First off, do you think RAID-0 is worth it?
second, which drives would you recommend for a raid setup?
Quietest drives for RAID-0
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
A lot of folks around here do not like raid 0. After all, you are using two drives instead of one, which usually means more noise. If one drive goes bad in a raid 0 array, your data is gone, and with two drives there is nearly twice the chance of a problem. In this zone people will use notebook drives for quiet despite the high cost, low capacity and reduced performance.
You should also be aware that if you must reset your bios to its defaults your raid array cannot be accessed until it is re enabled in bios. Usually that is trivial, but once I was careless enough to wipe out a system that way.
Having said that, if you do a lot of things with video, you will really notice the improvement in speed that raid 0 provides. A DVD is 4.3 gigs, and making one involves a lot of file creation and modification in relatively short steps that tend to keep you hanging around the PC. Benchmarks of file system performance show improvements of 50% or more. On the other hand, typical system benchmarks show little improvement.
It is also claimed by Seagate that a single platter drive has a lower chance of a head crash than a two platter drive, so a pair of single platter drives should not double the chance of a hardware failure. Likewise, higher capacity drives create more noise than the quietest low capacity drives. That is to say, two single platter 160's might only have a slightly elevated chance of failure and slightly higher noise level than a two platter 300 from the same series. In this case I am thinking about the new Seagate 7200.9 drives. I have two of the 160's in a raid 0 array. Muxing video and audio into a VIDEO_TS folder is about three times faster than it was on an older Seagate 80 gig PATA drive that many around here would consider to be very quiet.
Bottom Line: If you really need the performance raid 0 is great, but you can expect more noise.
You should also be aware that if you must reset your bios to its defaults your raid array cannot be accessed until it is re enabled in bios. Usually that is trivial, but once I was careless enough to wipe out a system that way.
Having said that, if you do a lot of things with video, you will really notice the improvement in speed that raid 0 provides. A DVD is 4.3 gigs, and making one involves a lot of file creation and modification in relatively short steps that tend to keep you hanging around the PC. Benchmarks of file system performance show improvements of 50% or more. On the other hand, typical system benchmarks show little improvement.
It is also claimed by Seagate that a single platter drive has a lower chance of a head crash than a two platter drive, so a pair of single platter drives should not double the chance of a hardware failure. Likewise, higher capacity drives create more noise than the quietest low capacity drives. That is to say, two single platter 160's might only have a slightly elevated chance of failure and slightly higher noise level than a two platter 300 from the same series. In this case I am thinking about the new Seagate 7200.9 drives. I have two of the 160's in a raid 0 array. Muxing video and audio into a VIDEO_TS folder is about three times faster than it was on an older Seagate 80 gig PATA drive that many around here would consider to be very quiet.
Bottom Line: If you really need the performance raid 0 is great, but you can expect more noise.
Whether or not it's worth it depends on whether the higher perforrmance outweights the higher cost and noise and lower reliability. Only you can determine that.
They're not the quietest drives, but I might recommend the Seagate's diver is using. They've got good performance, and they're probably more reliable than any similar drives, which could be important for raid 0.
They're not the quietest drives, but I might recommend the Seagate's diver is using. They've got good performance, and they're probably more reliable than any similar drives, which could be important for raid 0.
-
NeilBlanchard
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Greetings,
I've yet to see a credible test show that a RAID 0 array actually is faster in a way that matters in the real world. It might be a bit faster, but not by much -- and using a single Raptor instead would swamp the difference. Add to this, the very real increased risk that you will lose everything, due to a failure of just one of the drives...I can't say I would do it.
I've yet to see a credible test show that a RAID 0 array actually is faster in a way that matters in the real world. It might be a bit faster, but not by much -- and using a single Raptor instead would swamp the difference. Add to this, the very real increased risk that you will lose everything, due to a failure of just one of the drives...I can't say I would do it.
I disagree with the statement that Raid 0 has no beneficial effect in real-world usage scenarios.
