VOGONS


The Ultimate 486 Benchmark Comparison

Topic actions

Reply 80 of 280, by sebaz_ri

User metadata
Rank Member
Rank
Member
swaaye wrote:

Socket 5/7 boards support at least up to 3.52v because that's what many older CPUs need. This includes chips like K5, some Pentiums, and the Winchips.

So you can overvolt a 3.3v chip a little.

I overvolted the Pentium P54C-133 chip to 3.8v and is still running well

Reply 81 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The U4BC PDF, images, and text have been updated to reflect

1) An IBM 5x86C-133 running with a 66 MHz front-side bus
2) An IBM 5x86C-133 running with a 66 MHz front-side bus with branch prediction enabled
3) Retro Games 100's amazing AMD X5-200

For more information on running an IBM 5x86C - 100HF at 133 MHz, please refer to this thread; it is easily doable,
So you want a Cyrix 5x86-133?

EDIT: A major source of error was just discovered in the spreadsheet's averaging divisor for ALU calculations -- all numbers displayed on the ALU bar charts are 20-30% inflated! I suppose this makes sense, an AMD DX4-120's ALU should'nt be equivalent to a Pentium 100's. I've also realised that I wasn't weighing FPU and ALU results equally for the Overall bar chart. And too much weight had been given to PassMark tests #75-82. Each benchmark program should be weighed equally.

I'll have to fix this up later. Will advise when corrected.

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

Reply 82 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

All documents have been corrected. The results are not nearly as impressive anymore, but are perhaps more realistic. Let me know if you find any errors. Thanks.

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

Reply 83 of 280, by dirkmirk

User metadata
Rank Oldbie
Rank
Oldbie

Tell you what, its a damn shame they never made a POD 100, I got a few overclocked results.

From a few results I obtained

Cyrix 120 POD 83 POD 100

Quake timedemo 14.4 17.7 FAIL
Quake2 timedemo 9.8 12.5 FAIL
3d benchmark 90.9 77 90.9
PCPbench 10.4 9.5 11.6
speedsys 66.3 61.6 73.9
MP3 PASS PASS PASS(ASSUMED)

The cyrix is definately faster loading programmes, the POD has it moments with quake engined games, I was suprised the POD could run MP3s with windows media player whereas the cyrix could'nt cut it, Im going to throw the POD into a VLB motherboard and keep the cyrix in a pci system, in a way the POD might be better suited to the pci system, in reality no socket 3 cpu will probably run 3D acclelerated games or dos 640x480 graphics acceptably so its a moot point.

Reply 84 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

A Cyrix 5x86-120 will run full-scale mp3s in WMP without skipping. Ensure that you at least have these registers set: LSSER = 0, FP_FPAST = 1, and MEM_BYP = 1.

Note, the U4BC runs PcpBench in mode VESA Modus 100 (640x400 8bpp LFB). Your Quake scores all seem a bit low.

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

Reply 85 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

A new breakdown of which CPU's will be in the Ultimate 586/686 Benchmark Comparison (Socket 7). For instances where a duplicate frequency is noted (Cyrix MII @ 300 MHz), I may run a second comparison using a Super7 motherboard for this particular CPU.

A new breakdown of which CPU's will be in the Ultimate 586/686 Benchmark Comparison (Socket 7). For instances where a duplicate frequency is noted (Cyrix MII @ 300 MHz), I may run a second comparison using a Super7 motherboard for this particular CPU.

This is the current potential line-up,

A new breakdown of which CPU's will be in the Ultimate 586/686 Benchmark Comparison (Socket 7). For instances where a duplicate frequency is noted (Cyrix MII @ 300 MHz), I may run a second comparison using a Super7 motherboard for this particular CPU.

