kixs wrote on 2020-01-25, 07:38:

Performance VLB vs PCI with the same chipset and the same motherboard (486 VIP with 5x86-133) is the same. PCI isn't faster. Tested with ATI MACH64, S3 Trio64, Tseng ET4000/w32p.

Comparing PCI vs VLB on a VIP style motherboard is tricky as you are essentially penalising PCI - see Performance comparison of 486 motherboards with VLB-only, PCI-only, and PCI+VLB

However, you are right the difference is nothing to write home about unless you care about scores

For DOS performance pretty much every VLB/PCI card is good enough.

No, not really. Some VLB cards are quite bad for high-end 486/Pentium Overdrive.

My current VLB roundup for reference:

download/file.php?mode=view&id=75289

mpe wrote on 2020-01-25, 08:48:

kixs wrote on 2020-01-25, 07:38:

Performance VLB vs PCI with the same chipset and the same motherboard (486 VIP with 5x86-133) is the same. PCI isn't faster. Tested with ATI MACH64, S3 Trio64, Tseng ET4000/w32p.

Comparing PCI vs VLB on a VIP style motherboard is tricky as you are essentially penalising PCI - see Performance comparison of 486 motherboards with VLB-only, PCI-only, and PCI+VLB

However, you are right the difference is nothing to write home about unless you care about scores

I'm using VIP based on UMC chipset and I haven't really noticed any degraded performance on VLB or PCI. Maybe I should look into it more precisely.

The Serpent Rider wrote on 2020-01-25, 09:38:

For DOS performance pretty much every VLB/PCI card is good enough.

No, not really. Some VLB cards are quite bad for high-end 486/Pentium Overdrive.

VGA cards for general use... like S3 805, CL5428, Tridents... are pretty much the same in DOS. Only the last gen S3 pull ahead. The cards specifically made for Windows usage, like the ones based on Weitek Power9000/9100 chipset are really slow in DOS (DOS part uses OTI087x). Both VGA and PCI versions are the same.

. like S3 805, CL5428, Tridents..

Apparently Trident 9200 is quite horrible for everything.

are pretty much the same in DOS.

No, they aren't, but without powerful CPU it really doesn't matter.

So you're saying Trident 9200 is slow in DOS. How much slower then S3 805? As I only have Trident 9440 cards.

My tests were done with AMD 5x86-133 and PODP5V83 running at 100Mhz.

Didn't saw tests, only rumors about how horrible that card was. But if we take into consideration that 9400CXi is slower than common Cirrus Logic and S3 805 cards (with just 486DX2), 9200 would be even worse.

I've tested around 50 cards, mostly ISA and VLB and some PCI for comparision. Total 70 benchmarks as I've also done the combos with installed memory (1MB vs 2MB vs 4MB where possible). Of course only difference with installed memory is in Windows and only with chipsets that support memory interleaving.

Base system was PcChips M919 V3.4 VIP, 256KB cache, AMD 133 WB, 8MB and CF512MB.

I'll publish the benchmarks when I complete them... no idea when, but still many cards left to test.

I've done the usual:
3Dbench 1.0c, PCPbench, Doom, Quake, Sysinfo VESA speed, DrHard VGA speed, Windows Winbench 3.11 in 640x480/8-bit and 16-bit. Wanted to do more, but it takes a lot of time.

This is a current list of cards tested:

