ATA100/133 on BX boards

[dizzynoodle]

hello all,

would one benifit from using an ATA100 or ATA133 drive on either P2B or Asus P3B Mobos... i need to continue using mu existing machine while i make a new one [my cah flow hit real bad these 2 months!!]

So will i benifit from using ATA133 on the P2b or P3b-F mobo and which would be better to use ?

Also will REALLY benifit if i use a RAID controller on either of these boards ?

[cbman]

I personally think that it wouldn't make too much difference at all.

But what do I know

[Wombat]

ATA-33,66,133, etc only matter during the burst, when the data is coming from the HD's cache.
Even IBM says the gxp throughput is 20-25MB/s.

[dizzynoodle]

so i guess, i'm not really "losing out" on the 80GB maxtor 7200 rpm drive power by putting it on a BX board... so would i really then benifit by using a promise ATA100 card [with single drive not RAID setup] or not running on win2k ?

Also, would u say that there's a difference between the performance of the P2B or the P3B-F with the drives.. ?

[Gurm]

I've said it before and I'll say it again.

A single drive, even the fastest in the entire world, won't use the bandwidth of even an Ultra33 connection. In fact, most won't use the bandwidth of PIO4.

TWO drives, on the other hand, will burst at speeds that will flood the U33 bus - so keep 'em on separate channels.

If you intend to load up your channels, then yes U66 makes a difference. But that's it. U66->U100 is a pointless change. Absolutely, completely pointless. Anyone who thinks there is a difference is seeing driver optimization, not hardware speed changes.

- Gurm

[CHHAS]

The only scenario where you might be able to tell a difference is if you are using very large single files (huge bmp's or video files), most current IDE drives can transfer more than 33 MB/s, at least on the first part of the disk.

[The PIT]

One question I've got are they intending to increase drive speeds to 10k? This would help improve transfer rates although the drives would get hotter.

[fleabus2]

dizzynoodle:

I use a Promise IDE controller for my HDDs. I know that the speed advantage in doing so is minor but I’ll take anything to decrease the bottleneck. This isn’t an expensive uption.
To me, the main advantage of using a controller card is that it allows me to have my burner and DVD-ROM with channels of their own (no slaved drives) and avoid the conflicts that can come up with which drive on which channel and which is slaved to what etc to get DMA enabled properly. On my system each of my four drives has a channel to itself and DMA is enabled for all.
The only other thing I did was move my swapfile to a 1GB partition of its own. It is isolated by itself on the first partition [D:] of the HDD on the Promise’s Secondary Channel. It can fragment all it wants in there. Since I have 512MB system RAM, my swap file “in use” doesn’t get very active beyond what Win98SE needs for its own purposes.
Here are my HDTach reults and my current system:

HD Tach v2.61 (Unregistered demo)
SCAN >DEFRAG >RESTART >TEST.
==
Promise Ultra66 IDE0
30GB Maxtor +40 7200RPM ATA66 [UDMA4](C,E,F,G,H)
Model#: 53073U6
Random Access Time:10.3ms
Read Burst Speed: 56.5mbps
Read Speed Max.: 30385.0kps
Read Speed Min.: 23304.0kps
Read Speed Avg.: 29066.2kps
CPU Utilization: 4.7%
==
Promise Ultra66 IDE1
30GB Maxtor +40 7200RPM ATA66 [UDMA4](D,I,J,K,L,M)
Model#: 53073U6
Random Access Time:10.4ms
Read Burst Speed: 56.2ms
Read Speed Max.: 30381.0kps
Read Speed Min.: 24339.0kps
Read Speed Avg.: 29095.9kps
CPU Utilization: 4.8%
==
Win98SE clean
P3 800E slot 1
512MB PC133 @ 100MHz CAS2
Asus P3B-F i440BX

I’ll soon be building a new system (includes P4 2.0MHz; Asus P4T-E i850/478; 1GB RDRAM; 2x80GB Maxtor D740X ATA133 HDDs) and will use a Promise Ultra 133 IDE controller; not for the speed increase, from the above test one can see it is mostly a burst improvement, but for the convenience of avoiding that master/slave/DMA/where the burner should be hoo-hah.