This is the current potential line-up,

SOCKET 3

IBM 5x86c - 100 MHz (50 MHz x 2.0) - C
IBM 5x86c - 120 MHz (60 MHz x 2.0) - O
IBM 5x86c - 133 MHz (66 MHz x 2.0) - O

Intel DX4-WB - 133 MHz (66 MHz x 2.0) - O
Intel P24T - 83 MHz (33 MHz x 2.5) - C
Intel P24T - 100 MHz (40 MHz x 2.5) - O

AMD X5-ADZ - 133 MHz (33 MHz x 4.0) - C
AMD X5-ADZ - 160 MHz (40 MHz x 4.0) - O

SOCKET 7

Cyrix 6x86 - 75 MHz (75 MHz x 1.0) - U
Cyrix 6x86 - 83 MHz (83 MHz x 1.0) - U
Cyrix 6x86 - 100 MHz (50 MHz x 2.0) - U
Cyrix 6x86 - 110 MHz (55 MHz x 2.0) - U
Cyrix 6x86 - 120 MHz (60 MHz x 2.0) - C
Cyrix 6x86 - 133 MHz (66 MHz x 2.0) - C
Cyrix 6x86L - 150 MHz (75 MHz x 2.0) - C
Cyrix 6x86MX - 133 MHz (66 MHz x 2.0) - U
Cyrix 6x86MX - 150 MHz (60 MHz x 2.5) - U
Cyrix 6x86MX - 166 MHz (66 MHz x 2.5) - U
Cyrix 6x86MX - 200 MHz (66 MHz x 3.0) - C
Cyrix MII - 233 MHz (66 MHz x 3.5) - U
Cyrix MII - 250 MHz (83 MHz x 3.0) - C
Cyrix MII - 262 MHz (75 MHz x 3.5) - U
Cyrix MII - 292 MHz (83 MHz x 3.5) - C

Intel P54C - 75 MHz (50 MHz x 1.5) - C
Intel P54C - 90 MHz (60 MHz x 1.5) - C
Intel P54C - 100 MHz (66 MHz x 1.5) - C
Intel P54CQS - 120 MHz (60 MHz x 2.0) - C
Intel P54CS - 133 MHz (66 MHz x 2.0) - C
Intel P54CS - 150 MHz (60 MHz x 2.5) - U
Intel P54CS - 166 MHz (66 MHz x 2.5) - C
Intel P54CS - 200 MHz (66 MHz x 3.0) - C
Intel P55C-MMX - 133 MHz (66 MHz x 2.0) - U
Intel P55C-MMX - 166 MHz (66 MHz x 2.5) - U
Intel P55C-MMX - 200 MHz (66 MHz x 3.0) - U
Intel P55C-MMX - 233 MHz (66 MHz x 3.5) - C
Intel P55C-MMX - 262 MHz (75 MHz x 3.5) - O
Intel P55C-MMX - 300 MHz (75 MHz x 4.0) - O (extrapolated)
Intel Tillamook - 266 MHz (66 MHz x 4.0) - C
Intel Tillamook - 300 MHz (75 MHz x 4.0) - O
Intel Tillamook - 333 MHz (83 MHz x 4.0) - O

AMD K5 - 75 MHz (50 MHz x 1.5) - U
AMD K5 - 75 MHz (75 MHz x 1.0, if 1.0x supported) - U
AMD K5 - 83 MHz (83 MHz x 1.0, if 1.0x supported) - U
AMD K5 - 90 MHz (60 MHz x 1.5) - U
AMD K5 - 100 MHz (66 MHz x 1.5) - C
AMD K5 - 100 MHz (50 MHz x 2.0, if 2.0x supported) - C
AMD K5 - 117 MHz (66 MHz x 1.75, using 2.5x setting) - C
AMD K5 - 120 MHz (60 MHz x 2.0) - O
AMD K5 - 133 MHz (66 MHz x 2.0) - O
AMD K6 - 133 MHz (66 MHz x 2.0) - U
AMD K6 - 166 MHz (66 MHz x 2.5) - U
AMD K6 - 200 MHz (66 MHz x 3.0) - C
AMD K6 - 233 MHz (66 MHz x 3.5) - O
AMD K6 - 262 MHz (75 MHz x 3.5) - U
AMD K6 - 266 MHz (66 MHz x 4.0) - U
AMD K6 - 292 MHz (83 MHz x 3.5) - U
AMD K6 - 300 MHz (75 MHz x 4.0) - C

IDT Winchip - 200 MHz (66 MHz x 3.0) - C
IDT Winchip2 - 233 MHz (66 MHx x 3.5) - C
IDT Winchip2 - 240 MHz (60 MHz x 4.0) - O
IDT Winchip2 - 262 MHz (75 MHz x 3.5) - O
IDT Winchip2 - 266 MHz (66 MHz x 4.0) - O

