numeriK wrote:

With "Normal" (FPM) the board sees no L2 cache. With "EDO" the board acknowledges L2 cache.

1) Are you changing the physical RAM in the system when you are changing this BIOS setting?
2) Are you using cachechk to confirm that cache is disabled?
3) Have you tried toggling the L2 cache enable/disable feature in the BIOS?

I have a PC Chips M919 which forces L1 Write-thru cache instead of Write-back cache only when an AMD X5 is installed. My other (identical) M919 does not have this problem. This is the nature of PC Chips products.

feipoa wrote:

1) Are you changing the physical RAM in the system when you are changing this BIOS setting? 2) Are you using cachechk to confirm […]
Show full quote

numeriK wrote:

With "Normal" (FPM) the board sees no L2 cache. With "EDO" the board acknowledges L2 cache.

1) Are you changing the physical RAM in the system when you are changing this BIOS setting?
2) Are you using cachechk to confirm that cache is disabled?
3) Have you tried toggling the L2 cache enable/disable feature in the BIOS?

I have a PC Chips M919 which forces L1 Write-thru cache instead of Write-back cache only when an AMD X5 is installed. My other (identical) M919 does not have this problem. This is the nature of PC Chips products.

1) No, the physical RAM is staying the same the entire time. I'm only changing the BIOS setting.
2) No, not using cachechk, my confirmation is coming from speedsys 4.78 - is there an exe or zip of the latest cachechk that I can download? Is this it?: http://www.softlookup.com/display.asp?id=9029
3) Yes, toggling this does "Enable/Disable" the cache; however even with it "Enabled", if set to "Normal" DRAM, I get zero L2 cache readings.

How can I determine if my cache is actually changing from WT to WB? Both options are available to change in my BIOS as long as the External Cache is "Enabled".

1) You need to change the RAM when you change the RAM type in the BIOS. Sometimes, a PC Chips board will first appear to work with the wrong RAM setting, but the system will ultimately crash. This may explain why L2 cache is getting disabled - some automatic over ride by the BIOS on some of your settings. I know the M919 sets some settings automatically in the BIOS, like the PCI divider and sporatically L1 WB/WT mode.

2) Speedsys is fine. Ensure there is a speed drop after 256 KB, if that is in fact how much cache you have. Cachechk is faster to run though. Search for cachechk.exe The latest version is 7, but I've seen versions 4 and 5 as well.

3) Back to 1). If issue still occurs when placing in the correct RAM type, welcome to PC Chips.

X) You can determine this in Windows using CpuMark99 or WinBench99 - CPUMark. The overall performance is reported ever-so-slightly higher when L2 is set in WB mode.

@ feipoa

Perhaps you experienced this before, but what I'm noticing is that because my UMC chipset has the infamous "E" on it for EDO support, regardless of which RAM type is installed, when set to "EDO" in the BIOS the L2 cache works. Even with FPM installed, and the BIOS set to "Normal", L2 doesn't work.

I'm guessing this is another unforgettable PC Chips moment?... 🤣

numeriK wrote:

@ feipoa

Perhaps you experienced this before, but what I'm noticing is that because my UMC chipset has the infamous "E" on it for EDO support, regardless of which RAM type is installed, when set to "EDO" in the BIOS the L2 cache works. Even with FPM installed, and the BIOS set to "Normal", L2 doesn't work.

I'm guessing this is another unforgettable PC Chips moment?... lol

I have not run into this exact problem with the E chipsetted PC Chips boards. As was the case with PC Chips, board issues were never predictable from one board to the other.

feipoa wrote on 2012-11-02, 20:08:

numeriK wrote:

@ feipoa

Perhaps you experienced this before, but what I'm noticing is that because my UMC chipset has the infamous "E" on it for EDO support, regardless of which RAM type is installed, when set to "EDO" in the BIOS the L2 cache works. Even with FPM installed, and the BIOS set to "Normal", L2 doesn't work.

I'm guessing this is another unforgettable PC Chips moment?... 🤣

I have not run into this exact problem with the E chipsetted PC Chips boards. As was the case with PC Chips, board issues were never predictable from one board to the other.

Sorry to ressurect a dead thread.
I keep ending up here after building a 486 dx4 - 100 with a M919 board.
For those who come after me;
If you get a Cache stick (I got a new one of the remakes from the us.) You will need to set the ram type to EDO and the timings to 3.1.3 for it to work stable.

For reference my board is a DX-9700 1.0 or the pc-chips M919 ver 3.3b/f. Mine has the fake "write-Back" Plastic squares on the board.

My example also had a faulty serial 1.

I got caught by a European seller who thought this was a top of the line 486 board and sold it with 16 meg of edo and an amd 5x86-p75 ADW version.
I have a 486DX4-100 and 64 Megabytes of ram, A soundblaster 16, a Diamond viper PCI and a 4gb hdd. all i need to find now is the vesa extensions that are not shareware.

I spent a fair while trying to figure out exactly what was going on, and i thought to let anyone else who had this issue after 2022 and kept landing on this thread what the fix was.

Good travels all who come after me.