Cyrix system 100-GF CPU oc'ed to 120.
voltage set to 3.45

disappointing pcpbench scores even with the matrox
speedsys score better than amd-150

speedsys score 57.71

17.1 /vgamode pcpbench score.

The document Cyrix Register Enhancements Revealed would have been able to answer your questions on this. There is a chart on page two which lists out typical default settings on the Cyrix 5x86.

These are default on most motherboards,

USE_SUSP, WT1, LOCK_NW, USE_WBAK, DET_E, MEM_BYP = 1.

The document also points out that for DOS/Win9x/NT4/W2K you need to use a program loaded at boot to set the non-default settings.

Otherwise, you need to figure out how to modify the BIOS to set these registers. While modbin has the option to modify the AWARD BIOS register settings for the Cyrix 5x86, the HEX modification numbers didn't seem to match up to what I calculate they should be. There are also more register listings than I expected. I think there were also missing registers. For this reason, I have not gone deeper into this.

When I test the 120GP chip in this system.

ibmm9.exe shows stepping 1 rev 3

is that unusual ?

Yes, very unusual in my experience. This is the first Cyrix 5x86-120 I've seen with stepping 1, revision 5. Can you show a photo from the bottom side of the chip?

Also, from Cyrix 5x86 Register Enhancements Revealed, I noted "to get BTB working on Stepping 1, Revision 3 CPUs, it is necessary to disable LOOP, BWRT, and possibly RSTK".

feipoa wrote:

Also, from Cyrix 5x86 Register Enhancements Revealed, I noted "to get BTB working on Stepping 1, Revision 3 CPUs, it is necessary to disable LOOP, BWRT, and possibly RSTK".

Yeah I tried that. theres something else at play here.
Im having trouble getting either chips to install windows 98.
Dos is completely stable but windows 98 installation often fails.
Possibly a motherboard setting I've missed or my scsi interface having issues.

Ill take a pic of the underside when I get home from work.

feipoa wrote:

I am currently playing with a SiS board. I have witnessed this funny RAM speed slow down if the L2 cache is set to write-back mode and if the TAG is set to 8-bit. Did you put your L2 cache onto write-through mode? If that doesn't fix it, try putting 256 KB cache back and jumper the board for 256 KB cache.

I'm running into the same bizarre readings on my EISA SiS411/406 machine, but it only has 256k L2 cache. The machine currently has 8MB of RAM.

What happens if you try write-through mode for the L2 cache?

do you know how minimal some of these bios are?

do you mean the jumper setting on the MB or a possible bios setting?

heres a pic of the underside of the cyrix chip.

Even with no enhancements turned on and on the correct cpu voltages with DRAM at heavy/slow timings. Using either the GP or GF cyrix chips at there prescribed cpu frequencies like 100mhz on the GF and 120 mhz on the GP I still have issues installing windows 98.

Changing RAM does not help.
Increasing Wait states.
Turning Video cache on or off, Bios Cache, etc.
Slowing the DRAM timing to 3-2-2-2 etc.

Nothing works to help install Windows.

With windows 98 setup files on the hard disk I get "error your cd is scratched, try wiping it with a cloth" thats the general gist of the message.

Or the installation will just stop and get stuck at say 7% or 22%.
Seems to like those numbers.

I got windows 98 installed at one point and running well and the next day I turned on the PC I got a bunch of corrupt file messages and windows just wouldnt boot.

when installing windows 98 from a CD I do usually get further but often it still gets stuck.

Im thinking its a scsi issue or this motherboard is a piece of sh#$@

I have tried changing to an IDE drive but still I dont get far. The motherboard hardly recognizes any of the hard disks I have. Occasionally I'll get something recognized but it will be like a 200mb Connor 😀

Ill keep experimenting but Im not sure how much more time I can spend on this UM8810 before I am inclined to put an axe through it.

Must be nice to have a stable cyrix at 133 feipoa 😜

I have this reoccurring thought if I just had of gotten a MB8433 I wouldn't be having any of these issues.

it says: c m 1995 Cyrix USA G5L8541C

Ill contact the japanese dude and see if I can get any other cyrix chips from him.

