Live Performance Benchmarking Comparison for rFactor 2

IMPORTANT UPDATE - YOU MUST READ THIS BEFORE PERFORMING BENCHMARK OR EVEN PLAYING RF2 EVER AGAIN!

Before we get into the benchmark details, something EVERYONE needs to know and do (if possible) is to make sure their graphics card is running in PCI-e 3.0 x16 mode on their motherboard! We've previously been having differences of performance with the same graphics cards in different user systems when running rf2 and we now know it's due to some user systems running their graphics card in PCI-e 2.0 x16 mode or PCI-e 3.0 x8 (or even less). Massive credit and thanks to TechAde for figuring it out! Big double thumbs up to the man!

Things you need to know, for Intel users, PCI-e 3.0 can only be enabled if a) you're motherboard supports it (which it should as standard by now) and b) you have a 3rd generation (or greater) Intel CORE cpu (e.g. i5/i7-3000/4000 series cpu). But it's also not good enough to be just running in PCI-e 3.0. It also has to be PCI-e 3.0 x16 mode. Reason being that PCI-e 3.0 x8 has the same bandwidth as PCI-e 2.0 x16 and if you're graphics card is being limited in rf2 by PCI-e 2.0 x16, it will also for the 3.0 x8.

Also, this is from TechAde: "In order to enable PCIe 3.0 on Sandybridge-E/X79 board you need to run a tool from NVIDIA, available here: http://nvidia.custhelp.com/app/answers/detail/a_id/3135/~/geforce-gen3-support-on-x79-platform ". And additionally to that from Spinelli: "All IB-E and SB-E CPUs use X79 motherboards, and only X79 motherboards. Well, all X79 motherboards, like all IB-E CPUs, come with PCI-E 3.0 "out of the box". It's just the SB-E CPUs that need the registry update because only SB-E CPUs don't come with PCI-E 3.0 "out of the box"."

From the research i have done, it would seem that no AMD cpu currently supports PCI-E 3.0 which means if you have a fast enough graphics card that would benefit from PCI-e 3.0 x16 in rf2 and if you have an AMD cpu, then it's highly unlikely you will be able to enabled PCI-E 3.0. Sorry.

I have an i5-2500k and although my mobo supports PCI-e 3.0, i simply can't enabled it. The problem is that PCI-e is integrated into the cpu itself and anything below 3rd generation Intel CORE cpu's only have PCI-e 2.0 support integrated into them. It sucks but it is what it is.

Before you even look at the below benchmark, please check you're card is running in PCI-e 3.0 mode or else you are at high risk of losing at least 16% performance (if not more).

You can do so by downloading GPU-Z and checking the "Bus Interface". Here's random pic i've pulled from the internet showing you what GPU-Z says when your graphics card is in fact running in PCI-e 3.0 x16 mode.

GPU-Z download link here: GPU-Z download



Final note however is that not* all graphics cards will benefit from PCI-e 3.0 x16 versus PCI-e 3.0 x8 or PCI-e 2.0 x16 or lower. I have no means of easily being able to tell you whether you card will benefit from it or not but if you card is within 1-2 years old and is a high end card i would wager it being very possible. You can't look to other games for reference if it's required or not because each game (and benchmark) have different hardware requirements. Some games/benchmarks will perform the same with or without PCI-e 3.0 x16, however rf2 has real benefits if you card is powerful enough. As an example, the new GTX 970 card drops 13% fps performance when using PCI-e 2.0 x16 vs. PCI-e 3.0 x16.

[HR][/HR]
What is needed:

- rFactor 2 (Pre-paid or Demo version)
- Honda Civic 1.07
- Silverstone 1.14
- Fraps: http://www.fraps.com/download.php


Graphics Settings (copy exactly):






Car and Track Layout Selection:






Practice conditions (copy exactly):




Nvidia/AMD Control Panel Settings:



Set the equivalent for AMD Control Panel: 4xMSAA and "Override any application setting".

The reason for setting "Max pre-rendered frames" to "1" on the nvidia control panel is to ensure that the CPU does not bottlenecking performance. It'll lower you CPU utilisation, preventing any cores from clipping 100% during the benchmark. This doesn't happen on all systems but it can do for some and to avoid any chance of it happening completely, nvidia users should set "Max pre-rendered frames" to "1" on the nvidia control panel. Here's an example of the cpu utilisation effects of "Max pre-rendered frames" at default to "1": The top graph bottlenecked my performance and saw reduced average fps. Some user CPU's are clocked fast enough to avoid this problem from happening all together but you should still used it even outside benchmarking as it can only be a benefit to you, especially in other rendering conditions that demand more CPU.