RISE MP6 - 166 MHz (66 MHz x 2.5) - U
RISE MP6 - 166 MHz (83 MHz x 2.0) - U
RISE MP6 - 200 MHz (66 MHz x 3.0) - C
RISE MP6 - 233 MHz (66 MHz x 3.5) - O

SUPER SOCKET 7

Cyrix MII - 250 MHz (100 x 2.5) - C
Cyrix MII - 262 MHz (75 MHz x 3.5) - U
Cyrix MII - 266 MHz (66 MHz x 4.0) - U
Cyrix MII - 285 MHz (95 MHz x 3.0) - C
Cyrix MII - 300 MHz (100 MHz x 3.0) - O
Cyrix MII - 300 MHz (75 MHz x 4.0) - O
Cyrix MII - 333 MHz (83 MHz x 4.0) - O
Cyrix MII - 350 MHz (100 MHz x 3.5) - O
Cyrix MII - 375 MHz (83 MHz x 4.5) - O
Cyrix MII - 400 MHz (100 MHz x 4.0) - O

Intel P55C-MMX - 262 MHz (75 MHz x 3.5) - O

AMD K6 - 262 MHz (75 MHz x 3.5) - U
AMD K6 - 266 MHz (66 MHz x 4.0) - U
AMD K6 - 300 MHz (75 MHz x 4.0) - C
AMD K6-2 - 233 MHz (66 MHz x 3.5) - U
AMD K6-2 - 266 MHz (66 MHz x 4.0) - C
AMD K6-2 - 300 MHz (75 MHz x 4.0) - C
AMD K6-2 - 300 MHz (100 MHz x 3.0) - C
AMD K6-2 - 333 MHz (83 MHz x 4.0) - U
AMD K6-2 - 350 MHz (100 MHz x 3.5) - U
AMD K6-2 - 400 MHz (66 MHz x 6.0) - C
AMD K6-2 - 400 MHz (100 MHz x 4.0) - C
AMD K6-2 - 450 MHz (100 MHz x 4.5) - U
AMD K6-2 - 500 MHz (100 MHz x 5.0) - C
AMD K6-2 - 550 MHz (100 MHz x 5.5) - C
AMD K6-2 - 600 MHz (100 MHz x 6.0) - O
AMD K6-2+ - 400 MHz (66 MHz x 6.0) - U
AMD K6-2+ - 400 MHz (100 MHz x 4.0) - U
AMD K6-2+ - 450 MHz (100 MHz x 4.5) - C
AMD K6-2+ - 500 MHz (100 MHz x 5.0) - U
AMD K6-2+ - 550 MHz (100 MHz x 5.5) - C
AMD K6-2+ - 600 MHz (100 MHz x 6.0) - O
AMD K6-3+ - 400 MHz (66 MHz x 6.0) - U
AMD K6-3+ - 400 MHz (100 MHz x 4.0) - U
AMD K6-3+ - 450 MHz (100 MHz x 4.5) - C
AMD K6-3+ - 500 MHz (83 MHz x 6.0) - O
AMD K6-3+ - 500 MHz (100 MHz x 5.0) - O
AMD K6-3 - 500 MHz (100 MHz x 5.0, comparison with non + chips) - O
AMD K6-3+ - 550 MHz (100 MHz x 5.5) - O
AMD K6-3+ - 600 MHz (100 MHz x 6.0) - O

IDT Winchip2 - 233 MHz (66 MHx x 3.5) - C
IDT Winchip2 - 233 MHz (100 MHz x 2.33, using 5x setting) - C
IDT Winchip2 - 266 MHz (100 MHz x 2.66) - O

RISE MP6 - 200 MHz (66 MHz x 3.0) - C
RISE MP6 - 200 MHz (100 MHz x 2.0) - C

SLOT 1

Pentium II - 266 MHz (66 MHz x 4.0) - U
Pentium II - 300 MHz (75 MHz x 4.0) - U
Pentium II - 350 MHz (100 MHz x 3.5, if PII-400 mult. not underclock locked) - U
Pentium II - 400 MHz (100 MHz x 4.0) - C
Pentium III - 500 MHz (83 MHz x 6.0) - U
Pentium III - 500 MHz (100 MHz x 5.0, if 550 mult. not underclock locked) - U
Pentium III - 550 MHz (100 MHz x 5.5) - C
Pentium III - 600 MHz (100 MHz x 6.0) - C
Celeron - 400 MHz (66 MHz x 6.0) - C

