740ADFD Raptor 74GB vs Spinpoint T166 HD321KJ in Antec Solo

[syncmaster33]

I have an Antec SOLO with 3 Seagate drives, 2 200 gigs and one 300 gigs. The 300 (ST3300631A) is split into 2 with one partition working as my "os and app drive" and the rest is just storage. Well, I want something faster but I don't want it to turn my quiet computer into a coffee grinder. Now, I am not looking for a silent computer, but quiet is good.

SPCR liked the idle of the Raptor but the seeks were loud. I also read good reviews of the Spinpoint T166 series as well.

I can get the Samung for $77 and the Raptor for $100. I don't need a big drive since this will be just for my OS and applications. (Having a dedicated OS/App drive does help with speed doesn't it?)

I have a C2D 6400 with 2 gigs of ram and a Radeon 1950 Pro so I think that is pretty decent, but I still want more "snap".

Your thoughts please.

Anyone have a Raptor in a SOLO and are you happy with it?
Do you think I would notice a big difference in speed between the two?

[jdunning]

Anyone have a Raptor in a SOLO and are you happy with it?
Do you think I would notice a big difference in speed between the two?

I've got a 75GB Raptor in the elastic suspension in a Solo and have been very happy with it. I do have an Infrant NAS box in the same room, which is definitely louder than the Solo with 2 drives, but the Solo is almost inaudible with the NAS box off. Seeks on the Raptor are more noticeable than the 500GB WD that's also in the box, but only if I'm listening for them. And it seems to run cooler (though it's on the top, so it may have somewhat better airflow).

As for the speed, it's hard to say how much is due to the Raptor vs. 2GB RAM and the Core2 Duo, but it's definitely a snappy machine.

The only drawback to the Raptor is that it's only 75GB! I actually started out with the box dual booting Vista/XP, but I've started installing Vista apps on the XP partition because I'm running low on the Vista side.

[Nick Geraedts]

Assuming that your system has built in support for this - RAID0. It'll be a bit more pricey, but boy does it ever quicken the system. You can see my workstation specs below, and PhotoshopCS3 loads in 5 seconds flat (I've timed it with my watch). RAID0 is really the best option if you want that extra bit of responsiveness.

The raptor is seeing the end of it's useful life. Higher density platters are making up for the difference in rotational speed. The only thing Raptor drives still have going for them is random access time.

I did a comparison of my system with a friends. He's got the same motherboard, but two 74GB raptors in RAID0 instead of my two WD3200AAKS in RAID0. The benchmarks are below (Red=mine, blue=his):



As you can see - there isn't really too much real-world benefit, but you're paying a premium for the drives.

[syncmaster33]

I've consider Raid 0 and my motherboard supports it, BUT it's a software solution and not a dedicated chip. I've heard that makes a fairly substantial difference. I've also read (can't remember if it was Anandtech or Extremetech or...) where they don't really recommend it as the performance isn't that dramatic. That's why I was considering the Raptor.

The 74gig drive I think would be big enough as I don't dual boot and it is 4 gigs larger then my current boot partition. Yes it is almost full, but I've not been very good about uninstalling stuff I don't use.

I've also thought about just getting the larger Samsung drive and then quarter stroking it. I've read where this speeds things up to Raptor levels, but my current drive is such and I can't say it flies....

[snq]

I have both a Raptor (150 GB) and a T166 (500 GB) in the same case, and the T166 is inaudible compared to the Raptor. I wouldn't recommend it if you want a quiet PC, the seeking sounds are very loud.
This is in a P180 btw.

[Nick Geraedts]

Pretty much all onboard RAID solutions are "software" based, which means that they do the calculations through the CPU instead of a dedicated processor. For RAID0, 1 and 10, this isn't a problem at all since there are no calculations to perform. For RAID5, you might see a performance difference, but that's only in a VERY heavy loaded environment (pretty much never in a home-network system).

I can tell you with 100% confidence, two modern 7200RPM drives will outperform a single Raptor for any application. I just gave you two Raptors vs two WD 7200RPM drives - the numbers speak for themselves. There's no real benefit to the Raptors these days aside from bragging rights.

People tend to have a strange fear against using RAID systems... I'm not really sure why, but there's been a lot of negative commentary about this. Personally, I'll never build a system without RAID again. If the data is important, it's RAID1 or RAID5 - if not, RAID0. I had an older workstation (PIII based), and I did three successive upgrades to it.

