Hello! That has something to do with your hard disk cache. Because, when the data requested by the PC happens to be in the internal cache (typically 512k-2megs), it can be transferrred to the mainboard at full bus speed rate! That's an improvement of UDMA! And, the interfaces are always a little bit faster, then the disks are getting faster, and then they make a new interface standard again. That's how it always is, because, it is better to not to have problems with a slow interface.
So, either your benchmark doesn't manage to switch off the cache for its measure or it only shows you the highest bus speed rate. But, 33MB/66MB is definitely not a data rate any recent IDE HDD can achieve. Kind Regards, Stephan Hachinger ----- Original Message ----- From: Vachirasuk <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Cc: <debian-user@lists.debian.org> Sent: Sunday, March 19, 2000 3:36 PM Subject: Re: no merit from using DMA-66? > Hi, > > I think I am trying to measure data transfer rate from harddisk. I > have read UDMA-mini-HOWTO on LDP site and it also said "UDMA drives > will give you between 10 and 15 MB/s using UDMA mode 2 (33 MB/s) or 4 > (66 MB/s) enabled". If the DTR never gets higher than 33MB/sec then > why use DMA-66, since there is no performance improvement? How come > benchmark under Win98 gave me DTR of DMA-66 drive 2 times of normal > DMA-33 drive? Am I misunderstanding something here? > > Confused > > Vachi > > ----- > Vachirasuk Setalaphruk > ISE, Osaka University > > From: "Stephan Hachinger" <[EMAIL PROTECTED]> > Subject: Re: no merit from using DMA-66? > Date: Sat, 18 Mar 2000 21:35:53 +0100 > > > Hello! > > > > Are you really sure you're measuring the maximal bus speed? > > > > Because the bus speed (UDMA33=33MB/sec; UDMA66=66MB/sec) is only the maximum > > data rate which can be transferred. However, the magnetic disk can never be > > that fast, so that data rates of 16MB/sec are quite normal and not bad. > > 25MB/sec must be a very good hard disk. So, just don't worry, this high data > > rates can only be reached if the data comes from the HDD cache over the bus. > > > > Kind Regards, > > > > Stephan Hachinger > > >