6x86L = Low Power (low heat consumption) model + split voltage
6x86MX = MMX instructions supported
MII = Same as 6x86MX but select models support 75-100 MHZ FSB
U = underclocked setting
O = overclocked setting
C = correctly marked CPU in possession

Cyrix 6x86 and 6x86L support only integer 1x, 2x, 3x, and 4x multipliers.
Cyrix 6x86MX and MII support 1x, 2x, 2,5x, 3x, 3.5x, 4x, 4.5x, 5x
WinChip2 - 2A supports multipliers of 2.33x (using 5x setting), 2.5x, 3.0x, 3.5x, 4.0x, 4.5x

From reports else where, it seems unlikely to get a 266 MHz Tillimock configured properly w/cache, so I will see how far I can take my 233.

Last edited by feipoa on 2012-03-25, 10:46. Edited 16 times in total.

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

Reply 86 of 280, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Aren't the fatty market letters supposed to be 5's instead of 4's btw?
And will you be using the same benchmarks as you used for your 486 comparison?

P55C-MMX - 133 MHz (66 MHz x 2.0) - U P55C-MMX - 166 MHz (66 MHz x 2.5) - U P54C-MMX - 200 MHz (66 MHz x 2.5) - U P54C-MMX […]
Show full quote

P55C-MMX - 133 MHz (66 MHz x 2.0) - U
P55C-MMX - 166 MHz (66 MHz x 2.5) - U
P54C-MMX - 200 MHz (66 MHz x 2.5) - U
P54C-MMX - 233 MHz (66 MHz x 3.5) - C
P54C-MMX - 250 MHz (83 MHz x 3.0) - O
P54C-MMX - 266 MHz (66 MHz x 4.0) - C
P54C-MMX - 300 MHz (75 MHz x 4.0) - O
P54C-MMX - 300 MHz (100 MHz x 3.0) - O
P54C-MMX - 333 MHz (83 MHz x 4.0) - O
P54C-MMX - 350 MHz (100 MHz x 3.5) - O
P54C-MMX - 400 MHz (100 MHz x 4.0) - O

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 87 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Yes. Copy/paste must have caught me off guard! It is now corercted. Thank you for pointing that out.

And will you be using the same benchmarks as you used for your 486 comparison?

You have asked the same question I asked myself. I was thinking that a lot of the old benchmark programs used for the 486 comparison probably do not make use of MMX instructions, so the jump from non-MMX to MMX might not be so significant. I am open to suggestions for which benchmark programs to add or replace.

One of the reasons I might not use a Super7 board for this comparison is because, via ALi, it has fast chipset built-in cache (quite fast and very stable), and for VIA, it can have up to 2 MB of cache. Also, for ALi and VIA chipsets, I beleive some of the slower Cyrix 6x86 and Pentiums might not be supported. The problem with the large amount of fast cache is that it might incorrectly compare with the 512 KB of SRAM in the 5x86 motherboard. In general, the Super7 boards perform much faster than the TX boards.

If I use the 430TX chipset, I can plop in 64 MB of RAM w/512 KB cache and at least have more comparable results to the 5x86's I'm planning to include in the comparison. The issue with the TX board is that CPU's above 300 MHz don't get much respect when running at 75 and 83 MHz FSBs. I'd hate to botch things using 2 different boards. I'll probe my TX board to see if there is a hidden 100 MHz setting. While the Intel TX might not "officially" support 100 MHz, the motherboard's PLL might have it still.

The PII-400 is in there just for curiousity. I haven't yet decided which CPU to normalise the data to. But I am definately going to run the Cyrix 5x86-133 at 2 x 66 MHz to compare with the Cyrix 6x86-133. I'm really curious to see how much technology was removed from the 5x86 model.

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

Reply 88 of 280, by luckybob

User metadata
Rank l33t
Rank
l33t

YOU MUST give me the list of benchmarks you use. I want to add pentium pro and P-Pro overdrive data to the list. I had this idea back in august, but I dont have the processors to do it right. Socket 7 & slot 1 roundup

If its not too much to add, I have a DUAL socket 7 and 8 board that I want to add the data in. The socket 7 board has either 0,256 or 512kb cache. and I have 256,1m and overdrive p-pros'.

