VOGONS


First post, by kaputnik

User metadata
Rank Oldbie
Rank
Oldbie

Picked up a 233MMX CPU I want to install in my Gigabyte GA-586HX rev. 1.53 based Pentium rig.

According to the compatibility list, revisions 1.50-1.54 does not support this CPU. However, since the 166MMX is supported, even the older revisions of the board should be able to supply the needed dual voltage. Since the list was written, several BIOS updates has been released, got 2.9 installed on my board now.

Just thought I should ask before doing anything stupid; is this perhaps just a question of the compatibility list not being updated to reflect changes in newer BIOS revisions? Has anyone here tried faster MMX chips in older revisions of the GA-586HX board?

Reply 1 of 11, by elianda

User metadata
Rank l33t
Rank
l33t

I have a P166MMX running on a Rev 1.51 board and a K6 233 on a Rev 1.55 board.

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 2 of 11, by kaputnik

User metadata
Rank Oldbie
Rank
Oldbie

So, popped the 233MMX in and did some experimenting. Tested all 16 possible FSB speed/multiplier combos, and while the MMX chip boots just fine, it seems to interpret the 1.5 multiplier jumper setting as 2.0 (expected 3.5), and the 3.0 one as 2.5. Here's a table:

FSB	Multi   Freq   Note
50 1,5 100 50*2?
55 1,5 110 55*2?
60 1,5 120 60*2?
66 1,5 133 66*2?
50 2,0 100
55 2,0 110
60 2,0 120
66 2,0 133
50 2,5 125
55 2,5 133
60 2,5 150
66 2,5 166
50 3,0 125 50*2,5?
55 3,0 133 55*2,5?
60 3,0 150 60*2,5?
66 3,0 166 66*2,5?

Kind of odd behaviour, can it be explained, and perhaps even fixed somehow?

Also did some testing with Setmul and Speedsys. What's interesting here is that all the TR12 options (BPD, VPD, L1DX, CCD, DCD) seems to work, thought they weren't supposed to with MMX chips. Another table, using 66*2.5=166MHz as baseline:

No switches:      124.57
BPD: 106.46
VPD: 118.64
L1DX: 33.40
CCD: 68.09
DCD: 46.40
All of the above: 26.17

Not that I complain, but thought it was common consensus that the test registers were disabled on MMX chips other than PODs.

Reply 3 of 11, by clueless1

User metadata
Rank l33t
Rank
l33t

I think at one time it was thought that regular MMX did not do TR12, mainly because my POD worked while one other person's 233MMX did not. But since then some other people on here have come forward with their MMX cpu working. lvader has results of his 233MMX on the cachebench chart. I put his results on their own tab because he wired his FSB to some case switches and was able to get a much wider range of results. It's nice to see them all without being mixed in with the other results.

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 4 of 11, by lazibayer

User metadata
Rank Oldbie
Rank
Oldbie
kaputnik wrote:
So, popped the 233MMX in and did some experimenting. Tested all 16 possible FSB speed/multiplier combos, and while the MMX chip […]
Show full quote

So, popped the 233MMX in and did some experimenting. Tested all 16 possible FSB speed/multiplier combos, and while the MMX chip boots just fine, it seems to interpret the 1.5 multiplier jumper setting as 2.0 (expected 3.5), and the 3.0 one as 2.5. Here's a table:

FSB	Multi   Freq   Note
50 1,5 100 50*2?
55 1,5 110 55*2?
60 1,5 120 60*2?
66 1,5 133 66*2?
50 2,0 100
55 2,0 110
60 2,0 120
66 2,0 133
50 2,5 125
55 2,5 133
60 2,5 150
66 2,5 166
50 3,0 125 50*2,5?
55 3,0 133 55*2,5?
60 3,0 150 60*2,5?
66 3,0 166 66*2,5?

Kind of odd behaviour, can it be explained, and perhaps even fixed somehow?

Also did some testing with Setmul and Speedsys. What's interesting here is that all the TR12 options (BPD, VPD, L1DX, CCD, DCD) seems to work, thought they weren't supposed to with MMX chips. Another table, using 66*2.5=166MHz as baseline:

No switches:      124.57
BPD: 106.46
VPD: 118.64
L1DX: 33.40
CCD: 68.09
DCD: 46.40
All of the above: 26.17

Not that I complain, but thought it was common consensus that the test registers were disabled on MMX chips other than PODs.

I have an explanation and possible fix.
The multiplier settings on both P54C and P55C are determined by BF0 and BF1 pins, but P54C and P55C respond differently when the two pins are float.
P54C: both are pulled high.
P55C: BF0 is pulled low, BF1 is pulled high.

So if your board can only toggle the two pins between pulled low and float, lacking the ability of pulling them up, you will observe the described multiplier outcome. You can confirm this by measuring the voltage on the pins.

To fix it you can jump wire the floating pin of the jumper/switch to VCC3 with a 1k ohm resistor.

Reply 6 of 11, by kaputnik

User metadata
Rank Oldbie
Rank
Oldbie
clueless1 wrote:

I think at one time it was thought that regular MMX did not do TR12, mainly because my POD worked while one other person's 233MMX did not. But since then some other people on here have come forward with their MMX cpu working. lvader has results of his 233MMX on the cachebench chart. I put his results on their own tab because he wired his FSB to some case switches and was able to get a much wider range of results. It's nice to see them all without being mixed in with the other results.

