Rav wrote on 2023-04-03, 01:11:

I know my Cyrix 486 DX2 can do 100Mhz (it work fine with a 50Mhz FSB). But when I do that, I lose the cache.

Surely it it loses the cache it's not "working fine with a 50Mhz FSB". Which cache does it loose, the L1 or L2?

I am guessing it's a DX2-66, is it? Assuming so what happens with a 40 MHz FSB ?

red-ray wrote on 2023-04-03, 01:22:

Rav wrote on 2023-04-03, 01:11:

I know my Cyrix 486 DX2 can do 100Mhz (it work fine with a 50Mhz FSB). But when I do that, I lose the cache.

Surely it it loses the cache it's not "working fine with a 50Mhz FSB". Which cache does it loose, the L1 or L2?

I am guessing it's a DX2-66, is it? Assuming so what happens with a 40 MHz FSB ?

CPU is fine with a 50mhz fsb
L2 chinasram are not! It's why I want the multiplier mod instead

It's a DX2-66 yes
But my motherboard is that Acer A1GX crap and if I set to 40Mhz FSB it add a multiplier on memory and buses so everything is slower except the CPU (But the L2 work with that setting).
Setting it to 50Mhz bring back the other things (memory, ISA bus....) on par with 33Mhz FSB

speedsys (Memory throughput):
33Mhz bus, with cache = 23.86MB/s
40 Mhz bus, with cache = 16.9MB/s
50Mhz bus, without cache = 25.27MB/s

I did some extra research:
B14 from my Cyrix DX2 is not pulled up to VCC or down to vss (tested with A/B11). Maybe it use a different VCC pin (and the CPU have to be on the motherboard to be able to detect it, or probing all VCC pins individually).

Or if the IBM DX2 datasheet is right (Cyrix are supposed to be the same), unfortunately. According to the datasheet, the clock multiplier stuff is indeed pulled up to VCC BUT the B14 pin is NOT connected to it. B14 is supposed to be NC.
On the DX4, the pin is actually connected.

That could also explain why I could not get any measurement on eater VCC or VSS, if it's actually not connected...

Rav wrote on 2023-04-03, 01:11:

Hi there.

I know my Cyrix 486 DX2 can do 100Mhz (it work fine with a 50Mhz FSB). But when I do that, I lose the cache.

I was wondering if I can change the multiplier. The motherboard I have have no such control but from what I understand, I have to solder I wire on the socket between pin B14 and VSS, You think it will work on that specific CPU?

It is unlikely that DX2's have a way to change the multiplicator.
Later introduced and from DX4 backported 3 Volt compliant DX2's may have a way to change the multi from 2 to 3.

It is no limitation from the motherboard.

100 MHz is a heavy overclock for a DX2-66.
And 50 MHz is still a heavy overclock to the internal memory/cache system even if you can get it to operate in x1 mode.

Later die shrinks like in the 3 Volt DX4/100 are more tolerant to FSB overclocking and may be used as DX2/100 - mostly with slowing down L2 from 2-1-1-1 to 3-2-2-2.

Disruptor wrote on 2023-04-03, 06:07:

It is unlikely that DX2's have a way to change the multiplicator. Later introduced and from DX4 backported 3 Volt compliant DX2' […]
Show full quote

Rav wrote on 2023-04-03, 01:11:

Hi there.

I know my Cyrix 486 DX2 can do 100Mhz (it work fine with a 50Mhz FSB). But when I do that, I lose the cache.

I was wondering if I can change the multiplier. The motherboard I have have no such control but from what I understand, I have to solder I wire on the socket between pin B14 and VSS, You think it will work on that specific CPU?

It is unlikely that DX2's have a way to change the multiplicator.
Later introduced and from DX4 backported 3 Volt compliant DX2's may have a way to change the multi from 2 to 3.

It is no limitation from the motherboard.

The complete model I have is the TI486DX2-G66-GA. It is 3.45v so maybe there is hope, going to do more research tomorrow...

Rav wrote on 2023-04-03, 06:12:

The complete model I have is the TI486DX2-G66-GA. It is 3.45v so maybe there is hope, going to do more research tomorrow...

Ah, it's a DX4 backport then. Then I do not wonder why it runs with 50 MHz FSB at all.
Perhaps it was binned down to DX2-66 because some tests at 100 MHz have failed (1st candidate is L1 cache ofc.).

Perhaps you have success with 3.6 Volt or 4.0 Volt.

And, yes, you may try first whether you get your L2 cache working in write back mode at 66 MHz.
Cyrix 486 had write back L1 even in their DX, while Intel has introduced it in the later DX4 first.
So you may need to try both jumper settings.

Tools:
the cache graphs in SpeedSys 4.78
ctcm 1.7a from c't (heise): ctcm7 /nop

Disruptor wrote on 2023-04-03, 06:15:

Ah, it's a DX4 backport then. Then I do not wonder why it runs with 50 MHz FSB at all. Perhaps it was binned down to DX2-66 beca […]
Show full quote

Rav wrote on 2023-04-03, 06:12:

The complete model I have is the TI486DX2-G66-GA. It is 3.45v so maybe there is hope, going to do more research tomorrow...

