MSI GTX 970 mini review with rFactor 2 performance

It didn't sound like he was running two cards - just one (at least I hope it was just one )

 

Does anyone know how to do the cleanest fresh installation of rf2 without reformating and install a fresh os?

I can shake this constant 16% performance loss in rf2 only.

 

uninstall in control panels - delete c:\program files\rfactor2

delete ~\my documents\rfactor2

It still has your license stored somewhere - registry maybe ...

 

Out of intrest i did the same, and even when my sys is cleary disadvantaged and just a 290 non X which is not on a level of an GTX980 and a older gen gpu, your ~30% plus isn't really wowing in my opinion. Mantle would give her the rest.

View attachment 14365

 

Clean install of rfactor2:

Uninstall rfactor2 using the uninstall feature.
Delete the entire rfactor2 installation folder, and any rfactor2 related folder in your My Documents folder (to remove any user data files).
Next, delete the local license file: Navigate to your “C:\Users\<Your Windows Username here>\AppData\Roaming\.rFactor” folder and delete the files within (you can type this in your Windows Explorer address bar or navigate to it by first disabling “Hide system files and folders” in your folder settings (see windows documentation for more info)).
Make sure you have downloaded the most recent build of rfactor2 from the appropriate downloads page of rfactor.net and install it into a new clean folder. For rfactor2, using the Content Installer download is best because you can then test the game with minimal content.
Since you’ve deleted the license file in a previous step, you’ll need to reactivate your game. For rfactor2, run the launcher, select the Purchase tab, then select Reactivate.
Run rfactor2 with the minimum content and with NO 3rd party community created mods/addons/plugins (content you’ve downloaded elsewhere) and troubleshoot any issues with limited content.

Once you confirm the game is working fine, you can download additional content from the rf2 downloads page at rfactor.net

 

You and TechAde should both do the replay benchmark to know what your cards relative performance to one another really is.

 

ok, so i did all that (except deleting my license key and i can't find any rfacotr 2 related folder in My Documents) and no joy.

 

Impossible, the gfx cards are not running by themselfes, they are driven by two absolutly different bases and the replay thing is nonsense in my opinion as you already know. I'd sayed that for multible times and say it again, the replay performance is not as good as the live performance, for whatever reason and i'm not driving replays. You ever considered that there is maybe a incompatibilty or optimization issue with AMD's, or maybe it is just sys related but fact is that it doesn't replicates the real performance and i don't think it is right to misinform people when it comes to rf2 performance with AMD, because i don't think people should judge usability by the replay performance.

 

To know the true relative performance we would need to benchmark both GPUs in the same system. Only then can we be sure all other factors are equal.

That said, I'll give the replay a shot when I have a chance.

Sent from my SM-G900F using Tapatalk

 

I agree with speed. Live performance is what your after anyway. I think you'll just have to chop it up to rfactor 2 still not optimized. I still get flickering textures and black flashes with HD on.

Are you using MSI afterburner or some monitoring tool? I have a few games that it actually causes a performance loss.

 

I see one big difference between rF2 and the most other games I've tested. There's an option in MSI Afterburner to monitor the bus utilisation. I've tried to find a little more info about this feature but there is not much so I guess it's the amount of data transferred via the PCIe bus to the gfx card. Most games (tested PCARS, Star Citizen, Elite Dangerous, all DX11 btw) show something around 5 - 10 % bus utilization. rF2 shows something around 35 (No Ai, practice) - 60% (29 AI, race). The only game with similar results is DCS and like rF2, DCS has a 64bit exe and a DX9 gfx engine. In my opinion DX9 is the culprit.

 

Good observation, I run DCS too and I notice the same issues as well. I thought it was the PCI bus utilization is for monitoring how much bandwidth at the PCI bus I believe but I had it wrong. I can't find anything on it either.

 

Oops, lol, maybe i'm confusing all these GPU threads, lol. I thought you mentioned something about installs and uninstalls and not having the framerates you expected. I was going to say, try my GPU driver uninstall/install guide and see if that helps.

Think I'm getting too confused, lol!