Raid 0 significantly improves my application launch times. For instance, launching firefox improved from 16 seconds to 9 seconds. (The launch time is dominated by the antivirus on-demand scanning: if I turn off the antivirus software the launch time is under 2 seconds.)
The improvement in launch time applies to all the large applications I use, and in my view is well worth it. (I take regular backups, so from that point of view Raid 0 poses no extra risk of data loss.)
Raid 0 significantly improves my application launch times. For instance, launching firefox improved from 16 seconds to 9 seconds. (The launch time is dominated by the antivirus on-demand scanning: if I turn off the antivirus software the launch time is under 2 seconds.)
The improvement in launch time applies to all the large applications I use, and in my view is well worth it. (I take regular backups, so from that point of view Raid 0 poses no extra risk of data loss.)
Very true since Raid 0 only would help in transfers of large files. We're talking 100mb+. I've only seen improvements in Video editing when this sort of stuff is done, but for the most part people deal with very small files and latency is a huge effect there. So it's really not worth the riks to have such an unsafe raid array for those 1% or less time you spend transfering some large files etc.NeilBlanchard wrote: I've yet to see a credible test show that a RAID 0 array actually is faster in a way that matters in the real world. It might be a bit faster, but not by much -- and using a single Raptor instead would swamp the difference. Add to this, the very real increased risk that you will lose everything, due to a failure of just one of the drives...I can't say I would do it.
Neil,
In some recent tests of the Seagate 7200.9 160g drive the Raptor was only a bit faster. Storage costs with the Raptor are significantly higher. The new 150g Raptor is about $300 as compared to $90 for the Seagate.
Finally, I can tell you that in the real world it really does make a difference, if you are working with video processing. At least that is the only common single user application I know of where it makes such a big difference. But, I do a lot of that, and the time savings are in minutes, not seconds.
No doubt, if your real world consists of email, surfing the web, listening to MP3's, office work, and games Raid 0 will not make that great of a difference in performance, and other factors will favor not using it.
It is possible to speed up video processing by using two drives and sending the output to a different drive from the input. While the speed is very good, remembering where things are and getting the work to flow right is a PITA. Its also going to be the same noise and heat as two drives in Raid 0, but without the worry factor.
In some recent tests of the Seagate 7200.9 160g drive the Raptor was only a bit faster. Storage costs with the Raptor are significantly higher. The new 150g Raptor is about $300 as compared to $90 for the Seagate.
Finally, I can tell you that in the real world it really does make a difference, if you are working with video processing. At least that is the only common single user application I know of where it makes such a big difference. But, I do a lot of that, and the time savings are in minutes, not seconds.
No doubt, if your real world consists of email, surfing the web, listening to MP3's, office work, and games Raid 0 will not make that great of a difference in performance, and other factors will favor not using it.
It is possible to speed up video processing by using two drives and sending the output to a different drive from the input. While the speed is very good, remembering where things are and getting the work to flow right is a PITA. Its also going to be the same noise and heat as two drives in Raid 0, but without the worry factor.
-
Kung Fu Hung-Su
- Posts: 21
- Joined: Mon Nov 14, 2005 1:43 am
I can't believe no one here plays games...
For those that do, RAID 0 is a godsend for game load times. In multiplayer games where you may start running around as soon as the level as loaded for you, RAID 0 is even more of a godsend.
And I'm not concerned about hard drive failure, sure it's double the chance, but it's double only about 0.1%...
Even if it does fail, I'm sure you sensible people back up often anyway
For those that do, RAID 0 is a godsend for game load times. In multiplayer games where you may start running around as soon as the level as loaded for you, RAID 0 is even more of a godsend.
And I'm not concerned about hard drive failure, sure it's double the chance, but it's double only about 0.1%...
Even if it does fail, I'm sure you sensible people back up often anyway
Did you test this in a controlled manner?bkh wrote:I disagree with the statement that Raid 0 has no beneficial effect in real-world usage scenarios.
Raid 0 significantly improves my application launch times. For instance, launching firefox improved from 16 seconds to 9 seconds. (The launch time is dominated by the antivirus on-demand scanning: if I turn off the antivirus software the launch time is under 2 seconds.)