Happy trails,

[Gurm]

CHHAS:

The fastest IBM SCSI drives right now can only transfer 25MB/sec. sustained. And those are 15k drives.

No drive on earth can do 33MB/sec. sustained - and CERTAINLY not any IDE drives, which are more in the 15MB/sec. sustained range.

- Gurm

[dizzynoodle]

I'm trying to install the new 80GB drive on my P3B-F system thru Win2k diskmanager... and i keep getting an error that the drive is too big !!! at 74.55GB...surely as one partition... what's happening here ? I'm using the quick format opition as i don't think that i need to do a full format for the new HDD...

whatcha think ?

[Jon P. Inghram]

My Western Digital WD400BB sustains 21 megabytes/sec over more than half the drive, and bottoms out at 17 megs/sec. This is on a 440BX UDMA/33 controller. Burst speed hits 27 megs/sec. It does just tad better if I use the CUBX's onboard UDMA/66 controller but not enough to make the system stability issues worth the hassle.

[Patrick]

Unstable?

Jon, what's so unstable about the onboard CMD controller on the CUBX? Install the latest bios for the CUBX and the CMD controller operates as an ATA/100 controller. My 40Gb IBM 60GXP has a burst speed of at least 80Mbs/sec on the CMD controller. The scale only goes as high as 80Mbs on the HD Tach test.

[Gurm]

Patrick - no drive bursts at 80MB/sec. Honestly. I've got a couple of those drives - they don't burst anywhere near that. It's the driver faking you out. Sorry, man.

- Gurm

[Marshmallowman]

going from udma 33 to 100 will give a noticable difference.
While no drive really saturates a 33 bus
the decrease in latency, and the fact the ide controller is only take 1/3 of the pci bus bandwidth at 100 than at 33 produces a noticable diference.

I noticed the diffrerence, perhaps not in pure throughput but access's that use the drive cache certainly benefit from the increased bandwidth.

In fact I found going from udm33 to 60 is quite noticable itself in normal windows operations, loading programs which typicaly can be cached quite well by the drives cache..etc

you can get quite large burst speeds over an ata100 channel if the data is comming largely from the drive cache's, since most decent drives have 2Mb cache(8mb for wd se's woohoo). Despite what some "experts" say.
a 60Mb/s burst over an ata100 channel is quite possible, if it is comming from the drives cache, 80- might be pushing it tho

one thing going for ata133 is that also has *much* larger size limit for your drive size(48 bit adressing)

But if you have a slow system, the effects will not be that noticeable (BX's...etc)

[Wombat]

Originally posted by Marshmallowman
going from udma 33 to 100 will give a noticable difference.
While no drive really saturates a 33 bus
the decrease in latency, and the fact the ide controller is only take 1/3 of the pci bus bandwidth at 100 than at 33 produces a noticable diference.


Not to sound like Gurm, but.....ummmm, NO. The IDE controller speed shouldn't change PCI utilization one bit. Where would this supposed decrease in latency come from?


I noticed the diffrerence, perhaps not in pure throughput but access's that use the drive cache certainly benefit from the increased bandwidth.

The cache is the <I>only</I> part of the drive that might benefit from something other than ATA-33.


In fact I found going from udm33 to 60 is quite noticable itself in normal windows operations, loading programs which typicaly can be cached quite well by the drives cache..etc

Sounds like you're faking yourself out.


one thing going for ata133 is that also has *much* larger size limit for your drive size(48 bit adressing)

And this affects just about nobody right now. Most people's problems right now are with the addressing scheme of their FS, not their HD.

<br>

In other news, I was just reading the data sheet for the new 120gxp series (supposedly shipping next month). Average seek/latency times are 8.5/4.17, and sustained data rate is listed as 48-23MB/s, compared to 40.8-20.9 for the 60gxp. Nice! Well, nice, if they don't fail like crazy. I don't think that anyone will see these numbers, but I'm glad that they went up, rather than down, with the 40GB platters.