Example pic:


For AMD users: Below are instructions for enabling AMD's equivalent "Max pre-rendered frames" to "1". Please do this so we can avoid any cpu bottlenecking issues affecting your benchmark performance.

Benchmarking Instructions:

- After doing all the above, click "Race" in the main menu to enter the practice session.
- Click "Race" to enter the car.
- Press "I" key to enable AI driving.
- Start the FRAPS benchmark when crossing pit exit line.
- Stop the FRAPS benchmark after crossing the finish line.
- Perform at least twice to confirm consistency and post you highest average benchmark.
- Please state your PCI-e mode as well please (newly added)


Example results:

CPU: i5-2500K @ 4.3GHz
rFactor 2 Build: 860
Graphics driver version: 344.16
PCI-e mode: 2.0 x16 (newly added)

GPU: MSI GTX 970 Gaming 4G (manual overclock) - 1489 core/1847 mem (+311/+95)

Time: 67564ms - Min: 108 - Max: 165 - Avg: 138.121


Additional Note:

Please write your core and mem clock speeds in their load state. You can use tools such as GPU-Z, Afterburner, etc, to find this out.

 

CPU: i5-2500K @ 4.3GHz
rFactor 2 Build: 860
Graphics driver version: 344.11


GPU: GTX 770 (factory clock) - 1085 core/1750 mem

Time: 67174ms - Min: 77 - Max: 113 - Avg: 97.240


GPU: GTX 770 (manual overclock) - 1241 core/1852 mem (+156/+102)

Time: 67533ms - Min: 82 - Max: 121 - Avg: 103.505

 

PLEASE IGNORE - THESE ARE BAD RESULTS DUE TO USING PCI-E 2.0 X16 MODE BOTTLENECKING THE GRAPHICS CARD IN RF2.

CPU: i5-2500K @ 4.3GHz
rFactor 2 Build: 860
Graphics driver version: 344.16


GPU: GTX 970 (factory clock) - 1178 core/1750 mem

Time: 67ms - Min: 98 - Max: 145 - Avg: 123.077


GPU: MSI GTX 970 Gaming 4G - 1291 core/1750 mem (+113/+0)

Time: 67548ms - Min: 102 - Max: 153 - Avg: 128.901


GPU: MSI GTX 970 Gaming 4G (manual overclock) - 1489 core/1847 mem (+311/+95)

Time: 67595ms - Min: 108 - Max: 165 - Avg: 138.146

 

Think we might need some more 970 scores, I can't see what I've missed (if anything) because I'm getting averages of ~159fps at 1316core/1752mem on a 3570K@3.6, 344.11whql

 

I7 920 @ 3.5GHZ
Rfactor build 860
AMD driver 14.30.1005-140827a-175330E

GPU: Gigabyte AMD 7990 6GB 1000 core\1500 mem

2014-09-28 21:14:49 - rFactor2
Frames: 8421 - Time: 67923ms - Avg: 123.979 - Min: 101 - Max: 143

Ripper, your 770 is about on par with a 7970 ghz edition. Which is what my card is running at due to crossfire not scaling at all with Rfactor 2 even though both cores where utilized between 91 and 94 percent the entire run. Count in the fact that my I7 is first gen at 3.5ghz vs your second gen at 4.3ghz

How did your 970 score worst though? That should be like 150 to 160 average?

Now that I think about it. Maybe there is a little bit of scaling going on with my card. 10 percent maybe according to your bench.
I notice your core and mem speeds where higher than mine as well.

Of course I still feel rfactor 2 doesn't utilize properly. I should have 100 percent utilization.

 

Me and you avatar are sharing the same facial expression

Can you reset your nvidia control panel settings to default and then set up 4xAA for rf2 in there. Repeat the bench.

If not change, try removing your player.json file so it creates a new one and see if any change then.

Providing we are definately using the graphics settings, the only other real difference i see is our drivers. I'm using the 5 day newer 344.16 drivers.

 

This could be the difference between PCI 2.0 and PCI 3.0.

MJP is your board PCI 3.0?

Edit:
Then again maybe not, Ripper you said you had no issues in performance in other games correct?

 

I'm on PCI-E 3.0, so i don't think it could be this.

 

Didn't notice it before but it seems the car shadow wasn't being cast onto the track as it should, it was just a plain black rectangle. Only just noticed it in the replay monitor, tried with a fresh JSON and the car shadow's being cast as it should and I'm down 11fps lol. Put the previous JSON back in and the car shadow seems to be behaving as it should, not sure why it wasn't before!

So now I have....

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

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

Time: 66800ms - Avg: 148.129 - Min: 121 - Max: 169

 

That's better but you're still scoring much higher than me. 148 fps vs my 128.9.

Can you make another install and test from there by any chance?