The improvement in launch time applies to all the large applications I use, and in my view is well worth it. (I take regular backups, so from that point of view Raid 0 poses no extra risk of data loss.)
My IE load time went from about 10 seconds to about 2 seconds simply by defragging my drive. Firefox and IE both read lots of information when they start up (which is compounded by the anti-virus scanning) and highly fragged files will be the major cause of a slowdown. If you tested your Firefox load time after a clean install when you went to RAID, then your test is probably not valid unless you timed the non-RAID comparison after a clean install.
You get the same bandwidth bonus using RAID1, and you have redundancy to spare. Though I cant imagine that an extra 4 seconds on a game map would make that much of a differenceKung Fu Hung-Su wrote:I can't believe no one here plays games...
For those that do, RAID 0 is a godsend for game load times. In multiplayer games where you may start running around as soon as the level as loaded for you, RAID 0 is even more of a godsend.
And I'm not concerned about hard drive failure, sure it's double the chance, but it's double only about 0.1%...
Even if it does fail, I'm sure you sensible people back up often anyway
One more thing for everyone to think about. If your motherboard fails, at a minimum it will probably have to be relpaced by one that uses the same chipset, Raid bios and Raid drivers, if not the exact same motherboard, to gain access to the drives.
As for games, I could see how even four seconds could give someone a tactical advantage when playing against others by allowing the player to pick up a better weapon or position before others.
As for games, I could see how even four seconds could give someone a tactical advantage when playing against others by allowing the player to pick up a better weapon or position before others.
This is a recent test looking at Raid 0 and games:
The Quest for Lag Free Gaming
You want faster hard disk performance? Buy a raptor and be done with it.
The Quest for Lag Free Gaming
I think it sums it up. RAID 0 is not worth it. Unless you have a very specific task like video editing of large files or know for sure that a particular application responds well to Raid 0.Conclusion:
What did we learn from this little experiment? RAID 0 does indeed offer improvements, but not without a price. Frame rates with RAID 0 are generally lower. Most likely because RAID doesn't come for free - you'll have to give up a little bit of your precious processor time due to the controller's overhead. Games can utilize practically 100 % of your processor's time, so even a little overhead of 5% can and will cause games to run slightly slower. If you're trying to avoid stuttering, RAID may help but it won't do much. It's no substitute for real memory. It can also shorten load times, but the same increase (or better) can be gain by having more memory. That said, we can see there are actual improvements in load times with RAID 0, though not for all games. So, if you have a faster hard disk (such as the 10K rpm Raptors) but not on a RAID array, you probably get the benefits of higher transfer rates without the processor overhead.
You want faster hard disk performance? Buy a raptor and be done with it.
[quote="Pauli"][quote="bkh"]
Raid 0 significantly improves my application launch times. For instance, launching firefox improved from 16 seconds to 9 seconds.
The improvement in launch time applies to all the large applications I use, and in my view is well worth it. [/quote]
Did you test this in a controlled manner?
My IE load time went from about 10 seconds to about 2 seconds simply by defragging my drive. Firefox and IE both read lots of information when they start up (which is compounded by the anti-virus scanning) and highly fragged files will be the major cause of a slowdown. If you tested your Firefox load time after a clean install when you went to RAID, then your test is probably not valid unless you timed the non-RAID comparison after a clean install.[/quote]
Hi, Pauli. No, I did not do a proper controlled experiment. But I did not suffer the confounding factors that you cite. The measurements on the single disk were not impaired by abnormal fragmentation (I defragment frequently with the Windows tool, and these measurements were after defragmentation). The image on the RAID-0 was a clone of the single disk via Acronis 8.0, not a clean install, so the file sequence on disk should be identical, just spread over two drives. Moreover this experience is repeatable to the extent that I have similar qualitative results on two different computers, one a fast A64 running XP SP1, and the other a 1GHz Pentium III running W2k. (80G Maxtor IDE drives and Promise Tx2000 raid cards.)
