FPS Loss - Win some Lose some?

[coastaldriver 915]

Well this is not a complaint about Steves DX10 but a lesson learned about low end systems and FSX. After much consideration I obtained the DX10 Fixer and installed it on my system. 

Here is the outcome, it did everything it claimed, so 10/10 for that. The changes were great, particular the shadows which the Fixer even had showing in older Legacy aircraft quite well and I really love the cloud shadows on the ground and water. Overall it was wonderful and I was very satisfied BUT and here is the BUT, it killed my sim performance to the point of being almost unusable with DX10 selected. My FPS dropped from locked 30 to 6 and sometime even 2. There was no tearing or graphic flashes, just slow slow slow. 

I investigated a number of tweaks I had no considered before but they only marginally improved the loss of performance as measured by FPS, curiously it was worse from the VC but improved with the outside view. 

So near yet so far. The issue - while I had a 4.0GHZ CPU with 8 GB DDR RAM and a 2GB Video card, it was an older dual core Intel system and because it was an off the shelf system it had an older Intel G41 Video card and it was quite clear that FSX was overwhelming the GPU for the video card, end of story. So whilst DX10 did everything it said it would and the results were tantalising good not on this older system. So DX10 was deselected and I am back to DX9. 

Solution well first as I always thought would be the case there is only such much performance you can squeeze out of the older hardware and chips and I can thank the DX10 Fixer for really making this very clear. So the only solution if I want those goodies, a new PC with a high performance CPU and a high performance Video GPU.

 

Or is there a way to get things sped up again? I have the Buffer Pools fix in, the fibreframe rate and other similar config fixes. so I cannot think of anything else at the moment that seems sensible and plausible. 

[Chock 17,769]

You could try FSX Fiber Accelerator and FSX Booster Live, both of these (and yes, you can run them both at the same time) can and almost certainly will squeeze a bit more performance out of your machine. Now, they ain't free, but if you shop around, you can find them both for just over a tenner each (they are £12.99 on Just Flight for example and you might find them for a tad less if you look around further).

Now, a word about these utilities...

Technically, FSX Booster live is not really doing much you couldn't do manually with tweaks, but.. it does nevertheless allow you to do them on the fly, which is the real bonus of it, in that you can make a flight from a fancy payware airport with the slider dropped down to help frame rates, then once you get up to a bit of altitude and a bit of a distance away, where the sim does not have to draw that airport any longer, you can push that slider up (which is the 'Live' bit of it, i.e. no need to go into the options and stop flying and wait till the sim reloads things).

Fiber Accelerator on the other hand does a lot of that stuff on the fly automatically, so it is a bit more 'set and forget'.

Of course, no bit of software is going to turn older hardware into newer hardware, but I can absolutely confirm that these two things do work and will speed things up a bit on a system which might be struggling, and for about 20 quid (if you got both of em), it is not a bad interim solution, certainly costing less than a hardware change, and of course if you do get around to changing the hardware, you can still use those utilities to squeeze performance out of your sim, so they don't become redundant for it in that circumstance.

[SteveFx 258]

The first thing to figure out is why its so slow.

Uninstall the fixer libraries so it isn't involved  and then compare performance in native DX10 preview and  DX9.  

The only reasons for a huge differential in performance that I am aware of  are

1. converting all effects in the fixer,  (uninstalling fixer libs will back this out)
2. having 32 bit water textures. 
3. Different AA settings - you haven't said what you have.

Then try with the default Cessna at a default airport someone where you have no addon scenery - so perhaps Malaysia.  Don't initially compare in the VC as the VC shadows in DX10 require an extra pass of the aircraft model which could make a difference on a low end card. 

If you still see a difference change the water slider to 1.Max as the extra work in the more complex DX10 water shader could have an effect as well.  

I have just realised that you are trying to use cloud shadows as well - again that requires extra work in the GPU so I would suggest turning that off to initially compare as well.

Can you supply a name or order number as well please.

 

[ualani 4]

I have roughly the same system specs and have no trouble running DX10. 

[coastaldriver 915]

OK, thanks Steve et al. My order number was : STE170622-2733-12108 (Mike H).

I will work back the way suggested as I can see no reason why it should go so slow either. For example I have had no trouble with a high end payware model like the Captain Sim B707 into Capital Cities with live weather and heavy cloud etc. 

A bit of extra information - I am running the REX 3 PLUS Texture sets but not the REX weather engine for that I am using (FSXWX), I will check re the textures being loaded (bit size).

