430TX and the external cache (SP-586TB)

[RodH]

Hi,
I've been suffering no cache since reflashing my 430TX mobo (SP-586TB).
I've temporarily reverted to the previous BIOS and "Pipelined Burst" is now reported on the POST summary page, so there's no defect on the board and it looks like a BIOS issue.

The older BIOS was Award 4.51PG 586TX A.5 (supplied with the board), the newer one A.6 (from Rainbow)
Has anyone else had problems with TX_A6 (and solved them!) ?
Can anyone recommend a utility to check which chipset registers changed, or better still, allow me to change them from DOS so I can re-enable the cache after booting? I want to use the later BIOS for better disk and CPU support but it's a real shame to have no cache! I've also tried an unmodded A.6, and various memory & CPU combinations, all with the same results: A5 works fine, A6: "None"!
Thanks,
Rod

[RodH]

OK, Jan Stunebrink came to my rescue ... the A5 BIOS is now patched for K6 mobile "plus" chips and larger hard disks, and it works.

I've also found I can run the FSB at 83MHz (JP2:2-3, JP3:1-2, JP4:2-3). This overclocks the PCI & IDE which works OK for my cards/disks (i430TX requires PCICLK=CPUCLK/2) although that is too much for the mobo cache unless someone can point me to a faster version of Winbonds' W25P240AF-6 (hint, hint!).

I'll probably keep the faster FSB as it allows the CPU to run at 500MHz (6x on the K6-III+/450ACZ) and this CPU already has 256K cache onboard, and the onboard cache serves all main memory and at higher speed than the paltry 64MB the i430TX does with it's 512KB cache.
HTH
Rod

[KachiWachi]

RodH - Does your TX "slow down" when you enable the motherboard cache? See my posts on that here in the forum....one is in my sig, the other is a separate post.

[RodH]

I've tested it with the cache enabled and an 83MHz FSB and it fails memtest86+. Sometimes it just locks up altogether.

I've not compared with and without the cache enabled on a 75MHz FSB. Do you need that info?

Thanks,
Rod

[KachiWachi]

I would be interested in either 66 or 75MHz data...whatever the system will run at successfully.

It just seems odd to me that the system "slows down" by ~13% writing to memory when the motherboard cache is enabled. See the table entries for "DFI WA On Int" (102nS, 81.6MB/s) vs. "DFI WA On All" (113nS, 73.6MB/s).

Here is the link to the other post for reference -> http://www.wimsbios.com/phpBB2/viewtopic.php?t=5875

[RodH]

Here's a result from "Cachechk" cache enabled ...

L1 cache is 32KB--2085.1 MB/s 0.5 ns/byte (1014%)
L2 cache is 128KB--1392.7 MB/s 0.8 ns/byte (677%)
L3 cache is 512KB-- 430.8 MB/s 2.4 ns/byte (209%)
Main memory speed -- 205.5 MB/s 5.1 ns/byte (100%) [reading] 9.8 clks
Effective RAM access time (read ) is 40ns (a RAM bank is 8 bytes wide).
Effective RAM access time (write) is 87ns (a RAM bank is 8 bytes wide).

Here's the same, mobo cache disabled ...

L1 cache is 32KB--2085.1 MB/s 0.5 ns/byte (785%) (150%) 1.0 clks
L2 cache is 128KB--1386.5 MB/s 0.8 ns/byte (522%) (100%) 1.4 clks
Main memory speed -- 265.3 MB/s 4.0 ns/byte (100%) [reading] 7.6 clks
Effective RAM access time (read ) is 32ns (a RAM bank is 8 bytes wide).
Effective RAM access time (write) is 69ns (a RAM bank is 8 bytes wide).

Seems to me the mobo cache is slowing things down, both reading and writing with the 430TX !!!
HTH
Rod

[KachiWachi]

What FSB was this run at?

[RodH]

83MHz - before I'd run exhaustive tests with memtest86+

[KachiWachi]

OK...I was wondering why the numbers were higher than I expected.

It is still an interesting phenomena though. I'd like to try my machine with a 233MMX to see if it has something to do with the CPU, but I just have to get around to reconfiguring the machine for the test...some day...