Raid 0 significantly improves my application launch times. For instance, launching firefox improved from 16 seconds to 9 seconds.
The improvement in launch time applies to all the large applications I use, and in my view is well worth it. [/quote]
Did you test this in a controlled manner?
My IE load time went from about 10 seconds to about 2 seconds simply by defragging my drive. Firefox and IE both read lots of information when they start up (which is compounded by the anti-virus scanning) and highly fragged files will be the major cause of a slowdown. If you tested your Firefox load time after a clean install when you went to RAID, then your test is probably not valid unless you timed the non-RAID comparison after a clean install.[/quote]
Hi, Pauli. No, I did not do a proper controlled experiment. But I did not suffer the confounding factors that you cite. The measurements on the single disk were not impaired by abnormal fragmentation (I defragment frequently with the Windows tool, and these measurements were after defragmentation). The image on the RAID-0 was a clone of the single disk via Acronis 8.0, not a clean install, so the file sequence on disk should be identical, just spread over two drives. Moreover this experience is repeatable to the extent that I have similar qualitative results on two different computers, one a fast A64 running XP SP1, and the other a 1GHz Pentium III running W2k. (80G Maxtor IDE drives and Promise Tx2000 raid cards.)
On the topic above saying it could increase low times but decrease frame rates in games, that would only be if you were CPU-bound. In many games, you're videocard bound anyway, not CPU-bound. Also, using a PCI RAID controller would remove any possible load from the CPU anyway... not to mention, like if you have an NV4 board, isn't the RAID control done entirely within the NV4 chip anyway, not the CPU?
At any rate, I recently purchased 4 Seagate 160gb single platter drives with the intention of running a RAID 0+1 array. I haven't been able to find specifics on the arrangement in 0+1 with the NV4 driver, but I know like on the very large arrays we use at work, in 0+1 mode, you have maximum read performance because it will read from any of all 4 disks, which is astonishingly fast compared to only having read performance from a single disk or just a pair of disk. You have 4 separage disks for read activity.
Of course, the real question is if they have this properly implemented in the NV4 RAID controller. It would only seem logical to me. I mean, why would you intentionally sideline all read traffic to only the primary pair of disks? So who knows. Parts should show up tomorrow and with any luck I'll have it actually working and stuff within the next week or so. At the very least I know it will be a TON faster than my single Toshiba 2.5" 4200rpm drive in my current PC.
At any rate, I recently purchased 4 Seagate 160gb single platter drives with the intention of running a RAID 0+1 array. I haven't been able to find specifics on the arrangement in 0+1 with the NV4 driver, but I know like on the very large arrays we use at work, in 0+1 mode, you have maximum read performance because it will read from any of all 4 disks, which is astonishingly fast compared to only having read performance from a single disk or just a pair of disk. You have 4 separage disks for read activity.
Of course, the real question is if they have this properly implemented in the NV4 RAID controller. It would only seem logical to me. I mean, why would you intentionally sideline all read traffic to only the primary pair of disks? So who knows. Parts should show up tomorrow and with any luck I'll have it actually working and stuff within the next week or so. At the very least I know it will be a TON faster than my single Toshiba 2.5" 4200rpm drive in my current PC.
Well, they might if it's cheaper. When I investigated the Raid5 support on my DFI NF4, I discovered it was just there so they could claim to support Raid 5. No one would actually use it. It can only handle 3 drives, and performance is extremely poor. It seems there is little or no actual hardware support from the raid controller, just a crappy software solution tacked on for marketing purposes.AZBrandon wrote: Of course, the real question is if they have this properly implemented in the NV4 RAID controller. It would only seem logical to me. I mean, why would you intentionally sideline all read traffic to only the primary pair of disks?
0+1 doesn't take as much processing, though, and it doesn't seem like it would cost much to read from all four disks at once. So hopefully they didn't cripple it for the sake of economy.