S3 Trio64V+ 2MB PCI
Cardex Tseng ET4000/w32p 1MB VLB
ATI Mach64 2MB VRAM VLB NonUP
ATI Mach64 2MB VRAM VLB
ATI Mach64 2+2MB VRAM ISA
ATI Mach64 2MB DRAM ISA
ATI Mach64 2MB DRAM VLB
STB Nitro64 2MB ISA (CL-5434) V1.01 50ns
STB Nitro64 1MB ISA (CL-5434) V1.01 50ns
Diamond Speedstar64 2MB ISA (CL-5434) V2.02 50ns
Diamond Speedstar64 2MB ISA (CL-5434) V1.02 70ns
Orhid Kelvin64 CL-5434 2MB VLB 70ns
Orhid Kelvin64 CL-5434 1MB VLB 70ns
Diamond Stealth64 Video S3 968 2MB VRAM VLB 70ns V2.04
Miro 12SD S3 Trio32 2MB VLB 70ns
Jaton S3 805-P 2MB VLB 70ns
Diamond Stealth24 S3 805 1MB VLB 45ns V1.11
Spea V7-Mercury S3 928 2MB VRAM VLB 70ns
WD Bahamas64 S3 Trio64 2MB VLB
WD Bahamas64 S3 Trio64 1MB VLB
Diamond Stealth64 Video S3 868 2MB VLB DRAM 70ns V1.02
Diamond Stealth64 Video S3 868 1MB VLB DRAM 70ns V1.02
Diamond Stealth32 Tseng ET4000/W32P VLB 2MB 60ns V2.15
Diamond Stealth32 Tseng ET4000/W32P VLB 1MB 60ns V2.15
Tseng ET4000/W32P 2MB PCI
Tseng ET4000/W32P 1MB PCI
ATI Mach32 2MB DRAM VLB EXM229
ATI Mach32 2MB DRAM VLB EXM234
ARK 1000VL 2MB VLB 70ns 1WS
Cirrus Logic 5429 1MB VLB 70ns
Trident TGUI9440AGi 1MB VLB 70ns
Diamond Stealth64 Video VRAM S3 968 2+2MB PCI V1.01
ATI Mach32 2MB ISA DRAM EXM688DM1
ATI Mach32 2MB ISA VRAM EXM688VM1
Diamond Stealth64 S3 864 2MB VLB 60s 1.02H
Diamond Stealth64 S3 864 1MB VLB 60s 1.02H
Elsa Winner1000-ISA S3 805i 2MB ISA 60ns
Elsa Winner1000-ISA S3 805i 1MB ISA 60ns
Spea V7-Mirage S3 805 2MB ISA 60ns
Orchid Fahrenheit VA S3 801 2MB ISA 60ns
Diamond Stealth24 S3 805 1MB ISA 45ns V1.24
Tseng ET4000/W32i 1MB ISA
ATI Mach32 2MB VRAM VLB 70ns EXM195A
Hercules Dynamite VL Pro Tseng ET4000/W32i 2MB VLB 60ns
Cardex Tseng ET4000/w32p 2MB VLB 60ns
S3 964 2MB VRAM VLB 70ns
S3 864 2MB DRAM VLB 70ns
Matrox MGA Ultima 2MB VRAM VLB 70ns OTI-066 for DOS
Diamond Viper 2MB VRAM VLB 70ns OTI-087 for DOS
Diamond Viper 2MB VRAM PCI 70ns OTI-087 for DOS
Cirrus Logic 5428 1MB VLB 70ns
Cirrus Logic 5428 1MB VLB 80ns
S3 864 2MB DRAM VLB 60ns
Chips F64300 1MB VLB 70ns
Chips F64300 2MB VLB 70ns
Chips F64300 2MB VLB 70ns + cache
Trident TGUI9440 1MB VLB 70ns
Cardex Tseng ET4000AX 1MB VLB 70ns
Soyo S3 805 1MB VLB 70ns
Cirrus Logic 5430 1MB VLB 70ns
Cirrus Logic 5428 1MB VLB 70ns 0WS
Spea Mercury-P64 S3 964 2MB VRAM VLB 70ns
Digital S3 864 4MB VLB
ExpertColor S3 805 2MB VLB 70ns
Miro Crystal 20SV S3 964 2MB VRAM VLB 70ns
Miro Crystal 20SD S3 864 2MB DRAM VLB 70ns
Genoa Phantom64 S3 864 2MB DRAM VLB 70ns
ATI Mach64 2MB VRAM PCI 60ns EXM255
Orchid Fahrenheit 1280 Plus S3 801 1MB ISA 60ns
Tseng ET6000 4MB PCI

I'd love to see the results of those benchmarks. Any chance of uploading the raw results to a Google docs sheet and adding to it as you go?

@kixs
If you decide to do it you might as well do it in full master-style and add the T6100 😉

Every reason to believe the ARK Logic cards will only come 2nd to TSENG cards and tranding places with them in low res for top spot, based on numerous graphs I've seen before.