Another benchmark I would look into is one that will benchmark 16bit vs 32 bit software. I'll look for one if you dont have one in mind.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 89 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

On page 3 of this topic, I brought up doing the socket 7 benchmark comparison and listed out several CPU options. It was probably too soon after the 486 benchmark comparison to be of significant interest. So I am bringing it up again now since I have acquired a few more AMD processors and have since added them to the list.

I do not have a socket 8 motherboard, but would consider adding your results to the list if the settings and hardware are consistent. We'd need the same graphics card. I am likely going to use the Matrox Millenium G200 again since that is the fastest graphics card I can use in the socket 3 UMC motherboard (For the 5x86 CPUs). I'm probably going to limit that RAM to 64 MB and cache at 512 KB (if using the 430TX), or more if using a Super7.

The list of benchmarks used can be found in the Ultimate 486 Benchmark Comparison's PDF. Refer the tables. I haven't yet looked into additional software benchmark programs. I was waiting for suggestions :)

Benchmarks from about 1999 should be good, although old benchmarks are a pain to look for.

16-bit vs. 32-bit is a good idea. 32-bit is probably more useful. I figure I'll stick with Win98SE again.

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

Reply 90 of 280, by luckybob

User metadata
Rank l33t
Rank
l33t

I have a pci G200 as well. The socket 8 board is an intel 430HX. I have either 512mb or 256mb of edo simms for it. 60ns if memory serves, but i'd set it to what you have. The socket 7 board is also a 430hx board and i will use the same ram. The dual-7 board I have is a Tyan. They made it correctly and have the extra tag ram for full 1gb cache. I can turn this off if need be. (yay for server class hardware!)

I suggest NT 4 for the 32 bit tests as it is FULLY 32 bit.

For your slot 1 board, i'd make sure it was a 440fx chipset. they use 72 pin simms and dont have agp. so it will keep things the same. I have a dual-1 if need be.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 91 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The slot 1 is tricky. I only have a dual Slot 1 with the 440BX chipset (Dell workstation motherboard), a VIA board with the VT82C694X chipset, and another dual Slot 1 (Tyan S1832D) with the 440BX chipset. I know I had an old crappy Dell motherboard around here at one point which housed a Celeron 366 or 400, but it has completely disappeared.

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

Reply 92 of 280, by luckybob

User metadata
Rank l33t
Rank
l33t
feipoa wrote:

The slot 1 is tricky. I only have a dual Slot 1 with the 440BX chipset (Dell workstation motherboard), a VIA board with the VT82C694X chipset, and another dual Slot 1 (Tyan S1832D) with the 440BX chipset. I know I had an old crappy Dell motherboard around here at one point which housed a Celeron 366 or 400, but it has completely disappeared.

I have an asus p2b-ds which is 440BX and a tyan which is 440FX. And a tyan that runs a 440GX. So if I can get a hold of a few celeron chips I should be able to help out in the slot department.

It is a mistake to think you can solve any major problems just with potatoes.

Reply 93 of 280, by numeriK

User metadata
Rank Newbie
Rank
Newbie

I've been tinkering around this weekend... I figured I'd post my results of a few notable tests, please use in your comparison as you see fit.

==========================================

M919 v1.5 @ ~66MHz FSB
AMD 5x86 @ 205MHz (5v, 3x66MHz, L1 Read/Write @ 1/1WS, no L2)
128MB FPM @ 2-1-2

Speedtest 4.78 Results
Score: 69.75
Memory Bandwidth: 154.86
L1 Cache: 118.38
Memory Throughput: 60.33

3Dbench: N/A (freeze during test)
Win98SE: N/A (freeze on load)

==========================================

M919 v1.5 @ ~50MHz FSB
AMD 5x86 @ 200MHz (5v, 4x50MHz, L1 Read/Write @ 0/0WS, no L2)
128MB FPM @ 2-1-2

Speedtest 4.78 results
Score: 71.79
Memory Bandwidth: 151.10
L1 Cache: 125.04
Memory Throughput: 63.64

3Dbench: 90.9
Win98SE: N/A (freeze on load)

==========================================

M919 v1.5 @ ~60MHz FSB
AMD 5x86 @ 187MHz (5v, 3x60MHz, L1 Read/Write @ 1/0WS, no L2)
128MB FPM @ 2-1-2

