What HDD will fit a Core2 machine; Raptors?

[E_tron]

I'm building a new Core2 machine with an ASUS P5N32-E. I remember seeing a 36GB Raptor in action during early 2004; i liked what i saw . So, what is the new must have Raptor or equivalent? I'm thinking Raptors have gotten faster since that 36GB version. Is there one that wont break the bank; I really like price-performance.

 

[Automata]

I'm building a new Core2 machine with an ASUS P5N32-E. I remember seeing a 36GB Raptor in action during early 2004; i liked what i saw . So, what is the new must have Raptor or equivalent? I'm thinking Raptors have gotten faster since that 36GB version. Is there one that wont break the bank; I really like price-performance.

I'm not sure I fully understand what your saying.

Raptors are the fastest out there on SATA, check NewEgg for price specifics.

 

[E_tron]

Raptors are like 3.5 years old. Are the latest Raptors faster than those early ones?

 

[Rickster]

does the 36GB raptors have 16mb cache?

 

[Automata]

Raptors are like 3.5 years old. Are the latest Raptors faster than those early ones?

Yes, they stepped up the performance. Make sure you get the one with 16mb cache and not the 8mb cache, the 8mb one is the slower and older one.

 

[Automata]

does the 36GB raptors have 16mb cache?

Yes, the new ones do, check out NewEgg.

 

[Rickster]

Yes, they stepped up the performance. Make sure you get the one with 16mb cache and not the 8mb cache, the 8mb one is the slower and older one.

As far as I know the newer raptors have NCQ and the 16mb cache instead of 8mb. So does the 16mb cache improve the performance by alot?

 

[Automata]

As far as I know the newer raptors have NCQ and the 16mb cache instead of 8mb. So does the 16mb cache improve the performance by alot?

Not alot, but you can notice the difference. I've used both.

Talking about NCQ:
This can reduce the amount of unnecessary going back-and-forth on the drive's heads, resulting in increased performance (and slightly decreased wear of the drive) for workloads where multiple simultaneous read/write requests are outstanding, most often occurring in server-type applications. However, the current technology actually slows down HD access in certain applications, like games and sequential reads, because of the added latency induced by NCQ logic[1]. Native Command Queuing was preceded by Parallel ATA's version of Tagged Command Queueing (TCQ).

Hmm...seems to slow it down.

 

[jivetrky]

I remember seeing a benchmark showing that the new 74GB drives are a bit faster in transfers over the 150GB drive. 2 of them would be even better.

 

[Automata]

I remember seeing a benchmark showing that the new 74GB drives are a bit faster in transfers over the 150GB drive. 2 of them would be even better.

Two in RAID0...

 

[E_tron]

So, these things do benefit from RAID 0?

 

[jivetrky]

So, these things do benefit from RAID 0?

With RAID0 you increase the read/write speeds because it stripes the information across both drives. So let's say a file that was 1GB was being transferred to a single drive would take ~17 seconds to copy assuming a 60MB/s transfer speed.
With RAID0 you theoretically double that transfer speed because the controller can take that 1GB file and cut it in half and transfer each half to each drive at the same time. So I'll assume a decent 75% gain and say you could transfer at 105MB/s. It would only take ~9.75 seconds to transfer.

 

[Audioaficionado]

I already have one 8mb old school 74GB Raptor and it runs great.

I would love to try out that Intel matrix striped raid 0 with a pair of 74GB 16MB Raptors. NCQ would most likely be disabled if I have a choice.

My data however will always be on either mirrored arrays or controller card RAID 5 arrays. Matrix is good for boot/os/apps only IMHO.

 

[Nittenti]

I have 2 new 160GB Raptors (OEM discs from HP - see my other thread)

I had them up and running in a Intel Matrix RAID0 with 50GB from each - 180MB/s. As single discs they do pretty well to

 

[Rickster]

it really depends on what stripe you are using. if you transfer large files all the time and use a small stripe it would be slower than a non-raid array. i think most gamers would use 16k stripe if im not mistaken. im a noob of course please correct me.

btw whats up with the intel matrix raid thingy? can you not use other chips to run a raid array? like nvidia chips? whats so brilliant about matrix or what's so bad about others? and how good are the add-on raid pci-cards in comparison to the on-board ones. i heard that the good raid cards have a processor to calculate some sort of things and takes work load of the CPU when in raid. and if you use the on-board controller it takes up quite abit of cpu cycles. yes,no?

