Closed Stefos-art closed 2 years ago
So, more information might be good. 30fps at best ... with what? An outdoor level that you've made? One of the demo levels? If so, which one? What? Help them to help you.
It is interesting that so many are getting such different results with MAX. I have an older GTX 1080 ti and I get 100 fps on my test indoor levels. 50 fps and up on the test outdoor levels. I've yet to see frame rates before 50fps that I can recall. Perhaps a dip here? But I don't remember it.
Now, you've told us your video card and system RAM, but what about your CPU and operating system?
So, more information might be good. 30fps at best ... with what? An outdoor level that you've made? One of the demo levels? If so, which one? What? Help them to help you.
It is interesting that so many are getting such different results with MAX. I have an older GTX 1080 ti and I get 100 fps on my test indoor levels. 50 fps and up on the test outdoor levels. I've yet to see frame rates before 50fps that I can recall. Perhaps a dip here? But I don't remember it.
Now, you've told us your video card and system RAM, but what about your CPU and operating system?
Socket 1 ID = 0 Number of cores 6 (max 6) Number of threads 12 (max 12) Manufacturer GenuineIntel Name Intel Core i5 10400
Windows 10 pro
The test is an out door scene island showdown that comes with Max, but even empty scenes are slower than they should be, Indoor scene between 40 - 50 fps
In the Island Showdown, I am getting 56fps in most places. There was one place where it did drop below 50. However, I am running it on an ultra-wide monitor (3440 x 1440 screen resolution). On an empty, I get a flat 100fps and it's locked there only due to vsynch. On my test indoor levels (levels I've built myself), I get 100fps (again, at 3440 x 1440 resolution). With the Cellar game, I get a smooth 100fps throughout. With the Switch Demo, I get fluctuations between 60 and 100fps.
Currently, MAX is hit pretty hard by collision meshes and if Lee just used a lot of polygonal or hull collision meshes in the switch demo (I've not checked), then that can really hit your frame rates hard. There was another person getting bad frame rates on their own custom levels due to using polygonal collision meshes. I told him how to use custom collision meshes and his rates went up from 20 fps to 70 just be doing the switch.
Your CPU may be part of the bottleneck with MAX as it currently is. It's still alpha software and they are still ironing out bugs and speeding things up. There currently isn't any real culling that I am aware of to hide unseen polygons behind walls (though I could be wrong), so there's a lot they can still do to speed MAX up. Even so, I find it strange that I am getting so much better frame rates than you with my older video card on an ultra-wide display. Except my CPU is better CPY than yours (i9-10900K).
I am between 45 and 55 fps on a test map that contains more than 1500 3D objects, with the following configuration : Intel(R) Core(TM) i5 4460 - 3.20 GHz CPU (4 cores) NVIDIA GeForce GTX 1060 (6 GB) 16 GB Ram - SSD hard drive - Windows 10 Pro 64 bits
During the TestLevel, press TAB twice and visit the Shadows section. Use lower settings to see if it helps you gain FPS.
I've posted on Github before (and I'm not the only one) about tree shadow. This can dramatically slow down a PC, down to a 50% FPS drop, or more. I am almost always slowed down when I walk in the shade of trees.
Sometimes just turning to face the sun results in a very significant FPS drop. Our character goes so far as to walk on the spot. With trees, their shadows on the vegetation, it's even worse.
@dllllllllllb Assigning this to you so you can run some tests to reproduce this and have as much evidence and data ready for next week when the team embarks on some optimization work
I switched off shadows and got between high 30's in forest and high 40's - low 50's on the beach, still low but playable.
@dllllllllllb I will ensure our opt work solves this.
@Stefos-art Performance improved via object instancing work for this weeks build.
@Stefos-art Performance improved via object instancing work for this weeks build.