1) Move from single 1GHz to dual 1GHz CPUs.
2) Move from 1GB PC133 to 2GB PC133
3) Go from a single 40GB 7200RPM drive to two 80GB 7200RPM drives in RAID0.

The last was the one that really made all the difference in terms of responsiveness. It really is like night and day. The hard drive subsystem is the slowest part of computers these days, so every extra bit you put into it helps more. There's not really much point in upgrading parts that aren't the bottleneck, is there?

[syncmaster33]

Well, maybe I'll have to get 2 T166's and put then in Raid 0.

My only problem is this then. That makes for a big ass boot drive. Would it be better to just use the entire thing as one drive or can you partition a raid array? And if I partition it, would i see a performance hit if I am doing things on that non-os partition? Would it be trying to use the same thing twice?

Sorry for all the questions.....

[Nick Geraedts]

No worries about the questions. The more you know, the farther your money goes.

Indeed, you do get a massive boot drive when putting two large drives in RAID0. I've got a 600GB (640GiB) boot drive in my workstation, so yes, it's a bit overkill, but on the other hand - it's there to use.

Once the RAID array is setup in BIOS, it appears as one drive to the rest of the system. You can format it, partition it, and read/write to it just like any other drive. It'll even show up as a single disk in Disk Management under Windows.

I tend to stay away from partitions these days. I'd much rather have a separate storage/backup drive for critical data. Partitions are alright, but tend to be overused by some. If it's for purely static data, then sure - go ahead, since you only read/write to that occasionally, but putting programs on a separate partition actually decreases overall system performance.

Just a second word of warning (since somebody else is bound to bring this up) - RAID0 with two drives has twice the failure probability of a single drive setup. If you lose one drive, the data on both is gone. That being said - you should never keep a single copy of your important data at any time regardless. In practical terms, the reliability factor shouldn't play a role in your decision to use RAID0, since you're still running the risk of a single disk dying otherwise.

[syncmaster33]

Well, it looks like I'm going to get the 2 disks and do a Raid 0 setup. What I'm going to do with a app drive that big I have no idea. It's almost as big as all my drives now combined. I'm not too worried about the failure thing as I always keep my critical data backed up. If a drive goes down and I have to reinstall windows.....well, that will suck but it is always nice to have a fresh system. I also have an Acronis Image of my drive so it only takes 15 minutes or so to restore everything.

I have a couple more questions though.

I have read about needing to format in a 16cluster/stripe. What does that mean and how do I do it?

Also, is there anyway to install drivers without a floppy when installing xp pro?

[nick705]

Also, is there anyway to install drivers without a floppy when installing xp pro?

Yes, create a customised install CD with the SATA/RAID drivers slipstreamed - while you're at it, you can incorporate service packs, hotfix rollups and tweak the installation pretty much any way that takes your fancy.

http://www.nliteos.com/nlite.html

Incidentally, there's a wealth of info at StorageReview.com about all flavours of RAID - you might want to start here for a general assessment of the "benefits" of RAID 0 for general desktop use. There's also a good article at Anandtech with plenty of benchmarks.

YMMV of course....

[syncmaster33]

Thanks for the links. There is lots of information but now I'm undecided again. The article I alluded to in the first post is the one at AnandTech you linked. The first paragraph of the conclusion:

"If you haven't gotten the hint by now, we'll spell it out for you: there is no place, and no need for a RAID-0 array on a desktop computer. The real world performance increases are negligible at best and the reduction in reliability, thanks to a halving of the mean time between failure, makes RAID-0 far from worth it on the desktop."

and at Storagereview.com it sums things up with:

"To summarize, RAID 0 offers generally minimal performance gains, significantly increased risk of data loss, and greater cost."

Now I'm back to being frozen in headlights......

[nick705]

If it helps, I used to be a big Raptor fan, and I had one as my OS/apps drive in a Solo - I switched to a Samsung HD501LJ and any speed difference is negligible, whereas the seek noise difference is considerable.

