IDE channels vs performance - How to configure IDE devices?

[merlocka]

So, with my current PC (Asus A7V, 1GHz Athlon, 512MB PC133, WinXP) I was lazy and I have configured my IDE devices like :

Promise1 Master - n/a
Promise1 Slave - n/a
Promise2 Master - n/a
Promise2 Slave - n/a

VIA_IDE1 Master - System hard drive (Maxtor 40GB 5400RPM ATA100)
VIA_IDE1 Slave - Data hard drive (Maxtor 80GB 7200RPM ATA100)
VIA_IDE2 Master - CDRW (Lite-ON 24102B CDRW)
VIA_IDE2 Slave - DVD (Toshiba M1502)

I'm aware that a master and slave share the BW on a channel, so the ideal configuration would be to have only "master" devices and utilize the extra channels provided by the RAID controller on my mobo.

I've noticed no obvious performance improvements by switching to using the Promise controller for the hard drives (giving each device it's own channel). I figured that I would notice problems in CD-CD copying with both drives on the same channel, but I didn't really notice a difference (I didn't run any benchmarks or anything, just playing about with it).

Does anyone have any practical insight into this? Are there really any performance concerns with sharing "channels" between devices? I'm planning on upgrading my system and I'm trying to decide if I really need a Raid controller (just for the extra 2 channels) or if I should stick with my current (2 hard drive, 2 CD, onboard IDE) setup.

Any insight would be appreciated!

 

[thorin]

Basically with IDE it works like this. Only one drive can access the channel at a time so you can be reading/writting at the same time. If you go between drives on the same channel you get read-buffer-write if you go between 2 seperate channels with you get read-write. Noticible performance gains depend on system I/O, RAM, HD cache, etc, etc.... However seperate channels are definately the best bet.

Thorin

 

[merlocka]

Originally posted by: thorin
Basically with IDE it works like this. Only one drive can access the channel at a time so you can be reading/writting at the same time. If you go between drives on the same channel you get read-buffer-write if you go between 2 seperate channels with you get read-write. Noticible performance gains depend on system I/O, RAM, HD cache, etc, etc.... However seperate channels are definately the best bet.

Thorin

Thanks, I understand that the master and slave device cannot access the channel simultaniously. I guess my question is more specific, i.e. how much of a performance penalty...

So, I ran some ghetto benchmarks.

File size 250MB - copy from Slave to Master
same channel ~ 17 seconds
different channels ~ 15 seconds

File size 1700MB - copy from Slave to Master
same channel ~ 110 seconds
different channels ~ 90 seconds

considering i was timing it with the second hand on my watch, there is some uncertainty... but I expected more of a performance hit using both drives on the same channel.

Since these numbers indicate sustained throughput of about 16MB/s, ATA66/100 doesn't seem to be limiting it in either case. I imagine with faster drives the "different" channel test would begin to distance itself while the "same" channel test would hit a wall where it was bottlenecked by the channel.

 

[thorin]

The fastest drive on the market only pushes ~49.5MB/Sec at best, so you'll defintely never be in danger of saturating your ATA66/100 channels anytime soon.

What mode are your drives running? And do you have DMA enabled? (16MB/Sec is low).

Thorin

 

[Pariah]

"considering i was timing it with the second hand on my watch, there is some uncertainty... but I expected more of a performance hit using both drives on the same channel."

It won't really make that much of a difference. Yes, it is true that only one device per channel can be accessed at a time, but that does not mean you are limited to the throughput of one drive. If you have 2 drives capable of 30MB/s sustained, then you are not limited to 30MB/s when both are active, you can still potentially hit 60MB/s combined as has been exhibited with RAID 0 benchmarks using a single channel.

"The fastest drive on the market only pushes ~49.5MB/Sec at best, so you'll defintely never be in danger of saturating your ATA66/100 channels anytime soon."

ATA is up to around 57MB/s which is very close to if not already exceeding the real world limit of ATA66. Obviously if you have more than one drive on a channel, even ATA100 can potentially drop performance a bit, though it most likely won't be anything noticable or worth worrying about.

 

[merlocka]

Originally posted by: thorin
The fastest drive on the market only pushes ~49.5MB/Sec at best, so you'll defintely never be in danger of saturating your ATA66/100 channels anytime soon.

49.5MB/s sustained? That's pretty nice.

What mode are your drives running? And do you have DMA enabled? (16MB/Sec is low).

Thorin

Update - the first drive is a 40GB 5400RPM (not 7200)

They are running ATA66 (when connected to the VIA controller) and ATA100 when on the Promise controller (according to the BIOS). DMA is enabled.

Results from Sandra are 14290kB/s for the 40GB and 22816 for the 80GB. Those both seem low according to the Sandra comparisons. Hmmm. I wonder wasssup with this system.

 

[thomsbrain]

it doesn't make too much differnce with HD's, but optical drives get all weird when there's two on a channel accessing simulataneously.

 

[Viper96720]

If you don't access both drives simultaneously often I doubt the difference will be noticeable. Computer response probably be faster if you had your OS on the faster drive.