Looking forward to checking out!!! Thanks Stefos
Hi @Stefos-art, could you confirm if your performance issues have improved? Thanks :)
Hi @Stefos-art, could you confirm if your performance issues have improved? Thanks :)
I am afraid there is no improvement in FPS, however for some reason it plays better, though the fps read out says between 20 - 30 fps with low quality shadows. I used the map Island Showdown or for me Island Slowdown
@Stefos-art Would you be able to record a video showing the profiler whilst the issue occurs on Island Showdown? To enable the profiler, press TAB whilst in-game and it will appear
@LeeBamberTGC User reported not fixed, see latest comments
@Stefos-art Would you be able to record a video showing the profiler whilst the issue occurs on Island Showdown? To enable the profiler, press TAB whilst in-game and it will appear
@LeeBamberTGC User reported not fixed, see latest comments
There looks to be a further FPS slow down from the last time I tested the software https://youtu.be/YKYyK4hfDN0
@Stefos-art The values on the left are a little blurry to read in the video. Can you find a few slow sections of the same level and do a full-resolution screenshot of the left panel so I can see which element is making things slow down, thanks!
I reset the visuals before taking the snapshots here This one is with the default visuals:
@Stefos-art Thanks for these, I can see the values now. Can you make sure that when the app is launched, your system is using the 3060 card and not the integrated Intel graphics? It is definitely on the GPU side. Also, can you try running the island demo in 1920x1080 and see if you get 60fps, as this will allow us to test the pixel fill part of the pipeline, thanks!
My graphics display is set to 1920x1080 always, please see below all information on my computer resources being used during running of the game.
This is a better read out for the graphics card the last one recorded the drop when switching to the screenshot software.
Here is the GPU info
Here is a shot of MSI Afterburner during play
I'm quite puzzled by this because my RTX-2060 even at a resolution of 3440x1440 (21:9) does around 40/44 fps.
When gaming you could use GPU-Z for monitoring while playing, see if that gives some more clarity.
I'm quite puzzled by this because my RTX-2060 even at a resolution of 3440x1440 (21:9) does around 40/44 fps.
When gaming you could use GPU-Z for monitoring while playing, see if that gives some more clarity. Can you see something wrong here in GPU-Z?
As a reference, this is my GPU-Z file-reading while playing the demo
As a reference, this is my GPU-Z file-reading while playing the demo
I don't see much difference other than the clock speeds which reflect the different cards!!! I just don't get it
The GPU power draw shows a huge difference in my opinion compared to your graph. Also visible in the 6 and 8 pin power, while voltage is high.
What is causing this is something i'm interested in and will look for tonight, your card should easily outperform my card, certainly because you are not running the same crazy resolution as me. :)
Those readouts of yours show when your readings were flat line notice the almost straight lines at the end of your readouts, you need to take the snapshot while the game is running, you need to be fast at taking the snapshot as once you go to another software (snapshot) the graphics card slows down etc.
Yes i know but we see a whole history on those graphs, the numbers of a particular moment are not really that important.
Well, your certainly not the only one experiencing this problem. I've been reading a lot but haven't gotten any wiser.
Though -XMP must be turned on in the bios. -make sure that in the Nvidea control panel under 3d settings the power management setting is performance. (I don't know if there is also a powersaving setting in MSI afterburner/software)
For the rest, either wrong memory/settings or a cpu bottleneck.
@Stefos-art @earthling45 Thanks for delving deep into the hardware and monitoring side. Nothing leaps out at me as to the cause from the stats, and I agree it does look like both 2060 and 3060 cards are running MAX and show correct and expected values. The only thing I can guess is that our min spec is 970 while both 2060 and 3060 share some traits of the older GTX 960. Passmark clearly scores the 2060 and 3060 much higher than the GTX 960 so maybe their scoring is taking newer DirectX 12 metrics and newer techniques to get that better score?! MAX is currently DirectX 11 and does not chase any of the brand new 2019+ techniques that might provide high scores in commercial benchmarking tools. It is certainly a possibility that the whole XX60 range cuts back on something that MAX needs a lot of, and the XX70 range has enough to provide 60fps for its users.
We are doing performance work now, mainly to benefit our continuing VR work and to some degree integrated graphics performance. During this work, we are likely to find gains, and this should filter down nicely for XX60 cards.