Ah, that explains it. Quite sure I at least partly got that info from the discussion in your benchmarking thread 😀

lazibayer wrote:
I have an explanation and possible fix. The multiplier settings on both P54C and P55C are determined by BF0 and BF1 pins, but P5 […]
Show full quote

I have an explanation and possible fix.
The multiplier settings on both P54C and P55C are determined by BF0 and BF1 pins, but P54C and P55C respond differently when the two pins are float.
P54C: both are pulled high.
P55C: BF0 is pulled low, BF1 is pulled high.

So if your board can only toggle the two pins between pulled low and float, lacking the ability of pulling them up, you will observe the described multiplier outcome. You can confirm this by measuring the voltage on the pins.

To fix it you can jump wire the floating pin of the jumper/switch to VCC3 with a 1k ohm resistor.

Ah, that must be it, thanks! Got my jumper headers wired up to a bunch of SPDT switches used as SPST ones mounted on a 3.5" bay panel. Posted a few pics of the mod, if you're interested. At a glance it looks like it should be easy enough to implement if I got it right, will just have to replace the BF0 switch with a SPDT on-off-on one, and find some point on the board to connect it to VCC3 through that resistor, or just rewire the existing switch between VCC3 and ground 😀

Reply 7 of 11, by kaputnik

User metadata
Rank Oldbie
Rank
Oldbie
lazibayer wrote:
I have an explanation and possible fix. The multiplier settings on both P54C and P55C are determined by BF0 and BF1 pins, but P5 […]
Show full quote

I have an explanation and possible fix.
The multiplier settings on both P54C and P55C are determined by BF0 and BF1 pins, but P54C and P55C respond differently when the two pins are float.
P54C: both are pulled high.
P55C: BF0 is pulled low, BF1 is pulled high.

So if your board can only toggle the two pins between pulled low and float, lacking the ability of pulling them up, you will observe the described multiplier outcome. You can confirm this by measuring the voltage on the pins.

To fix it you can jump wire the floating pin of the jumper/switch to VCC3 with a 1k ohm resistor.

So, did the hack tonight, works like a charm 😀 Was some work with it though.

First I removed the CPU from the socket, located the VCC3 holes, put a sewing pin in one of them, and closed the socket to make contact. Started continuity testing all jumper headers I could find on the board. Nothing...

Began searching for some other point to tap in on the VCC3 plane. Quickly found that two pins of the unpopulated and undocumented header JP4 was connected to it, so decided to pull the mobo from the case, and populate the header. Did that, and put the computer back together again. Rewired my custom BF0 front panel switch, and patched in a 1kohm resistor. Used the same method with the sewing pin to find out which one of the BF0 jumper pins that is the floating one.

Only thing left to do now is writing a new frequency table for the front panel 😀

Reply 8 of 11, by lazibayer

User metadata
Rank Oldbie
Rank
Oldbie
kaputnik wrote:
So, did the hack tonight, works like a charm :) Was some work with it though. […]
Show full quote

So, did the hack tonight, works like a charm 😀 Was some work with it though.

First I removed the CPU from the socket, located the VCC3 holes, put a sewing pin in one of them, and closed the socket to make contact. Started continuity testing all jumper headers I could find on the board. Nothing...

Began searching for some other point to tap in on the VCC3 plane. Quickly found that two pins of the unpopulated and undocumented header JP4 was connected to it, so decided to pull the mobo from the case, and populate the header. Did that, and put the computer back together again. Rewired my custom BF0 front panel switch, and patched in a 1kohm resistor. Used the same method with the sewing pin to find out which one of the BF0 jumper pins that is the floating one.

Only thing left to do now is writing a new frequency table for the front panel 😀

Haha nice!

Reply 9 of 11, by RaiderOfLostVoodoo

User metadata
Rank Member
Rank
Member

Necro time!

I have Rev 1.52 of the GA-586HX. BIOS is the latest.
The manual states, that I can use MMX166 and MMX200:
qoSx2o3.png
Is this true? Do the normal jumper settings work for the MMX200?
Can I use it without modifications?

Reply 10 of 11, by GuillermoXT

User metadata
Rank Member
Rank
Member
elianda wrote on 2017-07-16, 13:14:

I have a P166MMX running on a Rev 1.51 board and a K6 233 on a Rev 1.55 board.

I have also a Rev 1.55 here with Pentium 133.
How do i have to set the board for the AMD K6 233 ANR and do I have to modify something?

My Retrosystems:
K6-2 400 running Win98SE
P1-133 with Win3.11
Tandon 286-8MHZ Running DOS 5.0 on XT-IDE Card

Reply 11 of 11, by Sphere478

User metadata
Rank l33t
Rank
l33t

Oh yeah, this is relevant to a thread I started the other day.

Re: Bf0, Bf1, Bf2 Multiplier Mod

Follow it, I’m hoping to eventually have a pcb that can be soldered to the back of the motherboard with a bunch of tricks built in to solve problems like this.

I am not familiar with this board or its versions. But images that I see look like it’s from the time right after socket 5 so it’s possible that some of these boards may be single plane and socket 7 (a oxymoron, I know) but anyway, if the manual for your version claims that it supports a non overdrive mmx cpu, then it should be dual plane.

If it’s dual plane then you can probably use whatever socket 5/7/ss7 cpu you want all the way up to a 400-500mhz k6-3+

But you’ll probably want to look into a patched bios for that.
DIY Bios Modding guide Jan Steunebrink k6-2+/3+ 128gb

But yeah, if it supports a 166 mmx then it will support a 233mmx.

If your multipliers are acting funny, then it’s probably because of the reasons mentioned in above replies. (Also see my 1st link)

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)