Seeing game freezes tied to voice activation

[100Coogn]

I've been getting freezes that last about .5 seconds, when receiving transmissions.  This was using DX11 version.
I'm going to try the DX9 now.

Coogan

[Mano]

I uploaded a film last night for 2 other bugs relating to buzzsaw terrain. I am pretty sure there were some mini freezes in my film. I will check it tonight and mark what time they occur and re post.
I will copy my dxdiag as well. I checked my msconfig  to remove anything from the start menu that might use resources.
All of my updaters check for updates are turned off as well.

 

[oboe]

'm getting pretty regular micro pauses in the game, and I am not currently using voice, and the screen freezes weren't tied to a transmission I received.  Tonight each time I got one, I checked Net Status and there is always a spike in Variance after I've experienced a game freeze/pause.  About 0.5 seconds in duration sounds right.  I took some screen shots of the variance graph but I don't know if Variance is tied to the pause or just accurately reports on a pause having just been experienced.

EDIT: Freezes are occurring in DX11, just played DX9 and didn't notice any

[Mano]

I have experienced mini freezes and they seem random.
I did not find a film yet that has mini freezes.......I will keep looking.
I did install a sound blaster card recently and it seems the freezes are less than before, but I am not really certain of that.

I attached a DXdiag in text form.

 

[Pudgie]

Last night I was running some system checks on my box after a stint of intense fighting at A1 while flying in a Spit MkV and experienced 1 1-2 sec screen freeze while playing. Opened Kapersky Internet Security 2015 and was looking around then it dawned on me that I hadn't made any updates to reflect the software change from the AHIII Beta versions of AHIII (Dx9 and Dx11) to the release versions of AHIII (Dx9 and Dx11) in the Trusted Applications settings under the Threats and Exclusions settings in the Additional tab so I went in and made these changes and removed the old syntax as both AHII and the Beta have been deleted.....then got to looking in the Performance tab under Computer Resources and noticed that the checkbox that instructs Kapersky to concede resources to apps when the CPU and\or disk usage are high wasn't checked so I checked this as well so Kapersky will bow out when AHIII is running and using the CPU\disk.

Will reboot my box to ensure all has taken effect then will run some testing.......err, playing............and will post back on the results.

 

[Pudgie]

Alright, just got done w\ running AHIII after making all the changes mentioned in earlier post....................

While in 1st pass from A1 to A2 to take on the cons defending it I had no issues at all...........all was smooth holding 78-79 FPS w\o any freezing, pausing or anything........excellent operations. After battling a heroic but useless attack against 4 cons (1 Yak3, 1 NIK, 1 P38 and 1 Spit MkXVI) in addition to ground fire me and my trusty Spit MkIX went down in flames but not before landing several good hits on the NIK as he was manoevering using a spiral dive trying to shake me off his tail but using my throttle to control the Spitty's rotational forces I was able to maintain the inside track and keep the NIK at 200 yrds and crossing into my gunsite (ended up at the deck crossing over the town at A2 where I got hit by ground fire)...........

So during the climbout enroute to A2 on the 2nd pass from A1 I started working my views as fast as I could to try to induce a screen pause (this would come thru the CPU to the GPU to be displayed to screen.....control inputs to be synched w\ the graphics frames along w\ the sounds) and I was able to induce a very short screen freeze approx. 5 sec in duration but after this I could not create another 1. Once back at the target I engaged the same cons and we battled it out again this time latching onto the Spit Mk XVI and manoevering for position in the vertical so I was having to keep visuals on the XVI as this was in his advantage and didn't quite keep aware of the ground and just when I had caught the XVI in a moment of weakness I entered an accelerated stall and couldn't pull up so I augered quite spectacular into the ground.

Here is a copy of the MSI Afterburner capture of the minute screen freeze from whipping the views around on my box attached below:

 

[oboe]

...and I was able to induce a very short screen freeze approx. 5 sec in duration...

Just saying, but I think a 5 second screen freeze is not "very short".   Did you mean 1/2 sec?

[Pudgie]

Just saying, but I think a 5 second screen freeze is not "very short".   Did you mean 1/2 sec?