In the meantime, does anyone have a handy chart that shows the GTX 960/970 alongside stats for the RTX 2060/2070 and RTX 3060/3070, it will be interesting to see what is cut back on (shader cores, memory, cache, etc).
There is another possibility i thought of last night. Obviously if a graphics card has more than one power connector it is advisable to use two separate PCI-E cables from the PSU. The PSU may have multiple rails to suply power to the system components.
@LeeBamberTGC I actually think that the RTX3060 is either configured wrongly or is not getting the needed power, it is underperforming quite a lot. :)
For me, i could try and see how the RTX performs on normal resolution if you want that, i think 40/44 at 3440x1440 is certainly very good already. This is with the Island showdown demo. :)
What I find strange is I am running a much older card, a GTX 1080 ti. And, yes, it is a "ti", so it's got that going for it. But at 3440 x 1440, on the Island Showdown demo, with the latest MAX builds, I get close to 60fps (58+) out on the beach area and 50+ in the jungle. I realize that your card is a "60" and mines an "80", but there's several generations difference between them. I would think you'd get a lot better than 40fps with any 3xxx card.
What I find strange is I am running a much older card, a GTX 1080 ti. And, yes, it is a "ti", so it's got that going for it. But at 3440 x 1440, on the Island Showdown demo, with the latest MAX builds, I get close to 60fps (58+) out on the beach area and 50+ in the jungle. I realize that your card is a "60" and mines an "80", but there's several generations difference between them. I would think you'd get a lot better than 40fps with any 3xxx card.
Now that I agree with...what I don't understand is how come all the stats show normal operation yet for some reason me and others are getting low FPS. I bought this rig because I needed a reasonably fast card with lot's of graphics memory for the 3D animation work!!! I sometimes think my old GTX 1060 worked better than this RTX 3060 sometimes!
Ok. I've tried the demo with a resolution of 1980x1080.
My system: AMD Ryzen 2600 16 gb ram RTX 2060 6gb win-10
The lowest fps is 60 in certain parts, obviously it is capped at 75 (my monitor settings) but i think it could go higher than 75 fps.
What I find strange is I am running a much older card, a GTX 1080 ti. And, yes, it is a "ti", so it's got that going for it. But at 3440 x 1440, on the Island Showdown demo, with the latest MAX builds, I get close to 60fps (58+) out on the beach area and 50+ in the jungle. I realize that your card is a "60" and mines an "80", but there's several generations difference between them. I would think you'd get a lot better than 40fps with any 3xxx card.
That is the difference between a high-end gamecard 80(ti) and a budget gamecard 50/60. Nothing strange about this and we should not forget that the 3060 is a RTX card, much lower power consumption, newer technique and capabilities compared to the GTX. I think a 3060 TI would be on par with your card in terms of performance. :)
https://en.wikipedia.org/wiki/GeForce_900_series https://en.wikipedia.org/wiki/GeForce_10_series https://en.wikipedia.org/wiki/GeForce_20_series https://en.wikipedia.org/wiki/GeForce_30_series
@LeeBamberTGC I think this is what you were looking for, i thought of making a chart but found that it already is nicely done in Wikipedia. :)
I know little about the mechanics of a computer Does this look to you that this card has more than 1 connector? I think I see 2 ...1 with 8 pins the other with 6?
There is another possibility i thought of last night. Obviously if a graphics card has more than one power connector it is advisable to use two separate PCI-E cables from the PSU. The PSU may have multiple rails to suply power to the system components.
Yes correct Stefos. :)
And if i'm seeing it correctly this is one cable with two connectors? Which PSU do you have, are you able to see the information?
Yes correct Stefos. :)
And if i'm seeing it correctly this is one cable with two connectors? Which PSU do you have, are you able to see the information?
Yes it looks to me 2 connectors 1 cable
Brand is kolink?