 

[jivetrky]

it really depends on what stripe you are using. if you transfer large files all the time and use a small stripe it would be slower than a non-raid array. i think most gamers would use 16k stripe if im not mistaken. im a noob of course please correct me.

btw whats up with the intel matrix raid thingy? can you not use other chips to run a raid array? like nvidia chips? whats so brilliant about matrix or what's so bad about others? and how good are the add-on raid pci-cards in comparison to the on-board ones. i heard that the good raid cards have a processor to calculate some sort of things and takes work load of the CPU when in raid. and if you use the on-board controller it takes up quite abit of cpu cycles. yes,no?

Cool thing about matrix is that you can divide the drives into separate partitions and then RAID0 those partitions. (though all that works through the Matrix Raid, you don't actually have to setup separate partitions) So if you partition out the best performing part of each drive (Not sure if that's the inner platter or outter platter) and stripe a few of those together, you have a better performing array for your OS and apps. Then the remaining space on the drives can be made into a separate array, which doesn't have to be RAID0, it can be a redundant setup, or etc.

 

[tuskenraider]

So, these things do benefit from RAID 0?

RAID0 doesn't make a drive better than it already is. In the end, whether you put a Raptor or a typical 7200rpm drive in RAID0, the benefit(if any, depending on use) will be the same percentage-wise. What will be different is that since the Raptor performs at a higher level, an increase of it's performance is worth a little more.

btw whats up with the intel matrix raid thingy? can you not use other chips to run a raid array? like nvidia chips? whats so brilliant about matrix or what's so bad about others? and how good are the add-on raid pci-cards in comparison to the on-board ones. i heard that the good raid cards have a processor to calculate some sort of things and takes work load of the CPU when in raid. and if you use the on-board controller it takes up quite abit of cpu cycles. yes,no?

It's Intel tech available only on their RAID chipsets. It's creation was to allow a person to create 2 RAID arrays(RAID0 and RAID1, or vice-versa) with only two drives, instead of four minimum. That way you could get some speed by using a RAID0 array and redundancy by using RAID1. Now some people use it as a psuedo performance increaser by making a small RAID0 array first, which uses the first part of the disk(fastest) and install their OS on it. Thing is, it is no different than if you were to make a small partition within a full RAID0 array that uses the whole 2 disks to do the same thing. With Matrix RAID though, you can still make another array(RAID0 or 1) with the unused space left on the disk, not just another partition on the same array. In this scenario, you have the potential to lose two arrays at once, but at the same time have one go bad and not the other. Pick your poison. As with all onboard controllers, they eat CPU cycles, but rarely more than 5%. Plugin controllers with their own processor definately provide benefits depending on the model. A good plugin card with processor and a chunk of RAM for caching will perform better, but at $300+. For most desktop users, this isn't worth the investment.

 

[Rickster]

So if you partition out the best performing part of each drive (Not sure if that's the inner platter or outter platter) and stripe a few of those together, you have a better performing array for your OS and apps.

Should be the outter platter if I'm not mistaken. More coverage of area on the outter circumference than the inner.

It's Intel tech available only on their RAID chipsets. It's creation was to allow a person to create 2 RAID arrays(RAID0 and RAID1, or vice-versa) with only two drives, instead of four minimum. That way you could get some speed by using a RAID0 array and redundancy by using RAID1. Now some people use it as a psuedo performance increaser by making a small RAID0 array first, which uses the first part of the disk(fastest) and install their OS on it. Thing is, it is no different than if you were to make a small partition within a full RAID0 array that uses the whole 2 disks to do the same thing. With Matrix RAID though, you can still make another array(RAID0 or 1) with the unused space left on the disk, not just another partition on the same array.

Ok, I understand what you've said. Just that I didn't know you could partition your drives so that you can install your OS on the first part of the disk/outter platter and then stripe individual partitions. I've not RAIDed anything in my life, I'm just reading up on some good info. I thought partitions make your hard disk seek slower?

 

[jivetrky]

Should be the outter platter if I'm not mistaken. More coverage of area on the outter circumference than the inner.

That's what I was thinking. On the outer part, the heads wouldn't have to move as far, so the access time should be better.

 

[Audioaficionado]

That's what I was thinking. On the outer part, the heads wouldn't have to move as far, so the access time should be better.

Access time is king. The faster rotational speed of the outer disk has much less effect.