Live Performance Benchmarking Comparison for rFactor 2

I'm just clarifying that people with Ivy Bridge-E CPUs (4960X, 4930K, 4820K) don't need any update, and that it's just the Sandy Bridge-E CPUs (3960X, 3930K, 3820K) that do.

This is a very specific and detailed thread with very specific and detailed analyzing, details, results, etc. so I thought that I should point that out. Plus, I'm trying to prevent any Ivy Bridge-E owners from being mislead into thinking they need some sort of update when they don't

 

Yeah, I've now edited the original post to make that clear, thanks.

 

Yeah, that just won't cut it i'm afraid. You're going to have to gut the soundcard out the other PCIE slot if it's preventing you're graphics card lane from working at PCI-E 3.0 x16.

Also, the OP has been updated with all the details of this very important discovery and added to the beggining of the thread to highlight the importance before conducting any future benchmarks and even playing rf2 ever again.

 

Or he can get an LGA 2011 socket CPU (Sandy Bridge-E, Ivy Bridge-E, Haswell-E) 40 PCI-E 3.0 @ 16x lanes (well except the 5820K that's been gimped by Intel to 20 or 24). SLI GPUs both running at PCI-E 3.0 @ 16x + PCI-E soundcard

Sorry, just had to rub that one in

On a more serious note though, PCI-E 3.0 @ 8X is equivalent to PCI-E 2.0 @ 16x, almost every benchmark ever done has showm differences of like 1-3 %, and even bringing it down to PCI-E 3.0 @ 4X - which is equivalent to PCI-E 2.0 @ 8X and PCI-E 1.0 @ 16x - only brought around 5-10 % differences. These rF2 results therefore blow my mind actually; it just doesn't seem right, especially if you look at the theoretical bandwidth of PCI-E 3.0 @ 8x / 2.0 @ 16x.

But sometimes it's more complex than that, and therefore the theoretical, "on paper" stuff isn't always the "final word".

 

This actually work one time, but then i can not replicate it in about 10 min when i close everything down ad try again. And first time when work radeon pro seems not working correctly like without any effect from radeon pro. Anyway thnx.

Try already 100 times , have no effect, thnx.

I am sure i setup everything correctly. And i uninstall CCC asumming they can maybe interfere, because both application have access to the same functions. So i want eliminated possibilities of this situation if they perhabs exsisting. Also i am sure i don't uninstall the driver, i am not that kind of noob As i write this i already try your suggestion without luck, i assuming something in the rfactor instalation interfere with radeon pro because any other game i have work normaly with radeon pro, but i am really not in mood to make fresh install only because for thist and you are right, something is really odd here Thnx anyway.

 

Avo1977 and LesiU (and anyone else who has access to a GTX 770), i'd still really appreciate it if i could get performance data from the card in this benchmark please. If you can also find out what PCI-e mode and speed the card is running at it would be of great help to me. Thanks!

In fact, my GTX 970 at PCI-e 2.0 x16 is performing at only 88% of what it should be compared to PCI-e 3.0 x16 users. Since my GTX 770 is anywhere between 67-77% the performance of a fully functioning GTX 970, i would predict that my GTX 770 is actually performing as it should be (i.e. is not bottlenecked by PCI-e 2.0 x16 or PCI-e 3.0 x8). This will be very interesting to confirm true or not and perhaps then we can then figure out by extrapolation what is the bandwidth threshold requiring PCI-e 3.0 x16 in rf2 specifically and can therefore know which cards (with more than a certain bandwidth level) will at minimum require PCI-e 3.0 x16 for rf2 to take full advantage of it.

 

Try it yourself, you should be able to force gen2 in the BIOS.

Sent from my SM-G900F using Tapatalk

 

oh yeah, lol.

 

Kamikaze, don't worry, RadeonPro does not interfere with AMD CCC, just like Nvidia Inspector does not interfere with Nvidia's control panel. Plus, there are some things in AMD CCC and Nvidia's control panel that are not in RadeonPro and Nvidia Inspector, like adjusting monitor resolutions, refresh rates, colour/brightness/gamma, enabling/disabling monitors, setting up triple screens, enabling 3D Vision (Nvidia), media player settings for videos/movies, etc.

