PCI-e 3 making a difference here as well

[Terry Rock]

About a year ago, I accidentally stumbled upon the performance increase in RF2 when using PCI-e 3.0 vs PCI-e 2.0.
Today, during some over-clocking experiments, I discovered the same is true for AC...though not on quite as large of a margin.
With significantly higher CPU clocks (in my case 4.5 GHz on an I7-3820 SB-e with PCI-e 2.0 enabled and all AC settings maxed out for testing) Points=11352, fps Ave=77, fps Min=60, fps Max=109, Var=2, CPU usage=45
The CPU was then reset to stock speed (3.6 GHz and PCI-e 3.0 enabled...again all AC settings max) Points=11461, fps Ave=78, fps Min=63, fps Max=127, Var=3, CPU usage=49.
The results were compiled after running three test per setup with reboots between each test.
Interesting to see that +900MHz...even without throttling yielded less max fps and smoothness with 2.0 vs 3.0 at lower speed.

 

[Billy Pilgrim]

Interesting indeed. And thank you for the info.
Where/how do you change PCi-e? Are there problems associated with changing this?

 

[Terry Rock]

Interesting indeed. And thank you for the info.
Where/how do you change PCi-e? Are there problems associated with changing this?

I have a Biostar TP-X79 motherboard which is socket 2011 version1.
The option to change PCI-e modes is in the BIOS.
I still however, have to employ the registry hack because not all boards had the feature turned on by default due to a timing issue.
Seek out Force_enable Gen3 and if you have a PCI-e 3 capable MB and GPU from that generation, you should be able to activate it.

http://www.trickmasterpc.com/force-enable-gen3-pcie-3.html

 

[RaceWasGood]

Very interesting. How do you test the individual games to get the before/after results?

 

[Terry Rock]

Very interesting. How do you test the individual games to get the before/after results?

I ran three 'baseline' test of AC's benchmark using (stock speed 3.6 GHz) at PCI-e 2.0.
Turbo was disabled to keep the processor from possibly ramping up above that 3.6 GHz.
I then went to 125 BCLK x 36 multiplier for 4.5 GHz still at PCI-e 2.0 and ran three O/C test again using AC's in-game benchmark while noting the results.
A reboot was performed after each bench.
I then went back to the stock 3.6 GHz. turbo again disabled... but with PCI-e 3.0 enabled and repeated the benchmarks with reboots in between test.

 

[RaceWasGood]

Wow, that was a way more indepth answer than i expected I was just wondering how you got the actual points, average framerates and such? Thanks!

 

[unknwn]

Can you post just pci-e 2.0 vs 3.0 results (with same clocks and else) so we could see a clear difference between two? I find the inclusion of different clocks kind of confusing as there are no clear results provided for each of combination. Thanks.

 

[Matuszczak Piotr]

Can you post just pci-e 2.0 vs 3.0 results (with same clocks and else) so we could see a clear difference between two? I find the inclusion of different clocks kind of confusing as there are no clear results provided for each of combination. Thanks.

What's confusing? Turning the PCI-e 3.0 has similar results as overclocking his CPU by 900MHz.

 

[Dan Costa]

Hello

Can you bring down the PCI-e 3.0 speed down to 8x and test against PCI-e 2.0 16x ?

I'm wondering if its the bandwidth or something else.

 

[Terry Rock]

Hello

Can you bring down the PCI-e 3.0 speed down to 8x and test against PCI-e 2.0 16x ?
I'm wondering if its the bandwidth or something else.

I don't believe I can do that on this motherboard (PCI-e 3.0x8)... without physically moving the GPU.
My board is configured with 2 PCI-e 3.0x16 slots upper and center and 1 PCI-e 3.0x8 below.
Unfortunately, my GPU will never fit into that lower slot without some type of case modification.
I do believe the results are clear though.
There is no other explanation for the difference in performance since nothing else was chanced between the baseline 3.6 GHz test at 2.0 and the baseline at 3.0.

 

[Insert Coin]

Odd you see such a difference in performance. Other tests show hardly any difference between PCI-e 1.1, 2.0 and 3.0, see here.

 

[unknwn]

What's confusing? Turning the PCI-e 3.0 has similar results as overclocking his CPU by 900MHz.

The point is that the OP didn't provide what kind of boost you get by increasing those 900MHz in the first place. Personally I haven't benchmarked how AC performance scales with increased CPU speed. In my opinion results provided here are more or less useless unless you dig for other benchmarks with similar hardware and AC to get the whole picture.

I would like to see a single value provided (boost in percentage) when going from PCI-E 2.0 to 3.0. Could someone test it? Thanks.
Also it would be worth mentioning what kind of GPU is tested as it should be significant factor.

 

[Terry Rock]

The point is that the OP didn't provide what kind of boost you get by increasing those 900MHz in the first place. In my opinion results provided here are more or less useless unless you dig for other benchmarks with similar hardware and AC to get the whole picture.
I would like to see a single value provided (boost in percentage) when going from PCI-E 2.0 to 3.0.

The benchmark results were clearly stated in the opening post at stock 3.6 GHz at PCI-e 3.0 vs the overclocked 4.5 GHz at PCI-e 2.0.
Logic should dictate that if the CPU at higher clocks is not throttling through heat production and no other parameters have been changed or adjusted, the increase in performance at the lower clocks could only come from one thing....the switch from 2.0 to 3.0.
There is no silver bullet.

 

[unknwn]

The benchmark results were clearly stated in the opening post at stock 3.6 GHz at PCI-e 3.0 vs the overclocked 4.5 GHz at PCI-e 2.0.
Logic should dictate that if the CPU at higher clocks is not throttling through heat production and no other parameters have been changed or adjusted, the increase in performance at the lower clocks could only come from one thing....the switch from 2.0 to 3.0.
There is no silver bullet.

I am not trying to disprove that there is a performance difference PCI-E 2 vs 3 as you shown.
I am trying to understand how much performance boost one can gain. Is it 5% or 25%? Such result was not provided yet. Which is what I am asking.
As in rF2 some claimed to get 25% boost, yet some barely anything noticeable depending on the setup.

Also why dont you mention which GPU was tested? It would make sense to think that the more powerfull GPU the more it is bottlenecked by slow PCI-E.

 

[Matuszczak Piotr]

As in rF2 some claimed to get 25% boost, yet some barely anything noticeable depending on the setup.

As You say results will vary, do the tests Yourself and You will get the best answear.

 

[unknwn]

As in rF2 some claimed to get 25% boost, yet some barely anything noticeable depending on the setup.

As You say results will vary, do the tests Yourself and You will get the best answear.

I cannot do the test as I dont have PCI-E 3.0 in the first place. Which is why I am asking.

 

[Terry Rock]

Also why dont you mention which GPU was tested? It would make sense to think that the more powerfull GPU the more it is bottlenecked by slow PCI-E.

I have a Gigabyte GV-N970WF30C-4GD.
My OS is Win10-64 running on a Samsung 840 EVO.
You don't need to sit there doing a bunch of percentages...just use the results from the first post.
If it helps you....great! That is the objective.
The information based off my result, is purely about helping guys who have the option but don't have it turned on.