Not sure if it would make a difference, but it might.

Do you have L2 cache installed?
I couldn't get WB cache for the CPU working in my 486 build until I installed some L2 cache, then it worked fine.
Sounds strange, but these old motherboards can be very picky.

I have 256 kB of L2 cache installed so that shouldn't be a problem. It's a nice board but I have a hard time figure some things out, even with a manual. I.e. to this day I don't know how to enable 133MHz on the AMD CPU. I tried to set the jumpers to 486 DX2 mode without success, then I tried to reconfigure jumper settings one at a time and everything I have managed to do was to set the CPU to 120 MHz. Unfortunately I have yet to buy a multimeter which would be quite handy in this situation.

Is the L2 cache set to WB or WT?

I made some tests some time ago and came to the result that you get best speed if only one of the caches uses WB. So if the L2 cache is already WB, leave the L1 cache at WT - this will give you better speed.

That's odd, MB should post regardless how the CPU cache writing policy is set, maybe that jumper which causes your MB not to POST is responsible for a completely different setting? Remove CPU from the socket and with a multimeter, in continuity mode, try to trace to which CPU socket pin that jumper connects.

I'll probably need a multimeter to solve this mystery. The L2 cache in WB mode does produce different results, mainly positive, but there are a few instances when the PC is slower (i.e. Doom). What is better - to set L1 to WB or L2 to WB?

Paar wrote:

If set the jumper on the system boots fine but I'm stuck with WT cache (according to CHCKCPU tool) even though I set L1 cache to WB in BIOS.

Check if you have set the TURBO jumper correctly

Paar wrote:

I'll probably need a multimeter to solve this mystery. The L2 cache in WB mode does produce different results, mainly positive, but there are a few instances when the PC is slower (i.e. Doom). What is better - to set L1 to WB or L2 to WB?

From the four possible combinations, I've got best speed when L1 is WB and L2 is WT. Posted some info on this here:

AMD 5x86 X5-133 (now with POD)

If you can't get L1 to work in WB mode, set L1 to WT and L2 to WB.

sonicx wrote:

Check if you have set the TURBO jumper correctly

AFAIK there are only TURBO jumpers for TURBO switch and TURBO LED. Both are ok, at least TURBO works when using L1 in WT mode. The FSB is lowered to 15 MHz when active so the CPU's speed is ultimately 45 MHz.

derSammler wrote:

From the four possible combinations, I've got best speed when L1 is WB and L2 is WT. Posted some info on this here:

AMD 5x86 X5-133 (now with POD)

If you can't get L1 to work in WB mode, set L1 to WT and L2 to WB.

So far I'm stuck with the latter option.

Bump! Refloating this thread 😁 I have the same board and also the same behaviour. Enabling WB jumper in the board it refuses to post. While enabling WT, it posts.

Any thoughts?

foxzflow wrote on 2024-02-18, 10:23:

Bump! Refloating this thread 😁 I have the same board and also the same behaviour. Enabling WB jumper in the board it refuses to post. While enabling WT, it posts.

Any thoughts?

Maybe try a newer firmware https://theretroweb.com/motherboards/s/dfi-g4 … pa-rev.-0-rev.a

foxzflow wrote on 2024-02-18, 10:23:

Bump! Refloating this thread 😁 I have the same board and also the same behaviour. Enabling WB jumper in the board it refuses to post. While enabling WT, it posts.

Any thoughts?

Did you try this setting for i486DX4 &EW?

If no, then please try it.

CoffeeOne wrote on 2024-02-18, 11:03:

Did you try this setting for i486DX4 &EW? 486wb.png […]
Show full quote

foxzflow wrote on 2024-02-18, 10:23:

Bump! Refloating this thread 😁 I have the same board and also the same behaviour. Enabling WB jumper in the board it refuses to post. While enabling WT, it posts.

Any thoughts?

Did you try this setting for i486DX4 &EW?
486wb.png

If no, then please try it.

Tried but did not help 😔

zuldan wrote on 2024-02-18, 10:29:

foxzflow wrote on 2024-02-18, 10:23:

Bump! Refloating this thread 😁 I have the same board and also the same behaviour. Enabling WB jumper in the board it refuses to post. While enabling WT, it posts.

Any thoughts?

Maybe try a newer firmware https://theretroweb.com/motherboards/s/dfi-g4 … pa-rev.-0-rev.a

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB!

Now I will have to dig a little bit about L1 and L2 cache settings (WB /WT) to understand which one gives better performance.

The board curently has 256kb cache installed and also 32Mb RAM.

Any suggestion or recommendation? 😜

256KB is enough for 64MB WT or 32MB WB in TAG7+1 mode. WB is better.
you can test both in cachechk

foxzflow wrote on 2024-02-18, 23:27:

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB! […]
Show full quote

zuldan wrote on 2024-02-18, 10:29:

foxzflow wrote on 2024-02-18, 10:23:

Bump! Refloating this thread 😁 I have the same board and also the same behaviour. Enabling WB jumper in the board it refuses to post. While enabling WT, it posts.

Any thoughts?

Maybe try a newer firmware https://theretroweb.com/motherboards/s/dfi-g4 … pa-rev.-0-rev.a

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB!

Now I will have to dig a little bit about L1 and L2 cache settings (WB /WT) to understand which one gives better performance.

