Massive FPS gains in rf2 using PCI-e 3.0 x16 with higher end cards!

How this affects triple screen performance is the next unanswered question on our hit list.

 

Are you overclocking that 2500, they can run 4.0-4.5 Ghz easy with proper cooling.

 

The reason he needs an upgrade is not because his cpu is too slow (not remotely, it fact it's more than plenty even at stock clock for rf2) but that he can't get his high end graphics card to run in PCI-e 3.0 x16 mode with a 2nd generation intel Core cpu (e.g. i5-2500). The reason is simply that only 3rd gen and up have PCI-e 3.0 support. Over the last 1-2 weeks on the forum, through benchmarking members of the community have noticed some serious performance inconsistencies between users with different bases but with the exact same graphics card (including the same clocks), observing as big as 33% of a difference in performance/average-fps between different systems with identical graphics cards. The reason is with respect to rf2 only, PCI-e 3.0 x16 vs PCI-e 2.0 x16 makes a difference. Now, when you look at reviews comparing different PCI-e speeds, they use common mainstream games (which do make up the majority of games) to see if different PCI-e speeds affect gaming performance/fps and they've pretty much all shown either a zero or near zero difference. Either rFactor 2 is the exception to this rule and/or it's gotten to a point with new gen cards that it makes a difference (at least in rf2). Even my GTX 770 loses 6-9% fps going from 3.0 to 2.0 x16 in rf2.

 

Yep. I run it at 4.2. I was surprised some time ago to find that the OCing the cpu did little to improve fps compared to OCing my gpu. Though that was with my older 560ti card.

 

Difference of performance=6-9%. Difference in money between 2nd generation and 4th genereation intel= 50%. no sense. and we're talking to go from 100 to 106-109 fps.....

 

For sure 6-9% wouldn't cut it. But 15-20%... pretty tempting. Thanks for your post but i think I'll wait until our "research team" comes to a conclusion. Thanks again to them for making the effort.

 

what we see here with Rf2 has not been noticed in any other game...secondly, as far as I have read, only people with really high FPS has noticed this increase. In that case I can imagine that PCI3 brings some "benefits"... Why the quotes...Do you notice going from 120 to 140 FPS? not sure... On the other side... if it's free why not... would you upgrade for going from 120 to 150 fps in just one game... not sure either. I would not get overexcited, at least not yet.

I want to see somebody with lets say 50 FPS and do the PCI2<>3 Mambo.... I suspect that there won't be any significant changes in FPS. If we get 30% here, that would be amazing. (even though technically very hard to explain...)

I think Spinelli announced he would test this scenario.... let's see.

 

I have just tested in some DX11 games/simcades/sims and there is no such increase, in fact < 1% in most cases. I'm going to test on some of the older ones that use dx9 to see if it has some bearings, but not until later on over the weekend.

rF2 however seems to benefit greatly at the high end with 1080p resolutions and less so with higher resolutions.

 

With one card and three screens i hover at around 100 fps in a big field but that's with my res dumbed down to 4xxx x 900. Also without full everything. At native 5040x1050 i think remember seeing 80s-90s. Not wanting to be greedy but i think another 20 fps would have some benefits.

Btw rf2 is the only game i play

 

There is no bang for buck don't waste your money yet.


Any buyers depends how good your chip is, my 2500K reached 5.1GHz.


I don't even see the value of Z97 and K chips anymore it is better to spend the extra on more graphics grunt.




example ( australian dollars )
====================


Intel 4690 / 4690K + $30

H97 / Z97 + $30

Stock cooler / Corsair H80 + $130

Look at the difference from 900MHz extra overclock above turbo , its pitiful. lol

http://www.bit-tech.net/hardware/2014/07/03/intel-core-i5-4690k-review/5

Mind you GOOD LUCK on getting 4.8GHz stable with just any chip lol
Looks at reviews compare a normal overclock 4.5Ghz .....the gains are miniscule.



for budget buyer looking for best performance the above Non K setup leaves you 190 more to spend on graphics which in turn will give more bang for buck then the overclock............simple.