I have a lot of ORBX FTX scenery including Global and Vector and ORBX Trees. 

I did the tests at Avalon (YMAV) in Victoria Australia because it is an ORBX AU area but a very basic airport with little in the way of objects. There was no AI and no ATC used and it was daytime. I tried a number of aircraft, good payware native FSX such as; FSX Default Beaver, the B737-800 FSX Default, the A2A Lockheed Constellation,  Captain Sim B737, the JBK DC4 and JBK Shorts S23 (FSX Portovers), the JF Mosquito and Milton Schupes D18S (FSX Native). There were absolutely no issues with any of the aircraft selected, VC texture or gauge displays or gauge operation (clickable spots), outside view was fine with some shadowing. I fly VC with pan-view via the mouse selected. All displayed shadows nicely and there seemed to be a slight improvement in overall texture clarity. I was very impressed with the shadows displayed in the old legacy JBK aircraft. Cloud shadows were evident but I had to bump up the water level to 2.0 X Med, to get it to display as suggested. All aircraft were selected to start and run at the same place at the same time of day with no active real weather being injected or used, standard sunny day with few clouds. Basically they were all as bad as each other and the FPS was pulled down to as low as 2.9 when in VC the best I could get back was a 6.9 when F11 or outside view was selected. There were no flashes or tearing evident in any textures or visuals but switching from VC to Outside and Back again could take 30 secs plus to reload the relevant textures but after reload they were fine. 

I have frames locked at 30 fps and deselected AA. I have no light bloom or aircraft shadows selected via FSX. I have my scenery levels set to those recommended by ORBX and 4096 texture load in the FSX. cfg. Traffic is less that 10% for all categories except road vehicles were it is only 2% and maritime traffic is set at 40% and airport activity at medium. Cloud draw distance is limited and thermals disabled as well.

The only tweaks (so called) I have in the FSX.cfg are: HIMEM=1, BufferPools=0, Job Scheduler Affinity Mask=14 and Texture Bandwith Multi set at 40. Other than that it is all settings selected or set by FSX based on my system. I might add that the REX setup is also based on selections for a dual core system for texture transfer not quad core. 

 

So as per your suggestions I will redo the same test but select a different area (non ORBX and a basic airfield with little objects other than default objects). Cocos Island because I tend to do a lot of simming in flying boats and amphibians.  Day time and Night time to see if there is a difference, I will use the above aircraft again.

First I will check the textures being loaded for water and REX. First for the water then to drop them back to 2048 sizes for all other categories, I will try with different settings for the water slider as well

Second I will check DX10 preview mode only with no DX10 Fixer input. 

Third I will check with DX10 Fixer selected. first with the lesser textures in REX and then back to bigger texture file size for day and for night. 

I will check with pre-loaded weather and then real weather to see if there is any change with and without cloud shadows. 

I appreciate the suggestions above from others but I will stay with the basic tweaks I have as I can see no advantage in a 3rd party interface for this stuff. 

This may take a while to do properly so will get back in probably several days time. 

[SteveFx 258]

ok, can you provide more details on your CPU and GPU

Are you saying that you run on the G41 built in video chip

I test most things on a Q6700 at stock with a AMD HD7750

[coastaldriver 915]

Okay first things first: 

CPU Is Intel E5400 Dual Core 8GB DDR3 and Video Card is Intel G41 (2GB). Yes I am aware it is low end for FSX and gaming generally but it is Windows 7 64 Bit with DX11 installed. (I confess to being a Unix/Linux person and suffer MS OS's of any sort). 

All the following conducted using FSX Default C172 at YCCS (Cocos Islands - Australia, thereby no ORBX scenery and default FSX scenery only) daytime. 

Tests; first I used lower end texture file loads, first to 2048 max then to 1024 max. I used the 512 kb Rex textures for water and held the cloud at 2048. Water effects low. DX10 was selected. Performance still slow, FPS varying from 1.9 to 11.9 generally about 6.0. Stuttering noticeable but shading operative and no tearing or unusual visuals. I did note a marginal pickup in FPS if fullscreen mode was selected as opposed to windowed mode and FPS would jump wildly from 1.9 to 11.0 and continuously, so no steady state achieved. There appeared to be some lag with respect to keyboard and mouse inputs and about a 2 FPS gain if mouse was not used in the VC. 

