Discussion in 'Test Server: Discussion' started by codeForge, Nov 5, 2013.

  1. Serpent808

    In a way, I disagree that it is bleeding edge, but I'm glad the issue have been fixed. Thank you for pushing forward with the Optimization. The overall performance is looking better and better.
  2. DOXaus123

    Whats the situation with 'crossfire' AFR?.. PS2 has always given my amd 6950 CF 'negative' scaling while in the 1st person view in cockpits of aircraft (due to the extra gui overlay i imagine) Does no one else with radeon crossfire suffer from this?
    I could work around it by forcing AFR with radeonpro which works better than AMD CCC profiles.. but now in the PTS this no longer works..only way to fix is it now is to turn off the HUD GUI = no more negative scaling. This also fixes the stutter in other vehicle 1st person cockpits.
    • Up x 2
  3. codeForge

    DOX, we have a fix for crossFire coming, but probably not as quickly as we'd both like!
    • Up x 1
  4. Frostbitten

    Sorry if this has been answered before... But what about Virtu MVP support? I would like to use integrated Intel HD 4000 for additional firepower.

    Some games are using it rather well, especially it's smoothing part.
  5. Serpent808

    I'm not getting Negative Scaling. I can't honestly say that I am getting 2.0x Scaling either. 80 FPS x 2 = 160 FPS, and this isn't true for all scenarios on the live server. On the PTS, I've been doing some runs with MLAA on, and with it off. I've seen my cards jump that high without Smoothing, but mainly at the Warp Gate. Flying, it still goes between 100 to 160, and some occasional 226 FPS (mainly if I'm angling the nose of the scythe towards the ground, or looking at the sky). It peaks that high. At the Palisades earlier today, there was a three-way between the Vanu, NC, and TR. TR held the Palisades, and I believe it was a 100+ battle with MFBT and ESF. With MLAA and Smoothing enabled, I had a FPS band between 45 and 60 FPS, just flying over the area. It would drop to 45 FPS for a few seconds, then pick itself back up to the mid 50s. With MLAA off, but smoothing still enabled, it dropped down to 55, and back up to 60. With Smoothing off, I can imagine that it will probably still be near or close 100 FPS.

    What is referred to as "Negative Scaling" means that, say for example you have a 7970 or RX9-280, and a 7990. You do a benchmark. The RX9-280 is a single Graphic Card, it's AMD, etc... AMD 7990 is a dual GPU, single PCB graphic card. The GPUs are in CrossfireX with one another on the 7990. Lets assume you're not using a AMD Beta Driver with Frame Pacing on it. If you were doing a benchmark in say Batman AO, the RX9-280 was getting 60 FPS at 1080p, in theory, you would think the 7990 should be getting 1.8, 1.9, 2.0 times that amount in FPS performance. So that comes out to 120 FPS, 115 FPS, 110 FPS roughly. Negative scaling refers to the performance of the 7990, in Crossfire, going below 60 FPS in this case. Direction is negative. So performance is going down in the negative direction instead of going up in the positive direction.

    A good question to ask you, DOX, is what AMD Catalyst Driver are you using? If you're using a WHQL Driver, they don't have any of the Frame Pacing Software on it. So you will be more subjected to runt-frames. This could lead to negative scaling. On the other hand, you should see some improvements with the Beta 13.11 v8.0 Driver, but it still doesn't address D3D9.0 with full support.
  6. Serpent808


    --AMD MLAA Enabled @ 61 FPS - Smoothing Enabled.
    --Look at that Eye-candy with Motion Blur disabled, High Settings with 6000m rendering distance, field of view at 70. 1080p.
    --Racer Chassis 3 @ 86 KPH.
    Look at how purty that picture is in CrossfireX with two RX9-290x's.

    How does this pertain to the PTS and SLI/CrossfireX? This is a glimps how awesome the picture quality AMD users will see with almost max settings, CrossfireX, and MLAA enabled. MLAA enabled just makes the eye-candy that much more nicer. Look at the edges of the rock, on the tank, and on the hills in the distance. Do you see a lot of stair-step lines? AMD users would much appreciate it if FPS performance would go up with MLAA enabled with the optimization.
  7. Loui5D

    What frames + temps are you getting with dual R290X's? i may just pop down to scan and drop some money on another one in a few weeks time.
  8. DOXaus123

    Yes its negative scaling, i dont have hard numbers and i use RP's adaptive for example

    singlegpu - 50fps in cockpit, 50fps in 3rd person
    crossfire - 34fps in cockpit, 60fps in 3rd person

    Yes im using beta 13.11b8, the issue has been there since ps2 beta. Seeing as you dont suffer from it with newer gpus, it looks like amds 'usual' crossfire profile abandonware for older archs like VLIW4/5 over their newer GCN.
    I dont think there is any framepacing support for dx9 as yet, radeonpro with DFC will deliver smoother framerates than anything amd will release anyway if your limiting frame rate/vsync.