 

I already have the mobo and a decent cooler (I already OC my cpu). Just need the 3xxx ivy bridge. Just trying to get all the grunt i can from my 780ti ! Plus i'll bet i can get $75 for my 2500. But... not saying im for sure going to do it. Let's see what the rf2 fps gain is when the test is done with triples. That's the only benchmark that matters to me

 

I run almost the same res as triples about 3x 1080p = 6220800 pixels... I have 3840 x 1644 = 6312960 pixels on a single screen, so no Multiview.... but you have about 1 million less pixels running at 5292000, so should give you a good indication

I'm just about to run some tests on rF2 at this resolution on a single 980...now I have a stable OC, I'll use the bench that DrR1pper has put on the live benchmark thread apart from the resolution and test 2.0 @ 16x vs 3.0 @ 16x... I'm not expecting > 10% difference to be frank, but lets see.... it still is relevant to this thread

 

$75 is cheap for any 2500K.

But all 2500K are not equal, mine does 5GHz stable ...that is worth dollars to someone that needs a 1155 processor.

I would want at least $150 for mine with new stock cooler ( $130US)

It still be faster in gaming then any $200 i5

http://www.ebay.com/itm/Intel-Core-...Fan-/191288733724?pt=CPUs&hash=item2c89b2701c

Cough ...cough and mine did 5.0GHz rock solid on 1.36v ( see Pic) hehehe
Plus that was on air ( CM Hyper 212+ ...not a Corsair H80i ) lol

Mine did 5GHz@1.36v with corsair H60 no worries but I never even liked pushing it that hard

I settled on 4.5GHz@1.28v myself

========================


Just saying I starting to question the value of gain vs dollars from overclocking almost exclusive for rF2 and a few other sims.

I think I can do better spending more on GPU next time.

 

I do wonder why this mostly happen just in rf2. Anyway.

Im on 2500k @4.5ghz stable. And as pointed out earlier. I can't get 3.0 with my gen.

So I do wonder if it's wise to upgrade gpu before an cheap upgrade to next gen mobo/cpu.

Well, My eyes are on a an gtx970, so it will be a huge improvement for me anyway. But missing out on those 20-30 percent.. grr!

 

Sorry guys!!! I was busy all day, then a friend came by and he wanted to play rF2 online for like 5 hours, lol (he's a "mainstream", console gamer but is in love with rF2's physics and FFB).

Anyways, I'll be doing all those PCI-E tests (triple screen, multiview, SLI-3D, 1xGPU-2D) tonight. Give me a few hours.

Thanks for the typo and result correction Dr.R.

 

I've got say thanks to you DrR1pper this PCI-e 3.0 x16 mode gave me Massive FPS gains in rf2, I use Lime Rock Park as my benchmark because i always get stutter there, if i don't achieve more than 80fps. PCI-e 3.0 x16 mode gave me a minimum 40fps boost, this plus Sim Commander4 has put rFactor2 back on top of my driving list.

 

Here are the results from last nights testing:

DrR1pper's benchmark at 3840x1644 resolution, 4xAA forces in NCP and pre-render frames =1

GPU core @ 1550Mhz, Mem @ 1900Mhz

PCI-E @ 2.0 x16
Avg: 89.875 - Min: 74- Max: 103
Avg: 86.068 - Min: 71 - Max: 98 (with FXAA in game)

PCI-E @ 3.0 x16
Avg: 103.173 (+14%) - Min: 88 (+18%) - Max: 113 (+9%)
Avg: 97.983 (+13%) - Min: 84 (+18%) - Max: 108 (+10%) (with FXAA in game)

The most important one for me is 'min' and the yield of 18% is pretty high... and that's just because of PCI 3.0. Initially my system wasn't set up correctly to reflect accurate results, but with the help from this and the live thread, I've sorted that out

Thing to note also, on the final test with FXAA on, PCI 3.0 @ x16 my vRAM was pushed to 3298MB... my 780 would have been screwed at that point...it only had 3GB!

