Can I use a 15,000 rpm cheetah?

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Supporting wife = 1 hard drive for you, 1 pair of shoes for me
8 hard drives for you, 8 new pairs of shoes for me
1 server for you, one diamond necklace for me.



supporting... sure thats what it is.
 
I'm pretty sure MOST PCI-X cards can fit into and function in normal PCI slots (since modern PCI slots can provide either voltage needed). Personally, I have both an Adaptec 29160 (SCSI) and an Adaptec 2810sa (SATA RAID) in PCI slots on my mobo right now.

As for the original poster, its not gonna be cheap even if you go the ebay route (I did), in comparison to a Raptor. To be honest I'd have a hard time recommending it after my own experience, just in the sense that the performance/price ratio is a bit crappy, and also SCSI cables are (apart from being expensive) typically very long, which has created quite the airflow restriction in my case.

As for the comment of putting PCI-X cards into PCI slots defeating the purpose, that really depends on the person's needs (or cough, purpose). Yeah, if you're going for speed you're gonna defeat the purpose, but if you're like me and just wanted to get data redundancy (by using RAID5 in my case), then it fits the bill. But even if you're going for a boost in speed, you can get over 100mb/s through the PCI system, which wouldn't be cutting off a lot of performance from using non-Raptor drives in RAID.
 
From what I've seen, SAS is the best way to go IF you are adding to your hardware.

I would get a PCIe card over a PCI SAS raid for a few reasons.

1. What else are you going to put in that PCIe slot? Save that PCI slot for audio.

2. The newer PCIe cards have higher throughput from what i've seen and in general are newer than the PCI variant.



Egghead had a fujitsu 74 gig drive (15K, 16 mb buffer) for 285. They just recently raised the price to 410 bucks each. I kick myself for not buying them.

I was going to buy 2 or 3 of them and run a raid0 set up with them. OR I might have just installed XP, Vista, and a variant of Linux on each one.

Really, its about cash. If you have the money to blow, buy it. I doubt you'd be disappointed. I have scsi drives from the EARLY 90s that still run. Those things are built like a rock.

Since you probably don't want to drop 1000+ on HDDs, sata is still fine. Unless you're having a few people always accessing that computer (it would be acting as a server) or moving LARGE files, you won't notice too much of a difference.

With the saved money, run a raid 0 for speed (or buy a raptor) and do a raid 5 for backing up your stuff.
 
it may be cheaper but it performs worse than a plain old Sata 150

Ultra 160 is perfect if you only run one SCSI disk - Ultra 160 cards are cheap and they have easily enough bandwidth for even the fastest SCSI drive.

If you want to do RAID then you will need Ultra320 or SAS - and you'll also need to get a pci-e controller as the pci bus is limited to 133MB/S so it will bottleneck. This is when SCSI starts to get expensive.

If you just want to run one very fast SCSI drive then my set up is valid, cheap and performs extremely well.
 
Since the only difference between your setup and mine is the actual drive (mine is 2-3 years older since I've had this setup for much longer), I guess your drive makes your setup "perform extremely well" whilst mine is only mediocre. However, my suspicions go beyond the drive itself, I have a very strong feeling that because the PCI's IRQ is a rather low priority, it's what causes the system to not be as responsive as it should.
 
I'm pretty sure MOST PCI-X cards can fit into and function in normal PCI slots (since modern PCI slots can provide either voltage needed).

Yeah, there should be more compatibility issues going the other way since PCI-X is 3.3v only.

As for the comment of putting PCI-X cards into PCI slots defeating the purpose, that really depends on the person's needs (or cough, purpose). Yeah, if you're going for speed you're gonna defeat the purpose, but if you're like me and just wanted to get data redundancy (by using RAID5 in my case), then it fits the bill

True but even in RAID 5 you are probably maxing out the bus when you do a read, especially if you have NIC's or a sound card going at the same time.

