Terrible performance on high end system - is SOE aware?

The measurements may make it seem worse than it is. It was a tower fight with about 190 people who were mostly crowded around the spawn room. It literally took me a week to find a spot where my performance on full ultra was that bad. To your points:

I have a power meter which measures 420W while running benchmarks on the CPU and GPU at the same time. So there is no issue here.

The CPU doesn't allow a higher overclock. It runs at 1.42V which is already outside the specifications and for most people frying material. 4.3GHz wasn't stable and 4.4GHz would not even boot.

The performance with my usual settings - ultra except for shadows and particles which are low - would net me 30+FPS in the a similar sitatuion while 50FPS - 140FPS is what I would see in a usual surrounding. Here is an example though I died quite often which is why you see the drops:


I am also going to provide new results after I am done upgrading to a GTX 780ti and a 5820k. I was lucky enough to win them in a contest and I am just waiting for the mainboard and RAM. I will also take a look at high shadows then.

its been like that since day 1. SOE doesn't care I guess. I told them months ago that the flack and explosions cause massive fps drops. Just turn down the effects quality to low. It drops my fps from 70 to 20fps

It really is peculiar, that you have such poor performance - i cant recognize that one bit.

What operating system are you using?

Is the game running on your ssd?

Do you use pagefile ?

Btw, when you say that its fine to effectively Draw 420 watt from an 550 watt psu, id have to disagree - that is dangerously close to its limits for constant load. When you get that 780 ti, it will Draw alot more power than the 670... aka time to get a new psu. If i were you, id get one with a good amount of headroom, as psu's tend to be the most effective at about 50% load.

Yeah i wouldnt use anything less than a 650W Seasonic or Delta or other high quality PSU with a 250W GPU. The 670 should be fine with the 550W but with an overclocked CPU and GPU boost you might not have enough power, despite only showing 420W total.

Thats about 100W more than i ever saw playing PS2 with my 660ti and Sandy Bridge i5-2320(95W TDP).

Your overclock and boost is probably straining even a good 550W PSU. Your voltages might be unstable if it womt even boot with that overclock.

The PS4 can't even run the game at 60fps

The PS4 is also about 1/2 as powerful or less than his PC.

I always write something under the graphs. This bit is mandatory to understand what is going on.

The jitter is so visible because I chose to take the frametime values instead of FPS from fraps. Those are more accurate but obviously offer room for misinterpretation. This is the same benchmark with FPS values.

My PSU is perfectly fine for the system. I measure a maximum of 275W while playing Planetside - 420W was the highest value while loading the GPU and CPU to the maximum. The PSU can provide 516W on the 12V rail. The difference in efficiency between running under 50% and 100% load is about 2% which is negligible.

The efficiency plays another important role. When I see 420W on the energy meter that doesn't mean that the PSU is providing 420W. You multiply the efficiency with the power draw of the PSU to get the respective value for the system itself which would be 420W * 0.93 = 391W.

Im well aware of that -_-'

Regardless of looking at the huge frametime spikes, or the huge fps drops, its not something that i in any way experience myself - Thus find it peculiar that you do. FYI, i consider spikes high, when they are at 33 ms or higher, which you have like all the time.

I did a Whole lot of testing on the subject myself a year ago, cause alot of people were complaining about performance etc, and i found that i had zero of the issues people complained about, aside of server desync.

http://i.imgur.com/TorJjD4.jpg

http://i.imgur.com/TvKKvkb.jpg

A couple of screenies from testing back then - as you can see, the frametime is spot on what it should be, despite heavy activity in the area (as seen both on screen, and on the map).

Regarding your psu, if you are fine with that, then thats all good i guess - id just expect an early death.

I have already written that the frame time spike happens every time I (and probably also you) die or simply open the map and that those you see during the benchmark are irrelevant.