Well I finally got my RAID 0+1 setup going. The hdtach benchmark looks kind of weird. In fact it's not really very good, averaging only about 55mb/sec for reads, but with these HUGE spikes here and there above 130mb/sec. SiSoftware Sandra benchmarked at 110mb/sec. Burst speed was 360mb/sec. I might post some screen caps later.
It was tempting to blow away the array and see how it does as a 4-drive RAID 0 for comparison, but I spent 4 days just getting all my software installed and configured and it would kill me to have to do that all over agian.
It was tempting to blow away the array and see how it does as a 4-drive RAID 0 for comparison, but I spent 4 days just getting all my software installed and configured and it would kill me to have to do that all over agian.
onboard raid cards are not full hardware raid. if they were full hardware solutions then there would be no need for drivers, you would see the raid array as an 'ordinary' disks from the operating system.
All the onboard raid solutions in consumer products to date (not counting the server products) need the CPU to do a lot of the work of the raid card.
-Quikkie
All the onboard raid solutions in consumer products to date (not counting the server products) need the CPU to do a lot of the work of the raid card.
-Quikkie
If you are CPU bound or not is of course dependent on which CPU you have and what settings you run. The test I linked to is using the NF4 board. So no you don’t escape the CPU penalty.AZBrandon wrote:On the topic above saying it could increase low times but decrease frame rates in games, that would only be if you were CPU-bound. In many games, you're videocard bound anyway, not CPU-bound. Also, using a PCI RAID controller would remove any possible load from the CPU anyway... not to mention, like if you have an NV4 board, isn't the RAID control done entirely within the NV4 chip anyway, not the CPU?
There is vast amount of information about raid @ storagereview. IMHO the best site for storage info. This thread discussing the latest raptor has some very interesting facts about raid: http://forums.storagereview.net/index.p ... 21621&st=0
I would suggesting reading the whole thread and look out for posts from Eugene; the head honcho at storagereview and a hard drive tester since many years.
Another article on Raid:
http://techreport.com/reviews/2005q4/ch ... dex.x?pg=1
The application oriented benchmarks do not show much improvement for Raid, but the non application benchmarks show dramatic performance increases.
Take your choice, and I am well aware that some people do not like "synthetic" benchmarks".
http://techreport.com/reviews/2005q4/ch ... dex.x?pg=1
The application oriented benchmarks do not show much improvement for Raid, but the non application benchmarks show dramatic performance increases.
Take your choice, and I am well aware that some people do not like "synthetic" benchmarks".
Well the main reason I went with RAID 0+1 was for reliability, not outright speed anyway. You can have any single drive fail and still be in business. Even once you're down to 3 drives, if another drive fails, there's only a 1 in 3 chance you're dead in the water. RAID 0+1 is definitely the most solid.
Of course, my master plan for the ultimate in reliability was dashed by combining a Seasonic S12-600 power supply with a DFI Lanparty motherboard...since they have compatibility problems.
And of course to make matters worse, I'm not 100% certain that if I switch motherboards that it will recognise the RAID array, which means I have to have backups before the mobo swap and hope and pray the new mobo would properly address my existing drive array. Odds of that happening? Probably slim. Option #2 is try a different power supply I guess.
Thanks for all the links above though, I'll definitely read though them. The whole point of my latest build was just to have a totally solid, fault-tolerant PC. I'm pissed like you can't imagine about the whole S12/DFI fiasco. Wish I'd have known to read up about it prior to buying the two.
Of course, my master plan for the ultimate in reliability was dashed by combining a Seasonic S12-600 power supply with a DFI Lanparty motherboard...since they have compatibility problems.
And of course to make matters worse, I'm not 100% certain that if I switch motherboards that it will recognise the RAID array, which means I have to have backups before the mobo swap and hope and pray the new mobo would properly address my existing drive array. Odds of that happening? Probably slim. Option #2 is try a different power supply I guess.
Thanks for all the links above though, I'll definitely read though them. The whole point of my latest build was just to have a totally solid, fault-tolerant PC. I'm pissed like you can't imagine about the whole S12/DFI fiasco. Wish I'd have known to read up about it prior to buying the two.