The board curently has 256kb cache installed and also 32Mb RAM.

Any suggestion or recommendation? 😜

I fully agree with rasz_pl's answer.
But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT)
So concentrate first on L1 WB.
Is the board really working in L1 WB? Please note that it is not relevant want you set in Bios for L1, it is done via jumpers. So make a screenshot of the output of speedsys.
Next step, when it is verified that L1 WB is active, stability test your system. Also important is to test, if writing to a floppy disk still works reliably.

CoffeeOne wrote on 2024-02-19, 20:10:

I fully agree with rasz_pl's answer. But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT) So […]
Show full quote

foxzflow wrote on 2024-02-18, 23:27:

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB! […]
Show full quote

zuldan wrote on 2024-02-18, 10:29:

Maybe try a newer firmware https://theretroweb.com/motherboards/s/dfi-g4 … pa-rev.-0-rev.a

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB!

Now I will have to dig a little bit about L1 and L2 cache settings (WB /WT) to understand which one gives better performance.

The board curently has 256kb cache installed and also 32Mb RAM.

Any suggestion or recommendation? 😜

I fully agree with rasz_pl's answer.
But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT)
So concentrate first on L1 WB.
Is the board really working in L1 WB? Please note that it is not relevant want you set in Bios for L1, it is done via jumpers. So make a screenshot of the output of speedsys.
Next step, when it is verified that L1 WB is active, stability test your system. Also important is to test, if writing to a floppy disk still works reliable way.

Great, thanks for sharing all your thoughts! Really appreciate it.

I just ran CHKCPU and it seems that CPU is correcly identified with WB enabled.

I am also running several test using speedsys and CACHECHK, confirming L1 cache and different performance settings related to RAM speed, external cache WT vs WB, etc.

Here some of my last screenshots after some testing.

One question. Why writing to floppy drive is a good / reliable for stability test?

Thanks!

Floppy uses DMA so if there's any glitch in memory or cache subsystems it will have errors.

foxzflow wrote on 2024-02-19, 22:18:

Great, thanks for sharing all your thoughts! Really appreciate it. […]
Show full quote

CoffeeOne wrote on 2024-02-19, 20:10:

I fully agree with rasz_pl's answer. But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT) So […]
Show full quote

foxzflow wrote on 2024-02-18, 23:27:

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB! […]
Show full quote

Thanks, I just updated the BIOS to the latest one (E5), and now it correctly detects 486DX4 in WB!

Now I will have to dig a little bit about L1 and L2 cache settings (WB /WT) to understand which one gives better performance.

The board curently has 256kb cache installed and also 32Mb RAM.

Any suggestion or recommendation? 😜

I fully agree with rasz_pl's answer.
But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT)
So concentrate first on L1 WB.
Is the board really working in L1 WB? Please note that it is not relevant want you set in Bios for L1, it is done via jumpers. So make a screenshot of the output of speedsys.
Next step, when it is verified that L1 WB is active, stability test your system. Also important is to test, if writing to a floppy disk still works reliable way.

Great, thanks for sharing all your thoughts! Really appreciate it.

I just ran CHKCPU and it seems that CPU is correcly identified with WB enabled.

I am also running several test using speedsys and CACHECHK, confirming L1 cache and different performance settings related to RAM speed, external cache WT vs WB, etc.

Here some of my last screenshots after some testing.

IMG_20240219_231057.jpg

IMG_20240219_231017.jpg

One question. Why writing to floppy drive is a good / reliable for stability test?

Thanks!

Your L2 cache and memory values are really bad. Did you try faster settings in the Bios?

CoffeeOne wrote on 2024-02-19, 22:29:

foxzflow wrote on 2024-02-19, 22:18:

Great, thanks for sharing all your thoughts! Really appreciate it. […]
Show full quote

CoffeeOne wrote on 2024-02-19, 20:10:

I fully agree with rasz_pl's answer. But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT) So […]
Show full quote

I fully agree with rasz_pl's answer.
But still L1 WB brings more additional performance than the L2 cache strategy (WB vs WT)
So concentrate first on L1 WB.
Is the board really working in L1 WB? Please note that it is not relevant want you set in Bios for L1, it is done via jumpers. So make a screenshot of the output of speedsys.
Next step, when it is verified that L1 WB is active, stability test your system. Also important is to test, if writing to a floppy disk still works reliable way.

Great, thanks for sharing all your thoughts! Really appreciate it.

I just ran CHKCPU and it seems that CPU is correcly identified with WB enabled.

I am also running several test using speedsys and CACHECHK, confirming L1 cache and different performance settings related to RAM speed, external cache WT vs WB, etc.

Here some of my last screenshots after some testing.

IMG_20240219_231057.jpg

IMG_20240219_231017.jpg

One question. Why writing to floppy drive is a good / reliable for stability test?

Thanks!

Your L2 cache and memory values are really bad. Did you try faster settings in the Bios?

Sorry to not post before, I have been dealing with a I/O write problem on the CF card that I was using. I just replaced it and now it is fine.

Going back to the cache settings, I can not get better results than the ones published above 🙁
Also, I found another great post in the forum about 486 benchmarks, and are not so better than my results.

Anyway, here are the available settings from the BIOS motherboard.

Any suggestions?

Thanks!