look into Crossfire as well. 3-way multi-GPU (regardless of SLI or Crossfire) however poses an issue as the slots configuration even on X99 / 5960X would change to X16 / X16 / X8.

#5016602

k3vst3r

2015-02-19 12:21

how are you tri firing with the Hero VI ? 1x295 and 1 x 290 ? because you have to disable all the pcie x1 ports to use the 3rd pcie 16 slot port at its max of x4 or it will only work at x1

It has three 16x slots, basically using 16x slots that run at 8x pci-e 3.0. Then the bottom slot which is 4x pci-e 2.0 only. It works but I know it's suffering performance hit for doing so.

#5016604

stevevnicks

2015-02-19 12:21

I'll probably look into Crossfire as well. 3-way multi-GPU (regardless of SLI or Crossfire) however poses an issue as the slots configuration even on X99 / 5960X would change to X16 / X16 / X8.

yea three way x16 x16 x8 will skew the test ... unless everything was set at x8 for tri setup ? i dont know funny one that one 🙂 unless try 16x x16 x8 vs 8x 8x 8x to see the difference ? i done some testing years ago using the Asus M3A32-MVP DELUXE wi/fi Editon and 4x 3870 i dont know why 3d mark says only 1 card but it had 4 cards in there. http://www.3dmark.com/3dm06/10054832 http://www.hardwareheaven.com/community/threads/crossfire-x8-x4-and-gaming.227980/

#5016616

k3vst3r

2015-02-19 12:40

Technically could show it as 16x 16 x8, then as 8x 8x 8x on same pci-e 3.0. This would give base figure to move forward with. Then reduce it down to pci-e 2.0, then further down to gen 1 pci-e. With 3 cards though it's same for everyone, their limited to 40x pci-e anyway. So it would reflect real life tests anyway.

#5016621

Pale Rider

2015-02-19 12:46

You're preaching to the choir, Hilbert. The people that come to Guru3D are the very people that crave that extra 3 to 5% gain and will do whatever it takes to get it. Regardless of what you or anyone else says. We've all been raised to demand those extra few performance points from birth onwards. It's like telling a slalom skier that those few extra milliseconds really don't matter. Or a drag racer the same thing. We crave speed above all else in this hobby. We're obsessed with performance. Other than that, a good read. To be chuckled over and quickly forgotten surely. But, it is good to stretch our minds every once in a while, though. I'm curious as to how well bridgeless crossfire works and when/if it can be implemented with Nvidia cards.

#5016627

stevevnicks

2015-02-19 12:56

It has three 16x slots, basically using 16x slots that run at 8x pci-e 3.0. Then the bottom slot which is 4x pci-e 2.0 only. It works but I know it's suffering performance hit for doing so.

so where do you plug your sound card? i have same motherboard if you use the 2 16 slots at x8 and 3rd at x4 then all the PCIe x1 ports are disabled so the 3rd 16 slot can run at x4 ?

#5016629

TheDeeGee

2015-02-19 13:01

more or less 2.0 x8 is the same as 1.1 x16

I run at 8x Gen3, but i guess that's less than 1% difference compared to 16x.

#5016631

stevevnicks

2015-02-19 13:01

You're preaching to the choir, Hilbert. The people that come to Guru3D are the very people that crave that extra 3 to 5% gain and will do whatever it takes to get it. Regardless of what you or anyone else says. We've all been raised to demand those extra few performance points from birth onwards. It's like telling a slalom skier that those few extra milliseconds really don't matter. Or a drag racer the same thing. We crave speed above all else in this hobby. We're obsessed with performance. Other than that, a good read. To be chuckled over and quickly forgotten surely. But, it is good to stretch our minds every once in a while, though. I'm curious as to how well bridgeless crossfire works and when/if it can be implemented with Nvidia cards.

scrub that lol sozz read it wrong 🙂