Raptor questions
I recently got a WD Raptor to try out and now have some questions:
- Must you use their cable and a legacy molex power connector for them to work correctly? I tried with several different SATA cables and a SATA power connector and it refused to work without errors and was extremely slow.
- Are there any tricks to squeeze more out of them as single drives?
- Are they really worth it or would 2 regular SATA HD's be just as good for everyday use in a RAID-0?
- How does the ATTO below look for a SINGLE DRIVE on an SI 3112 onboard controller?
0
Comments
Gigabyte GA-7N400 Pro 2 Motherboard
XP2600
2 x 256Mb Corsair PC2700
Onboard Promise Raid controller.
Sata Power connectors.
Imo, they perform a lot better than most IDE drives.
ATTO Benchmark:
Did you ever try benching them as single drives? I always test individually prior to putting 2 drives in a RAID-0 to make sure each is running well. I am also concerned about mine not running with the SATA power connector.
Never had a prob with the power connectors, they were already fitted on the Antec Truepower 430W PSU.
Are you using PSU convertor leads? Maybe they're faulty, although that's doubtful.
Dunno, really, what the prob may be there.
Found the problem. It is actually the SATA headers on my mobo. I tried hooking it up to my HPT SATA RAID card with the SATA power hooked up on the drive and all is well. Now I'm ticked about the headers on my mobo.
Dan,
I ran it on Sandra and got 38,000 and an average seek time of 5.6ms so I guess it works.
In my experience, being someone who's run many different RAID 0 arrays, including one with a pair of PATA WD 1200JB's, I can say that the Raptors are considerably faster to anything I've previously run in a RAID 0 environment. SATA or not. They do however also really shine as independent drives.
So don't feel you have to make the jump to RAID 0 to get the most out of them. On the contrary, the RAPTORS are at their best (IMO) when run as single drives.
Anyhow, I've never been very impressed with the Raptors (at least the 36GB ones). Why? Well, for starters, here's an ATTO of 2 160GB Maxtor DiamondMax 9 8MB/7200RPM/ATA-133 drives on SATA converters in RAID 0 on a HighPoint RocketRaid 1540:
<img src="http://www.short-media.com/forum/attachment.php?attachmentid=1590&stc=1>
SATA can be wired and run by BIOS as a PCI-rate base bus process for storage, think this is what happened. It is a shortcut used on non-server boards way too often. Bus rate itself can be limited, plus use of main RAM for buffering can limit actual end-to-end benching if RAM is not fast and fairly hugely present, even with better boards that have a more dedicated SATA pipe in them.
I'd almost BET the card has buffers (RAM like moules or wiring) builtiin while the onboard depends on main RAM. So main RAM load over time is less than it would be for onboard, the buffers absorb it and can let the drives burst from and possibly back to card if buffers are enough in size. REAL good cards can use RAM modules that are faster than normal RAM, and have large RAM style buffers. They will not be as fast as the GRAM used on real fast video cards, but....
To get best RAID 0 performance on SATA, you need a multichannel SATA card and each side of mirror on a seperate drive channel, or bus gets double-loaded with both main writes to target and writes to source. then you get bus inneficiencies that can wipe your SATA advantage. The first SATA cards were single channel cards.
This is similar to SCSI III performance differing alot from card to card and to embedded where a specifically storage bus is present versu where common bus lines are tapped into to be used, drives on different LUNs with chipsets that had different physical routes to drives by LUN had them basicially on seperate channels. Single LUN circuit used for both LUNS got double the load in mirroring or other RAID, the cards stunk on average ebcause RAID 0 writes both ways at close to same time very often and you get main changes triggering mirror writes to the copy.
Unless they happen to be Western Digital.
I have two SATA WD120's in a RAID 0 on another machine and I had a helluva job getting them recognised in the RAID controller. It didn't help that at the time Western Digital's tech help site was down.
After a lot of trial and error, I finally hit on the right jumper settings to get both disks recognised as a RAID 0. It was no jumper on one and position three on the other.
Apparently this is a quirk of Western Digital that has caught out a fair few people.
