Anyone can tell me why BIOS shows HDD 32-bit mode off?

stf92 · 09-16-2011, 08:02 PM

Hi: What's said.

jefro · 09-16-2011, 08:35 PM

I'll ask a dumb question. Don't you want 48 bit access?

stf92 · 09-16-2011, 09:09 PM

Quite a novice in these things. If 48 bits is possible I'd like it. Can smartctl tell me anything about the bus width?

cascade9 · 09-16-2011, 10:25 PM

I think you are thinking of LBA 48-bit jefro, and it isnt 48-bit 'access', its just 'adressing'.

Quote:

48-bit Logical Block Addressing (LBA) is a technology which extends the capacity of IDE ATA/ATAPI devices beyond a previous limit of 137.4 GB. This limit applies to IDE ATA/ATAPI devices only and not to SCSI interface devices. The original design specification for the ATA interface only provided 28-bits with which to address the devices. This meant that a hard disk could only have a maximum of 268,435,456 sectors of 512 bytes of data thus limiting the ATA interface to a maximum of 137.4 gigabytes. With 48-bit addressing the limit is 144 petabytes (144,000,000 gigabytes).

http://www.48bitlba.com/

I have no idea why your BIOS says 32-bit mode is off stf92. To be honest, I'm not sure I recall ever seeing anything about 32-bit for HDDs in any BIOS I've played with.

You have tried setting it to 32-bit, right?

*edit- while I remember, I should say I have heard of 16/32-bit transfers for HDDs. I just dont think I've sen it myself (if I did it was a long, long time ago)

Quote:

16-Bit and 32-Bit Access

One of the options on some chipsets and BIOSes is so-called 32-bit access or 32-bit transfers. In fact, the IDE/ATA interface always does transfers 16 bits at a time, reflecting its name ("AT attachment"--the original AT used a 16-bit data bus and a 16-bit ISA I/O bus). For this reason, the name "32-bit" access or transfer is somewhat of a misnomer.

Since modern PCs use 32-bit I/O buses such as the PCI bus, doing 16-bit transfers is a waste of half of the potential bandwidth of the bus. Enabling 32-bit access in the BIOS (if available) causes the PCI hard disk interface controller to bundle together two 16-bit chunks of data from the drive into a 32-bit group, which is then transmitted to the processor or memory. This results in a small performance increase.

Note: Some BIOSes (or add-in controller cards) may automatically and permanently enable this feature, and therefore not bother to mention it in the BIOS setup program.

http://www.pcguide.com/ref/hdd/if/id...cc32Bit-c.html

stf92 · 09-17-2011, 12:06 AM

Perhaps the footnote on the last quote applies to my case, for I do not have any BIOS option to change bus width (16/32 bits). Thanks for the links. It seems the second link addresses all the disk parameters the BIOS shows. For this machine and disk, I have

Code:

WPcom          0
LBA            On
Blk mode       On
PIO mode       4
32-bit mode    Off

The disk is a 10GB Western Digital. A question: are not disk data transfers made by DMA? And if so, why PIO?

business_kid · 09-17-2011, 03:43 AM

hdparm -I /dev/sda might answer your question.

It could also be the wrong driver for your chipset. A favourite error is to have generic pci enabled as well as your own chipset. But generic loads first. So then, generic has the chipset, and drops back to 1st gear if your chipset is not generic.

I compile in my own m/b chipset into the kernel, and do not compile generic support into my kernels. You may be able to get around it with some kernel option. I don't know.

stf92 · 09-17-2011, 09:59 AM

Thanks a lot.

cascade9 · 09-18-2011, 03:51 AM

Quote:

Originally Posted by stf92

Perhaps the footnote on the last quote applies to my case, for I do not have any BIOS option to change bus width (16/32 bits). Thanks for the links. It seems the second link addresses all the disk parameters the BIOS shows. For this machine and disk, I have

Code:

WPcom          0
LBA            On
Blk mode       On
PIO mode       4
32-bit mode    Off

The disk is a 10GB Western Digital. A question: are not disk data transfers made by DMA? And if so, why PIO?

Modern HDD transfers are DMA, PIO is older. If I recall correctly you could be forced back into PIO mode from comptibility problems, mainly between the BIOS or controller and HDD or between HDDs and/or optical drives on one channel (eg, 1 x HDD 'A' on PATA channel '0' will run at UDMA 2 (Ultra ATA/33), if you install HDD 'B' on PATA channel '0' as well, then both HDDs will drop back to PIO4).

If that info in [code] tags is from a linux command readout, it could be BIOS comptibility problems or software (like business_kid suggested). If its from the BIOS info, its probably comptibility problems and nothing to do with drivers.

BTW, there are is also 'Single/Multiword DMA' which is between PIO and UDMA. It can apear in BIOS or from the OS as a PIO drive. I dont think that is what is happening here, all the 10GB IDE drives I know of were at least UDMA 2.

jefro · 09-18-2011, 12:35 PM

It is a very old standard. http://en.wikipedia.org/wiki/32-bit_disk_access It is for those old OS's. I'd guess it could still be relevant to very old kernels. Maybe zipslack under dos.

Since you have lba enabled and you are in pio 4 mode, don't worry about the 32 bit mode, you don't need it.

stf92 · 09-19-2011, 12:13 AM

Thanks guys. All very interesting information.

Quote:

Since you have lba enabled and you are in pio 4 mode, don't worry about the 32 bit mode, you don't need it.

And thanks for this too.

business_kid · 09-19-2011, 02:03 AM

PIO 4 runs at about 3 MB/S, that's the only hitch. It is a fallback standard, available since the 1980s. By the 1990s it was gone, and retained only as a disaster situation. Test using

hdparm -tT /dev/sdx

I could get 40MB/s with crappy 486 or 586 cpus in the 90s. Chipset driver is the no. #1 suspect. No 10GB disk needs PIO 4 - my guess is ata 3 or 4 at least.

stf92 · 09-19-2011, 02:46 AM

I ran hdparm -tT:

Timing buffer cache reads: 128MB in 3.66s = 34.97 MB/s
Timing buffered disk reads: 64 MB in 4.07s = 15.72 MB/s

I also ran hdparm -I: It says disk supports ATA 2,3,4,5. Likely used: 6.

cascade9 · 09-19-2011, 07:14 PM

15.72MB/sec would be typical of UDMA33/ATA-4 or maybe UDMA66/ATA-5 (depending on which benchmark you are using).

Its a bit slow for a UDMA100/ATA-6.

stf92 · 09-19-2011, 09:08 PM

Then it looks like I shouldn't care. But I think business_kid made a point here. How is it possible that CPU is in charge of disk I/O?

business_kid · 09-20-2011, 02:15 AM

Quote:

Originally Posted by stf92

Then it looks like I shouldn't care. But I think business_kid made a point here. How is it possible that CPU is in charge of disk I/O?

That wasn't what I _meant_ to say. I meant to say the motherboard chipset controls the hard disk at one end, and that giving it the right driver improves the speed.

IMHO, 15MB/s is crap. PIO 4 is CRAP. If the board is 486 or greater, it should be capable of twice 15MB/S at least. The famous Via chipset with the so-called 'hardware problem' did 42MB/s
and I bought one of those in 1998, IIRC. That was not the best/fastest out then. These all ran on 33mhz pci and you probably have at least 66mhz.

PIO4 is saying "I can't talk to this hard disk, so let's use sign language." Motherboard chipset.
Motherboard chipset. . . .