Very nice. Did you try a CPU which uses 16KB in writeback mode?

Did you also try the card's original drivers?

The card didn't come with any drivers, so I had none to try.

It looks like the card has the potential to work with 16KB L1 WB CPUs. I tried it in my Asus 486SV2GX4 with an AMD 5x86-133 in L1 WB mode and it seemed to work fine. However, it seemed to hang in my ASUS 486SP3 with the same CPU, though I didn't really bother to mess with the BIOS settings to see if I could get it to work. It also behaved somewhat erratic with a Cyrix 5x86 in the 486SV2GX4, though I'm not sure yet if that's the motherboard's fault or the card's. My guess is it will work in some configurations and not others.

Take everything with a grain of salt though as I'm still trying to get my 486SV2GX4 configured properly. I'm still not 100% sure everything is set up right. For some reason, utilities like CACHECHK and SPEEDSYS are showing really slow memory bandwith, like around 8-12 MB/s. I think there's something wrong with my BIOS or Jumpers related to memory and cache. CHKCPU shows the CPU is in L1 WB mode, but CACHECHK isn't properly identifying my CPU and sometimes shows L1, L2, and L3 cache.

So no issue using an 8 GB partition in DOS with Am5x86-WB and 486SV2GX4?

I recall the 486SV2GX4 having issues with the Cyrix 5x86-WB and SCSI, so I don't think this test will reveal much.

8-12 MB/s sound abnormally slow. Are you using the forum patched BIOS that fixes the L2 cache bug, which will kill your DRAM speed?

I was using a 2GB partition on a 4GB disk w/ DOS 6.22 and everything seemed fine with the AMD 5x86-WB and 486SV2GX4. I later switched over to the BIOS with the L2 patch, and it seemed to improve things, however memory still seems a decent bit slower than on my other 486 boards.

Do you know what memory/L2/L1 speed I should expect to see with the 486SV2GX4? Also, I noticed SPEEDSYS and CACHECHK seem to spit out very different results on most boards that I test them with. Does anyone know why this is?

My 486SV2GX4 is a 1.8 revision, which doesn't have official jumper settings for the AMD 5x86-WB. I tried following this guide: ASUS VL/I-486SV2G(X4) REV. 1.8 conversion and jumpers guide and it seems to be mostly working. However, a lot of programs still won't properly identify the CPU which has me thinking maybe my jumper settings are wrong, or maybe only rev 2.0 and later fully enable the AMD 5x86?

I think the jumper settings for v1.8 and v2.0/2.1 are different. I am using the v2.0/v2.1 jumper settings.

What cachechk results do you record at 160 MHz?

At 160 MHz, with L1:WB, Cachechk shows:
L1 read = 164.9 MB/s
L2 read = 74.5 MB/s
RAM read - 51.1 MB/s
L1/L2/RAM write = 82.9 MB/s

I am pretty sure these were taken before I made the L2 BIOS fix for this motherboard. Did you grab my fix, or one of the others floating around VOGONS?

I don't normally record Speedsys memory metrics. I'd have to pull a dozen cases out of the closet to find my VL/I-486SV2GX4 system to run more tests.

The question as to why the Speedsys results differ from the Cachechk results was talked about a few years ago. I think mkarcher or one of the other 2 gurus answered it, but alas, I forgot the answer.

Does the DTC 3274VL work fine at 160 MHz on this motherboard? Have you tested it in Windows 9x or 3.1? On the Adaptec AHA-2842A, another VLB SCSI card, there is a jumper which needs to be specifically set for use with L1:WB CPUs. Such a jumper isn't present on the DTC 3274VL. Performance metrics in DOS and Windows at 160 MHz would be nice to have. I wonder if the DTC's onboard SRAM adds any benefit, or just slows things down some when using modern SCSI HDDs (which already have 8 MB of flash)?

Yeah my board is only giving me ~141MB/s on L1, ~25MB/s on L2, and ~15MB/s on main memory, and that's after the BIOS update. This is with the AMD X5 at 160MHz WB. Something seems off...

What Win95/Win3.1 benchmarks would be useful? I might have time to run some stuff this weekend.

I didn't stress test the DTC super hard, but it seemed stable after running multiple programs from the HDD, CD, Floppy, even after performing some soft reboots with ctrl-alt-del. The DTC with the Cyrix could run a few programs of the HDD or CD, but would usually lock up eventually or hang after a reboot.

willmurray461 wrote on 2025-04-27, 05:01:

It looks like the card has the potential to work with 16KB L1 WB CPUs. I tried it in my Asus 486SV2GX4 with an AMD 5x86-133 in L1 WB mode and it seemed to work fine. However, it seemed to hang in my ASUS 486SP3 with the same CPU, though I didn't really bother to mess with the BIOS settings to see if I could get it to work.

If you mean the ASUS PCI/I-486SP3 (Saturn based, no VLB slots), you're right. This board does not like newer processors with WB cache. I use an AMD Am486DX-4 NV8T in it.
If you mean the more popular ASUS PVI-486SP3 (PCI slots and one VLB slot), all of your processors should work fine.

willmurray461 wrote on 2025-04-29, 03:02:

Yeah my board is only giving me ~141MB/s on L1, ~25MB/s on L2, and ~15MB/s on main memory, and that's after the BIOS update. This is with the AMD X5 at 160MHz WB. Something seems off...

What Win95/Win3.1 benchmarks would be useful? I might have time to run some stuff this weekend.

I didn't stress test the DTC super hard, but it seemed stable after running multiple programs from the HDD, CD, Floppy, even after performing some soft reboots with ctrl-alt-del. The DTC with the Cyrix could run a few programs of the HDD or CD, but would usually lock up eventually or hang after a reboot.

Our values are starkly different. Did you compare the v1.8 board to the v2.x board to see how many physical jumper differences there are? Did you try putting your L2 cache on Write-through mode? It may be the BIOSes' "auto" option under L2. Also, you may try using 256K double-banked.