 

Ahum, PCIX 16x did quite alot 145 avg to 179 which is I think a 23% performance increase.

 

Ya man. If you research it almost everyone, from pros to ameteurs, from forum to forum, from website to website, all say that PCI-E 3.0 @ 8x / 2.0 @ 16x has sooooo much bandwidth, our games and GPUs aren't close to saturating it yet especially with just a single GPU and/or a single screen, look at these tests, look at those tests, almost no difference, you're worrying about nothing, 3.0 @ 16x is just for future proofing ATM, and on and on....

I feel like "crashing" all those threads and showing them our results here to shut everybody up, lol.

 

Ok just tested on a total fresh separate install

Core 1316MHz (my stock boost clock)
Time: 66706ms - Avg: 148.562 - Min: 121 - Max: 169

Core 1278MHz (Sentri's clocks)
Time: 66878ms - Avg: 146.371 - Min: 119 - Max: 166

Doh just started reading the lastest posts about pci-e 2/3 and 16/8x etc. I always did wonder about about that in spite of all the tests done by other sites for a single gpu at the most popular resolution (1080p) it made absolutely no difference whatsoever.

 

DrR1pper, your first post says that Intel 3000 CPUs are fine, but Sandy Bridge-E (3960X, 3930K, 3820K) CPUs need the registry fix, as Techade pointed out.



I would like to report that I enabled realroad (your test has it disabled/set to static), time scale (day/night/light cycling), mechanical failures, tyre wear, and fuel usage. The settings made no differences to your benchmark's framerates. I enabled and tested the settings individually and all at once.

I also set rF2 to 64 sounds (what I normally use, but I use 32 for the benchmark as that is default) and set it to 5.1 instead of stereo (2.0). Again, no difference in framerates in your benchmark. I even enabled all the session and audio option changes simultaneously, but still no framerate difference in your benchmark.

I just thought i'd let everyone know in-case anyone was wondering.

 

In my case I have PCI-E 2.0 16x option only

 

Thanks for prompting me. Btw, are you sure the CPU needs that or it's the Sandybridge-E/X79 board that needs it?

 

More fuel for the Gen2/3 debate....

CPU: i5-3570K @ 3.6GHz
rFactor 2 Build: 860
Graphics driver version: 344.11

GPU: MSI Gaming GTX 970 - 1316 core/1752 mem

Core 1316MHz (my stock boost clock)
Time: 66706ms - Avg: 148.562 - Min: 121 - Max: 169 - PCI-E GEN3
Time: 66722ms - Avg: 126.450 - Min: 100 - Max: 150 - PCI-E GEN2 FORCED IN BIOS

 

Very nice. You regained 15% fps performance!

I think i'll make a fresh thread highlighting the importance of checking this for users with reference to this thread.

 

My Sandybridge-E (3930K) on X79 needed it.

Sent from my SM-G900F using Tapatalk

 

Opps, i meant to pose the question as "Btw, are you sure the CPU needs that or it's the Sandybridge-E/X79 board that needs it?" So i know exactly what to say in the OP regarding this.

This is what i've added atm:

Is that ok/sufficient?

 

X79 motherboards themselves are PCI-E 3.0...

- Sandy Bridge-E (3960X, 3930K, 3820K) is basically the 6/12 core/thread version of Sandy Bridge (eg. 2600K, 2500K). (Well, the 3820K has 4/8 cores/threads, not 6/12, but i'm just trying to simplify things here)
- Ivy Bridge-E (4960X, 4930K, 4820K) is basically the 6/12 core/thread version of Ivy Bridge (eg. 3770K, 3570K). (Well, the 4820K has 4/8 cores/threads, not 6/12, but i'm just trying to simplify things here)

IB/IB-E is the successor to SB/SB-E. IB/IB-E comes with PCI-E 3.0 "out of the box", while SB-E needs the registry update, and with SB (non-E) PCI'E 3.0 just isn't possible.

Therefore, just Sandy Bridge-E needs the registry hack. Not Ivy Bridge-E. Again, they both use X79 motherboards, and all X79 motherboards are fully PCI-E 3.0 compatible "out of the box", the CPU you throw on that motherboard (IB-E or SB-E) is where the difference lies.

Get it?