Advice wanted on a sticking issue with rail sim

[bharrison]

Hi,

I'm looking for some advice on a sticking issue I have on Rail Sim. I've noticed that on the S&D route and the Paddington to Oxford routes mainley the sim stick for a few seconds then carries on again. It happens mostley between Paddington and reading on the GW mainline And at various points on the S&D route I noticed it on the delivery day senario.
I've installed both upgrades. Which I was advised to do when I mentioned this to the people Ihad purchased it off. My operating system is windows Vista

My pc Spec's are below

Processor: 2.0GHz
Memory: 2GB RAM
DirectX Version: DirectX 10
NVIDIA GeForce 8500 GT
HD is 222GB

Thanks
Ben

[ightenhill]

Can you see or hear a lot of Hard drive activity at this point.. If so defrag (not with the windows one though unless your in safe mode) try diskeeper8.

PS what processor type?

[Acorncomputer]

Hi bharrison

I think you are referring to 'stutter' and 'freezing' which is a subject that has been discussed extensively on this forum and there are many threads that include something about it. The effects seem vary from person to person so what is right for one person does not work for another.

Try looking at this thread as a starter http://forums.uktrainsim.com/viewtopic. ... sd=a&hilit

[bharrison]

Ightenhill you have a Pm.

Ben

[sjbaker34]

Forget diskeeper8, try perfectdisk 2008 a lot better in my opinion. cheers Steve.

[ightenhill]

Forget diskeeper8, try perfectdisk 2008 a lot better in my opinion. cheers Steve.

probably not much in them, they all do the job.. just avoid using the built in windows utility..it doesnt do the job

[PaulH2]

Forget diskeeper8, try perfectdisk 2008 a lot better in my opinion. cheers Steve.

probably not much in them, they all do the job.. just avoid using the built in windows utility..it doesnt do the job

Well, it does, but is much more limited than commercial utlities.

I would tend to agree that there isn't much in it between the various utilities, but Diskeeper's prices have started to get pretty high over the past couple of years, and if you have a 64-bit system you'll have to go for at least the Pro version.

[bharrison]

Has anyone manged to stop Rs Stuttering yet? Thats why i Gave Up on it. i found it off putting.


thanks

[jkxx74]

I wrote a bit on this in another thread but will try to elaborate a bit more here.

The issue of stutter in RS arises for a variety of reasons, some related to current computer architectures and the nature of the sim itself, rather than deficiencies in the program itself. (I feel RSDL has been unfairly blamed for something that's not quite their fault here..)

In terms of software, all of the suggestions above will probably help with RS running a bit more smoothly. That said, let's look at the way an FPS handles content vs the way something like Rail Simulator does:

* In an FPS the development team has pre-created the entire level complete with textures, buildings, and other content
* The level is usually quite small, confined to an area less than a square mile in many cases
* Although vehicles are sometimes available, players walk on foot, requiring an effective screen refresh speed of less than 5 mph when running\
* The smaller level about can fit into the available RAM on most PCs available today

Now let's compare to RS:

* RS is completely open, so we the users can place as much content as our computers will handle (potentially)
* Trains in RS travel at really high speeds, exceeding 100 mph for trains such as the HST. Terrain moves past at a really fast pace in this case
* The routes in RS are many miles long, with the 'short' ones easily covering 20 miles in length and much more if the area of the routes is considered
* There is no way an entire route can fit into the 1 - 4 GB of RAM available in today's PCs
* RS has to calculate shadows and other features in real time as the users can produce infinite varieties of route content

And so on, but the point is that a comparison to other games isn't really fair. What RSDL can do is require a 64-bit OS (to address more than about 3 GB of RAM, the effective limit in 32-bit systems) and request that the user install 8 GB of RAM or more, then cache the entire route's contents into that RAM upon loading a particular scenario. The only other alternative to this is something like the gimmick below -

http://techreport.com/articles.x/16255 - That drive is supposed to deliver read rates at up to 400 MB/sec when used in a RAID 0 striping configuration.

So, yes, RS can be made to run much smoother but at a much higher cost to the end-user.

Cheers,

jkxx

[paulz6]

Ben,

I think you could really do with another 1GB of RAM. Also looking at your machine spec, I don't think you are going to get impressive performance.

On my idle (but with long up time) Vista system, running a web browser and the task manager only, my machine is using the best part of 1GB, with a 1.5GB page file. Running a heavy AI Ox-Padd scenario, the memory usage reaches 1.8 - 1.9 GB and the page file increases to 3GB. The rest of the free memory is taken up the the cache.