But judging from your screenshots you seem to be running fine with a frame rate target of 60. Though looking at the bio lab screenshot this is not quite the amount of players I had during the benching session which was about twice as many

I agree that the 200 ms spikes are irrelevant (cause thats obviously just a load spike) - but i doubt that you die once every 5 seconds, on each of those 33+ ms But you say that you experience the spikes when dying, and opening map - what are the biggest spikes during actual gameplay?

Honestly, ive always run the game perfectly with sli, so it has always puzzled me what were causing peoples issues. When i started playing the game, i was playing on an old i7 920, which was overclocked to 4 ghz, but it was showing its age, and had a hard time due to its low singe core performance, so didnt give me more than like 40 fps - after upgrading it though, i cant say that ive ever had fps / frametime related issues.

Regarding the screenshots, they are as said taken 1 year ago, when i was playing on the same base setup as now, but with 2x gtx 680's - i cant remember precisly how many there were on each screenshot, but i think the biolab was 24 vs 24, and the Tower 48 vs 48

I need to find out what I have to avoid doing first. Will take a while.

As expected scoping seems to be bad so I played a run as max. The result is still much worse than expected:
http://imgur.com/a/m9avo - two spikes are due to taking a screenshot and dying though.

This could be due to borderless windowed though fullscreen looks like this:
http://imgur.com/a/gX6eV - with one spike due to dying.

Overall I would guess that it is simply the GPU which is limiting and therefor handling explosions not too well. That's why I will test the GTX 780 TI next.

What power supply company have your worked for... ??? A power meter read peak power. 420W is 420 watts not 0.93 and if your are talking about power factor that is an efficiency statement for the PFC...

The correct answer is 451.6129 watts down the front end of an over loaded PFC.... It is an assumption on your part that the 12 volt rail is not taking a dive. Unless you have a high frequency SCOPE on the power line it could easily be going out of regulation.

The fact that you hit 420 watts means the transformer in side your power supply is saturated for the most part.

When your power supply blows up and cooks your mother board don't cry to us....!

what's the relationship between frametimes and real fps? i mean, like if have 30 fps and 20 frametimes or 40 fps and 50 frametimes what changes? is it inversely proportional? high frametime=less fps?

You can convert both: 1000 / frametime = FPS.
33ms -> 30FPS
20ms -> 50FPS
10ms -> 100FPS and so on..

However when Fraps measures FPS then it literally measures how many frames happened each second. So you have one FPS value for each second. With frametimes Fraps measures the time between each single frame.

So the frametime measurements give you information on framedrops even if they took a few milliseconds.

_________________________

Regarding the GTX 780TI: The benchmarks I just did with the new card indicate that borderless windowed is part of the framedrop problem but not all of it. The framerate is higher but the drops are consistent.

"As expected scoping seems to be bad so I played a run as max. The result is still much worse than expected:
http://imgur.com/a/m9avo - two spikes are due to taking a screenshot and dying though.

This could be due to borderless windowed though fullscreen looks like this:
http://imgur.com/a/gX6eV - with one spike due to dying.

Overall I would guess that it is simply the GPU which is limiting and therefor handling explosions not too well. That's why I will test the GTX 780 TI next. "


Smagjus that may very well be the case - i see my gpu's going from like 40% to 70% load when explosions and stuff like that is going on, so doesnt sound unlikely at all

But as others have mentioned aswell, i would strongly encourage you to get a bigger psu before taking that 780 ti into use!

look up "flip size que planetside 2" on google. try 1 or 2 see what happens. and please turn your shadows to low or off, thats just obvious.

As an nvidia user, it is set to 1 as standard with the latest drivers

I have heard your advice. If it dies you will be the first to know and I will not say I have not been warned.

Anyways, the GTX 780 TI is running fine. The power consumption is surprisingly only slightly higher than the GTX 670. That is most likely because the 670 was overclocked and the bios was modded to allow higher voltages and power targets.

I read 310W during Planetside aswell as 310W running cudaminer (altcoin mining client) in benchmark mode