    As far as MLAA goes, and's mlaa is pretty subpar, imo fxaa is better and smaa is even better. That said you have R9 290's? you should be downsampling anyway by adjustings ps2's renderquality in useroptions over the default 1.0
  9. BITES

    Did this but all that did was basically restore frame rates to similar levels on live.
    Overall the average is slightly lower but you definitely notice less large drops or super high peaks.

    So its about average ... will retest a few bits later.

    System is :
    CPU: Intel i7 3770K @ 4.7ghz, unparked
    Mobo: Gigabyte Z77X-UD3H (F20c Bios)
    Video: SLi 2x Asus 660Ti OC - GPU@1019Mhz, Memory@1627Mhz, Boost@1097
    Ram: 16gb G.Skill 2400Mhz
    Display: Asus MX299Q (29" Wide@2560x1080
    Audio: Creative 7.1
    Misc Data/Driver Info: Nvidia@327.63, Physx@09.13.0725, Windows 7 Ultimate 64bit
  10. SniperCzar

    Speaking of which, when are we going to get the option to turn off the darn built in FXAA which blurs everything and costs another 15% of performance? I'm sure people who do these kinds of AA hacks would appreciate not running blurry AA underneath clean AA. I think there was an INI hack at some point but that was removed for some reason :p... any input Mr. Ryan?

    Maybe while you're talking to NVIDIA about SLI optimization you can ask about some of their nicer AA technologies too :D
    • Up x 1
  11. SurgeonX

    OK, gave this a good run tonight, and with the OMFG: PU01 release, you're spot on...
    It feels super smooth.
    The FPS was pretty much locked at 60 with smoothing and V-Sync on, and it just felt really fluid.
    There were still a few dips, again I mainly notice it when turning as Infantry on the ground, but it's not a real issue.
    It was quite a decent fight at the Crown, about 3 hours before the official test, and during gameplay I never noticed it dropping anywhere below 30.
    Checking out the FRAPS data, the FPS does regularly dip down to 20, but this always happens when I die and the camera floats up to the Killed By dialog.

    I'll definitely continue playing with Smoothing and V-Synch on, because it just feels so god damn silky :D

    Here's the data from about 70 minutes on Indar at The Crown...


  12. Serpent808

    With Smoothing, 60 to 61 FPS just about everywhere. It drops down to 45 for a really short time, in a massive bio lab fight, but it spikes up and back to 55 or 60 fps. Without smoothing, looking at 100 to 160 fps. Temps are at 95 degs c on both. Max is probably 96 degs c. I want to say like 70 to 90 fps in a bio lab with 100+ players in it without smoothing.

    There's an issue with the RX9-290x. Idono if it's really an issue, but the core frequencies have a tendency to throttle. I've seen it more often when the temperature goes past the 95 deg C mark. I've turned off Overdrive and Ultra Low Power State (ULPS), but it still occurs, and the change in Mhz occurs more frequently with ULPS off. I think it has something to do more with the temperature than anything else. With Ultra Low Power State on, the buffer GPU, GPU 2, will still push 100% load at 1000 Mhz Core, but the primary GPU, GPU 1, will throttle. It won't throttle sporadically. It will drop to like 850 MHz, stay constant, than spike up or down a little. I suspect that the GPU sense so much additional power is needed to perform it's task, and that's why it spikes.

    In benches that don't utilize multiple GPUs, the Primary GPU, GPU 1, will actually stay at 300 MHz while the secondary GPU, GPU 2, will be at full load and core frequencies. The Primary is hooked up to my monitor. On MSI Afterburner, it shows that GPU 2 is doing 100% load. In GPU-Z, it shows GPU 1 doing the 100% load at 1000 Mhz, and GPU 2 is doing nothing. I found this interesting. I've seen a similar situation occur with World of Tanks recently, and the CrossfireX mode was set to disabled. MSI Afterburner registered on the 2nd GPU as doing a 100% load, and not the Primary Card. Most likely, the 2nd GPU is producing frames, sending it to the first GPU, and the first GPU sends it off to the monitor.