Also to note, I have yet to find any gain at all between PCI 2.0 and 3.0 on DX11 games/sims whatever. and I tried to test on some DX9 games, but could not get fraps to work with them last night.

 

Benchmark Results (2D-1xGPU and 3D-SLI)

2x GTX 780 Ti @ 1163 MHz core, 1750 (7000 effective) MHz memory, both GPUs at simultaneous full 16x PCI-E (3.0 or 2.0)
i7-4930K @ 4.2 GHz

I did tests with two different graphics settings. I used the "official" benchmark settings as the base gfx settings; anything changed from those settings are listed below. I chose to include the second test (Test B) because the "official" benchmark settings are resulting in very high framerates from most people ( 125 - 190 avg fps) and I therefore felt the need to include a more demanding test. Test B's graphics settings give very consistent runs; I did many tests to verify this.

3D framerates are listed as total frames rather than per eye; this makes for a better comparison since we are comparing the total frames-per-second that the GPU/GPUs can render, regardless of which monitor/eye it sends those frames to.

[HR][/HR]
Test A - (triple screens [5760 x 1080], multiview)

2D-1xGPU

PCI-E 3.0 @ 16x:
Avg: 75 - Min: 60 - Max: 86

PCI-E 2.0 @ 16x (equivalent to 3.0 @ 8x):
Avg: 58 - Min: 45 - Max: 69

FPS Increase:
Avg: 29% - Min: 33% - Max: 25%


3D-SLI

PCI-E 3.0 @ 16x:
Avg: 134 - Min: 110 - Max: 154

PCI-E 2.0 @ 16x (equivalent to 3.0 @ 8x):
Avg: 96 - Min: 74 - Max: 112

FPS Increase:
Avg: 40% - Min: 49% - Max: 38%

[HR][/HR]
Test B - (triple screens [5760 x 1080], multiview, 8xAA, 12AM, rain)

2D-1xGPU

PCI-E 3.0 @ 16x:
Avg: 52 - Min: 42 - Max: 59

PCI-E 2.0 @ 16x (equivalent to 3.0 @ 8x):
Avg: 43 - Min: 34 - Max: 52

FPS Increase:
Avg: 21% - Min: 24% - Max: 13%


3D-SLI

PCI-E 3.0 @ 16x:
Avg: 97 - Min: 78 - Max: 110

PCI-E 2.0 @ 16x (equivalent to 3.0 @ 8x):
Avg: 76 - Min: 60 - Max: 90

FPS Increase:
Avg: 28% - Min: 30% - Max: 22%

[HR][/HR]
NOTE: The SLI scaling when using Nvidia 3D Vision is absolutely fantastic and near perfect. Every PCI-E 3.0 @ 16x SLI result (avg, min and max) shows an 80-86% framerate improvement. However - as good as that result sounds - that number (80-86 %) is actually misleading - it should be much closer to an amazing 100%. Why? Because the SLI tests were done in 3D mode and 3D takes a bit of extra power on it's own, while the single-GPU tests were done in 2D mode (3D Vision fully disabled) and therefore did'nt have to deal with the slight, extra GPU demand of 3D.

 

Thanks so much to all you guys doing these tests. I would call the gain with PCI-e 3.0 when using triple monitors substantial. 20-30% for Spinelli in 2D single card. Mid to high teens for Mr. Pix with his slightly different test. (Holy cow that 980 kicks butt!)

One thing I'm curious about is the FOV setting in the test. The screen shot shows "default". What does that mean? How would that differ for Mr. Pix at 3840x1644 and Spinelli at 5760x1080?

I think there is a new (or maybe used... Don't shoot me Durge!) Ivy Bridge in my future.

Edit... As I ponder I notice Spinelli using Multiview which would certainly tip the scale in Pix's favor. Granted, it's not a test of card v card but rather 3.0 v 2.0. But I need some reason for not kicking myself for buying the 780ti early!