Does speedsys say that it is running at correct frequency?

Are you sure your cache is working?

Sphere478 wrote on 2022-11-27, 21:55:

Does speedsys say that it is running at correct frequency?

Are you sure your cache is working?

Yes,is working at 75mhz

About cache, i tested with cachechk and detect the 128kb

I checked ram and cache speed and look normal, L1 cache faster than L2 and L2 than ram

Perhaps it isn’t running in write back mode?

Maybe FSB differences are to blame

If you only used half your cache sockets that could also lead to slower performance sometimes

About fsb 32x2 for dx2, 25x3 for overdrive and no idea the toshiba laptop

But even with lover fsb, 3x25 i believe must be faster then 2x33

But i never use a overdrive CPU i ask because no idea if this cpus maybe have some handicap etc

About WB WT etc no idea,PS/2 bios is very limited with options

But i tested diferent cache modes in dx2 board and no difference or marginal

Sphere478 wrote on 2022-11-27, 22:09:

If you only used half your cache sockets that could also lead to slower performance sometimes

I read on internet that 128<>256 difference is marginal in performance

sadly i cant test 256 i only have 128

I checked benchmark from people but no info from overdrive i find at least

I don’t own one of those CPUs myself but from what I can see I would expect the performance to be similar (exact?) to a regular DX4 in the same motherboard

Obviously you cannot run a DX4 in a five volt motherboard though Without an adapter

See if you can set it to 33mhz x2 heck 33x3 may even work… 100mhz

The overdrive and normal dx4s came in 100mhz flavors.

theelf wrote on 2022-11-27, 22:09:

About fsb 32x2 for dx2, 25x3 for overdrive and no idea the toshiba laptop

But even with lover fsb, 3x25 i believe must be faster then 2x33

Agree, try to overclock the dx4 to 100mhz where it will shine.

I have commented on this in the past.

Your benchmarks match my experience.

Intel recommended OEMs not use the DX4-75 in new desktop systems for a reason and that reason is that although a dx4-75 is a good upgrade for an obsolete system that lack a 33mhz FSB bringing said system up to dx2-66 performance .
The 33% difference in FSB and video/hd bus speed more than makes up for the 11% gain in clock.
Laptops commonly used the 75mhz chip purely to save on power and systems designed for the dx4-75 were usually optimized to make best use of the limited bus, even in ideal circumstances the dx4-75 was not an upgrade over a dx2-66

Intel had proposed all the way back in 1991
486dx2
486dx3
486dx4
Monikers alongside a proposed 30mhz FSB

Desktop OEMs were loath to adopt the strange 30mhz FSB so the 75mhz DX3 died on the vine, had this not occurred newer systems could have gained a small boost from a 75mhz dx3

rmay635703 wrote on 2022-11-29, 05:49:

Agree, try to overclock the dx4 to 100mhz where it will shine. […]
Show full quote

theelf wrote on 2022-11-27, 22:09:

About fsb 32x2 for dx2, 25x3 for overdrive and no idea the toshiba laptop

But even with lover fsb, 3x25 i believe must be faster then 2x33

Agree, try to overclock the dx4 to 100mhz where it will shine.

I have commented on this in the past.

Your benchmarks match my experience.

Intel recommended OEMs not use the DX4-75 in new desktop systems for a reason and that reason is that although a dx4-75 is a good upgrade for an obsolete system that lack a 33mhz FSB bringing said system up to dx2-66 performance .
The 33% difference in FSB and video/hd bus speed more than makes up for the 11% gain in clock.
Laptops commonly used the 75mhz chip purely to save on power and systems designed for the dx4-75 were usually optimized to make best use of the limited bus, even in ideal circumstances the dx4-75 was not an upgrade over a dx2-66

Intel had proposed all the way back in 1991
486dx2
486dx3
486dx4
Monikers alongside a proposed 30mhz FSB

Desktop OEMs were loath to adopt the strange 30mhz FSB so the 75mhz DX3 died on the vine, had this not occurred newer systems could have gained a small boost from a 75mhz dx3

Hi! yes, after many test, im almost sure the problem is the FSB

I force a intel DX4-100 to 3x25 in another desktop PC I have, and perform similar. I REALLY did not believe a 3x25 will perform almost same than 2x33

Sadly the PS/1 is locket at 25mhz bus, at least I cant see any jumper for change to 33mhz fsb, i will need to accept a DX2 performance and enjoy the IBM in this way!

