If you go beyond the cacheable limit you can just use a RAMDRIVE (I believe some of those allocate memory from the top to the bottom, thus you can make it grab all the uncached memory) and put the virtual memory pagefile there. Even uncached, memory should be much faster than some HDD, period correct or not (as they'll be limited by the ATA interface/PCI bus anyway)

Half-Saint wrote:

Just checked the modules, there's 4C4M4E8 -6 chips on one module, 4LC4M4E8 -6 on the other and GM71C17403CJ6 on the other two. Hope that helps.

That's EDO.

idspispopd wrote:

feipoa wrote:

OK. How much memory does GLQuake use (irregardless of the resolution)?

Quoting from the system requirements: "16MB of RAM (+ quite a bit of Virtual Memory)"
Having 32MB should already be quite comfortable.

When I load Windows, the Task Manager indicates that more than 32 MB of memory is already in use. For this scenario, is having a system with 32 MB of memory still quite comfortable, or would having more memory be more comfortable?

alexanrs wrote:

If you go beyond the cacheable limit you can just use a RAMDRIVE (I believe some of those allocate memory from the top to the bottom, thus you can make it grab all the uncached memory) and put the virtual memory pagefile there. Even uncached, memory should be much faster than some HDD, period correct or not (as they'll be limited by the ATA interface/PCI bus anyway)

Have you undertaken a performance analysis using this technique? It would certainly be interesting to compare it with a) using half the memory uncached w/SCSI HDD swap, and b) using less memory such that all is cached. Ideally, the RAMDRIVE swap method should be faster than b).

idspispopd wrote:

feipoa wrote:

OK. How much memory does GLQuake use (irregardless of the resolution)?

Quoting from the system requirements: "16MB of RAM (+ quite a bit of Virtual Memory)"
Having 32MB should already be quite comfortable.

If anyone cares - the VQuake readme makes a reference to needing 20MB under Windows 95 for "alias models". VQuake can run in DOS as well though. This is 1997 software as well however so it's not surprising.

feipoa wrote:

How much memory does GLQuake use?

I ran a little test in Windows 95c. In a system with 128 MB of RAM, the Resource Monitor reports 62.3 MB of RAM allocated at startup and 92 MB being allocated when GLQuake loads. Similarly, Windows NT 4.0 had 34 MB of RAM allocated at startup and 64 MB allocated when GLQuake loads. I decided to see how reducing the total system memory impacted the GLQuake timedemo score. The results are as follows:

Total RAM
128 MB = 25.3 fps
64 MB = 25.3 fps
32 MB = 25.0 fps
16 MB = 23.6 fps (swap file activity causes pausing in game)
12 MB = 22.6 fps (swap file activity causes pauses in game)
8 MB = Windows gave some error and would not boot.

Half-Saint wrote:

kixs wrote:

Half-Saint wrote:

I found two more 32MB 72-pin SIMMs in my stash and I'm wondering, what the advantages of having 128MB in a 486 system would be? Back in the day, I only had 8MB in my 486DX-40 and it worked rather well.

Are these FPM or EDO? As not many 486 chipsets would support EDO.

Just checked the modules, there's 4C4M4E8 -6 chips on one module, 4LC4M4E8 -6 on the other and GM71C17403CJ6 on the other two. Hope that helps.

These are EDO. On what motherboard are you using them?

Shuttle HOT-433 w/ AMD 486DX4-100 😀