huh? please dont tell me they make "rounded" usb cables.. thats about as lame as makin a "rounded" network cable =)

I was referring to the size of the cable, like how big a rounded SATA cable would be

i've seen pics of a serial ata cable from comdex or something, anyways it's a real thin wire, like about the size of 2 pair 18 gauge speaker wire.

on that strange hdd suggestion by webmasta33, if you just made a hdd with 2 sets of heads you could have them be like raid0 for some (long read) operations, and also have multi-seeks for short operations so that they could both access the entire drive just have 2 of them doing it

That was I was saying... I guess I just didn't word it well.

yeah, i think it would be possible if you put one arm in the front of the drive and one in the back, and made the drive longer than normal. that would rule.

see multiple heads on the same platter is pointless. The way data is written to the platters is already pretty much as efficient as it can get. Spinning at even 5400 rpm shows that the effectiveness of 2 heads trying to read the same sectors on a platter wouldn't work. at 7200 or 10 or 15k RPM its even more pointless. You multiply the risk of a crash by 2, and the latency in just the controller of the board controlling 2 heads instead of 1 to find a certain data sector on the drive would be bad.

it would be a lot of work, a lot of hardware, and very little reward.

120 GB WD = $192
60 GB WD = $118x2 = $236

Factor in the added complexity of putting double the heads on the 120 GB and I am not seeing the price/performance benefit.

On the other hand if you had the second head in there as a backup to the primary, as in it sits there parked until the primary breaks, that might be worthwhile. And the complexity wouldn't be as high since it doesn't have to control both heads at the same time. That might be a good enough level of backup in situations where raid 1 isn't justified.

well you would need a secondary spindle also. Since many drive deaths are caused by spindle failure, head crashing isnt as much of an issue. Normally if a head crashes its cause it lands on the disk platters, and damages the disk platter.

bear in mind the wd 120gb is 20-30% faster than their 60gb ones

its gonna rock once they come out with 10k ide drives and put 4 direct channels on each board.. that would be some sweet raid 5.. and cheap to boot, even 0+1.

abit and epox are shipping boards with 4 channel ide raid and two more normal channels, but they don't support raid 5, and I can't see too many boards having onboard ide raid 5 anytime soon, SATA will come out before that.

10k ide will be cool, especially if WD gets in on it

well, once serial ata becomes mainstream i see no reason why they wouldnt throw in raid 5, its so much better than 0 and cheaper than 0+1.

whoa whoa whoa... pull the mis information bus over to the side of the road and come out with your hands up!


Raid 5 is "so much better" than RAID 0? Explain that one to me!?

RAID 5 is SLOW.. wicked slow on write times, and terribly intensive on your controler cards ram, if it doesnt have ram.. then the write speed will be about where a single drive would be and you will have terrible CPU usage.

Better as in redundant? Yep I will give you that... But speed HELLLLL NOOOO

If RAID 5 is important enough to you then you can afford SCSI. If not then put up with RAID 0. Frankly IDE RAID is pretty pointless - the seek time is the killer on desktop disk speed and RAID of any flavour does nothing to improve seeks. The only way to improve them is to buy a faster disk. SCSI anyone? :)

excuse me "mr. allknowingone" :P ... but you slam raid 5 in one aspect it and try to make it seem bad in ALL aspects. its read spead is a tad slower than 0, which in my book.. i like, and i will love the trade off a little bit of write speed for redundancy.. im sure alot of ppl would. in fact you would be stupid not to, but thats just my opinion.

the point, which im sorry escapes you, is that it is faster than a single IDE disk.. no it isnt SCSI, but this thread isnt about neither is it even compareing to SCSI. we all know SCSI is better, pointing that out is infact pointless. and again, you take one aspect, blow it out of proportion, and try to make the entire thing look bad. bad show.

just like all the raid versions.. there are specific uses for all them.

RAID5 is good for stuff that is mainly read time ( IE: web servers, etc...) But it should not be used in DB servers as the Read/Write lag kills performance.

How are you getting stats on Raid5 that says its write speed is just slightly slower then 0? On a high end SCSI card with megs of ram... Yep thats right. On an IDE card with no ram and no real hardware layer interface ( Most promise cards are all software level)... They will/would SUCK at RAID 5 simply cause there is NO write cache on the car.


1/0 and RAID 3 are favorites for people who have serious Read/Write access and want redundancy. Duplexing is of course the best of all but you have a 50% loss of storrage space. RAID 3 on a nice SCSI card is VERY nice, since you can put all the parity data on one specific drive and have it on its own SCSI channel. that way you dont have to wait for writing parity data to the same platters you are writing your real data on.

Jas, I am not a knowitall... I just have worked on RAID systems in professional settings for roughly 10 years. I am the Anti-RAID5 guy after I have seen how it has screwed up the speed of many DB's and file servers.

Redundancy is nice, but on a home PC should not come at the cost of performance.