Ok, that is definately a 2x/3x only Cyrix 5x86-120 chip.

Are you using this manual to set your jumpers? Is this the board you have?
http://www.elhvb.com/mobokive/Archive/Ecs%20- … ual/um8810p.zip

I am refering to settting the L2 cache to write-through mode in the BIOS. I can see in the manual, that there is an option to disable the write-back mode in the BIOS. It is in the "Memory Cache" BIOS category.

Does your board pass a round of MemTest86+ v4.00?

Yeah thats the manual I use and the settings I use.

Yep thats the board.

Are we talking about two separate things here?

The slow RAM is on the AMD System.

Do you think I need to change from WB to WT to get the cyrix system stable? Are you using WT on your cyrix systems?

We could be talking about different things... There's multiple threads on 486 PCI systems at the moment :)

I run L2 in write-through mode to increase the cacheable range. I also found write-through cache works more reliably in 512K/1024K double-banked cache configurations.

The IODATA label indicates it is possibly a 4X chip. How did you come to the conclusion that the chip is 2X/3X? Was it tested in 2X mode? Did I miss one of the posts?

feipoa wrote:

What happens if you try write-through mode for the L2 cache?

The memory throughput numbers jump up. I'll post the numbers on my EISA build thread later.

Anonymous Coward wrote:

The IODATA label indicates it is possibly a 4X chip. How did you come to the conclusion that the chip is 2X/3X? Was it tested in 2X mode? Did I miss one of the posts?

From the markings on the bottom of the chip. The IODATA "X4" is probably short for "DX4".

NJRoadfan wrote:

feipoa wrote:

What happens if you try write-through mode for the L2 cache?

The memory throughput numbers jump up. I'll post the numbers on my EISA build thread later.

:)

So what's the trick in deciphering the markings?

Anonymous Coward wrote:

So what's the trick in deciphering the markings?

I was being deliberately vague. Announcing it [here] will only make them harder to find and more expensive.

finally stability.

I have set the L2 cache to "disable WB" in bios but im not sure if thats changed anything, how can I tell if its definitely on WT?

I think what gave me stability was setting the HCLK PCICLK to 1 1/2

However now I have issues turning on cyrix enhancements.

What I was using previously and feipoa's defaults both crash the system.

defaults:

RSTK_EN 1
BTB_EN 0
LOOP_EN 1
LSSER 0
WT1 1
BWRT 1
LINBRST 1
FP_FAST 1
MEM BYP 1
DTE_EW 1
IORT 000
USE_WBAK 1

previously working:

RSTK_EN 1
BTB_EN 0
LOOP_EN 1
LSSR 0
WT1 1
BWRT 1
LINBRST 0
FP FAST 1
MEM BYP 1
DTE_EW 0
IORT 000
USE_WBAK 1

Could you remind me which frequency you are using to test the chip at, and which voltage?

Try it at 100 MHz, 3x33.

You should not need to halve the PCI clock at 3x33 MHz. If you are using 40 MHz (and 1:1), your stability issues are probably due to your cache timings being too fast, or some PCI card can't cut it at 40 MHz.

RSTK_EN 1
BTB_EN 0 (this is default)
LOOP_EN 1
LSSR 0
WT1 1
BWRT 0 (I think step1/rev3 chips might not work with BWRT - I forget now)
LINBRST 0
FP FAST 1
MEM BYP 1 (this is default)
DTE_EW 1 (this is default)
IORT 000 (this is default)
USE_WBAK 1 (this is default)

Try using chkcpu16, CTCM, and CTCM7 to determine the L2 cache WT/WB mode.

120mhz cpu
40 mhz bus

3.45 volts

The UMM8810 manual states that if you wish to use a Cyrix 5x86 120 then you need to set the PCI divider to 1 1/2.

When I do that every cyrix enhancement even turned on individually seems to crash the system.

Windows 98 is running surprisingly fast even without the enhancements.

I'll...

Pull the Yamaha sound card out.
Drop the Bus to 33
Drop the Voltage to 3.3
Return the hclk pciclk to 1

and retry the enhancements ...

Leave the voltage at 3.45 V.