I then deselected cloud shadows, rain effects, and no AA but kept DX10 and legacy shadowing. Slight marginal improvement but nothing significant. I probably could have lived with this but the FPS would fluctuate between 3 and 12 in the space of seconds and a constant steady state was not achieved. 

I should add that I have only ever aimed to have have a smooth sim performance and avoid highly detailed textures etc for the sake of smoothness and sim performance. I don't chase FPS as long as it is smooth and have avoided most tweaks as being unworkable in a 32 bit system for reasons understood by most. 

No DX10 Fixer but FSX selected DX10 preview only produced some odd texture results or blocking especially for the prop or dynamic textures like smoke, this cleared after about 1 minute and reverted to normal, some shadowing but odd cockpit texture striping. Performance was slightly better but still not smooth and about 60% lower than normal. 

I did note and investigate Monitor performance (LG 23M45 wide screen) as well. There does seem to be some communication issues between the PC hardware and the Monitor, indicated by an occasional loss of signal popup when screen changes were made by FSX, for example from windowed to full screen mode. Sound on or off had no effect on performance. 

I am running a Logitech 3 joystick and FSUIPC, they have not been finely tuned to work with each other. And there can be lag at times between the joystick button selection and event activation in FSX. Same with the keyboard in FSX. I have not been able to resolved these issues and calibration usually does nothing but set it up. 

All drivers for all hardware are up to date as is MS Win 7 for patches and updates. 

I then uninstalled DX10 and reverted to my FSX setup with max texture load 4096, no DX10 and water effects 2.0X medium. Performance returned to normal, i.e., averaging 28-30 FPS no stutters and smooth performance, VC and outside screen. FSX loaded faster, (normal startup is 2 mins and after flight parameter selection, FSX can be up and running in as little as 1 min 30 Secs, which I regard as acceptable. 

Steve I do not think it is a DX issue per se, excepting forcing FSX to use DX10/11 which does put an extra graphic load on the system, my gut feeling is that is a hardware/software communication issue and therefore the hardware (I/O) addressing setup and the culprits are the monitor, keyboard, mouse and joystick, which it is I have not been able to define. 

If I can run FSX with a high end addon at 30 fps smoothly with dynamic weather then the PC has the capacity to use DX in my view but forcing it to go to DX seems to cause an issue and exposes the other issues namely the hardware input/output management by the BIOS and Win 7. 

 

Still If you have any other views I would appreciate it but I can get shadows in the VC  and up the performance but not a smooth running outcome. 

[SteveFx 258]

Ok, so you get 28-30fps with motherboard graphics - what do the device entries in your fsx.cfg look like?   I would have expected that on very low end graphics that DX10 would in fact perform better but this is a somewhat extreme case.  Have you never considered buying a low end AMD/Nvidia GPU?

[coastaldriver 915]

SteveFX -Discovered a doubled up Display Device Entry one setting for Trilinear (My Default) and the other for Anisotropic. I removed one entry and will see what happens now.  Other than that no other devices in the cfg other than a primary device with a GUID reference and a sound device with a GUID reference. 

Yes investigating a new AMD/Nividia GPU simply because Intel no longer support the G41 (works fine) by I think my PC was one of those HP combos from some time ago so everything was matched with HP stuff. Nothing against HP like their hardware, always been robust and long lived for me. 

I will see what happens now no DX10 and DX10. Other than that all I can say is I think your little program did what it was advertised to do, so no complaints but the performance issue is just one of those puzzles you get in the PC world from time to time. And I did not purchase it to get a performance boost, I actually expected a small hit, but I must confess I really do like the visuals of the shadows in DX10. 

 

Cheers

[coastaldriver 915]

SteveFx, FWIW, no change, FPS is a little higher but unstable (fluctuating between 1.0 and 16). Anyway I have returned to DX9 for the time being and off to get a decent video card (NVIDIA), It definitely is the Intel G41 GPU that is the source of the problems as indicated by the monitor reporting the ocasional loss of signal (milisecs). So wasted enough of your time on this issue. It seems that Intel card is actually about 15 years old in terms of design and it just does not like being asked to process DX10 graphics from a 32 bit software program. Lessons learned.

Still think DX10Fixer is a great product and would recommend it to any FSX user (except those with a low end system.)

 

[SteveFx 258]

Ok, the things to watch out with an off the shelf system and a replacement GPU are power consumption, card height, card length and whether it occupies a single or double slot.

[coastaldriver 915]

Thanks for the advice. Cheers