Driver File Omissions - Check It Out

[Rags]

Brian, there is a big difference between functionality and speed. Yes, if the systems are identical and the system was setup properly, the functionality will be the same. But monolithic drivers are faster for kernels designed to take advantage of them. The memory footprint is also smaller, and is measurable. As far as that article goes, I don't have the time or even feel the need to point out every mistake in there. There are quite a few misleading things that he posted and some of them are flat out false.

Rags

[Brian R.]

Have you a reliable reference?

[Butcher]

In what way are separate files supposedly more reliable than monolithic drivers? The only time I can see it being an issue if is your separate file is significantly newer than the monolithic file, and this doesn't happen with sane installs of windows as monolithic files only contain MS drivers, which are all updated at the same time.
If you're worried about files being corrupted and causing reliabilty issues you're wasting your time with your current setup, you should have RAID5 HDDs to make sure you don't get file corruption and if stability is a big issue for you, you should be running NT or 2k.
On the other hand not using a monolithic file is probably not a great deal of a hit... each file you rep[lace out of it probably adds about 100ms to your boot time, not realy a great deal.
That is unless monolithic drivers can load all at once, rather than it just being an archive.
If they load all at once, there is a significant saving in using a monolithic driver - up to maybe 1 mb of memory per driver. In addition to the fact that lots of separate drivers running is much slower than just one big driver handling everything (due to context switching).

[Gurm]

The individual driver files approach takes up SUBSTANTIALLY more space, and runs slower in 99% of configurations.

Period.

End of story.

This has been debated in numerous places before and it turns out that whether or not you like the idea of monolithic files, they are a reality and actually work as advertised.

- Gurm

------------------
Listen up, you primitive screwheads! See this? This is my BOOMSTICK! Etc. etc.

[Brian R.]

Butcher, the only way I am looking at it is from a 1st principle approach. The more complex something is, the more likely it is to fail. If a driver file fails, it fails. If vmm32.vxd fails, alot of drivers fail. Might be terminal until you can rebuild it or whatever. I have never seen as situation where a performance gain comes without some cost.

Don't get me wrong, I'm not strongly advocating this issue. I'm only passing on an interesting opinion of a friend of mine. His opinion and logic made sense to me. Now that I'm getting additional feedback, I'm seeing the other side of the picture and getting a broader prospective.

One thing I am seeing is that there are more than a few people with both opinions.

In the mean time, I have removed the files and allowed vmm32.vxd to do its job on my Win98 computers.

FYI, my computer runs W2K.

Thanks for the opinions

[This message has been edited by Brian R. (edited 22 September 2000).]

[Gurm]

For what it's worth, I have tried it both ways. When not using the monolithic file, I have run into MASSIVE stability problems.

This arises because when you use the monolithic file you have matching file versions all the time. Yeah, if the file gets corrupted (and how that would happen I have no idea since it's in memory ALL THE TIME) you're SOL.

So if you have 100 little files you're 100 times more likely to have a fatal error - all the files in VMM32 are required files which will cause system instability should they go south.

Add to that the fact that you're 100 times more likely to accidentally "upgrade" one of these files (happened to me several times) and cause a version mismatch... and it just isn't worth it.

And... Brian... just because I'm blunt and tend to overstate a little doesn't mean I'm wrong... now does it?

- Gurm

------------------
Listen up, you primitive screwheads! See this? This is my BOOMSTICK! Etc. etc.

[Brian R.]

Wrong or right is not the issue. The finality of your opinion was. No problem.

[Wombat]

Brian,
I'm not trying to nitpick here. I understand your view. It just seems to me that you are the type of person that would like to know what is actually going on.

>I understand they end up with the same functionality in the end after the vmm32.vxd file has been loaded

Functionality, maybe. However, separate files are slower. With a monolith, you just move a pointer variable. With separate files, you have to go through the formality of opening a file handle, maybe spawning a thread or two along the way.

Also, if you're gaming, every meg of RAM counts.

[Brian R.]

Wombat - No problem. Please clarify this issue as much as you like. I just want to understand the situation so that I can make an informed decision.

I am hoping to get someone with a different viewpoint in the discussion. I can't believe this issue is so clear cut. If you search the internet, there is great disagreement.