See if you can replicate the performance and stability problems at 40 or 50 MHz FSB. Your issues with HOT-433 mimic my experience in general. HOT-433, 4DPS, and M919 are all flakey like this.

feipoa wrote on 2023-03-01, 12:52:

See if you can replicate the performance and stability problems at 40 or 50 MHz FSB. Your issues with HOT-433 mimic my experience in general. HOT-433, 4DPS, and M919 are all flakey like this.

Testing at lower ("normal") FSB-frequencies was my idea, too. I will do this a bit later - the next days I will test the ECS-board, if I have time for it...

Both SiS-boards I have are until now really stable at the conditions shown above, also they "feel" in Windows faster than the HOT-433.

BTW, my best 3DMARK99-score with the HOT-433 + GF2MX was 411 (3D) and 469 (CPU). Sadly, almost every test @ 180MHz scores different at a wide range of about 30-50 points below this (= very flakey), even the CPU-results in Sandra98 remain (more or less) stable.

I don't understand why 3dmark99max works with D3D but D3D games do not.

this was my brief experience with 486+gf2 as well - ok in DOS, very flaky in windows accelerated 3D graphics APIs.

pshipkov wrote on 2023-03-02, 02:28:

this was my brief experience with 486+gf2 as well - ok in DOS, very flaky in windows accelerated 3D graphics APIs.

I am affraid, at least the HOT (maybe better: COLD)-433 is unstable not only with the GF2, but with other VGAs, too (tested with a Matrox Mystique 220 today once again @ FSB 60 -> freeze of the rotating cube of the Dx6.1-interface in 1-2 seconds).

Could the instabilities be due to power delivery of these newer GPUs?

chrismeyer6 wrote on 2023-03-02, 14:19:

Could the instabilities be due to power delivery of these newer GPUs?

The Matrox Mystique is not a very new or fast GPU (even it was more common on Pentium-based boards).
Not sure, but maybe the HOT-433 has problems using EDO-RAM (at least at FSB > 33/40 MHz).
It will cost more of time to find this out...
The massive drop of the CPU-frequency is absolutely strange and scarry - it seems like a "throttling" of a Pentium 4 for 486...

EDIT: am I right (or wrong), that the lower CPU-frequency is caused by a drop of the FSB?

gonzo wrote on 2023-03-02, 19:45:

EDIT: am I right (or wrong), that the lower CPU-frequency is caused by a drop of the FSB?

We can't say. There are many ways to effectively slow down a 486 system. Dropping the FSB is possible, but it needs to be done with care. A processor like the 5x86 has a "clock multiplier" that generates the 133/160 MHz internal clock from the 33/40 MHz FSB clock. This clock multiplier circuit can't cope with random changes of the FSB frequency, the FSB frequency needs to be adjusted slowly.

Modern 486 clock generator chips can slowly adjust the FSB clock, and the is used by many late 486 "green" motherboards, when power management considers the board "idle". Try disabling power management in the setup to prevent sudden drops of the FSB.

Other ways to slow down a 486 system, that are commonly used to implement a "de-turbo mode": L2 cache can be disabled or the front-side bus might be blocked (using HOLD) for regular time intervals.

gonzo wrote on 2023-02-08, 18:22:

Fortunately, the ZIDA works great at FSB 60 MHz (FSB:PCI = 1:2)

How did you manage to enable 1:2 divider? JP18? Does this mode work for 50fsb? In my case no at any fsb.

pshipkov wrote on 2023-03-02, 02:28:

this was my brief experience with 486+gf2 as well - ok in DOS, very flaky in windows accelerated 3D graphics APIs.

pshipkov, I just made an upload of pictures for another 486-system @ 200 MHz and a GF 2 MX-400, yo can take a look here: Re: 3 (+3 more) retro battle stations

feipoa wrote on 2023-03-01, 12:52:

See if you can replicate the performance and stability problems at 40 or 50 MHz FSB. Your issues with HOT-433 mimic my experience in general. HOT-433, 4DPS, and M919 are all flakey like this.

feipoa, I just made an upload of pictures for another 486-system without stabilty-problems in Windows @ 200 MHz and a GF 2 MX-400, you can take a look here: Re: 3 (+3 more) retro battle stations

overdrive333 wrote on 2024-02-16, 10:37:

gonzo wrote on 2023-02-08, 18:22:

Fortunately, the ZIDA works great at FSB 60 MHz (FSB:PCI = 1:2)

How did you manage to enable 1:2 divider? JP18? Does this mode work for 50fsb? In my case no at any fsb.

Sorry for my late answer, I am not very active in the last time.
About your question: yes, it is JP18, as there is no BIOS-option.
In addition to this, for an FSB > 33 MHz you MUST set in advance (still under 33 MHz) the BIOS-option "Cache Write Cycle" = 3, and the options "L2 Cache Tag Bits: 8 bits" as well "L2 Cache Policy: Write Thru"
At FSB 60 MHz, you should use only 32 MB of RAM, and in best case only one module.
Good luck.