I haven't personally used RAID-0 in any meaningful sense, apart from setting it up to see how it all works and running HDTach benchmarks (which I gather aren't altogether reliable for RAID-0 anyway). The arguments against it seem quite convincing to me though, backed up by solid data, and the "pro" arguments seem largely based on people's impressions that their system "feels" faster (which may or may not be placebo). Whichever way you look at it though, RAID-anything introduces another layer of complexity and a potential point of failure, which IMHO is reason enough to avoid it unless the benefits are easily quantifiable and worthwhile.

I do think there's some benefit in having the OS/apps and user data on two separate physical drives though, as it means both can both be accessed concurrently which should speed things up, and it obviously makes life easier in the event of a system reinstall.

3) Go from a single 40GB 7200RPM drive to two 80GB 7200RPM drives in RAID0.

The last was the one that really made all the difference in terms of responsiveness. It really is like night and day. The hard drive subsystem is the slowest part of computers these days, so every extra bit you put into it helps more. There's not really much point in upgrading parts that aren't the bottleneck, is there?

I'm fairly certain you'd have seen a similar improvement by upgrading from the old 40GB drive to a single 160GB one - the increased capacity and consequent data localisation (meaning higher transfer speeds, shorter seeks) would have played their part, not to mention newer and better mechanicals...

[Nick Geraedts]

Incidentally, there's a wealth of info at StorageReview.com about all flavours of RAID - you might want to start here for a general assessment of the "benefits" of RAID 0 for general desktop use. There's also a good article at Anandtech with plenty of benchmarks.

The most of the benchmarks on there are for "general system performance", so to speak. Many of those benchmarks are not hard-drive limited, so the performance benefits that you'll see are minimal.

@nick705 - You're right in that most benchmarks regarding RAID0 are generally negative, but I'm wondering if anyone here can say that they've copied a 150MB file to the same drive and not seen the "file copy progress" window show up? I did so last night - Ctrl+C, Ctrl+V - poof. New file. The benchmarks from HDTach for RAID arrays are a little spotty at times (they claim I get 40MB/s on my RAID5, while I've copied files to/from the array at over 100MB/s before).

If people want to work together to give some standard testing procedure amongst ourselves, we could work that out. I use a program called TeraCopy to copy large files between my computers or hard drives. It's useful because you can restart the file copy at the end, after you've changed the copy buffer. It'll also give you an average file copy speed at the end of the run. What do we all say? I really do believe that the responsiveness of systems with RAID0 are far higher than those without - especially when dealing with concurrent I/O operations.

[nick705]

It's an interesting idea, but I'm not sure how you'd establish an effective common methodology - for instance, people will have different OS/apps setups, different hardware, different amounts of RAM and RAM-consuming running processes which will affect buffering, and various makes and models of HDD in various states of fragmentation, so finding a way of not comparing apples with oranges is a big ask.

It's true that some of the Anandtech benchmarks are not necessarily HDD-limited, but in a way that's the whole point - if they're a fair reflection of overall "general system performance" it demonstrates that, by and large, the speed of the HDD subsystem is *not* usually the main performance bottleneck, and RAID-0 has nothing much to offer *unless* you have unusual usage patterns which benefit purely from massive STR.

I'm not saying that such scenarios don't exist (video editing springs to mind), but even then you might get more benefit from having independent source and destination drives which can read and write simultaneously (unlike copying from one location to another on a RAID-0 array). Beyond that, and for "normal" use, disk accesses seem to be generally too random to gain much advantage.

[Nick Geraedts]

It's true that some of the Anandtech benchmarks are not necessarily HDD-limited, but in a way that's the whole point - if they're a fair reflection of overall "general system performance" it demonstrates that, by and large, the speed of the HDD subsystem is *not* usually the main performance bottleneck, and RAID-0 has nothing much to offer *unless* you have unusual usage patterns which benefit purely from massive STR.

I'll agree with you on that - when I'm sitting here typing this post, or browsing the forums, RAID0 isn't going to help my overall system speed. However, when I'm copying large files or loading large programs, I do notice the difference. It is definitely a question about bottlenecks, but the strange thing is that people seem to neglect the 60MB/s (I'm being generous) bottleneck of the harddrive subsystem when it comes to reading and writing files.

There's a reason why large corporate databases are stored on RAID10 as opposed to straight RAID1. The added striping gives far better performance over single disk implementations. They still need the redundancy, but they'll get that performance boost that they need.

