Dead or no fan? 😀

There is a fan and it is working

Marcin1199 wrote on 2024-08-05, 10:08:

There is a fan and it is working

Is it stock fan or third-party?
If it third-party - Pentium Overdrive PODP5V83 - Detected as Pentium 33 and underperforming - Compaq Deskpro XE
There is photo for such modification

Fan is oryginal, I attach photo

Here is the reference to 486dx2-66. Do I good understand, that it is working on 18 Mhz?

Try to test it with SPEEDSYS 4.71 for pure DOS session

I attach results and bios setting too.

The memory speed is very slow, just like the Turbo switch is pressed. Are you using a GMB-486SPS motherboard? Where is its J5 Turbo Switch? According to the manual, 2-3 should be shorted to enable Turbo mode.

zwrr wrote on 2024-08-05, 12:40:

The memory speed is very slow, just like the Turbo switch is pressed. Are you using a GMB-486SPS motherboard? Where is its J5 Turbo Switch? According to the manual, 2-3 should be shorted to enable Turbo mode.

Yes, it is BMB-486SPS. According to the manual J5 should be open for P24-T, but I changed it to 2-3 and now it is a way better! Thank you for that advice!

However, benchamarks still looks similar. Any other ideas?

I made a mistake/typo above - my mainboard is GMB-486SPS. But anyway - in manual there is an information for P24T that J5 should be empty. I put jumer in 2-3, performance rises but after few resets it come back to the standard (slow), so I changed to 1-2 and the effect was the same. So I am not sure whether this jumper J5 gives the right effect for POD83. When performance rises I get 8 FPS in Quake 1. When it slows down, I have just 3 FPS.

Did this board really run “fine” with a 486dx2?

It sounds like you are having an intermittent problem which may be caused by damage, a loose component or coorsion.

I would still run the turbo switch issue to ground and ensure your integrated fan is making proper contact for the chip to detect and run 2.5x and not 1x
as a drop from 8->3 is exactly what would be expected if the cpu is thinking it has a bad fan and drops to 1x / 33mhz

Some aftermarket fans are too small to trip the fan detection circuit, other times a “marginal “ fan won’t get detected either.

There is a mod you can do to spoof the presence of a fan if your having issues

Likely not related to your issue but the p24t had a black eye because some 486 boards required an interposer to work and others had cache issues for one reason or another.

If you truly exhaust all options you should test the chip and memory benchmarks with the cache setup in various ways including fully disabled.
I would also put a switch or momentary switch to jp5 in case something with that jumper is marginal.

You may need to investigate for damage and deoxit or hit with a heat gun or soldering iron if anything is messed up

I haven't made a deep test on 486dx2, but as I remember, it worked rather fine.

I tried to enable and disable internal and L2 cache, but if it is disabled, it works even more bad. It is set on 128kb, internal (write-back) and L2 (write-thru). Is it ok?

Even if performance slows down, CPU-Z shows me 83Mhz so probably it is not a fan problem

Can you change l2 to write back and l1 to write through ? Both to write back?

Also the turbo jumper is meant to be possible to change on the fly, when it slows down change the turbo setting.

Another possibility is a bios issue but that may not be possible to update.

Lastly I would drop the amount of memory to 8 or 16mb, more than that can also be problematic

Would it damage the CPU to run without the fan attached, to check whether it would run slower or not without it - to establish if it's running 83 rather than 33?

There's something just so OFF about the memory access within that 16KB L1 cache. So very squiggly and weird.

The L2 cache, being 128KB, can only cache 32MB of RAM (this may depend on the set associativity and stuff, it's been a while since I've dealt with all that) so cache misses can occur quite frequently in your case. Cache misses will really reduce performance. Reduce the RAM to 32MB and see what happens.

Also of note is that while 84MB would *logically* be possible using a combination of 2x32MB, 1x16MB, and 1x4MB SIMMs, 84MB is not a supported configuration according to the manual.

Marcin1199 wrote on 2024-08-05, 09:22:

Hi, […]
Show full quote

Hi,

Maybe someone could help me, because I have no idea why it looks like that. My POD83 has almost such performace as 486 DX66 when I measure benchmark. Even if I compare it with the POD83 reference in CPU-Z, it looks much slower.

I have win98 SE and Voodoo 3 3000 PCI, Sound Blaster 16 ISA and 86 MB EDO RAM. Win98 has problems with smooth work.

I attached photos of CPU-Z.

Hi.
Oh my. 86MB of EDO RAM?
First of all, if you have really EDO RAM, then you will have problems. The chipset only supports FPM RAM. It is also 72pin, but different.
So if indeed there is some EDO RAM equipped, remove it.
You can make high res pictures of all your RAM modules, then we can tell you, what is EDO and what is FPM.
But 86MB???
With only 128kB cache that is beyond the cacheable area. Normally with L2 WT scheme and 128kB, you should have a maximum of 32MB of RAM, when you use L2 WB, then most likely only 16MB are cacheable by the L2.
So very likely you have the wrong RAM type and you use too much.
In the meantime, another Vogons user wrote the same about the amount of RAM.

Thank you for advices. I have made a few attepts with my RAM configuration (just 16, just 32, 2x32) and with WT/WB, but it haven't had impact to increase performance as expected. I attach foto of my RAM. Yes, it is 86 mb, because I wanted a lot to launch Win XP 😀 this was one of crucial reasons why I decided to buy POD83 😀 this experiment was the success! 😀

Do you think that RAM could have impact on CPU Speed benchmark?

It is very weird for me, that CPU-Z identifies my POD as 83 mhz, but in benchmark CPU Speed has such low level.

I attach one more foto