Brand is kolink? yes
If i'm correct this PSU has two cables for PCI-E, so it should be possible to adjust it and not draw all the power from one cable. But you are not familiar with this so my advise is to ask someone who is, or if your system is still in warranty the advise is of course to go that route with the problem you have because the graphicscard is obviously not performing as it should.
If i'm correct this PSU has two cables for PCI-E, so it should be possible to adjust it and not draw all the power from one cable. But you are not familiar with this so my advise is to ask someone who is, or if your system is still in warranty the advise is of course to go that route with the problem you have because the graphicscard is obviously not performing as it should.
Thanks for the help, I will look into it!!!
Hi Stefos-art, Usually if a PC is not getting enough power it tends to crash a lot or even shut down, that is not to say your GPU degradation is because of power shortage but it is worth looking into.
RTX 3060 Ti demands 215 Watts at full load, PSU power efficiency drops over several years, so if your PSU is for example over 10 years old, a 650 Watts could have the equivalent power rating as a 450 Watts PSU or worse. But this is just a general example. It all depends on the tech & circuitry component inside a PSU.
You also have to take in consideration of CPU power draw, whether you have any mechanical drives (DVD, Hard Drives etc..) how many powered USB ports, these all soak up wattage.
here is a basic breakdown of watt usage by modern components
you can get a rough idea at what your system is drawing here https://www.bequiet.com/en/psucalculator
It is also worth checking if Windows 10 - 11 as hardware-accelerated GPU scheduling turned on, by default windows as it off System-->Display-->Graphics-->Default graphics settings turn on Hardware-accelerated GPU scheduling
hope this helps
If i'm correct this PSU has two cables for PCI-E, so it should be possible to adjust it and not draw all the power from one cable. But you are not familiar with this so my advise is to ask someone who is, or if your system is still in warranty the advise is of course to go that route with the problem you have because the graphicscard is obviously not performing as it should.
Thanks for the help, I will look into it!!!
Hi Stefos-art, Usually if a PC is not getting enough power it tends to crash a lot or even shut down, that is not to say your GPU degradation is because of power shortage but it is worth looking into.
RTX 3060 Ti demands 215 Watts at full load, PSU power efficiency drops over several years, so if your PSU is for example over 10 years old, a 650 Watts could have the equivalent power rating as a 450 Watts PSU or worse. But this is just a general example. It all depends on the tech & circuitry component inside a PSU.
You also have to take in consideration of CPU power draw, whether you have any mechanical drives (DVD, Hard Drives etc..) how many powered USB ports, these all soak up wattage.
here is a basic breakdown of watt usage by modern components
you can get a rough idea at what your system is drawing here https://www.bequiet.com/en/psucalculator
It is also worth checking if Windows 10 - 11 as hardware-accelerated GPU scheduling turned on, by default windows as it off System-->Display-->Graphics-->Default graphics settings turn on Hardware-accelerated GPU scheduling
hope this helps
Hey Thanks for all the info!!!!!
@GraPhiX-Guru GraPhiX-Guru @earthling45 @LeeBamberTGC @MonkeyFrogStudio I would like to thank you all for helping in this problem of mine, YES it was the power cable, I switched around one of the cables so now both power inputs are on separate cables and I am now hitting the 50's FPS on Island Showdown, now obviously there should be more speed but with the further work that Lee is doing we should all see even more improvement in FPS....thanks again to all..............I am a very happy little camper!!!!!!!!!!!!!!!!!!!!!!! @LeeBamberTGC Very strange all the numbers seemed right for running GameGuru however even though all the numbers looked ok it all came down to a lack of juice so hardware. I will be in contact with those I bought the rig from and asking them why they added the card on 1 cable....I think they should have known better!!!!
Stefos
@Stefos-art Wow, what an amazing thread!! Big thanks to all who helped here, and definately one for me to remember next time someone has mysterious performance issues :)
I have a NVIDIA GeForce RTX 3060 12 GB of Graphics memory, 16 RAM system memory, however I am lucky to get 30 fps best and around 20 fps lowest. I don't look to have a problem in other apps, though I don't really play games. I am a 3D animator and don't look to have problems with those apps I use.