My best guess is that it's not properly configured or some defect - did you check the turbo jumper? Sometimes it needs to be closed.

Otherwise I only have a few 286 motherboards from Octek and I quite like them.

Yeah I checked the turbo button and closing it made the board even slower when WB was disabled and froze everything when WB was enabled and didn't even boot. That is suspisious as it should be ignored at worst.

I have checked this mirror of Octek Australia's FTP server and there is no mention of any manual (at least useful). I guess this board is doomed forever 😁. Only hope is if someone has this manual in printed form.

Can't open the docs on my phone... But look here:
https://www.elhvb.com/mobokive/Archive/Octek/ … 2010/index.html

I know about that site. The manual file mentioned there is only small doc file with jumper settings for Cyrix CPU. It doesn't even mention what board revision it is for.

I have this MB in repair process so I hope I'll be able to benchmark it against my DataExpert MB with Am5x86 that doesn't support cache WB (related cpu pins are not connected)

I guess leaked battery? I see that a lot on these boards. Otherwise it's a very nice motherboard, especially with the PS2 mouse mod. The only downside is general flakiness when enabling WB mode but that's chipset's issue.

Have you tried different sets of cache chips and did this have an effect on the stability issues?

Did you try 256K L2 cache (8 x 32Kx8) to see if cache interleave improves performance?

Sorry, I wasn't specific enough - I meant L1 WB mode, otherwise L2 WB mode was fine for me. Another issue is that there's no way to set L1 to WB and L2 to WT in the BIOS (for this UMC chipset in general) which is the preferred combination when using L1 WB CPU - overall system performance is higher. Both AMI and Award BIOSes don't have this feature. I think I saw this option in Phoenix BIOS but I haven't tested it yet. I'm currently using board with SiS chipset.

Yes, leaked battery.
I fixed it that KBD works now but I'm not sure if I routed all dissolved traces correctly. Please can someone with this good MB make a detailed photo of the area I shown below?

The board has strange issue that sometimes it fails power on - no POST codes comming. Then I have to use HW reset switch and it makes it boot every time, at least yet.

I'm a bit dissapointed about cache sockets. I read in spec the MB can do 512kB and feel god until I looked closer that it requires ultra rare 128k chips instead 64k chips that I have because one row of sockets are short ones that can hold only 32k chips 🙁

I started with intel 486DX33 with S3 Trio 64 VLB and onboard IDE and it seems to run stable at fastest timings, WB cache.
Doom: 15,29 FPS
Quake: 4,1 /1,6 FPS (320*200 / 640*480)

I also plan tu burn newer BIOS as mine is from 1994 and I found latest from 1996 at TheRetroweb. Mine MB HW rev is 1.01

majestyk wrote on 2025-03-01, 06:09:

Did you try 256K L2 cache (8 x 32Kx8) to see if cache interleave improves performance?

What means cache interleave? There's no such option in SETUP and I never seen this for cache. As it runs at 0WS I don't see any meaning for cache interlevaing. My current config ix 9*32k chips.

RayeR wrote on 2025-03-02, 05:48:

majestyk wrote on 2025-03-01, 06:09:

Did you try 256K L2 cache (8 x 32Kx8) to see if cache interleave improves performance?

What means cache interleave? There's no such option in SETUP and I never seen this for cache. As it runs at 0WS I don't see any meaning for cache interlevaing. My current config ix 9*32k chips.

Not all 486 chipsets support L2 cache interleaving. Here are some hints to be found:
https://forum.vcfed.org/index.php?threads/doe … terleave.44473/

It´s generally not configurable in BIOS setup, but enabled automatically when 2 banks cache are populated.
The "ASUS_PCI-I-486SP3G" (Saturn chipset) for example is said to perform better with two banks (2 x 4 x 32Kx8 = 256K) cache than with one bank (4 x 128Kx 9= 512K).
This might be an issue in cases when there are 4 x 28-pin and 4 x 32-pin sockets on a mainboard...like the Octek Hippo 10.

