sync data transfer only affects SCSI devices... nothing happens with IDE drives
DMA mode is also called "bus mastering"... it allows the IDE controller to talk to the drive directly instead of using the CPU to relay data, speeding up data transfer by freeing up CPU cycles... Win95's DMA support only worked best with Intel chipsets. Some cheap CD-ROM drives did not function properly with it

Information needed:
What chipset does the motherboard use?
Are you using the latest chipset drivers?

MVP3... i guess its just all via chipset and/or driver's fault. no doubt dma support in those early chips are buggy, may or may not be solved with proper 4in1 driver versions.

Yes, EPOX MVP3C/2 :

♪ Uses VIA MVP3 AGPset.

♪ Latest BIOS version, but

♪ No additional drivers installed - using vanilla Win95c installation.

Disabling DMA makes the game run smoothly. (Only the movie sequences and music initaition are affected.)

More info :

never want to bother with early via boards again... at least not until the 694.
not only those chipsets are slow and buggy, but also boards from different brands can fuck up in different ways...

you will need to install the via 4in1 drivers... the older version... it is available on their site... do not use the latest one, it will slow the system down

default win95 does NOT come with any via chipset drivers and it will default to the generic bullshit drivers.... you should know better that old windows never really came with a comprehensive driver library... you always needed to download and install drivers for all the hardware

DMA support worked OK on VIA boards, you just need to install those drivers!

Thanks for the info! I'll download that.

TheMAN wrote:

you will need to install the via 4in1 drivers... the older version... it is available on their site... do not use the latest one, it will slow the system down

default win95 does NOT come with any via chipset drivers and it will default to the generic bullshit drivers.... you should know better that old windows never really came with a comprehensive driver library... you always needed to download and install drivers for all the hardware

DMA support worked OK on VIA boards, you just need to install those drivers!

you are right, they work with certain old versions of drivers but would crash with later ones, i forgot the exact version but i guess you should look into those released before 2001.
but odd bugs may still occur even if drivers are successfully installed, they are just no match for intel chipsets in both performance and stability. anyway, the mvp3 board above seems to have a 686a southbridge which is more modern and expected to be less buggy than the original 596b, so maybe it would work.

Actually, as I have posted above, there's nothing wrong with the board or it's association with games anymore (even without installing any chipset drivers). Problems cropped up when I enabled the DMA, so just wanted to know more about the DMA option which is available in the Device Manager section. And just wanted to know why enabling DMA caused hiccups, and when it's not, even when no drivers installed, everything's fine?

It's stable in all games I have tested.

Only the movies and music initiation had breaks when the DMA option is checked. Removing it does away with this problem.

TheMAN wrote:

DMA mode is also called "bus mastering"... it allows the IDE controller to talk to the drive directly instead of using the CPU to relay data, speeding up data transfer by freeing up CPU cycles... Win95's DMA support only worked best with Intel chipsets. Some cheap CD-ROM drives did not function properly with it

This is only half way correct, since PCI Bus Mastering capabilities on Intel chipsets were introduced with the PCI IDE ISA Xcelerator Southbridges (PIIX). Now if you look what chipset was out using a PIIX in August 1995 it boils down to the 430FX (Triton I) only with the very first revision of the PIIX allowing only identical transfer modes for master/slave.
But DMA support itself is available already in early IDE controllers for ISA systems. Usually with a more capable hard disk drive you get Multi Word DMA 2 support. It hasn't to be a Bus Mastering DMA at all and is also worth to be activated in Win95. (I know of no ISA IDE Controller using ISA Bus Mastering).
I would expect from the default Win95 IDE driver that it requires to have identical transfer modes for master/slave. Also check your transfer settings in BIOS and what is set really with the atbus tool (c't) before booting up Win95.

Edit: Here is written that Win95 waits for the transfer to end, even when using DMA: http://www.pcguide.com/ref/mbsys/buses/types/pciIDEBM-c.html

This can be a good explanation for the effect you see, since DMA transfer blocks might be larger that using PIO, such that it 'pauses' longer -> stuttering.

elianda wrote:

Edit: Here is written that Win95 waits for the transfer to end, even when using DMA: http://www.pcguide.com/ref/mbsys/buses/types/pciIDEBM-c.html

This can be a good explanation for the effect you see, since DMA transfer blocks might be larger that using PIO, such that it 'pauses' longer -> stuttering.