 

speed1, TechAde and rogue22. Let's first make sure were on the same page. It is my understanding and interpretation that the entire purpose of the replay benchmark threads were to find out the relative card performances to one another. This has nothing to do with wanting to know the live performances. All those doing the replay benchmarks simply want to know how much more in percentage one cards performs vs another in rf2. There was however two major assumption made with this benchmarking method that was never properly tested to confirm if true or not. That is, firstly the cpu performance was not bottlenecked in the replay benchmark method and secondly the relative performance ratio (i.e. the performance difference in percentage) between cards would translate proportionately to live gameplay.

I think we've been able to confirm both of these assumptions are true. In the test both me and speed1 have just conducted in this thread: http://isiforums.net/f/showthread.p...een-AMD-Nvidia?p=309268&viewfull=1#post309268

There have been concerns that the replay benchmark method brings the cpu into question (bottlenecking) and so a solo lap of a track in the live replay would alleviate any of those doubts in the end result. But what we've found in our mini test is that relative performance raito's between cards scale perfectly (99%) from a replay benchmark to a solo live benchmark.

In short, it means we can rely on the performance ratio between cards in a replay benchmark with 99% certainty that translates in live performance ratios (or at least when comparing me and speed1).

Replay benchmarks don't cause my cpu cores to go over 25% on my i5-2500k @4.3ghz and so i see no reason to think it might skew a replay benchmarks result on how well our gpu performs. Ofc, without being able to test every single persons computer in this same way me and speed1 have, we won't be able to know for sure at what point the cpu starts bottlenecking replay benchmark performance. But if i had to guess, based on my cpu utilisation, it's only going to be fore people on core 2 duo/quad and lower type cpu's where cpu utilisation is hitting/spiking 100%.

 

I really don't know which one is the non optimized, rf2 or still the AMD driver but fact is that my R290 is performing very well compared to what it was like before the 14.4 AMD driver in combination with rf2. The point is comparing GFX cards or gpu's in general driven with different bases is a bit off in my opinion. You can't say card X is better as card Z when they actually are driven by different bases, and as i'm aware of i'm the only one posting about a R290 which is driven by an outdated base. How this can be a base to compare the Quality and performance capability. I really would love to see someone would put a R290 in an Intel sys of this gen to see what it would look like, not to mention what it would look like with Mantle in use which is actually a DX killer, as example in BF4.

I'm not using MSI afterburner or any monitoring tool and i haven't any performance loss in any game i tryed, rather the contrary, it is just great, and that with the old base it is driven on. Apart from that i don't play games, i bought BF4 just to see Mantle performing. The game is BS, to much stress for me old man.

 

To be fair, i tryed the replay benchmark config ( 20AI-Silverstone Int.-Grid Position last ) live again and the result was similar to the replay benchmark, and for whatever reason it wasn't before.

 

A sample of two isn't a lot to base a 99% certainty prediction on but generally I'd agree.

Sent from my SM-G900F using Tapatalk

Edit: wow, seriously ninja'd!

 

So long as you are confident that the base is not bottlenecking the benchmarking, i don't see any reason why not.

In fact, when you look at validated scores on 3d mark for example, the gpu score is very consistent for the same gpu but the overall and cpu score can vary wildly (200-300% differences).

This is not the case in older benchmarks from 3dmark vantage and before where graphics performance was scalled almost proportionately with cpu speed. I think that day has well and truly past for 90% of games (and example of one that has not would be Arma).

You're right and there is an assumption made there (by me) too ofc. But as i stated (maybe i edited it in whilst you read the first post), we would need to do this test with different cpu speeds and see at what speed this ratio between replay and live benchmark breaks away. That would be the surest way to know at what clock speed the replay benchmark method switches from gpu to cpu dependency. In fact, this can be done on anyone's system.

(Ofc the easiest alternative is to trust the cpu utilisation graphs telling you if you're bottlenecking or not.)

 

What do you mean?

 

I had a difference between the results when driving the replay benchmark as such and the same config live before, and now there isn't any difference. This was what i was originally talking about that my results with the replay where more bad as when i was using the benchmark config for a live session to test. It doesn't relates to our little test.