Not sure, but shouldn't matter so long as your used 1920x1080 resolution in rf2 for the live benchmark.
Anyway back to basics, everything else I play the bandwidth does not affect me, thats why I cant understand the issue here. I can get over 60 frames off one card so I cant justify the cost of at the very least a 3rd Gen i5. Although an E series chip would be cool!
Yes, 1.1 x16 per card has little to no significant difference in fps/performance in almost all other games/benchmarks. Take for instance 3d mark FireStrike benchmark, there's a vid on youtube showing pci-e 1.1 vs 2.0 vs 3.0 performance on a high end card and variances are less than 5%. Don't forget that 60 fps off is 38% performance lost from 2.0 x16 (what your cpu and mobo can achieve in single mode) but you're actually losing 90 fps which is closer to 48% off because the normal should be 3.0 x16. Factor in that SLI scaling is not perfect, and a single GTX 980 in pci-e 3.0 x16 mode beats two GTX 980's in SLI in PCI-e 1.1 x16/x16. And yes, you would need and E series chip for the 32 lanes or more of PCI-e 3.0. If rf2 is something you use and take seriously, i would definitely upgrade the base system if i were in your position. You don't need the latest E series cpu and mobo's either.
1. rF2 suffers quite a bit in anything other than PCI-E 3.0 @16x. Regardless of single card or SLI. There are benchmarks from a ton of people, and everyone confirmed this. You can test yourself as well. 2. Unless you're playing in 3D mode, then SLI is just killing your performance even more as rF2 in anything other than 3D mode (NVidia 3D Vision) has worse framerates than even a single card (negative scaling) 3. If rF2 isn't scaling the game to full-screen then when you are playing with a resolution different than your actual screen's resolution, then use the settings in your NVidia control panel (you mentioned them in your post) to make your GPU scale the image. GPU scaling does introduce input lag.
The resolution change was only to run the benchmark spin, its back to glorious uber res now So pci 3.0 at 8x still suffers? Really odd.
Yes it would, because PCI-e 3.0 x8 is the same as 2.0 x16. And based on the reported benchmark results, it would appear that the PCI-e 2.0 x16 (the same as 3.0 x8) bandwidth becomes a bottleneck for graphics cards in rf2 from around the GTX 760 performance level cards and upwards. For instance, my GTX 770 saw a 7-9% fps boost from PCI-e 2.0 x16 to 3.0 x16. My GTX 970 saw a 16-18% fps boost, GTX 980 owners a 26%, etc. But the fact that you're using and i5-2500k (which only has up to 16 lanes of PCI-e 2.0 available) means you can only get your 2x GTX 980's to run at PCI-e 1.1 x16 each (equivalent to PCI-e 2.0 x8 each). This is quite a bottleneck. During the rf2 beta days and before i new about pci-e lanes (thanks to forum members), i remember being very confused why some people we're reporting SLI did not scale well for them because it appeared to scale very well for me (if i recall correctly something like 80+% and in it was in 2d Spinelli). Now i realise why that was the case for me, i had an i7-920 cpu which has 32 lanes of PCI-e 2.0 available and i also had a 2x PCI-e 2.0 x16 slot in SLI motherboard which meant both my GTX 460's were able to run without being pci-e bottlenecked.
There has to be another explanation, there's no way a 970 can be bottlenecked by pci 3 @ 8x Here's a test with other software showing similar result's to rf2 albeit to less effect. They mention the way they have implemented post processing http://www.techpowerup.com/mobile/reviews/NVIDIA/GTX_980_PCI-Express_Scaling/1.html The problem seems to not come from the speed/power of the gpu saturating the pci bus, it comes from the software constantly streaming data from cpu to gpu which saturates the bus. They seem to think it's likely to become more prevalent with future console ports.