Ah, yes, this is starting to make sense to me, with the problem I had when I enabled the DMA option. Thanks, elianda.

The 686A southbridge on my ASUS K7M is probably the best chip onboard because the AMD 751 is shit for sure. Actually the 751 does remind me of VIA northbridges with its broken AGP and horrid memory performance. 😀

I haven't seen any weird DMA problems with it in 98SE or ME. Stock Windows drivers seem fine in those OSs.

elianda wrote:

This is only half way correct, since PCI Bus Mastering capabilities on Intel chipsets were introduced with the PCI IDE ISA Xcele […]
Show full quote

TheMAN wrote:

DMA mode is also called "bus mastering"... it allows the IDE controller to talk to the drive directly instead of using the CPU to relay data, speeding up data transfer by freeing up CPU cycles... Win95's DMA support only worked best with Intel chipsets. Some cheap CD-ROM drives did not function properly with it

This is only half way correct, since PCI Bus Mastering capabilities on Intel chipsets were introduced with the PCI IDE ISA Xcelerator Southbridges (PIIX). Now if you look what chipset was out using a PIIX in August 1995 it boils down to the 430FX (Triton I) only with the very first revision of the PIIX allowing only identical transfer modes for master/slave.
But DMA support itself is available already in early IDE controllers for ISA systems. Usually with a more capable hard disk drive you get Multi Word DMA 2 support. It hasn't to be a Bus Mastering DMA at all and is also worth to be activated in Win95. (I know of no ISA IDE Controller using ISA Bus Mastering).
I would expect from the default Win95 IDE driver that it requires to have identical transfer modes for master/slave. Also check your transfer settings in BIOS and what is set really with the atbus tool (c't) before booting up Win95.

Edit: Here is written that Win95 waits for the transfer to end, even when using DMA: http://www.pcguide.com/ref/mbsys/buses/types/pciIDEBM-c.html

This can be a good explanation for the effect you see, since DMA transfer blocks might be larger that using PIO, such that it 'pauses' longer -> stuttering.

say what you will, but I was making a generalization 😀
Those old ISA IDE bus mastering controllers weren't really supported by MS anyway, so anything using the "Standard IDE/ESDI controller" drivers will have the DMA checkbox grayed out... I'm sure with the installation of the specific 3rd party drivers, DMA mode will work

but the topic of discussion is the standard issue OSR2 drivers... MS's drivers really didn't support much other than the PIIX/PIIX3 controllers... remember, OSR2's venture into the market was limited, it was OEM only and MS designed OSR2 around major OEM's requirements (Dell, Compaq, HP, IBM, etc)... all of them were using intel chipsets (and motherboards), and being an intermediate product, it only made sense to support only what was really needed.... besides that, DMA mode in 1996 was still "experimental" and everyone was afraid of data loss/errors... it probably did occur in less than 1% situations, but the fear was blow out of proportion... things changed with Win98... first of all, it was found that DMA mode wasn't such a big deal after all, and it was 99% compatible with existing systems... second of all, drivers matured and further hardware support was added into Win98... DMA mode really worked properly out of the box with Win98 for the most part

as for different transfer speeds, it does work... I experienced this myself... I had an IDE toshiba cd-rom on the same controller as the zip drive... the zip drive only supported PIO3 and the toshiba supported DMA... enabling DMA mode allowed the CD-ROM to be in DMA while the zip drive stayed in PIO

Also, can I ask you why you have your K6-3 running at 66x5.5? You can run it at 66x6 by setting the multiplier setting on the motherboard at 2.0... the K6-3 internally links 2.0 to 6.0... with that, you can be running at 400mhz

and I believe there is a modified BIOS available for your board which provides better K6-3 support (I already flashed my Shuttle 569 with one and its already working better)
http://web.inter.nl.net/hcc/J.Steunebrink/k6plus.htm