[nick705]

It is definitely a question about bottlenecks, but the strange thing is that people seem to neglect the 60MB/s (I'm being generous) bottleneck of the harddrive subsystem when it comes to reading and writing files.

Well, the proof is in the figures - yes, in theory the HDD subsystem is the slowest link in the chain by several orders of magnitude, but *in practice* it doesn't normally seem to be the main bottleneck. I'd imagine that with decent prefetching and cacheing algorithms, the OS can normally compensate pretty effectively for purely mechanical HDD limitations, at least for typical desktop usage.

I don't think large corporate databases are a fair illustration - server drives are typically accessed in a completely different way from a desktop or workstation PC, and that really is comparing apples to oranges.

I take your point about the speed of copying large files from one place to another (*provided* the array is reasonably unfragmented and there's enough free RAM to buffer a decent amount of data and avoid heavy seeking), but unless you're in the habit of constantly shuffling vast amounts of data around your HDDs, the question remains as to whether this represents a major real-life benefit in most circumstances (if you're just moving the files on the same partition as opposed to copying them it's a non-issue anyway of course).

For what it's worth, I accept there are at least some situations where RAID-0 provides measurable benefits, but at least from my own POV they're few and far between enough to make the downsides much more relevant. I'm not talking about "doubling the chance of data loss" (which is a bogus argument IMO), but the sheer inconvenience - an array is tied to the controller it was created on, and you can't just transfer a data-filled HDD from one PC to another, which is something I've done so many times I've lost count. You lose the capability of SMART monitoring, at least with most (all?) cheap motherboard controllers, you can't monitor drive temps, and most of all, it's just one more thing to go wrong, which as a fully paid-up member of the KISS society I'd much rather avoid unless there's a compelling reason not to...

[Nick Geraedts]

Hard drive prefetching is actually only done in Vista. The "prefetch" in XP has little to do with actually prefetching files, but rather the order in which to read the files for a specific program. I can't say much about *nix systems, since I don't know how they deal with this, but that's exactly what SuperFetch does in Vista.

Corporate Databases are a different ball game - but they're proof in the pudding of the advantages during high I/O operations.

Your comment about the speed requirement of the RAM for a buffer - only 8-16MB is really needed, depending on the cache size of your drives. From my personal testing with TeraCopy, setting the buffer equal to the hard drive cache size provides the best overall transfer speeds. I tend to make a lot of Windows unattended installation discs (for testing and such), and the final stage in the process is copying the source files into an ISO. The speed benefit from RAID0 helps me there.

I'll agree with you about the KISS principle - I try to apply it wherever possible. Being the debating guy that I am... I'll try to tackle a couple of your points (not trying to downplay them - just provide my opinion). I would only use RAID0 on a system drive or a scratch disk - never a data storage disk. Anything that's worth keeping goes on a backup disk or something with redundancy. As for SMART monitoring - the only part of that I tend to look at is the temperatures. I remember reading an article that SMART data only really helped about 50% of the time in detecting a hard drive failure. For me, the good ol' finger test on the hard drive case does the trick for me.

I will agree with you on one thing though... figuring out which drivers I needed for my motherboard RAID support in Windows was a major PITA. It took me a few times to get that right. Thankfully, all my current systems are based on the 965 chipset, with the same ICH8R controller.

[SoopahMan]

RAID0 is a contested solution for single-user desktop systems. StorageReview.com has a lengthy article on why they think it's a dumb idea. I used to use RAID0 with 2 36gb Raptors and noticed an improvement in speed especially when loading Photoshop, but the experts say I didn't. So, word to the wise.

As for hearing the 74gb Raptor itself, you won't hear it in suspension. I have 2 in a P180 case and hear nothing.

However, the Raptor runs hot. You will need active cooling if you intend to make heavy use of it, and and if you put it in a RAID array then 2 Raptors are firing together, heating each other up... fan needed. In my P180 I used a big 120mm fan running slowly. It keeps the Raptors at 39C, a very acceptable temp. I can't hear the fan.

