Firestrike, PCI-E 1.0 and Physics score

Kaapstad · 31 Jan 2014 at 20:17

Got a real odd one here

I was running a single 290X @1200/1600 with a 4930k @4.7 on the Firestrike bench and got some really strange Physics scores when switching between PCI-E 1.0, 2.0 and 3.0

3dmark Firestrike PCI-E 1.0 x16

Physics Score 17760

http://www.3dmark.com/fs/1637834

3dmark Firestrike PCI-E 2.0 x16

Physics Score 15141

http://www.3dmark.com/fs/1637900

3dmark Firestrike PCI-E 3.0 x16

Physics Score 15090

http://www.3dmark.com/fs/1637972

My memory setup.

The question is why is the Physics score so much higher when using PCI-E 1.0

I only get this problem with the Firestrike bench, if I run 3dmark11 Physics test everything is normal.

Thanks.

8 Pack · 31 Jan 2014 at 21:55

Honest answer I am not sure. I have never tested myself because obviously GT's would be incredibly low with PCI X 1.0. PCI X 2.0 tends to be best for overall when benching Nvidia and PCI 3.0 When benching AMD.

You tried dual channel. this is better for all tests in FSX and FS.

Looking at your mem timings I would try work on TWCL. With those Sammys (Dom Platy 2400) You should be able to go TWCL 6 easy at 2400.

Rroff · 31 Jan 2014 at 22:09

Could be a queue depth thing, where newer pci-e revisions wait for more data in the queue before commiting, which would give overall better performance, but penalise specific scenarios but I'm just guessing.

Kaapstad · 31 Jan 2014 at 22:12

8 Pack said:
Honest answer I am not sure. I have never tested myself because obviously GT's would be incredibly low with PCI X 1.0. PCI X 2.0 tends to be best for overall when benching Nvidia and PCI 3.0 When benching AMD.

You tried dual channel. this is better for all tests in FSX and FS.

Looking at your mem timings I would try work on TWCL. With those Sammys (Dom Platy 2400) You should be able to go TWCL 6 easy at 2400.

Thanks for quick reply

I will have a look at TWCL

I will also give dual channel a go next time I run FS and FSX.

wazza300 · 31 Jan 2014 at 22:22

could it be handing more usage over to the cpu and restricting the gpu in the combined test due to pcie1? cant remember if the physics score is based on combined or not

does seem weird tho

8 Pack · 31 Jan 2014 at 22:34

No physics is just CPU only mate. Combined is both.

I think with less stress on the CPU from not having to allow massive bandwidth to the graphics cards it can a lot more onboard cache and thus processing power to the Physics test and hence a better score. I will test myself when I get time Monday maybe. have just set up DC2 for some more LN2 action.