    You're probably right about AMD abandoning the older architectures. I've ran with 2 6990s from XFX on PS2 when it first came out, in CrossfireX, and there is a difference, a slight difference, in the picture quality.

    DFC locks your FPS, but your Frame time changes. It actually increases your Frame time by 10 ms. So if you're at 60 FPS, or 16.667 ms, it's actually around 26.667 ms.

    I've got a lot of mix feelings about SMAA and FXAA. I've seen reviews showing that SSAA is actually better. SMAA didn't look any better on Crysis 3 with a 1600p monitor. MLAA just acts as post processing and some additional 8x AA. I forget the exact number. Just that MLAA was something discovered by Intel, and AMD paid Intel for the technology... I don't know. I think the image quality goes up with MLAA. I know the MLAA quality on the AMD 6000 series is lacking a little in comparison to the 7000 Series, or RX9.

    I like the idea. Do it, do it!!! Let us turn it off so members can get back 15% of their performance. :cool:

    I hope MLAA isn't considered as a Hack...
  13. SniperCzar

    SLI is working much much better for me now. Prior to the latest tweaks playing with smoothing on would hitch every few seconds for about 1/3 of a second, now there are a few late frames but overall not extremely noticeable. Frame smoothing off now is better than frame smoothing on just a few days ago, so that's some real progress!

    Here's a short video I took tonight. Look in particular at the middle portion where I slide up the gravlift a couple times, even on the 30FPS youtube video you can see some choppy motion with smoothing off.

    Original ShadowPlay footage (500MB!) - 2 11.07.2013 -

    I have another 45 minutes of footage I took at the Crown fight tonight with a complete CPU/GPU usage overlay I can upload later if anyone's interested.
  14. Serpent808

    Ya I can see it a little. Looks like frames are being dropped when you're about to land on the pad. You can see it for each time you go on the lift, you hover, than when you're about to land, there's a sudden chop.

    Noticed that you're GPU loads are a little off between GPU 1 and 2. what was it like 65% and 71%, 79% and 75%. Seems a little off for Default.

    You're load time is a little slow too. Is that due to the GPU in SLI loading up, or a Storage Driver thing?
  15. S7rudL

  16. Nate

    codeForge, don't know how much hassle this would be but could you change from PlanetSide2.exe to PlanetSide2Test.exe. This would put all the Nvidia SLI (not sure about AMD) users back to their non-customized 3D settings for the game. Our live customized profile using AFR2 wouldn't mess with your testing.
  17. codeForge

    Nate, that would also unfortunately disable all the optimizations that nVidia has put into their driver for SLI.

    I see what you're saying. I'm hoping we can better inform players on their state in the near future, to remove some of this frustrating guesswork. Stay tuned.
    • Up x 3
  18. SkattoNollFem

    Are there any news to how the SLI is coming along? It really makes a big difference to me when it comes to performance.
    • Up x 1
  19. OliTF1

    A late comment but I think it should be point out. If it's not news, it certainly important:

    IMPORANT: I get the same stutter everyone here gets, but turning off the HUD removes it completely. Whether shooting or not, in warpgate or huge battle... no matter: No HUD = Perfect and smooth FPS.

    Try it guys and comment on the HUD effect. For one, I'm 100% sure the SLI problems and in the HUD, not in the "game" or it's 3d graphics.

    • Intel i7 quad
    • GTX560 ti SLI
    • Everything on high + low shadows + no smoothing (trying to meet me 120 hz monitor)
    P.S: PU01 and 02 clearly ruined it for me, but I bet the devs can find the source of the problem in that pesky HUD and fix it. We just need to help them find the cause. I love PS2!
  20. OliTF1

    Here is a nice little in-game test devs should have a look at. There might be hints of the cause of the problem in the phenomena below:

    If you want to "see" the stutter in the game, try spinning around (works great with a controller) looking at the floor or in a clear field. The faster you go, the easier to see: your viewmodel arm and gun actually twitch, like ticks/spasms as the SLI stutter occurs.

    Further more, if you switch your HUD on and off while doing this, you'll see your viewmodel twitch come and go as you activate and deactivate the HUD.

    The source of the problem most likely is elements of the HUD. I suspected ammo count/display after PU01, but there was a slight change in PU02 that lets me believe the ammo bar might not be the culprit.