I think you are right about SLI pci-e 3@x8x8, fps gain SLI 3@ x16x16 vs 3@ x8x8 = 3-5% when gaming, benchmark 3DMark is 22 000 vs 20 000 and I think that another explanation is rF2`s pure multi GPU support, GPU loads are fine (95-99%) but rF2 just don`t deliver fps what is needed to good game experience. Keep also in mind that Otta56 reports only 14fps with Titan 3-way SLI when rF2 default Nvdia Inspector profile and game settings are used.
Ya, it's relly weird. Only in rF2. Most game only show 2-5% differences from PCI-E 3.0 @16X and 3.0 @8x / 2.0 @16x, but for some reason - and I have no idea what that reason is - rF2 shows quite large differences. We have seen many improvements in the 10-20% range, I even saw improvements in the 30% range, and I think the largest i ever saw was 42%. I believe I was looking at minimum and average framerates, I tend to only pay attention to those two while, for the most part, ignoring maximum framerates. It must be something very, very specific to the ISI engine. If you look at the benchmark results (I'll find them for you when i get to my PC) you'll see all sorts of tests from many people showing all supporting the same conclusion. There are tests for single GPU, SLI, 2D, 3D, single screen, triple screen. All come to the same conclusion.
15fps oh dear, Im hopeful when the sim comes out of beta that nvidia/amd will create working profile's. I get 35fps on default sli profile and it's jitters beyond playable.
From what I have read in the above article it is engine specific but the bad news is it's going to become more commonplace. I'm hoping the next gen Intel's will have 32 lane's instead of forcing us down the enterprise path.
I've been giving this a little thought. As I see it which admittedly isn't a very educated view is... If the engine is in a constant state of communication between cpu/gpu and given the fact that people believe that rf2 is gpu dependent as opposed to cpu dependent (I play enough latest aaa title's to realise the graphic fidelity of rf2 shouldn't be gpu dependent), "tasks" must be offloaded to the gpu. This explains the bandwidth saturation and maybe the lack of sli performance. As I understand it in a regular sli set up card 1 renders a frame then card 2 renders a frame which should give decent scaling. I'm guessing the extra "task" the gpu is being given isn't supported by the existing sli setup. With the way the new consoles work and lazy ports we can expect more bandwidth issues in the future, I'm hoping this can and will be resolved with and sli driver update but the suspicious side of me tells me we might see something like sli 2.0
There's two issues, the fact that rF2 gains a lot when using PCI-E 3.0 @ 16x, and the fact that rF2 has negative scaling with SLI (I think if you turn off both reflection options then you may get rid of the negative scaling and even some SLI improvements). A start would be to disable SLI when playing rF2, that should give you a big bump in performance.
Absolutely, its literally unplayable as default for me in sli, I'm just trying to understand why, unfortunately the conclusion I have come to makes me think it won't improve
The ting with SLI is that if you have 60hz monitors it is easy to get 60fps with max settings but if you hate input lag like I do you go for 120hz monitors and it is not possible to get 120fps with HDR, multiview, road environment reflection etc on. With my system I get round 200fps in AC, pCars, iRacing etc. with high game settings but with rFactor2 I must make huge compromises to get game experience like it was back in 2000. I have said this many times before, it is soon 2015, no multiview and no multi GPU support do not belong to modern gaming and if you are asking when this is going to happen you get message from admins and all f..b... to shut the f.... o...
The difference with 120hz and 60hz vsync is tremendous. The lag is so bad at 60hz, that it effects my lap times. 120hz with vsync on the other hand, does exactly what it supposed to do, eliminate tearing and stutter. I can feel it instantly now that I have been using 120hz for more than a month. I just got new GTX 970s cards in SLI and I can maintain 120 FPS with no reflections and FXAA on...just about everything else on its highest setting (no AA, except FXAA). rF2 looks good, but I know the reflections issue is a SLI driver problem (or at least not something that ISI and even Kunos cannot address with AC). Eye candy is nice, but high frame rates on fast monitors is more a priority for me.
V sync causes the lag not 60hz I made a little video, my monitor isn't regarded as good for input lag but it's not going to cause me trouble http://youtu.be/g0nPwSoC2hE http://youtu.be/5Q1V1YPJddA