Good catch, Oboe.............

I had put the period behind the 5 like this "approx. 5. sec" when I typed it the 1st time then removed it after noticing it but got fooled by the period after "approx.".......................

Yes that should have been .5 sec.

Also doesn't help if I'm not wearing my glasses when I'm typing either....................... .

 

 

[JunkyII]

Just saying, but I think a 5 second screen freeze is not "very short".   Did you mean 1/2 sec?

My screen freezes are about a second.....still need to record and see if it shows in film viewer.

[hitech]

Pudge thanks for the screen shot, I noticed gpu % was spiking just before freeze, you gave me a thought , I am wondering if it could be a shader bug causing a very long frame.

HiTech

[Pudgie]

Pudge thanks for the screen shot, I noticed gpu % was spiking just before freeze, you gave me a thought , I am wondering if it could be a shader bug causing a very long frame.

HiTech

No problem, Hitech.

Just so you know I've been working on my side of this as well. I've been doing some research and have made a setting change on my box to see if it helps w\ this as well. When I looked at that graph it said to me that my GPU had ceased work and was starting to down clock due to not having received the interrupt request from the CPU to display frames and that work was in the mem cache to get......

What I've done is disabled hyperthreading on my I7 5820K CPU to force Windows to better parse threads across the 6 physical CPU cores instead of hyperthreading threads down fewer CPU cores w\ the thought that by doing this using the 6 physical CPU cores instead of using say 3 of the 6 physical cores due to hyperthreading the CPU will actually become more efficient as from all that I've read it is my conclusion that once a CPU has at least 4 or more physical CPU cores present then hyperthreading at the consumer level doesn't deliver the CPU efficiency as it would have on a CPU w\ less than 4 physical cores due to the nature of how hyperthreading works relative to the avg consumer computer workload, including playing games, and the way in which Windows 7 allocates threads to a CPU core.............

So far from my testing I have noted my CPU did show to gain some efficiency as noted using Performance Monitor reading the threads processed from before to after using the exact same test procedure that I used (went from 1131 to 1301). So far when flying AHIII this morning since I did this I haven't encountered 1 screen freeze to date but I'll need to put more run time in before I make any definitive call on this but I certainly do like how my box is running since I've done this and the MSI AB graphs do show the CPU usage % did go up approx. 12%-15% from prior usage of 9%-12% (21%-27% now so my CPU is more actively involved while the game is running which should help w\ reducing\eliminating the screen freeze\pause that I have noticed if the issue was due to CPU processing as I DO run a LOT of component monitoring softwares in the background on my box).

 

Anyway if you need something else from me just let me know. Always glad to help.

 

[ccvi]

Maybe I'm misreading this, but isn't short increase before dropping to ~0 in the GPU% at the end simply related to alt-tabbing out (or quitting), and the freeze is somewhere earlier in the data?

[Pudgie]

After playing most of the day yesterday in at least 3-4 instances I didn't witness 1 screen freeze occurring regardless of the numbers online or whether the bulk of the fighting was at low alt where the graphics load would be highest.........

But after I had gotten shot down and was on the runway at A1 and had just started up the ole Merlin (prop was starting to spin up so there was visual movement in the graphics frame sequencing), a Bish player cued their mic and lo and behold I saw the screen pause for a split second then resume.

This is the 1st time that I could directly tie voice activation to a screen freeze\pause. Never saw another 1 after this even though several players were cueing their mics and speaking frequently........... I also demonstrated that this can happen due to other activities so this is not directly tied to voice activation but this is a process that is seen as an input thru the game software to the CPU to the GPU..........

Gonna do some checking in my X7's control panel and in Windows to see if there is some recording setting that may be influencing this on my end but after working w\ this I'm becoming more convinced that the interrupt signal from the CPU to the GPU that more work is available in the mem cache to retrieve AND it's time to flip finished frames to display is somehow getting delayed so the GPU hesitates until this interrupt is received and the work is then transferred to the GPU then the GPU resumes rendering\flipping operations. Why this is happening in this manner is the $64.00 question IMHO.

Maybe I'm misreading this, but isn't short increase before dropping to ~0 in the GPU% at the end simply related to alt-tabbing out (or quitting), and the freeze is somewhere earlier in the data?

