Fraps and Aces High

[Ack-Ack]

Um... that's weird.  I thought it was really low quality and very big hit on the computer's performance, and not at all intended for recording games, as its intended for capturing the Windows desktop.  Would you explain how you configured it?

Also, now that you're here, Chalenge earlier said:

Can you elaborate?

It is a resource hog but so is Fraps and from my experience, Camtasia actually has a smaller resource footprint that Camtasia.  As for quality, it all depends on what you've selected for your settings when you set up your recordings.

Here is Camtasia's resource page that's of great help when using this program.

Camtasia Studio Learning Center

ack-ack

[Chalenge]

I didnt realise it was Flash based. Thats going to be a problem. With a DirectX game like AH I think you are much better off sticking to FRAPS but at least there is a trial version of Camtasia for you to try.

[Skuzzy]

<snip>This is actually quite useful information.  So what you and Skuzzy are essentially saying is that Aces High is CPU bound, while WoW is more GPU bound.  

That is not what I said.

To me the computer's performance while playing a game such as WoW or Aces High is directly related to the framerates achieved (and consistency of said framerates).  Saying Aces High is more demanding on my "computer" when I can get it to run at nearly max settings at 60 FPS (synced with my monitor) while I can't do the same in WoW is to me disingenuous.  Because to me, the computer is the entire system, not just one aspect of it that you seem to be measuring against.

I will say it very clearly.  Aces High is more strenuous on your entire computer system than WoW is.  WoW is network, sound, and CPU light compared to Aces High.  Graphically, they place different loads on the video card making a direct comparison difficult.

Check the CPU usage for each game, you will probably find Aces High is using around two times more CPU than WoW is (depending on the situation in each game).

Our graphic tradeoffs have more to do with the visual depth than anything else.  To allow you to see 17 miles in any direction requires a lot more graphic time than the shorter distances in WoW.  Wow is all about close up details, as you are almost always close to the objects in the frame.  It will be more graphically challenging than Aces High, but far less CPU instensive.

Fraps does not place that much of a load on the graphics card.  It simply copies the frame buffer from the card to the system.  Writting the data to the hard drive is where most of the herky-jerky behavior will occur in Aces High as that suspends the CPU during the writes.

You really cannot use a synced frame rate to derive any load differences.  Turn of vsych to get a better idea of frame loop load differences.

[BoilerDown]

That is not what I said.

Then one of us doesn't seem to know what it means to be CPU-bound vs GPU-bound.  Or are you saying that WoW is neither-bound?  Obviously something is the limiting factor for fps in WoW, and in WoW I can make the fps drop lower than I can in Aces High.  So what is it?

I will say it very clearly.  Aces High is more strenuous on your entire computer system than WoW is.  WoW is network, sound, and CPU light compared to Aces High.  Graphically, they place different loads on the video card making a direct comparison difficult.

Italics and underlining mine. The comparison is simple.  If AH has higher loads on those three parts, yet WoW can be made to run slower on the "entire computer system" than AH, something other than "network, sound, and CPU" must be the culprit.  I would bet on the GPU, because there's not much left to bet on.

Our graphic tradeoffs have more to do with the visual depth than anything else.  To allow you to see 17 miles in any direction requires a lot more graphic time than the shorter distances in WoW.  Wow is all about close up details, as you are almost always close to the objects in the frame.  It will be more graphically challenging than Aces High, but far less CPU instensive.

I think we're at agreement other than agreeing that we're agreeing.  Are you playing a "my game is harder than your game" game?  Cause I'm not a WoW dev, I don't care.  Other than I want to know why Fraps takes a much bigger chunck out of the gameplay experience in AH than it does in WoW.

Fraps does not place that much of a load on the graphics card.  It simply copies the frame buffer from the card to the system.  Writting the data to the hard drive is where most of the herky-jerky behavior will occur in Aces High as that suspends the CPU during the writes.

Again this is neatly attributed to... let me put it this way... WoW has more CPU-headroom than AH, so "suspending the CPU" wouldn't affect WoW as much.  

What if I had a quad-core instead of a dual-core?  How many threads does AH attempt to use?  Would Fraps be able to do its work (in an XP environment) with a quad-core without suspending a CPU that AH is depending on?

You really cannot use a synced frame rate to derive any load differences.  Turn of vsych to get a better idea of frame loop load differences.

I was under the impression that while you're at frames less than the vsynch of the monitor (60 fps in my case), suspending vsynch doesn't actually do anything.

[Skuzzy]

Then one of us doesn't seem to know what it means to be CPU-bound vs GPU-bound.  Or are you saying that WoW is neither-bound?  Obviously something is the limiting factor for fps in WoW, and in WoW I can make the fps drop lower than I can in Aces High.  So what is it?

Game performance is not something easily determined when viewing desparate parts of a computer system.  Aces High, nor WoW has to be either CPU or GPU bound.  It will depend on the system.

Italics and underlining mine. The comparison is simple.  If AH has higher loads on those three parts, yet WoW can be made to run slower on the "entire computer system" than AH, something other than "network, sound, and CPU" must be the culprit.  I would bet on the GPU, because there's not much left to bet on.

