KT7 ATA66 performance in 98SE?

[otheos]

I installed 98SE on my non-RAID KT7 on a Quantum LM10+ (ATA66) on the primary ATA66 (VIA) controller. this is the ONLY device attached (no CD/Floppy). I installed 4.25a +Viagart and tested with hdtach2.61. Max burst is 37MB/s (it was 59MB/s on the A7V the same HD). Any ideas??

[Peter M]

The burst performance of VIA IDE controllers can be screwed up bigtime by BIOS not setting it up right.

Check with your mainboard vendor if there are updates ... the issue has been discovered quite a while ago, so there should be.

Regards, Peter

[OPPAINTER]

Maybe you need to put a check by DMA. In device manager/disk drives/(your hd/settings.

[OPPAINTER]

Also when you put in 4in1 did you check off the box for DMA?

[otheos]

yes DMA is enabled....

[otheos]

OK, I've looked everywhere I could and didn't find anything! The burst speed will not go higher than 38MB/s no matter what I do.

I remember I had the EPoX G2 (SS7) which used the same 686A southgate and had the same problem untill Epox came with a new BIOS and fixed it. But this only came about after people complained (and epox listened). I doubt Abit will do the same....

[Peter M]

The Epox G2 uses the 596B south ... but that one has the same IDE controller part inside as does the 686A.

Epox improved IDE performance twice - first they fixed the handling of UDMA66 drives on 586B, 596, 596A UDMA33 controller to at least give good UDMA33 performance, and then much later they fixed the 596B UDMA66 controller init to give full speed.

Tried complaining to Abit yet? After all, the problem and the solution are long known, and at the retail price they charge, one could expect some service.

Regards, Peter

[otheos]

Thanks for the details Peter. However I doubt Abit provides support, they rather let their suppliers do so (like asus/gigabyte and unlike Epox).

Any chance we can nail this problem ourselves (with a little help from you perhaps)?

Thanks again, you're a good man.

[Peter M]

Since I don't have hands on an Epox G2 or G5 board, I can't investigate what exactly they changed in the IDE controller configuration registers from the oldest to the latest BIOS. That's what would be needed for reverse engineering that - use the original IDE-performance-sucks-totally BIOS, dump the entire PCI register set to a file, then use the intermediate well-it-does-UDMA33-properly-now BIOS and dump again, and then do the same with the latest all-is-fine-now BIOS. Then compare whatever relevant settings have changed (using VIA's data sheets), and then try to set exactly that manually on the Abit board.

That's how it could be done, but while I certainly have the skills and tools, I don't have the hardware.

Regards, Peter

[otheos]

Thanks Peter,

I'll take my chances and see what I can do. I'll let everyone know if I fix it.

I wonder though, the KT7 is such a popular m/b someone else should have noticed the same surely. Could it be my setup (before I go digging my G2 Bios)?

All input welcomed.

[otheos]

I got it!
In the SoftMENUIII the Enhanced Chip Performance setting when set to enabled limits the burst speed to 30MB/s. Disable that and back in business (58MB/s). I did not notice other system performance changes by setting it to Disable, I will report if I do.

[Amorpheous]

thanks for the info otheos. i too am running a kt7 and was wondering why my udma66 drive's performance wasn't up to par.

[Peter M]

erm, well, maybe that sets the PCI bus priorities to be too CPU centric, which would then hamper other PCI bus masters' performance.

Regards, Peter

[otheos]

I'll see to it Peter, thanks.
Extra bonus from disabling Enhanced Chip Performance is that Fast Writes are now enabled (a bug in UL bios does not allow both features at the same time).

[radman3d2]

Otheos, what program are you using to measure the performance? I have a ATA66 HD and MB but I have only measured it under Sandra (12005). My MB is a DFI K6BV3+/66. Thanks. Oh, and where can I get it?

[This message has been edited by radman3d2 (edited 11-21-2000).]

[otheos]

I use HDTACH 2.61. It's a shareware and without the license will only do read measurements. If you do a search in any search engine for it it will come up. www.tweakfiles.com (http://www.tweakfiles.com) will have it under benchmark too.