VOGONS


The Ultimate 486 Benchmark Comparison

Topic actions

Reply 160 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

I do not know of any 486 motherboard which was designed to work with a 60-66 MHz FSB. My Shuttle HOT-433 boards also are unable to run reliably with a 66 MHz FSB. Initial tests indicated that 60 MHz seemed OK.

Of the two examples with reliable 66 MHz FSB operation, the board designs did not implement discrete component buffers. I am not sure if this is related or not as both of the working samples had small PCB footprints as well.

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

Reply 161 of 280, by skv400

User metadata
Rank Newbie
Rank
Newbie
feipoa wrote:

I do not know of any 486 motherboard which was designed to work with a 60-66 MHz FSB. My Shuttle HOT-433 boards also are unable to run reliably with a 66 MHz FSB. Initial tests indicated that 60 MHz seemed OK.

Of the two examples with reliable 66 MHz FSB operation, the board designs did not implement discrete component buffers. I am not sure if this is related or not as both of the working samples had small PCB footprints as well.

Thank you.

My hot-433 is not special. Common motherboard.

I dispointted can't use fsb 66 on my motherboard.

Youtube Channel : https://www.youtube.com/channel/UCeaVPa3VMe9nUGDtFtVzvDQ

Reply 164 of 280, by magicmanred

User metadata
Rank Newbie
Rank
Newbie

That's fantastic that you used one processor to emulate multiple cpu's at different mhzs for these benchmarks.
I figured I post/ask this here since you seem quite experienced in this area...

I will be building an Asus PVI-486SP3 soon and have two CPU's at my disposal. Just waiting for parts to begin.
It's coming with a genuine Intel 486DX4 100mhz chip, and I opted to try/buy a spare AMD 5x86 P75 (486DX5 133) Model AM486DX5-133W16BGC (Lower volt than ADW variant).

It's been a while since I've tinkered with Socket 3's.
If my memory serves me correctly, these toggle between 2 and 3 multipliers. Where some AMD chips see 2x as a 4x.
And different boards have different jumpers for FSB's usually from somewhere in the range of 25mhz up to 50mhz.

According to my motherboards manaul page 3...
https://www.dropbox.com/s/n29olmnecbpondp/pvi … 486sp3.pdf?dl=1
...I will need to set JP7 to pin 1&2 (3.45v). The only other option is 3.6v. Is the 3.45v setting safe on the AMD 5x86 133 BGC chip that runs at 3.3v and the i486DX4 100mhz that runs at 3v?

I see the CPU jumper settings in the manual for the intel 486DX4 100mhz &EW chip on page 6 (3v chip, so I suppose the board knows to supply the correct volt) but I'm confused as to what setting to use for the AMD 5x86 133 chip. I'm assuming it's the same as the 486DX4 100mhz. On page 7 there is an identical configuration for the AM486DX4=SV8B setting.

The board supports 25mhz, 33mhz & 40mhz clock/fsb (no 50 😒)
With the board set to 33mhz, the i486DX4 100mhz should read 100 (3x33mhz). I suppose if I wanted to try 120mhz, I'd set it to 40mhz (3x40mhz) and set the VLB wait state (page 10) to "1 Wait" (Nice that it has this option as most boards have issues on VLB over 33mhz FSB. I'm assuming there will be a slight performance hit on the VLB this way though).

Now if the above is true, then how does the AMD 5x86 reach 133mhz? Or does it just know automagically to use a 2x multi as the 4x as the AMD's do even though it's set to 3 (assuming I leave the jumpers alone). And if THAT is true, then I would assume to attempt 160mhz I'd just change the FSB to 40mhz as I mentioned above? (4x40mhz).

Any help is appreciated as I don't want to start the build off with smoke & fireworks (although that may be a glorious to watch).
-M

edit:
Upon further research, I've realized that I have rev. 1.8 of the board. Rev. 1.22 has a 50mhz option for FSB.
Does anyone know if a 50mhz FSB is achievable on rev. 1.8 by copying the jumper config settings from the 1.22? Just curious in the event that I want to tinker with different FSB's once I figure everything out.

Reply 165 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

According to the Intel DX4 spec sheets, that CPU's nominal operating voltage is 3.3V. For AMD DX5, it is 3.45 V. The only AMD DX5 I see listed at 3.3 V is the AMD-X5-133SFZ.

Set the Intel DX4-100 to 3.3 V. Set the AMD DX5 to 3.45 V. However, I have never had issues running the Intel DX4-100 at 3.45 V.

Sometimes, you can get away with running the VLB at 40 MHz without adding 1 wait state. You just need to try it.

The AMD runs at 33 MHz x 4 = 133 MHz. This CPU takes the 2x multiplier request (jumper on the MB) and multiplies by 4x instead of 2x. All multpilication is done inside the CPU. If you set the MB jumper to 3x, the AMD DX5 will run at 33x3 = 100 MHz. There should be one jumper on the MB which sets between 2x and 3x.