Main memory isn't just used by application data and the O/S, it is also used to cache data and as a backup memory area for the GPU. (I think the GPU may use the main memory similar to how the O/S uses a swap file - it there is not enough graphics memory to contain the textures then it has to resort to using the main memory - perhaps someone who knows what they are talking about could expand.) You should look at least 50% more memory than your application data requires in my opinion.

My PC is a Q6600 (2.4 GHz), 4GB RAM (only 3GB recognised on 32bit Vista), 2x8600GTS (2x256MB) system. Running in non-SLi window mode (I use two monitors for development), I just get acceptable performance in my opinion. I don't get major stutters but the frame rate can get a little choppy in places.

Your CPU clock speed is not as quick. Also the 8500 only has about half the performance of a 8600.

I think those who get decent frame rates have at least 8800 cards (about twice as quick again). I've also heard someone running a 64bit 8GB machine noticed quite a difference.

[bharrison]

I've changed my Machine since I oringinally put this post up.

My new pc Is much quicker and has 2Gb ram and a dual core proccessor.


Thanks

[dipper6]

Regarding Defrag programs, Iu have tried diskeeper, Perfect disk and a couple more, but my all time favourite is
O&O Defrag.
It can defrag in any of 5 methods unlike the others.

STEALTH Method
The STEALTH method differs from the other methods in that it is much faster and uses much less main memory for the defragmentation.
The STEALTH method has been designed with large file servers in mind. It can, however, be used to great effect on workstations and other computers.

SPACE Method
The SPACE methods offer you a great solution for heavy fragmentation. It also allows you to consolidate your data so that the amount of contiguous free space is as large as possible and further fragmentation is prevented.
The SPACE method is the best one for your initial defragmentation and for defragmentation in the background, as long as there is enough free space on the hard disk and the number of files is not too great. Otherwise, you should consider using the STEALTH method.

COMPLETE/Access Method
The COMPLETE/Access method defragments your files and reorganizes your file structure. Although this method is slower than the STEALTH and SPACE methods, it guarantees maximum system performance when your files are being read.
The COMPLETE/Access method is suitable for servers and workstations. Please bear in mind that this method uses a lot of central memory due to the reorganization of your system. If this is likely to be a serious problem for your system, we recommend you use the STEALTH or SPACE methods.

COMPLETE/Modified Method
The COMPLETE/Modified method defragments your files and also reorganizes your file structure. Although this method is slower than the STEALTH or SPACE methods, it guarantees maximum performance in the reading of your files.
The COMPLETE method is particularly suitable for databases and file servers. It is important to bear in mind that this method requires quite a lot of main memory due to file reorganization. If this is a serious problem for your system, we recommend you use the STEALTH or SPACE methods.

COMPLETE/Name Method
The COMPLETE/Name method defragments your files and also reorganizes your file structure. Although this method is slower than the STEALTH and SPACE methods, it guarantees maximum performance for the read access to your system.

The COMPLETE/Name method is particularly suitable for system drives. Please bear in mind that this method needs a lot of main memory due to the reorganization process. If this is a serious problem for your system, we recommend you use the STEALTH or SPACE methods.

http://www.oo-software.com

[paulz6]

What RSDL can do is require a 64-bit OS (to address more than about 3 GB of RAM, the effective limit in 32-bit systems) and request that the user install 8 GB of RAM or more, then cache the entire route's contents into that RAM upon loading a particular scenario.

It sounds a little unnecessary to load up the entire content of the route into RAM. You would need a lot of RAM for a large network.
It requires a little multi-processor thinking, and since most new computers come with two cores these days, it is probably the way forward.
If you are accurately modeling an entire networks rail operations in real time, then all the rail network will have to be loaded into RAM. You only need to load up the scenery tile data around the camera though. Having a separate tile data manager thread would help. All it has to do is monitor the operators location and load up and drop data tiles in advance as the operator moves location. The tile data manager would be pre-loading the data requiring blocking synchronous disk reads. The block doesn't matter, the real-time processing threads can keep doing what they are doing without the synchronous wait, as long as the tile data manager thread loads the next tile data in advance of the real-time threads wanting to use them. I think this could be efficiently done without swamping the bus in a game of cache-line ping-pong, which is the real performance killer of real-time multi-processing systems.