VOGONS


Modern graphics on a 486

Topic actions

Reply 60 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++
ElectricMonk wrote:
leileilol wrote:

Isn't it obvious enough PCI is the bus of concern in a "modern graphics" thread?

Except it's a 486 system, which had more than just PCI, depending on the motherboard. ISA, EISA, and VLB ring a bell?

The first post indicates that I have tested these cards, Matrox G200, Matrox G450, Voodoo3, Riva TNT, Riva TNT2, GeForce 6200, Voodoo Banshee, GeForce2, 3DLabs Permedia2, S3 Trio3D, ATI Rage Pro, and ATI Rage 128, all of which are PCI.

Plan your life wisely, you'll be dead before you know it.

Reply 61 of 371, by leileilol

User metadata
Rank l33t++
Rank
l33t++
ElectricMonk wrote:

Except it's a 486 system, which had more than just PCI, depending on the motherboard. ISA, EISA, and VLB ring a bell?

12215635.jpg
Go actually read the thread, and notice how there are no ISA, EISA VLB Voodoos, Radeons and Geforces and stop assuming I was born yesterday with your lecturing of common knowledge and irrelevant hardware experiences to back up your 'expertise'.

apsosig.png
long live PCem

Reply 62 of 371, by ElectricMonk

User metadata
Rank Member
Rank
Member
leileilol wrote:
http://cdn.meme.am/instances/500x/12215635.jpg Go actually read the thread, and notice how there are no ISA, EISA VLB Voodoos, R […]
Show full quote
ElectricMonk wrote:

Except it's a 486 system, which had more than just PCI, depending on the motherboard. ISA, EISA, and VLB ring a bell?

12215635.jpg
Go actually read the thread, and notice how there are no ISA, EISA VLB Voodoos, Radeons and Geforces and stop assuming I was born yesterday with your lecturing of common knowledge and irrelevant hardware experiences to back up your 'expertise'.

I haven't been lecturing you. i asked a question, you responded with your typical smarmy attitude, so I pointed out that 486s didn't always come with PCI slots.

But have fun being condescending. That's all I ever see you post, anyways...

Reply 63 of 371, by rgart

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

With all the extensive testing of drivers for GeForce2, ATI Rage 128 VR, ATI Rage 128 Pro, and RIVA TNT, my MB-8433UUD will no longer function when a Cyrix 5x86 CPU is installed. Am5x86 and POD chips still work fine in this motherboard. I've seen something similar happen once before where an M919 would no longer work with a Am5x86 in Write-Back mode. I think the problem in this case is somehow related to the PCI implementation and how the Cyrix 5x86 handles it. MemTest passes OK.

I am not sure what to blaim the cause on. I did notice that when using the GeFroce2, that the BIOS pages redrew very slowly. The BIOS pages also loaded slow with the RIVA TNT, but not nearly as slow.

The symptoms are such that I receive a BSOD in WinNT4 upon boot or Win95c will hang on loading at a point just before the welcome screen. Upon switching to an Am5x86, everything loads fine. Switching to a different MB-8433UUD and keeping the Cyrix 5x86, everything works fine. I guess I'll be switching to another MB-8433UUD to continue testing ATI Rage 128 drivers for working NT4 drivers on a Cyrix 5x86.

The MB-8433 motherboard has somehow been damaged from testing so many modern video cards?

How many MB-8433 boards do you have laying around?

=My Cyrix 5x86 systems : 120MHz vs 133MHz=. =My 486DX2-66MHz=

Reply 64 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Yes, I think something related to all the testing and off/on cycles have upset one of the boards. I have several MB-8433UUD boards. The PCB revision numbers do not always agree with which PCB changes have been made to the board though. My inventory of this board is as follows,

MB-8433UUD v1 - 1 unit
(no version marking, so presumably version 1.0)
This board needs to have the PCI bus frequency reduced irregardless of the set frequency. For example, I need to set FSB-to-PCI multiplier at 2/3 or 1/2, even if using a 33 MHz FSB. As such, this board is defective.

MB-8433UUD v2 - 2 units
I have two of these. One has chipsets with 9622 datecodes, the other with 9534 datecodes. Even though both are version 2.0, the one with earlier datecodes clearly has some slight changes to the PCB. The earlier version 2.0 board has chipsets which are even older than those found on my version 1.0 board. I have not extensively tested the earlier v2.0 board because the chipsets are quite old in comparison. It is the later v2.0 board I have which no longer functions with Cyrix/IBM 5x86 CPUs. This board is now defective, however it is a good board to use with the Am5x86-160 or POD100 provided that you use 256 KB of double-banked cache. If you increase the cach to 512 KB, you can no longer use the fastest cache timings. In fact, you need to use the slowest cache timings.