Paar, how did you jumpered the Vcore for AMD 486DX4? I have also rev 1.01 and I see jumpers doesn't match the manual. I can find only JP46, JP48. And only JP48, when closed, have some effect - but I got only 2,5V...

UPDATE: Ah, it seems the MB has some kind of dual-voltage CPU autodetection. It tricked me that when CPU is not inserted in the socket it fires 5V there. I draw some part of VRM schematic. JP48 is connected to Vout and TL431 reference. You can attach ext. resistor to lower output voltage downto 2,5V. JP46 is connected to GND and TL431 reference. Now I put in ST486DX4V100 and measured 3,48V.

UPDATE2: With Am5x86 I have available option L1 write back/through in SETUP but in fact it seems run always WT mode 🙁
When jumpered CPU type as 486DX4 before, even this option was disabled (forced to WT). Despite this it runs a bit faster than my DataExpert MB. But I cannot boot at 40MHz FSB, I need increase Vcore to 3,6-3,7V.

I finally found how to switch L1 cache mode:

WB/WT# CPU pin (B13) from the CPU socket is connected directly to JP37.4 and JP37.5 is at +5V so adding a jumper to position 4-5 enables L1 Write Back cache while removed jumper falls back to Write Through (there’s 10k pull-down to GND).

If JP37 is jumpered to WT mode the WB/WT option in SETUP is disabled, forced to WT.
If JP37 is jumpered to WB mode the WB/WT option in SETUP is enabled but the choice has no effetc (WB is permanently enabled). Can be checked e.g. by CHKCPU.EXE

L1 WT mode

L1 WB mode

Have you done more tests? Any stability issues? Try running Blood in demo mode for a 15 minutes, this games always was picky about system stability.

Too bad you cannot easily switch to L2 WT mode while having L1 in WB mode. The performance would go up significantly. I have seen an option to do so in Phoenix BIOS for an ECS board with this chipset but haven't tried it yet.

I run only doom and quake timedemo yet. Quake is also sensitive e.g. on unstable OC.

Can you describe how did you jumpered your MB? We have the same HW rev. And it seems this MB doesn't support changing modes of L2 cache so I think I have the best settings. I will try OC at 160MHz and wait for 512k cache. I will finish article on my website soon...

Sorry I don't remember what jumpers I used to enable L1 WB mode, maybe try to check the manual I have uploaded to RetroWeb. Otherwise check datasheet for WB capable CPU and check what pins are needed for WB mode. Then you can trace the jumper in question with multimeter.

You could try BIOS for this motherboard to see if you can switch L2 cache mode. The board in question uses different (newer) I/O chip so I'm not sure it will work.

As I.wrote before I already tracked JP37 responsible for L1 wb/wt setting. Now L2 is in question. Can someone confirm that used UMC chipset is L2 wb capable for sure? There's no datasheet.

Btw you don't longer own this MB?

It is mentioned in the motherboard's manual that the chipset runs L2 cache in WB mode. I guess we have to believe it.

I sold the motherboard as it was too unstable for my tastes. But that was because I run WB CPU in it, which was unstable and slower than competition (especially SiS chipset) and because of the fact that I couldn't run L2 in WT. I think I would recommend the board only if you want to use WT CPU. It doesn't make a big difference anyway, in my tests the IntelDX4 WT is just about a few percent than the IntelDX4 WB, in single digits.

Aha, I didn't read manual thoroughly, as it's scanned cannot search in it, so I assume L2 WB support. As there's no option to switch L2 mode I couldn't do more here...

I think that something else went wrong just with your MB (bad solder joint/via/trace...) and it's not generic problem of this MB. You also mentioned some issues with long POST and HDD controller but mine is OK. Ididn't tried CF yet but 10GB PATA HDD works fine. I didn't face any stability problem with L1 WB yet and it gives me 3,5FPS bonus over WT in Doom (+8%) so it's not negligible speed up. I guess it works as it should...

If it's stable and you're happy with the performance increase, than I would call it a success.