The comparison is not that simple and that is the point.  This is why I keep stressing "Aces High places a higher load on the entire computer than WoW does".  WoW can be made to run slower if you pump up the graphic details.  It (WoW) is all about the video card.

EDIT:  I meant to say something but got side-tracked.  A graphics engine can draw a frame in many different ways.  The different ways places more strain/load on one part of a video card, versus another part.  By moving the load around on a video card, the same video card might struggle with one game, but royally kick butt at another game.  This is one of many reasons why you simply cannot compare one game to another.

I think we're at agreement other than agreeing that we're agreeing.  Are you playing a "my game is harder than your game" game?  Cause I'm not a WoW dev, I don't care.  Other than I want to know why Fraps takes a much bigger chunck out of the gameplay experience in AH than it does in WoW.  Again this is neatly attributed to... let me put it this way... WoW has more CPU-headroom than AH, so "suspending the CPU" wouldn't affect WoW as much.  

I am not playing any game.  This has nothing to do with "ours is bigger than his".  FRAPS uses CPU which is going to hurt our performance more than WoW will be hurt, as we use a higher percentage of CPU than WoW does, in most cases.  I told you how to test it yourself.  You do not have to believe me.

What if I had a quad-core instead of a dual-core?  How many threads does AH attempt to use?  Would Fraps be able to do its work (in an XP environment) with a quad-core without suspending a CPU that AH is depending on?

You can have 400 cores, it will not matter.  If the program cannot fit into the cores cache, then the core will have to access memory to fetch the next instruction/data.  At that time, everything else that needs access to memory is stopped.  If the hard disk is being written to, all the cores must stop if they cannot continue within thier cache.  If your network card is a bu master (most are today), then when it needs to access system RAM, the CPUs are dead.  External write access to system RAM invalidates the internal CPU cache.

Games are horribly large executables with even larger sets of data.  It is rare (not sure it is even possible) to be able to contain much of a frame loop within the confines of a single cores cache.

Aces High is one of the few natively multi-threaded games on the market.  We use 2 cores, if available (AMD users are out of luck here due to issues within the AMD CPU itself).

I was under the impression that while you're at frames less than the vsynch of the monitor (60 fps in my case), suspending vsynch doesn't actually do anything.

Incorrect. With vsync enabled, it can cause blocking issues within the frame draw that can extend the frame draw time.  It will depend on the video card.  When testing, it is generally best to disable vsync just to be sure.

[Chalenge]

A faster CPU is the way to go versus the number of cores (I know... I said that already). i7 has other advantages being the later socket so it has things like native SATA3 which I believe helps out FRAPS too.

[TequilaChaser]

Aces High is one of the few natively multi-threaded games on the market.  We use 2 cores, if available (AMD users are out of luck here due to issues within the AMD CPU itself).

<--- starts scratching head ----

I thought all that was done and over with and why noone needed the AMD dual Core Optimizer patch anymore.....

I never have had any trouble with any of my 4 Dual Core AMD's I have built and tested or flown AH on

just saying ( and have been able to max out my FPS with hires text pkg and 70 or 75 frame rate depending whether iwas in XP 32bit(70 )  
o r Win7 64bit ( 75 )  

edited: ( I am not counting this piece of junk PC I am currently using which I have recently thrown together  which I mentioned in the following thread http://bbs.hitechcreations.com/smf/index.php/topic,303290.0.html   .... my main PC is waiting on a warranty replacement MB from ASUS where the PC User performed a ID10T error while doing my quarterly cleaning of it )

anyways, sorry if I sidestepped the conversation

[Skuzzy]

AMD supplied a work around, which basically keeps the game on one core.

[TequilaChaser]

so the work around did away with the having to use the AMD Optimizer patch I take it???

sorry if I am mixing things up, I was just trying to go from memory

[Skuzzy]

It is a slightly different approach, but essentially solves the problem the same way.  It still does not work with every version of the CPU, but it works for most.

[Ack-Ack]

I didnt realise it was Flash based. Thats going to be a problem. With a DirectX game like AH I think you are much better off sticking to FRAPS but at least there is a trial version of Camtasia for you to try.

The videos made from Camtasia aren't flashed based unless you choose to convert them to that format after recording. The ones I've done are in .mpg or .wmv format.  If he's using the trial version it may be limited to only being able to convert the video captures to flash format.

ack-ack

[BoilerDown]

Game performance is not something easily determined when viewing desparate parts of a computer system.  Aces High, nor WoW has to be either CPU or GPU bound.  It will depend on the system.

Obviously.  But I was interested and talking about my particular system (the specs were posted previously), not "in general" computers with any conceivable combination of CPUs and GPUS.  It was also a comparison between WoW and AH, with AH being more CPU bound and WoW being more GPU bound (on my computer).

If the program cannot fit into the cores cache, then the core will have to access memory to fetch the next instruction/data.  At that time, everything else that needs access to memory is stopped.  If the hard disk is being written to, all the cores must stop if they cannot continue within thier cache.  If your network card is a bu master (most are today), then when it needs to access system RAM, the CPUs are dead.  External write access to system RAM invalidates the internal CPU cache.

