VOGONS


First post, by Woolie Wool

User metadata
Rank Member
Rank
Member

I have a Matrox Millennium MGA 2064W set aside for a future socket 8 workstation build. I've tested it in my Athlon rig and it works nicely. It has 4MB of VRAM but I see daughterboards available on eBay in the $60 range (about three times the cost of the card itself!) to add 6MB additional VRAM. Would I see much benefit from this addition playing software rendered games on a 180 or 200 MHz Pentium Pro with 64MB EDO RAM? It could probably do 640x480 in Quake, but not far above that without using a Voodoo1/2, at which point the Matrox's RAM becomes irrelevant. MechWarrior 2, maybe I could get up to 800x600? What about the Windows NT desktop? Would I need more than 4MB to get to 1600x1200?

wp0kyr-2.png CALIFORNIA_RAYZEN
1wpfky-2.png REDBOX
3q6x0e-2.png FUNKENSTEIN_3D

Reply 1 of 7, by Unknown_K

User metadata
Rank Oldbie
Rank
Oldbie

The extra RAM won't help speed, it just gives you more color depth at higher resolutions. I used an 8MB Matrox Millenium on a Dual PPro system for years. Forget what the max resolution supported is, but the Matrox website should have the manual in PDF.

Collector of old computers, hardware, and software

Reply 2 of 7, by uzurpator

User metadata
Rank Newbie
Rank
Newbie
Woolie Wool wrote:

Would I need more than 4MB to get to 1600x1200?

for 1600x1200 and 32 bit color:
1600 x 1200 x 4 = 7.680.000 bytes, so 8 megs. For 16 bit color 4MB is sufficient.

Die ewigkeit ist hier und jetzt.

Reply 3 of 7, by Woolie Wool

User metadata
Rank Member
Rank
Member

So my max resolution comes out to around 1152x864 without the daughterboard. Well I guess building even an old workstation doesn't come cheap!

At least when I build it it will really be something distinctive--dual Tualatins, 486DX4/100s, and even my socket A build are as common as weeds, but nobody except me would build a simulacrum of a $12,000 Windows NT workstation from 1996 converted into the ultimate 1996 DOS gaming rig to slay all others.

wp0kyr-2.png CALIFORNIA_RAYZEN
1wpfky-2.png REDBOX
3q6x0e-2.png FUNKENSTEIN_3D

Reply 4 of 7, by shamino

User metadata
Rank l33t
Rank
l33t
uzurpator wrote:
Woolie Wool wrote:

Would I need more than 4MB to get to 1600x1200?

for 1600x1200 and 32 bit color:
1600 x 1200 x 4 = 7.680.000 bytes, so 8 megs. For 16 bit color 4MB is sufficient.

I would double the RAM requirement though in order to allow for double buffering. Without double buffering the GUI will likely have visible tearing and flickering which can be ugly to live with. Might also be slower.
I've never had this situation with a Matrox card though so I don't know how well they handle it.

Reply 5 of 7, by Scali

User metadata
Rank l33t
Rank
l33t
shamino wrote:

I would double the RAM requirement though in order to allow for double buffering. Without double buffering the GUI will likely have visible tearing and flickering which can be ugly to live with. Might also be slower.
I've never had this situation with a Matrox card though so I don't know how well they handle it.

Prior to Windows Vista, there was no double-buffering in the Windows GUI. It uses a system of dirty rectangles and redraws directly onto the frontbuffer.
The reason for this is probably because most video cards in the early days had barely enough RAM for a single buffer in higher resolutions.
Double-buffering a GUI is also much less efficient. Things changed with Vista because 3d accelerators were now standard, so you could assume lots of texture memory and accelerated blits via the texturing hardware.

Some applications implemented their own double-buffering though. This was done by creating a bitmap in system memory, drawing to that, and then drawing the entire bitmap to your window at once.
But since it is done in system memory, the actual amount of VRAM has no effect.

http://scalibq.wordpress.com/just-keeping-it- … ro-programming/

Reply 6 of 7, by shamino

User metadata
Rank l33t
Rank
l33t
Scali wrote:
Prior to Windows Vista, there was no double-buffering in the Windows GUI. It uses a system of dirty rectangles and redraws direc […]
Show full quote

Prior to Windows Vista, there was no double-buffering in the Windows GUI. It uses a system of dirty rectangles and redraws directly onto the frontbuffer.
The reason for this is probably because most video cards in the early days had barely enough RAM for a single buffer in higher resolutions.
Double-buffering a GUI is also much less efficient. Things changed with Vista because 3d accelerators were now standard, so you could assume lots of texture memory and accelerated blits via the texturing hardware.

Some applications implemented their own double-buffering though. This was done by creating a bitmap in system memory, drawing to that, and then drawing the entire bitmap to your window at once.
But since it is done in system memory, the actual amount of VRAM has no effect.

I'm confused then. I have an old Compaq LTE5380 laptop that I use with NT4. It has 1MB video memory and a 1024x768 screen which performs abysmally in it's native resolution (1024x768x8), with visually annoying traits in how it renders the GUI. If I lower it to 800x600 then the GUI becomes completely solid and renders cleanly. (I don't know what video chipset that laptop uses)

I feel certain that I've seen the same flickery rendering issues when overtaxing the available RAM on S3 Trio/Virge cards, but to be fair it's been years since I would have last seen it.

Maybe the issue is that overtaxing the RAM inhibits the card's 2D acceleration features?