Post pic of your clock gen. Maybe hidden settings

Have you compared to the benchmark results here? Honestly doesnt seem too far off from what I'd expect, maybe im wrong: https://www.complang.tuwien.ac.at/misc/doombench.html

My 486 dx4 @ 120 gets about 1660 realtks after bios tweaking etc. But obviously results can widely vary due to innumerable hardware combinations, how the weather is that day, etc. Which is why HW from this period is so great 😀

Also doom is not the all be all for benchmarking. What do other tools tell you about various subsystems? Screenshot from speedsys for example

drosse1meyer wrote on 2022-11-30, 23:14:

Have you compared to the benchmark results here? Honestly doesnt seem too far off from what I'd expect, maybe im wrong: https://www.complang.tuwien.ac.at/misc/doombench.html

My 486 dx4 @ 120 gets about 1660 realtks after bios tweaking etc. But obviously results can widely vary due to innumerable hardware combinations, how the weather is that day, etc. Which is why HW from this period is so great 😀

Also doom is not the all be all for benchmarking. What do other tools tell you about various subsystems? Screenshot from speedsys for example

Yes, I already compare with this doom benchmarks and 3dbench too

About doom, i get 2498 realtiks more in the line of a DX266

speedsys hangs when detect hdd, i think did not like the 120GB HDD im using right now

Sphere478 wrote on 2022-11-30, 22:09:

Post pic of your clock gen. Maybe hidden settings

Tomorrow i will do, but i will not touch this machine at all, is a very nice computer, i will not change crystal etc. Just what IBM decide in jumpers+bios

theelf wrote on 2022-11-30, 23:45:

Yes, I already compare with this doom benchmarks and 3dbench too […]
Show full quote

drosse1meyer wrote on 2022-11-30, 23:14:

Have you compared to the benchmark results here? Honestly doesnt seem too far off from what I'd expect, maybe im wrong: https://www.complang.tuwien.ac.at/misc/doombench.html

My 486 dx4 @ 120 gets about 1660 realtks after bios tweaking etc. But obviously results can widely vary due to innumerable hardware combinations, how the weather is that day, etc. Which is why HW from this period is so great 😀

Also doom is not the all be all for benchmarking. What do other tools tell you about various subsystems? Screenshot from speedsys for example

Yes, I already compare with this doom benchmarks and 3dbench too

About doom, i get 2498 realtiks more in the line of a DX266

speedsys hangs when detect hdd, i think did not like the 120GB HDD im using right now

Sphere478 wrote on 2022-11-30, 22:09:

Post pic of your clock gen. Maybe hidden settings

Tomorrow i will do, but i will not touch this machine at all, is a very nice computer, i will not change crystal etc. Just what IBM decide in jumpers+bios

Confused... in the first post you said you're getting 38 FPS which is about 1950 realtks... 2498 realtks is ~ 30 fps

drosse1meyer wrote on 2022-11-30, 23:57:

theelf wrote on 2022-11-30, 23:45:

Yes, I already compare with this doom benchmarks and 3dbench too […]
Show full quote

drosse1meyer wrote on 2022-11-30, 23:14:

Have you compared to the benchmark results here? Honestly doesnt seem too far off from what I'd expect, maybe im wrong: https://www.complang.tuwien.ac.at/misc/doombench.html

My 486 dx4 @ 120 gets about 1660 realtks after bios tweaking etc. But obviously results can widely vary due to innumerable hardware combinations, how the weather is that day, etc. Which is why HW from this period is so great 😀

Also doom is not the all be all for benchmarking. What do other tools tell you about various subsystems? Screenshot from speedsys for example

Yes, I already compare with this doom benchmarks and 3dbench too

About doom, i get 2498 realtiks more in the line of a DX266

speedsys hangs when detect hdd, i think did not like the 120GB HDD im using right now

Sphere478 wrote on 2022-11-30, 22:09:

Post pic of your clock gen. Maybe hidden settings

Tomorrow i will do, but i will not touch this machine at all, is a very nice computer, i will not change crystal etc. Just what IBM decide in jumpers+bios

Confused... in the first post you said you're getting 38 FPS which is about 1950 realtks... 2498 realtks is ~ 30 fps

Is fastdoom not regular doom. For the benchmark you ask i need to use regular 1.9c doom since this chart is using this version

I believe now must be the new standar since regular doom is obsolete

theelf wrote on 2022-12-01, 00:14:

drosse1meyer wrote on 2022-11-30, 23:57:

theelf wrote on 2022-11-30, 23:45:

Yes, I already compare with this doom benchmarks and 3dbench too […]
Show full quote

Yes, I already compare with this doom benchmarks and 3dbench too

About doom, i get 2498 realtiks more in the line of a DX266

speedsys hangs when detect hdd, i think did not like the 120GB HDD im using right now

Tomorrow i will do, but i will not touch this machine at all, is a very nice computer, i will not change crystal etc. Just what IBM decide in jumpers+bios

Confused... in the first post you said you're getting 38 FPS which is about 1950 realtks... 2498 realtks is ~ 30 fps

Is fastdoom not regular doom

I believe now must be the new standar since regular doom is obsolete

Ah OK. IDK then, I've always just used Doom High in Phil's Dosbench suite.