Games are horribly large executables with even larger sets of data.  It is rare (not sure it is even possible) to be able to contain much of a frame loop within the confines of a single cores cache.

Aces High is one of the few natively multi-threaded games on the market.  We use 2 cores, if available (AMD users are out of luck here due to issues within the AMD CPU itself).

This is good to know.  So a faster hard drive sub-system might be more important for Fraps-ing AH than more cores on the CPU?  Or are you saying it might be impossible to make it work better?

Also I just got an email back from a real human at the Fraps tech support.  Not a lot of progress yet, but they did say:

If you have access to another machine with a quad (or higher) core processor, you may want to try using Fraps on it instead.  We have designed Fraps to use up to two cores that are the least utilized by the machine, so having extra idle cores while playing the game can help Fraps to capture without impacting the game performance.

Incorrect. With vsync enabled, it can cause blocking issues within the frame draw that can extend the frame draw time.  It will depend on the video card.  When testing, it is generally best to disable vsync just to be sure.

I'll have to do some new tests then.

[Skuzzy]

A faster drive subsystem will help everything.  The faster the data can be written to the drive, the shorter the times where memory is not available to anything else.

Faster system RAM can also help, but that is a rather controversial topic as most people do not know how to get thier computers configured to maximize performance when using higher speed RAM.

While there is a lot to it, the basic reason my home computer can boot to the desktop in 4 seconds flat is due, in large part, to its faster RAM.  The HD is just a Seagate 500GB NS model.

[BoilerDown]

Got another email from Fraps:

Thanks for your reply.  As an example, I have tested Aces High on a high end machine with a quad core Intel i7 processor and was able to capture 30 FPS clips at 2560x1440, with the game running at 30 FPS solidly.  Alternatively, if I dropped the video resolution down to 1920x1080 (1080p) I could run the game at 60 FPS while still capturing at 30 FPS (with the Lock Framerate option disabled in Fraps).

From what I saw of the CPU performance meters on the system at the time I was playing the game, i noticed quite a high use of 2 physical cores, so having more than 2 will most likely provide the best performance boost for capturing this kind of CPU intensive game.

Some good work from them doing the testing I can't do myself without additional hardware.  Granted I have no idea what the rest of their system was, but this gives a good indication to me that a quad-core CPU will help out, as Aces High wants two threads, and Fraps wants two threads.

The Sandy Bridge CPU is officially released in less than an hour from now, and to support it I'd need a new motherboard that supports faster hard drive I/O and memory than I'm using now.  I think that covers my bases and I'm overdue for an upgrade anyways.

[BoilerDown]

So this past weekend I finally installed my new motherboard (Asus P8P67 Evo), CPU (2600K), and memory, reinstalled XP and drivers from scratch, re-did the tweaking to remove unneeded processes (mostly using advice from http://www.blackviper.com/WinXP/servicecfg.htm ), got the new (3.2.8) version of Fraps and Aces High installed, and started testing a bit.  Only my optical drives are plugged into the potentially faulty sata ports, so that isn't a factor in these tests.

For certain its much improved.  I've been capturing full screen lossless at various framerates and the game-play framerate never drops below 30 while recording.  30 is pretty slow you say?  Well I agree, but I noticed some really strange behavior.

First my results aren't directly comparable to my old results because of the newer Fraps.  The changelog for the new versions includes the following:

"3.2.8 - 3rd Feb 2011
- Greatly reduced CPU usage while recording on most configurations"

So maybe making the Fraps people aware of the problems with Aces High led them to tweak their code?  Who knows?

Anyways, what I found is that IF I capture at an un-locked framerate of 15 or more, Aces High always runs at a multiple of the framerate that I'm capturing at.  So what happens if I set the capture framerate to 30 is that the game will run at 60 until I get into some heavy action, then drop down to 30 and stay there.  It may fall back up to 60 if I leave the area that's hitting the framerates hard.  If I set the capture framerate to 15, as soon as I hit some action that drops the gameplay framerate below 60, it falls down to 45 and stays there except one time I hit a rather dense dogfight and it dropped down to 30.  If I set the capture framerate to 20, it falls to 40, a capture framerate of 24 was tested to fall down to 48.

But if I set the capture framerate to 14 or lower, the gameplay framerates change dynamically as you'd expect, lower in high action areas and synced at 60 in low action areas.

When not capturing this setup is good enough to keep the game at 60 FPS at all times, but I haven't tried unsyncing to the monitor frequency yet to see what happens.

Anyways, while improved, this problem makes things still not quite right.  I'm guessing I have plenty of CPU horsepower now, but either my hard drive or my video card is too slow to keep it at 60 FPS while capturing at all times.  But I can't justify the cost to upgrade those as well at this time, I'll just have to capture at 14 or 12 or something.

Btw, turning off lossless doesn't seem to have a measurable effect, so I keep it on.  I haven't tried half-size yet, my experience is that half-size looks like crap and takes up nearly as much hard drive space as full-size anyways.