 

Avg: 145.375 - Min: 122 - Max: 164

CPU: I7 4790k 4Ghz
GPU: Zotac GTX 970 1227mhz Core/1753mhz mem
rFactor 2 Build: 860
Graphics driver version: 344.16

 

Can you provide your the time(ms) as well please.

Also are you 100% sure that your gpu clock is at 1227 during the benchmark?

My performance is less than yours or MJP's.

 

MJP, would you mind repeating the bench at Sentri's clocks of 1227/1753 please.

Here are my scores vs both of yours:

1227mhz core/1753mhz mem
Min: 100 - Max: 149 - Avg: 126.399 (86.7% your Sentri's performance)


1316mhz core/1753mhz mem
Min: 103 - Max: 154 - Avg: 130.022 (87.8% your MJP's performance)


So you two are getting correct relative performance scores per your clock speed settings to one another but i'm losing out 12-13% performance. Hmmm

Either something is wrong with my hardware (but i don't think it is given how 3dmark scored perfectly) or somewhere in rf2 we have a different setting.

I did perform my benchmarks with a brand new clean install of rfactor 2 so all files are default. Would you guys mind doing me a massive favour of installing a second copy of rF2 860 and bench it please?

 

Technically speaking, NVidia users using "pre-rendered frames" @ 1 is not equivalent to AMD as most AMD users will be left using the default pre-rendered frames (most likely 2 or 3). AMD users can only adjust pre-rendered frames using 3rd party software like Radeon Pro (Nvidia Inspector equivalent, well actually even more powerful than NVidia Inspector but that's for a different discussion).

Should probably advise everyone to use full, 100% default settings except 4xMSAA. Not exactly fair if NVidia users using MPRF @ 1 while AMD users have it at default (probably 2 or 3).

P.S. Only 1 car, no opponents, only 4xMSAA, benchmark scores are going to be high as hell and possibly even CPU limited (especially for people that have well scaling high-end multi GPU setups). You could at least make the weather settings be set to the fullest, harshest rain conditions , but now I'm just picking on you

 

CPU: i7-4770K @ 4.3GHz
rFactor2 Build: 860
Graphics driver version: 344.11 WHQL

GPU: EVGA GTX 980 SC default clocks (Core 1240MHz, Est. boost 1342MHz, Actual 1354/1366MHz, Mem 3506MHz)

Run #1
Frames: 11758 - Time: 67377ms - Avg: 174.511 - Min: 141 - Max: 199

Run #2
Frames: 11511 - Time: 66004ms - Avg: 174.399 - Min: 142 - Max: 199


GPU: EVGA GTX 980 SC +100MHz (Core 1340MHz, Est. boost 1442MHz, Actual 1441/1454MHz, Mem 3605MHz)

Run #1
Frames: 12278 - Time: 67501ms - Avg: 181.894 - Min: 147 - Max: 208

Run #2
Frames: 12358 - Time: 67657ms - Avg: 182.657 - Min: 148 - Max: 209


GPU: EVGA GTX 980 SC +125MHz (Core 1366MHz, Est. boost 1467MHz, Actual 1479/1491MHz, Mem 3630MHz)

Run #1
Frames: 12461 - Time: 67688ms - Avg: 184.095 - Min: 149 - Max: 210

Run #2
Frames: 12328 - Time: 67455ms - Avg: 182.759 - Min: 145 - Max: 210 (slightly low min & avg?)

 

TechAde, am i right in saying that the reason you writing 2 actual core speeds during the bench is because your gpu clock is jumping between them? If so, this isn't normal. Is that also why you're doing 2 repeats because i see you benchmarks are not as consistent as mine (99.96% at least each time).

If so, have you tried increasing the Power Limit and/or Core Voltage in afterburner to see if it stabilises the core speed during use?

 

Yes, it generally starts at the higher clocks then drops a few MHz. I've seen this happening all NVIDIA GPUs for years, never thought anything of it.

During the OC runs I upped the power/temp limits to their max in Inspector, they were default for the stock runs. I've not touched voltage.

Edit:

I had a fiddle in Inspector to see if I could stop the clock drop happening. Increasing 'Power and Temperature Target' to 125% and moving the temperature slider to 91°C made no difference at all.

However, adding +37.5mV seems to have done the job, at stock clocks boost now seems to be locked @ 1380MHz.

I'll re-run the benchmarks when I get a chance.

 

DrR1pper, max pre-rendered frames....post #14. Not fair/even ATM with AMD...

 

i have i7 2600k gtx 680 4 gig driver 344.11. not over clocked. have fraps could some1 tell me how i do this testing pls. i have never done this type of thing but would like to know how good i have it with what i have.

 

Hmmm, I never thought about this at all. I'm going to take a peak in radeon pro.