But even if you're going for a boost in speed, you can get over 100mb/s through the PCI system, which wouldn't be cutting off a lot of performance from using non-Raptor drives in RAID.

RAID 1 or 0 maybe not but a RAID 5 read would probably be pushing it with any decent HD.
 
newegg has an LSI 8 port 4X pci-e 128MB cache SAS/SATA controller on sale for $350 from $650 normally. I just got one of those, and the upside is that it also can do SATA drives so you can mix and match both SAS and SATA so have SAS drives for apps and SATA for data.
 
Do you have anything else significant on the PCI bus? I'm pretty sure that USB uses the PCI bus, so if you've got lots of USB devices perhaps that would be an issue.

What results do you get from HD Tach?
 
"True but even in RAID 5 you are probably maxing out the bus when you do a read, especially if you have NIC's or a sound card going at the same time."

Precisely my point, it is maxing out the bus but I don't care, my RAID5 is for data security, not speed.

"RAID 1 or 0 maybe not but a RAID 5 read would probably be pushing it with any decent HD."

Yes although I didn't spell it out, I meant RAID0 since that is what the OP was asking about. You can set it up on almost any motherboard these days, and he was interested in a low-cost performance solution after all.
 
Do you have anything else significant on the PCI bus? I'm pretty sure that USB uses the PCI bus, so if you've got lots of USB devices perhaps that would be an issue.

What results do you get from HD Tach?

Well, PCI bus is one thing and IRQs are another. For example, each PCI slot on my mobo supposedly has its own IRQ. You say USB uses the PCI bus, that's the first I'm hearing about it, so if you could provide a link or something to back that up, I'd love to read up on it and learn something new. As for USB devices - mouse, keyboard are the only permanent devices.

Results from HD Tach are:
Normal curve from just over 60mb/s falling to just under 45mb/s.
Average read: 54.4mb/s
Random access: 6.0ms
Burst Speed: 79.4mb/s
 
[quote="Fedor]Well, PCI bus is one thing and IRQs are another. For example, each PCI slot on my mobo supposedly has its own IRQ. You say USB uses the PCI bus, that's the first I'm hearing about it, so if you could provide a link or something to back that up, I'd love to read up on it and learn something new. As for USB devices - mouse, keyboard are the only permanent devices.

Results from HD Tach are:
Normal curve from just over 60mb/s falling to just under 45mb/s.
Average read: 54.4mb/s
Random access: 6.0ms
Burst Speed: 79.4mb/s[/quote]

I was wrong about USB - I picked up this from a forum talking about a USB PCI card - which is obviously a bit different.

Apologies for the FUD - you only have to worry about any other PCI cards.

My HD Tach results are:
SCSI 15K:
Normal curve: 93 to 63
Average read: 80.9mb/s
Random access: 6.0ms
Burst speed: 118.2mb/s

Raptor:
Normal curve: 70 to 53
Average read: 64.9mb/s
Random access: 8.0ms
Burst speed: 120.6mb/s

So your drive is performing at a pretty low level - what model is it?
 
It's the second gen Cheetah, its like more than 3 years older than your model like I mentioned before. It's performance is as it should be for that model. However, what is supposed to make a big difference in terms of system responsiveness is the fact that it has a far lower seek time than 7200rpm drives (and not as much but Raptors too) - which is something you can see even in your Random Access result. And my drive gives me the same Random Access as yours, so you'd think that the system would be just as responsive. Personally I don't really see a difference between my setup and that of my bro-in-law who uses a Raptor (comparing both systems from a fresh install, since I did both).
 
Your random access time is the same as mine because they are both 15K - the speed of the rotation pretty much dictates random access time.

But your average STR is far lower than my drive and even the Raptor - this is where newer, more densely packed drives gain an advantage.

So I'm not surprised that your drive is not stellar in terms of performance. You would see an increase in performance with a Fujitsu MAU/MAX or a Maxtor Atlas II (these are the best SCSI drives for single user usage), or indeed a Raptor (which is totally tuned for single user usage).
 