The increase in speed for a RAID set vs straight drives is negligable in almost all desktop access patterns yet you still seem to think it's amazing and then accuse me of blowing things out of proportion. Let me restate this since you missed it, access time is key, being able to pull 30 extra megs down in a second is irrelvevant as almost all reads are for a few k to a few megs. Having a few ms off your access time makes more difference than RAID in pretty much all circumstances. The only way to get that is to spin the disk faster. IDE is, when it comes down to it, bottom of the line and kept there for a reason - most people want cheap storage not good performance. If you're happy with IDE that's fine, but don't try to make out that RAIDing it or serial ATA or any other IDE gimmick will make it fast, it won't.
The point about SCSI is this - people are spending a crap load of cash on these IDE RAID setups etc. when they could just bite the ****ing bullet and buy SCSI for the same and get much better performance. Half the reason they do this is because people go round saying things like "2 disks in RAID0 is twice as fast!!" which is just BS. Then they buy it and either are dissapointed or convince themself it's really fast to avoid looking like a moron. Thus the cycle continues...

ok, ill take a bow from this one.. :D

although, they sell ide cards that have cache on them. i know ide is no where near SCSI in terms of performance. i just think its being knocked too hard. most ppl cannot afford SCSI. so ide raid is the best it can get.

Better to save your cash and just run normal IDE IMO, IDE RAID is just a way for a few card makers to make a fast buck.

but if it's integrated into a board, no reason not to use it.

if I never get a board without raid on it, I'm going to get a 15k rpm 18gb hdd. Which may be happening very soon, the iwill and abit boards neither have raid

any mATX boards with ultra160?

there are 1u boards with scsi, but no matx's

im willing to bet those cost a bit much..

BUT if you need cheap capacity, and you have a raid card... WHY NOT?

Can't the highpoint controller used for most onboard raid be used as a regular ide controller? So, it's really a question of is it worth using it for raid in a desktop?

Aren't we at the point where hdd capacity has surpassed what is needed by the average user? I know people always think that, but honestly how much more bloated can windows get? No matter how big the drive is we can still find ways to fill it, but if you have 100+ gigs of space are you really feeling pressed to upgrade? The smallest drive you can even buy now is 20 gigs, and excluding my machine that is more than the other 3 pcs here combined. The only reason I have more is for video work, otherwise my old 8.4 gig is fine spacewise. But I would upgrade it because it's slow by comparison to a new drive. My point being that faster, quieter, more reliable seem like the best selling points now (side note: this makes the new ibm drives all the more mind boggling). I would love to have a smallish 10k rpm drive for the system and a slower drive for general storage, without having to go the scsi route.

I have an 18G IDE for bulk storage and an 18G SCSI for my 3 OSes and most of my programs. It works well and I don't feel short of space in particular.

ok... We all know that scsi has lower seek times and a bunch of other things that make it a lot better that ide.
I'll grant u all that.

<Disclaimer>
I do not work for any hw companies.
Opinions are my own.
I like SCSI.
I'm not advocating any hardware specifically
sorry for the long post
</Disclaimer>

the one thing that IDE has it's the only cheap solution we have vs scsi. IDE Raid makes things a little better but has these down sides

- More CPU Util
- Doesn't help seek times

Down side to scsi raid (other than cost)

- Doesn't help seek times

ok... for general reference scsi has ~ half the seek time of ide. I'm not being exact and i didn't go try to find the fastest seek times for both tech. this is just avg times.


we have scsi which is relatively expensive, but has very fast seek times and when combined into an array hands down the fastest sustained xfer rates.

ide is cheap... has relatively slow seek times and does fairly well during sustained xfer esp in an array.

SCSI = Performace platforms
IDE = everything else

some price comparisons... I didn't go find the largest SCSI u160 drive cause that's a moot point as far as expense. I did go find a 120gb for ide only for a price comparison. everything else is based upon some ppls perception that u don't need that much space. so i found the lowest price for some of the better hardware out there. notice i didn't claim the best.

seagate 18.4gb 10k rpm $145 1mb cache
fujitsu 18.4gb 10k rpm $153 8mb cache

adaptec aha 19160 u160 $147
mylex accelraid a352-2-32nb u160 raid 32mb cache $598
adaptec ada-2100s u160 raid 32mb cache $393

$292 for 1 drive and 1 controller
$683 for 2 drives and 1 raid controller

wd 120gb 7200rpm $177
seagate cuda 20gb 7200rpm $60
wd 20gb 7200rpm $62

HighPoint RR100 $35
Tekram DC200 $27
Promise FT100Tx2 $69
Adaptec 1200A $85

$177 for 1 drive (generally all mobos have ata100) or
$60 for 1 drive (for those who don't need 120gb)
$147 for 2 drives and 1 raid controller


ok it's 2 drives for the smallest array u can have.
i took the price of the cheapest controller so that way u have at least some kind of intro price to raid.

as far as serial ata goes, it's hard to say one way or the other... It'd be pure speculation on my part. There is the point that u have to run at much higher frequencies to get the same amount of data over a serial connection as u get over parallel. however, it'd seem really foolish to try to bring in a new tech that'd be slower. It has been shown that serial can compete with parallel (i.e. rambus) when used creatively (i.e. the quad pumped bus or whatever they call the bus for the P4). The only thing that comes to my mind as being a problem would be the cable length. I'm not sure how well high frequencies are preserved over long distances but then again how high will the frequency be for serial ata? it still has to travel at most 18(?) inches.


Just some things to think about.

Cyph3r

P.S. sorry about the long post

Pretty timely article over at hardware central

http://www.hardwarecentral.com/hardw...inions/4109/1/