On the bottleneck discussion, the hard drive IS a bottleneck but only in narrow situations:
* Writing temp files
* Writing/reading the system Paging file
* Writing/reading the Photoshop or other app scratch files
* Managing streamed video from an HD TV Tuner card

My answer to this of late has been to keep my drives as follows:
RAID1 C drive - 2 cheap drives - Windows, Program Files, Registry
RAID5 E drive - 5 cheap drives - Media
Single disk F drive - 1 superfast drive (Raptor) - Page file, scratch disk, and Timeshifting TV files. I also record shows to this disk then move them when done, resulting in one nice contiguous file on the protected RAID array instead of a fragmented mess like it is when first recorded.

It's resulted in my system running very smoothly, and not losing any data either which is rather nice. It also keeps from burning up drives with scratch files you don't care if you lose anyway.

[nick705]

Hard drive prefetching is actually only done in Vista. The "prefetch" in XP has little to do with actually prefetching files, but rather the order in which to read the files for a specific program. I can't say much about *nix systems, since I don't know how they deal with this, but that's exactly what SuperFetch does in Vista.

OK, I guess I should have said "read-ahead caching" rather than "prefetching," but I understand most modern OSes have this capability to at least some degree, although the exact nuts and bolts I'll leave to better heads than mine. The point I was making is that for typical desktop/workstation setups, where the same data tends to be accessed repeatedly, the OS can smooth over a multitude of HDD sins to the point where other system components tend to be the main limiting factor, which seems to be confirmed by most (relevant) benchmarks.

I remember reading an article that SMART data only really helped about 50% of the time in detecting a hard drive failure. For me, the good ol' finger test on the hard drive case does the trick for me.

Yes, I remember reading the same article, but IIRC it also mentioned that a bad critical SMART value corresponds with a very much higher chance of imminent HDD failure - in other words, SMART vary rarely provides a false positive although it often provides false negatives. Not something you'd want to rely on 100%, but useful nonetheless - if you *do* get a SMART warning at least you know the writing's on the wall and you can prepare accordingly.

Maybe not a dealbreaker either way, but I do like my comfort blankets.

I will agree with you on one thing though... figuring out which drivers I needed for my motherboard RAID support in Windows was a major PITA. It took me a few times to get that right. Thankfully, all my current systems are based on the 965 chipset, with the same ICH8R controller.

There is that, but what I was really getting at is the problem of moving an established array from one setup to another. You can't rely on a different controller being able to see the array, even if it's a similar chipset, so you more or less have to assume you'll need to create a new array and restore from backup, which isn't fun if you're dealing with hundreds of GB of data. Not a problem if the array is used purely for OS/scratch/temp use or whatever, but if for example you have 2x320GB drives in RAID-0 as in your sig, it seems a bit of a waste of capacity, and smaller/slower drives would compromise the performance you were presumably looking for.

I suppose at the end of the day the obvious answer is to try it for yourself - regardless of the benchmarks, if you "feel" your PC is faster (or if it actually is for your purposes), and the downsides don't matter, I guess it would be worthwhile. I'm unconvinced *most* users would see much benefit however...

As for hearing the 74gb Raptor itself, you won't hear it in suspension. I have 2 in a P180 case and hear nothing.

I think that shows how much one person's perceptions can differ from another - when I had a Raptor suspended in a Solo, its seeking was still easily the loudest noise emanating from the PC, to the point where it became a major irritant whenever Windows did any of its periodic behind-the-scenes housekeeping. I honestly didn't notice any performance hit switching to a Samsung HD501LJ - maybe Windows booted fractionally slower, but that could have been my preconceptions and it never seemed worthwhile getting the stopwatch out to be sure.

[Nick Geraedts]

I suppose at the end of the day the obvious answer is to try it for yourself - regardless of the benchmarks, if you "feel" your PC is faster (or if it actually is for your purposes), and the downsides don't matter, I guess it would be worthwhile. I'm unconvinced *most* users would see much benefit however...

Yup - everyone has their opinions. I'm a Windows junkie, and now a RAID junkie... haha.

The "noise level" of what people consider quiet varies greatly. I consider my systems at home quiet enough, since they're not a bother. I don't mind the very slight background hum of a few fans - in fact I prefer it, since it's a queue that the systems are still up and running! Secondly - I'm usually listening to music, so as long as the system doesn't contribute to the overall noise level of the room - I'm happy.

In comparison, the SPCR lab is a whole other ball game. At times, you feel like you're somehow going to disturb the people next door by your walking - that's the difference in ambient noise level compared to at home. It's kinda eerie sometimes... (sorry Mike )