Yes, set the FSB jumper to 40 MHz to run the CPU at 160 MHz. I have a VLB system running at 160 MHz.

I suspect the board layout is similar. You may need to solder on a jumper header for 50 MHz. I cannot say for sure though because I've not looked at this issue on this particular MB, but on other MB's, there is usually a wire which needs to be removed, or a jumper header which needs to be soldered on to realise the missing FSB settings.

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

Reply 166 of 280, by Elia1995

User metadata
Rank Oldbie
Rank
Oldbie

I have a Texas Instruments 486DX4-T100 which gets detected as a DX2 🤣

Currently assembled vintage computers I own: 11

Most important ones:
A "modded" Olivetti M4 434 S (currently broken).
An Epson El Plus 386DX running MS-DOS 6.22 (currently broken).
Celeron Coppermine 1.10GHz on an M754LMRTP motherboard

Reply 167 of 280, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
magicmanred wrote:

I will be building an Asus PVI-486SP3 soon and have two CPU's at my disposal. Just waiting for parts to begin.

The board supports 25mhz, 33mhz & 40mhz clock/fsb (no 50 😒)
With the board set to 33mhz, the i486DX4 100mhz should read 100 (3x33mhz).

Iirc, ASUS PVI-486SP3 had undocumented jumper settings and (dunno which revisions) some supposedly were able to clock the FSB down to something like 8MHz 😁

It was documented on this page http://winhistory.de/more/386/xpmini.htm.en
Just do a search for "486SP3" and it will come up.

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

Reply 168 of 280, by melbar

User metadata
Rank Oldbie
Rank
Oldbie

Thanks, feipoa, for the 486 benchmarks from 2011 and the 686 benchmarks from 2013.

I really appreciate all the work and it's quite useful for all vogons people.

According the fact, i will get also a 486DX2 system, actually i had no purpose to get a system before the pentium era...

So, despite all these data from feipoa, i've calculated the percentage differences (with data from feipoa's Overall chart) of some CPUs (not all of feipoa's cpu's but these i am interesting ...).

Attachments

  • 486_Pentium.jpg
    Filename
    486_Pentium.jpg
    File size
    203.42 KiB
    Views
    9035 views
    File license
    Fair use/fair dealing exception

#1 K6-2/500, #2 Athlon1200, #3 Celeron1000A, #4 A64-3700, #5 P4HT-3200, #6 P4-2800, #7 Am486DX2-66

Reply 169 of 280, by clueless1

User metadata
Rank l33t
Rank
l33t

@melbar -- thanks for sharing your analysis. I spent a lot of time in feipoa's PDF files, I find them invaluable!

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 170 of 280, by elianda

User metadata
Rank l33t
Rank
l33t

I find it quite difficult to understand the numbers in the chart melbar.
If you take the basic value of the systems performance shouldn't it be exactly 100% in the white cells in the yellow block?
Also what do the red values mean? (and what are the percentages in the eights column?)

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 171 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Version 2 of the Ultimate 486 Benchmarks will be a lot more exhaustive like the 686 benchmarks. I'm still in preparation for it, but haven't started it yet. I'll fix some minor issues with version 1 and add 386 upgrade to 486 CPUs, add up to Pentium 100 and 386 chips in general. Socket 3's will be tested on the MB-8433UUD instead of M919. I'll also try to add every oddball CPU configuration for 486. I know AMD had a load of them.

Elianda, it looks like he's comparing the relative "overall performance (from U4BC)" of select CPUs. Take the AMD X5 as an example, the i486DX2 is 47% slower, the P75 is 14% faster, the K6-2-500 is 734% faster. Another example, take the P166MMX, the X5-133 is 59% slower; the K6-2-500 is 243% faster.

I'm not sure where he obtained the values for the K6-2, MMX233, MMX166, P166, P150, P100, P66, P75, and P60. He would have had to take all the individual Ultimate 686 Benchmark data values and normalise them to the P100 (divide their values by the values of the P100). This is required because the 486 data is normalised to the P100 and the 686 data is normalised to the P233MMX. I'm assuming he has done this? Alternately, he has rerun all the 486 benchmarks himself with the CPUs mentioned above ( K6-2, MMX233, MMX166, P166, P150, P100, P66, P75, and P60).

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

Reply 172 of 280, by luckybob

User metadata
Rank l33t
Rank
l33t

so you will need to rename this thread to the penultimate 486 benchmark comparison then.

I've gotten rid of most of my non-ibm 486 hardware, but if any data from 486 class ps/2 systems would be useful, say something.

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

Reply 173 of 280, by melbar

User metadata
Rank Oldbie
Rank
Oldbie

You're right feipoa. I've taken data from the 686 benchmark (which is normalized @ PMMX233, file: 686_Overall_and_Quake1.png) and from the 486 benchmark (which is normalized @ P100, file: Overall_486.png).