Firstly, you saying that the access time is the same cause they are both 15k drives is like saying that a 2litre engine car will always perform the same as another. Seek times for 7200rpm drives vary a lot for example (over 3ms difference is easy to find), just take a look at any database.

But it is access time that dictates system responsiveness, not some sequential read speed you get from HD Tach (cause with Windows and a pagefile, its a constant onslaught of small reads and writes and loads of seeking - a Raptor should be slower hands down for responsiveness). Therefore, I am convinced that it is because the CPU treats it as a low priority that causes Windows to not be as quick as it should, imho.
 
Firstly, you saying that the access time is the same cause they are both 15k drives is like saying that a 2litre engine car will always perform the same as another. Seek times for 7200rpm drives vary a lot for example (over 3ms difference is easy to find), just take a look at any database.

But it is access time that dictates system responsiveness, not some sequential read speed you get from HD Tach (cause with Windows and a pagefile, its a constant onslaught of small reads and writes and loads of seeking - a Raptor should be slower hands down for responsiveness). Therefore, I am convinced that it is because the CPU treats it as a low priority that causes Windows to not be as quick as it should, imho.

The RPM dictates to a large degree the access time - that's what I am saying. Of course there will be differences between one 15K and another 15K, but it is only a matter of 0.5ms between the top 15K drives.

Both random access time and STR are important, but so is the firmware. Raptors are highly optimised for desktop usage (where the next file is probably close to the current one), whereas SCSI drives are more oriented to multi-user scenarios (where the next file could be anywhere on the disk). Therefore the advantage that SCSI drives have in terms of access time and STR is eroded by the inferior (in terms of desktop) firmware.

That's why you see better performance from the Raptor - it is better optimised for the environment it lives in. Sure it has a worse access time, but it has a better STR and more relevant firmware.

BTW - I believe you had a Seagate Cheetah? Just in case you didn't know - I believe that you can set that to 'Desktop mode' with SeaTools Enterprise that adjusts the firmware to a single-user environment.
 
Heh, I still find it amusing that you feel the need to be so patronising. But, moving on....
"but it is only a matter of 0.5ms between the top 15K drives."
Yes, exactly. Modern, top 15K drives. Not all drives, and of course if you start comparing today's fastest 15K drive in comparison to say mine which is 5-6 years old, then performance is gonna differ.

As for SCSI drives being orientated towards serving more smaller requests than say a Raptor, yeah I know that. But does that mean a 15K SCSI drive should be slower than a Raptor in a single-user scenario? I wouldn't think so, its a far more advanced piece of hardware.

As for the Desktop mode as you put it, yeah I noticed that setting a while back. My Cheetah was dying so I used SeaTools to check it, and while I was at it discovered those settings. Yeah, the Desktop mode is on now. Out of curiousity I went ahead and disabled it, and interestingly HD Tach didn't change one bit.

Now I'm basically trying to figure out whether to get a new 15K drive (like maybe the perpendicular Cheetah 15K.5 - MONSTER transfer rates!) or to just get a Raptor.
 
Your random access time is the same as mine because they are both 15K - the speed of the rotation pretty much dictates random access time.

But your average STR is far lower than my drive and even the Raptor - this is where newer, more densely packed drives gain an advantage.

So I'm not surprised that your drive is not stellar in terms of performance. You would see an increase in performance with a Fujitsu MAU/MAX or a Maxtor Atlas II (these are the best SCSI drives for single user usage), or indeed a Raptor (which is totally tuned for single user usage).

maxtor15K2.png

Interface limited (Ultra160 card kills drives, not using it anymore)
 
Wow, and I thought I had problems. Your drive's potential is obviously being slaughtered there. I don't really follow, what's with that constant ~68mb/s? I mean, ethel clearly doesn't have your problem, seeing as how his curve starts at 93mb/s and his average is higher than your max. May I ask you:
1) What brand and model of U160 card is it?
2) Is it in a regular PCI slot or PCI-X?
3) If it's in a regular PCI, what other PCI cards do you use?