Speedtest 4.78 results
Score: 62.74
Memory Bandwidth: 139.20
L1 Cache: 106.49
Memory Throughput: 54.22

3Dbench: 83.3
Win98SE: successful load
3DMark99Max: 15 3DMarks, 5 CPUMarks, 1.3MegaTexels/sec Fillrate, 1.5MegaTexels/sec Fillrate w/ Multi-Texturing
Test settings: Default benchmark, 800x600x16-bit

DOOM timedemo
demo1: 47.58fps
demo2: 48.42fps
demo3: 45.93fps

QUAKE timedemo
demo1: 16.0fps
demo2: 16.7fps
demo3: 15.1fps

==========================================

I haven't even got around to overclocking the TNT card within Windows yet, so the 3DMark results can likely get higher.

If anyone has any performance tips for these UMC boards please pass along! Thanks!

8433UUD v2 | AMD 5x86 @ 180MHz (60MHz x 3, 30MHz PCI) | 64MB EDO | TNT 16MB PCI | SB AWE64 ISA | Win98SE

Reply 94 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The start page of this thread has been updated such that the charts for ALU, FPU, and Overall performance are displayed directly in the web browser.

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

Reply 95 of 280, by vetz

User metadata
Rank l33t
Rank
l33t

How were Doom and Quake tested? With or without sound? This is not stated anywhere.

From my testing with Doom, going directly into DOS skipping loading HIMEM, EMM386, etc, etc gives 1-2 extra FPS. Having sound loaded takes another 1-2 FPS.

3D Accelerated Games List (Proprietary APIs - No 3DFX/Direct3D)
3D Acceleration Comparison Episodes

Reply 96 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I'm pretty sure the sound card driver was not initiated. I installed Windows 98SE, but I booted directly into DOS without going into Windows. I think if you first boot into Windows, then "restart in MS-DOS mode", then the DOS sound card drivers are loaded, however, I did not do this. So for all DOS tests, sound was not enabled.

I don't recall if HIMEM gets loaded with the default Windows 98SE installation when booting directly into DOS. I suspect I had HIMEM loaded though because I like to use it as an initial means to test the extended RAM. Unfortunately, my gear is all packed away now.

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

Reply 97 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

A new PDF report and the Overall-Score PNG image were replaced.

This change only affects the IBM 5x86c-133 (2x66) for Quake 1 and Overall scores. I accidently had sound enabled for this test and have now disabled it.

EDIT: I reverted the files back after realising they were right the first time. The increase in Quake 1 scores noted recently was because I fogot the Rage 128 was still in the computer instead of the desired Matrox G200.

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

Reply 98 of 280, by elianda

User metadata
Rank l33t
Rank
l33t

I would like to add some more benches regarding ISA vs Local Bus VGA card speed, as the 486 time introduced VLB/PCI as local bus for the masses.
I used a ECS UM8810PAIO V3.2 with 256 kB WT cache for benching ISA and PCI cards with various 5V 486 CPUs.
The benchmark I used is Doom as it is more real world as 3dbench f.e. (No Sound, -timedemo demo3, just himem)
So here are 20 data points:

localbus.png

The small numbers are the PCI or ISA bus speed.

So first of all fast ISA cards as the TSENG ET4000 can keep up with a PCI card until 486DX-33. For the 486DX-40 it requires already a massive ISA bus overclock, like 20 MHz as shown. With a 486DX2-66 the TSENG ET4000 performs like the PCI card with a 486DX-40.
With an average ISA VGA card the performance is already about 2/3 of a PCI card with a 486DX-25 and doesn't rise anymore after a 486DX-33.
So even if you own a fast ISA graphics card starting with a 486DX-40 a local bus card is much faster.

A 386DX-40 performs comparable to a 486DX-20.

Retronn.de - Vintage Hardware Gallery, Drivers, Guides, Videos. Now with file search
Youtube Channel
FTP Server - Driver Archive and more
DVI2PCIe alignment and 2D image quality measurement tool

Reply 99 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Nice work. That is quite the deviation in slopes. Its a good thing users with a VL or PCI bus weren't using a 486-DX1 - what a waste that would have been. Seems like most VLB systems were fitted with a DX2 or a DX4. Most PCI 486's were fitted with a DX4 or DX5.

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