MB-8433UUD v3 - 2 units
One version 3 board came stock with a socketed RTC, the other board had the RTC soldered on. One of these boards is being used as my IBM 5x86-133/2x system. It works well with my 1024K cache modification on a 66 MHz FSB. The other version 3 board I noted as possibly unreliable, but I am using it now for testing and it seems OK thus far.

MB-8433UUD v3.1 - 1 unit
I used this board for years without fail in a 24/7 server. Then one day the diskette controller (Super I/O chip) stopped functioning. I can put in an ISA diskette controller card, which worked with Win9x and DOS, however NT4 does not appreciate its presence and has issues with diskette access. I eventually need to cut off the old super i/o QFP chip and replace it. I have the replacement chips on hand, but no opportunity to replace it.

So I have 6 boards on hand, but only 2 seems fully functional. These boards were practically given away on eBay back in 2001.

Plan your life wisely, you'll be dead before you know it.

Reply 67 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Here are some charts of the results thus far. I still have one mystery card to try out.

With these three socket 3 systems, I am wondering which 3dfx, openGL, and Direct3D games would now be considered playable when using GPU acceleration. My target frame rate is 23 fps+

GLQuake_Cyrix5x86-133.png
Filename
GLQuake_Cyrix5x86-133.png
File size
8.89 KiB
Views
2549 views
File license
Fair use/fair dealing exception
GLQuake_AMD5x86-160.png
Filename
GLQuake_AMD5x86-160.png
File size
9.29 KiB
Views
2549 views
File license
Fair use/fair dealing exception
GLQuake_POD100.png
Filename
GLQuake_POD100.png
File size
9 KiB
Views
2549 views
File license
Fair use/fair dealing exception
Last edited by feipoa on 2014-10-16, 04:15. Edited 2 times in total.

Plan your life wisely, you'll be dead before you know it.

Reply 68 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I wouldn't consider Quake II playable at these frame rates, but here are the results anyway.

QuakeII_Cyrix5x86-133.png
Filename
QuakeII_Cyrix5x86-133.png
File size
8.59 KiB
Views
2549 views
File license
Fair use/fair dealing exception
QuakeII_AMD5x86-160.png
Filename
QuakeII_AMD5x86-160.png
File size
8.91 KiB
Views
2549 views
File license
Fair use/fair dealing exception
QuakeII_POD100.png
Filename
QuakeII_POD100.png
File size
8.67 KiB
Views
2549 views
File license
Fair use/fair dealing exception

Plan your life wisely, you'll be dead before you know it.

Reply 69 of 371, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie

Interesting to see the AMD-160 outperform the CX5x86-133 in Quake not that it really matters, is their a way to see the minimum framerates as the average doesn't represent reality, the slowdowns are huge in Quake on a 486.

Reply 70 of 371, by leileilol

User metadata
Rank l33t++
Rank
l33t++

GLQuake doesn't slow down as much since it doesn't update dynamic lights by default (unlike Quake2 which does). Dynamic light update = CPU eating + Slow bus transfer texture upload = 8048ouch.

gl_flashblend ultimately determines the 486's framerate fate 😀

apsosig.png
long live PCem

Reply 71 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Setting the gl_flashblend to 0 decreases the frame rate. It is set to 1 by default.
With sound enabled and gl_flashblend enabled at 800x600x16, the score is 24.0 fps
With sound enabled and gl_flashblend set to 0, the score is 21.0 fps.

Plan your life wisely, you'll be dead before you know it.

Reply 75 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++

There was only one or two GF2 driver versions which worked on the Am5x86, but the frame rates matched that of the Voodoo3 in GLQuake. My biggest complaint with the TNT and GF2 on the 486 was that the BIOS menu was redrawn slowly and text in DOS moved more slowly. Is there a way to correct this? PCPBench and 3Dbench ran fine though.

Plan your life wisely, you'll be dead before you know it.

Reply 76 of 371, by smeezekitty

User metadata
Rank Oldbie
Rank
Oldbie
feipoa wrote:

There was only one or two GF2 driver versions which worked on the Am5x86, but the frame rates matched that of the Voodoo3 in GLQuake. My biggest complaint with the TNT and GF2 on the 486 was that the BIOS menu was redrawn slowly and text in DOS moved more slowly. Is there a way to correct this? PCPBench and 3Dbench ran fine though.

I wish I had one of those cards or I would give it a test.

Can you determine if the cards really are slow in text mode or if the BIOS routines are just slow?

I have an EGA card that is dog slow in text mode in certain machines through the BIOS routines but reasonably fast through direct video mode writes.

Try landmark? http://dosbenchmark.wordpress.com/otherbenchmarks/
It is certainly not the best bench around but it will give you comparative video speed in text mode using direct memory write.

If direct memory writes are still slow, chances are that the card runs in graphics mode all the time and the text mode is emulated on the core and not very well.

Reply 77 of 371, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Unfortunately, I have had to put away my test bed for the time being. I'll try your suggestion when I have time to work on this again.

Plan your life wisely, you'll be dead before you know it.