Ah, it's a DX4 backport then. Then I do not wonder why it runs with 50 MHz FSB at all.
Perhaps it was binned down to DX2-66 because some tests at 100 MHz have failed (1st candidate is L1 cache ofc.).

Perhaps you have success with 3.6 Volt or 4.0 Volt.

And, yes, you may try first whether you get your L2 cache working in write back mode at 66 MHz.
Cyrix 486 had write back L1 even in their DX, while Intel has introduced it in the later DX4 first.
So you may need to try both jumper settings.

Tools:
the cache graphs in SpeedSys 4.78
ctcm 1.7a from c't (heise): ctcm7 /nop

I did notice that I get WT for L1 if the bus is 33, and WB if the bus is 40 or 50.
For the L2, it's WB for 33 and 40, then it's non existent for 50

Changing the CPU voltage won't change a thing for the L2 I think. It was already at 3.8 (oups). I set it back down to 3.45.

66Mhz(33fsb):

More will follow in a few minutes

80Mhz(40fsb)

100Mhz(50fsb)

40 Mhz seam to give no real advantage over 33 (because of the memory drop)
50 Mhz, Maybe I should just use that and forget about the cache... Until I can get a DX4+

Will take the board out tonight and see if I can measure something between B14 and VCC while the CPU is on the board, and if so, Will connect B14 to VSS and see if that change something.

Your memory performance looks like you have used autoconfig in BIOS, haven't you?

I don't give too much on the cache write strategy of ctcm, because sometimes it shows inplausible results.
But your transfer rates don't really look fast.
I'll start my DX2 machine later this day.

Disruptor wrote on 2023-04-03, 17:43:

Your memory performance looks like you have used autoconfig in BIOS, haven't you?

That's the only choice with that board. There is no option to adjust memory.
I plan to try to fix that eventually (when I get a eprom program plus some eprom that fit... Find board with same chipset and try them, I did already extract the video BIOS so I will also have to patch replacement BIOS to add the video rom in it).

Rav wrote on 2023-04-03, 17:46:

Disruptor wrote on 2023-04-03, 17:43:

Your memory performance looks like you have used autoconfig in BIOS, haven't you?

That's the only choice with that board. There is no option to adjust memory.
I plan to try to fix that eventually (when I get a eprom programer plus some eprom that fit... Find board with same chipset and try them, I did already extract the video BIOS so I will also have to patch replacement BIOS to add the video rom in it).

Rav wrote on 2023-04-03, 17:46:

Disruptor wrote on 2023-04-03, 17:43:

Your memory performance looks like you have used autoconfig in BIOS, haven't you?

That's the only choice with that board. There is no option to adjust memory.
I plan to try to fix that eventually (when I get a eprom program plus some eprom that fit... Find board with same chipset and try them, I did already extract the video BIOS so I will also have to patch replacement BIOS to add the video rom in it).

Don't say that.
Can you post a picture of that board please? We need to identify the chipset.
Which BIOS do you use? AMI or Award or something else?

Disruptor wrote on 2023-04-03, 18:36:

Don't say that. Can you post a picture of that board please? We need to identify the chipset. Which BIOS do you use? AMI or Awar […]
Show full quote

Rav wrote on 2023-04-03, 17:46:

Disruptor wrote on 2023-04-03, 17:43:

Your memory performance looks like you have used autoconfig in BIOS, haven't you?

That's the only choice with that board. There is no option to adjust memory.
I plan to try to fix that eventually (when I get a eprom program plus some eprom that fit... Find board with same chipset and try them, I did already extract the video BIOS so I will also have to patch replacement BIOS to add the video rom in it).

Don't say that.
Can you post a picture of that board please? We need to identify the chipset.
Which BIOS do you use? AMI or Award or something else?

Here are the documentation : https://stason.org/TULARC/pc/motherboards/A/A … 6-G-A1GX-2.html
Note that 50Mhz fsb is not on the documentation nor on the board silkscreen, found it by mistake because the 40mhz jumper setting was corroded and not making contact initially (JP11 not connected = 50Mhz)

Bios is "Acer"

Ok, then you may give c't chip a try.
ftp://ftp.heise.de/pub/ct/ctsi/ctchip34.zip

Wait, I don't see that it has support for the ALI 1429 ???

Disruptor wrote on 2023-04-03, 18:51:

Ok, then you may give c't chip a try.
ftp://ftp.heise.de/pub/ct/ctsi/ctchip34.zip

Wait, I don't see that it has support for the ALI 1429 ???

Right, no Ali chipset on there. It can only read some of the cyrix cpu registers.
I also checked other solutions like mrbios. (Same, no support for ALI M1429G)

I did peek into a [different] acer chip datasheet that I found : http://66.113.161.23/~mR_Slug/pub/datasheets/ … ALi/M1489-2.pdf

And I disassembled the BIOS.

I can confirm that the Lock Register stuff is the same (page 37 of said datasheet).

And see no trace of messing with index register 1Bh, which if is the same in M1429 than M1489, would mean that it would use default of slow timing for read and write, disable RAS only refresh, disable onboard parity check and 4T for CPU time slot (I don't know what that latter is).

Next step would be to try to unlock by writing 03h to port 22h and C5h to port 23 then write 1Bh to port 22 and 0Ah to port 23 then bench if it change anything...