The problem is, to take a correlation from the pentiums (even the higher ones, from the 686 Benchmark) to the low-end 486, since the 686 benchmark start already with a high-end 486 AMD X5-133.

So, i've taken your data as master, blue box from 686 benchmark and green box from 486 benchmark (see figure below).

Be aware of these 486 values (written red italic in the yellow box). They are simple calculated with cross-multiplication with the lowest value from the 686 benchmark (the X5-133).
Example: X5-133 [686-value: 31.6 ; 486-value: 71.4]

i486 DX2 [486-value: 38.0]

38.0 / 71.4 x 31.6 = 16.8

So, it's only a relationship from the differences (in percent) from one benchmark to another.
Then i've used for the my personal chart (the 5 columns at the right side only values from 686 benchmark (inlusive the NOT 100% correct values from the low-end 486)).

I know that it's not 100% correct anymore, cause you can see:
Example:
Now i'll get the i486-DX2 (compared to the P100): -64.9%
The correct value with data from the 486 benchmark:
(38.0 - 100.0) / 100.0 = -62.0%

So i have an error of 2.9% regarding this example.

Attachments

  • 486_Pentium_2.jpg
    Filename
    486_Pentium_2.jpg
    File size
    199.96 KiB
    Views
    9003 views
    File license
    Fair use/fair dealing exception

#1 K6-2/500, #2 Athlon1200, #3 Celeron1000A, #4 A64-3700, #5 P4HT-3200, #6 P4-2800, #7 Am486DX2-66

Reply 174 of 280, by feipoa

User metadata
Rank l33t++
Rank
l33t++

Did you check a few other states for error bars as well?

luckybob: still the ultimate 486 benchmark comparison, but there will be more non-socket 3 data for cross-comparison purposes.

This melbar got me thinking, though... I can't imagine there being any interest in comparing a P166MMX to an AMD X5-160. They are in different ballparks entirely. So what is a realistic good stopping ground for comparisons with the 486's? Normally, I was thinking P75, P90, and P100 were all of interest for comparison. But Melbar is comparing chips up to an AMD K6-2, which is pretty far reaching. Is a P133, AMD K5-133, and Cyrix 6x86-133 a better stopping ground for cross comparison? Or is a P100 sufficient?

Luckybob: What PS/2 system ya got? And do you have a P60 or P66 system setup? I almost got a decent board from cpuworld for cheap, but I always seem to chime in too late.

[by the way, this project will be years later - still lots of 386-to-486 work to do.]

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

Reply 175 of 280, by luckybob

User metadata
Rank l33t
Rank
l33t

I wish i had a p60 board. I have a 486-25, 486-66, and p90. If i had $150 burning a hole in my pocket i could get the one on ebay...

I also have a 386 system with a 486 socket upgrade for giggles.

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

Reply 176 of 280, by melbar

User metadata
Rank Oldbie
Rank
Oldbie

I am interesting to compare the 486, the slow ones and fast ones against the Pentium's. I agree with you, that normally it does make only sense, to compare up to the P100.

Since i have built a 'time-machine' with my K6-2 500 (after watching Phil's famous video), it happens that i've collected also some pentium hardware.
We have all seen, that with these cache trick's, it is nearly impossible to emulate an 486DX2-66 or -80 or even an DX4-100 (see the speadsheet of clueless1).
Some VIA Cyrix cpu's can be set next to this speed, but i'm excluding these 'rare' variants.

That's the reason why i have include the K6-2 500, cause i have some marker for my K6-2 machine (when usine the cache-tricks): 386-SX33, 486SLC/DLC-40, i486-SX25, i486-DX33

It is nice that in your 486 comparison, all the different 486 are inside, but when you want to compare the low-end, medium and high-end pentiums (which are quite easy to achieve and cheap), then you might understand my approach.

From the comparison, i've already shown, i've calculated the error comparing the 686 chart values vs. the 486 chart values.
I've hide all the data from faster than P100.
As you can see, there is no error (only one decimal after comma shown) for the slower cpu's. But when you compare up to your P100, then you can see the difference...

Attachments

#1 K6-2/500, #2 Athlon1200, #3 Celeron1000A, #4 A64-3700, #5 P4HT-3200, #6 P4-2800, #7 Am486DX2-66

Reply 179 of 280, by gloomkeep

User metadata
Rank Newbie
Rank
Newbie
Intel486dx33 wrote:
I am confused. So what is the conclusion for the fastest 486 build ? What cpu ? What ram ? What motherboard ? […]
Show full quote

I am confused. So what is the conclusion for the fastest 486 build ?
What cpu ?
What ram ?
What motherboard ?

This post should answer all of your questions: The Ultimate 486 Benchmark Comparison