Hi Ccvi,

Don't look at the end of the graph data as that data is when I exited the game. Note the large spike in the frametime in the middle of the graph that lines up w\ the GPU clock rate drop that lines up w\ the FPS drop that lines up w\ a highlighted down spike in GPU usage %.....that is the result of the screen freeze created while I was whipping the views around as fast as I could....doesn't mean that the view whipping around was the direct cause but the increased amount of data input into the game client software to get processed thru the CPU to then get sent to the GPU to process caused some kind of delay that caused the GPU to stop momentarily at that time as the large frametime spike clearly shows (recording the graphics frame sequencing of my Fury X and any enlongated spikes above the baseline represent a slow down in frame sequence timing). If you look closely down to the CPU usage % you will see a small highlighted shift in CPU usage % that also lines up w\ the GPU frametime shift...............

I have provided another snippet of the same graph where I used MS Paint to highlight this..............

This is why I run MSI Afterburner in the background (among a host of other component monitoring software) while I'm playing to record my vid card's performance. Using this software to monitor my Fury X I have captured and noted every instance of a screen freeze and to date they all have shown to interrupt the GPU in this exact manner, some not as severe as this 1 was but all affect in the exact same pattern. This data along w\ some other reading\studying up is the basis of my assessment of what's going on. Hitech is in a much better position to make use of this data than I am and is also why I posted a snippet of this so that they\all can see what I see to help them out.

This is also why Nvidia and AMD are big on the FCAT process (mostly driven by Nvidia at the outset) and using frame timing to TS and monitor and optimize GPU graphics performance.

Hope this helps.

 

[Pudgie]

Pudge thanks for the screen shot, I noticed gpu % was spiking just before freeze, you gave me a thought , I am wondering if it could be a shader bug causing a very long frame.

HiTech

Hitech,

Here are a couple of snippets attached below of MSI Afterburner graphs of my box running AHIII Release 3 Patch 6, 1 using Dx9, 1 using Dx11.........

In both of these I had AMD's Power Efficiency setting enabled (this would be enabled in Radeon Settings global driver settings by default....I had been running this disabled since the Beta days due to issues of GPU being excessively downclocked causing stuttering when running the Beta Dx11 version since Patch 22 forward) as noted as PE On.......

Both versions were run over TT in my Spitty Mk IX w\ AHIII client settings identical to try to maintain an apples-to-apples game load scenario as best could be done.

With the Dx9 version I couldn't get the game to exhibit a screen pause and you can also see that the Dx9 version worked the GPU and CPU much more aggressively as noted in GPU clock speed, temp and usage % as well as CPU temp and usage %.

With the Dx11 version I was successful in getting the game to exhibit a screen pause and you can see that the Dx11 version caused the GPU and CPU to downclock as noted in GPU clock speed, temp and usage % and CPU temp and usage % w\ the GPU downclocking more aggressively which caused much more stuttering as noted in the GPU FPS graph line difference between the Dx9 and Dx11 versions.

With all this in mind, would it be reasonable to think that the screen pausing\freezing may be a function of Dx11 interaction w\ the AHIII client? Is anyone who is using VR getting screen freezes using Dx11?

Also w\ the AHIII client software being the exact same as used in both Dx9 and Dx11, why would Dx9 work the GPU\CPU more than DX11? I would think that this would be opposite from my vantage point......... Probably need to do a little reading up on the differences between MS Dx9 and Dx11 to satisfy my curiosity.....

Anyway I put these out here in hopes this could help you out in solving this issue.

 

[Pudgie]

Just got done running the Dx9 version of Patch 7 and all went like clockwork.................no screen pauses, freezes, stutters, nothing but butter................

Had Resource Monitor running in the background monitoring my CPU and all was looking darn good............

Core0 was at 54%-57% usage, Core1 was at 78%-89% usage, the other 4 were no more than 5%-10% usage to come out to a total of 25%-27% total CPU usage (hyperthreading disabled).

Now that's how I want to see this I7 5820K CPU get utilized by AHIII where the 2 cores the client asked for it gets!