Edit - Same goes for you ethel, could you elabourate on your U160 card and PCI configuration?
 
Heh, I still find it amusing that you feel the need to be so patronising. But, moving on....
"but it is only a matter of 0.5ms between the top 15K drives."
Yes, exactly. Modern, top 15K drives. Not all drives, and of course if you start comparing today's fastest 15K drive in comparison to say mine which is 5-6 years old, then performance is gonna differ.

As for SCSI drives being orientated towards serving more smaller requests than say a Raptor, yeah I know that. But does that mean a 15K SCSI drive should be slower than a Raptor in a single-user scenario? I wouldn't think so, its a far more advanced piece of hardware.

As for the Desktop mode as you put it, yeah I noticed that setting a while back. My Cheetah was dying so I used SeaTools to check it, and while I was at it discovered those settings. Yeah, the Desktop mode is on now. Out of curiousity I went ahead and disabled it, and interestingly HD Tach didn't change one bit.

Now I'm basically trying to figure out whether to get a new 15K drive (like maybe the perpendicular Cheetah 15K.5 - MONSTER transfer rates!) or to just get a Raptor.

Hey, I'm just trying to help - sorry if you think I'm being patronising. I don't know what stuff you know or don't know, just trying to help you as you were saying you have a SCSI drive not performing.

The 15K.5 has incredible STR as you say, but at storage review they still rate the MAU / MAX and the Atlas 15K II as better for most tasks, especially single user:

http://www.storagereview.com/articles/200609/ST3300655LW_1.html

Another good review that you might find useful is this one of the 150GB Raptor where they basically say that the 150GB Raptor is superior to even the latest SCSI drives in a lot of single user scenarios (because of the firmware). This may explain why your older generation SCSI drive feels less responsive than a Raptor:

http://www.storagereview.com/articles/200601/WD1500ADFD_1.html

My personal experience of a 74GB raptor vs a latest generation 73GB SCSI is that the SCSI 'feels' a lot snappier as my Windows drive. But goddammit, they are both fast - I'd never go back to 7200 rpm :)
 
Wow that graph certainly shows some kind of problem on the interface. Those Atlas 15K II drives are nice.

Fedor: I have no other devices in PCI slots and the card I have is the 29160LP.

I have a couple of ideas that *might* help:

1) A lot of problems like this are to do with the quality of the cable and/or termination - if you can get your hands on one, try and get a good quality U320 cable with built in termination.
2) Are there any other devices on the PCI bus that may be consuming bandwidth? If so, remove them and see if that makes any difference.
3) If you are sharing PCI bandwidth, try upping the PCI Latency timer in the BIOS. Here's a link:

http://www.adriansrojakpot.com/Speed_Demonz/BIOS_Guide/BIOS_Guide_02h.htm
 
I realise you are only trying to help, but I've been trying to stress the fact that if you go for a _cheap_ SCSI setup, then it wont be the latest generation drive. Mine Cheetah is the X36 one - although it wasn't known as such it was effectively the 15K.2. At the time of purchase, .3 was the latest so I got one generation back to keep costs low. And as such the performance that I get is to be expected. What I was really getting at is that the Raptor can beat down on a latest-1 gen SCSI drive even today, and you dont have to worry about SCSI card, big a** cables, terminators and so on, and the Raptor would probably end up being cheaper. That's why I was having a general go against going with SCSI on the "cheap".

I've seen that storagereview review from ages ago. They were such a good site, its really sad that they aren't keeping up with the times anymore. As you can see in that review, the Fujitsu MAX series isn't even there - I was looking at maybe getting one of them but I can't find a single review anywhere on the net. As for the second link, it's quite interesting, thanks. I'm glad at least you are here to defend my attacks on SCSI 😛