The raptors are very optimnized though for desktop performance. So if I want faster I go scsi as for what I use a high performance disk subsystem for they end up being much faster. But for normal desktop type crap like opening word.... surfing the web etc... most people couldn't tell the scsi from a fast ide. The scsi is smoother for me. The temp files and internet temp and swap/pagefiles are all on seperate disks trying to divide the i-o across the channels and drives. But for normal crap the machine feels fast but its not like its turbo charged or anything. The only time I felt a noticable differance was with a bunch of fast scsi drives in raid-0 where the access time was down around 2ms and the transfer rate was kissing 400,000. Then yes.... you just sit there and grin because its so startlingly fast even for normal stuff when the disk is defragged right it was just kick as*.
Tex
At the end of the day, it boils down to user preference, is the user willing to pay so much more just for that little bit extra speed? What's a realistic economy ratio for speed vs space? It's all user preference. I love my 36Gig Raptors, but they cost me a lot of dosh. It wasn't an economic buy, I just wanted the best at the time, no matter what. To me, that mentality works, but for others it doesn't.
But like folk have pointed out, they're are some nice alternatives to the Raptors at the moment.
And my desk sits with a 7 foot tall 19" rack filled with computers 3 feet behind me, filled wioth dual cpu servers with big and fans and scsi racks with big fans so unless your raptors are really really loud you still couln't heasr them.
I meant both computers running and you can use a kvm to switch back and forth. I have won more money from extreme techy guys on bets like... I have five computers in the room here now, all on a kvm... You can switch back and forth easy with all the same apps on each machine. They range from dual opterons to 3400+ Athlon 64's to dual 2100 XP's dual p3 1000's to a single 1700+ on a kr7. All on a gigabit LAN. Allhabe a gb of ram but one. The game is you can not benchmark but can do any normal activities like surf the web... open documents or spreadsheets etc.. I have no games but using Photoshop or Office apps etc.. You have be able to pick out all 5 box's and corerctly by telling me what cpu's are in them. Everybody laughs and says they are in but no one has taken any of my money yet.
Never even yet had anyone get more then 3 right. Normal is one or two right. A fast scsi disk can blur which is whcih. For normal desktop playing there isnt a huge differance if everything is setup right and you have lots of memory. For web surfing and normal stuff you couldn't of told the old dual p3 server from the new dual opteron. Gaming is a differant story. Just like me running Oracle its pretty easy to pick them out by memory and disk subsystem. But for a normal home user doing normal desktop chores a dual opteron is fun but its just not that much faster in real life for most real uises that a home user would use it for, and I'm a benchmark whore. But new toys are fun. But trying to justify the need for them isnt so easy if your honest.
I don't game. I couldn't tell enough differance between a nvidia gforce 5200 FX then a 32mb radeon 7000. Not in what I do. And i'm sitting in fronbt of a nice 21 inch hitchi monitor. But I don't game and that would show up big time there. But to see the sales guys at comp usa and fry's etc.. selling grandma and grandpa a new video card and saying they need the new $350 video card makes me want to puke. A normal user doen't need anything like that either IF they don't game or do image or video editing or something beyond the realm of the normal home user.
The two computers here which are normaly picked as the fastest both have scsi raid and the second one is either the sloest or next to the slowest cpu's Its the old dual p3 with 64/66 slots. Everyone pays all this money at times to overclock and push a machine untill its unstabel to gain an edge and the disk subsystem is always the slowest link for any non-gaming activities. For non gaming a user would be better off getting a fast disk subsystem then a $400 video card.
Tex
And well said Spinner. I managed to secure a couple of used Raptor 36 Gig drives, so didn't pay full whack for them, and I'm pleased as punch with them.
I notice my ATTO test result was only a little higher than the ATTO test for the two Maxtors posted here. But it's that little bit that makes the difference
The main thing is, I can actually notice a difference in everyday use, which to my mind is the main criteria.
And I've just added two x 512 Crucial PC3200 sticks and an XP3200 to the system, it fairly flies along now
And that's it, for now, next upgrade will be looking towards a 64 Bit system with a PCI Express vid card but I'm going to do my best to at least use this system for a year. We shall see
Absolutely, and I tell that to everyone I meet who asks my advice. The primary bottleneck in the modern day PC is still the hard disk, and I can't see that changing any time soon.
Hence why for the last 3 years my primary rigs have always used a RAID 0 array.
And here is my single maxtor. They hit 115,000 in raid-0. Are dead quiet and I got 120gb for 90 bucks. Hard to beat.
Tex
Sweet
Anything running on the system during that test?
At the time I ran the ATTO test for the benchmark I posted above, there was all kinds of stuff loaded, all the usual Office stuff, grafix programs and about a dozen games loaded. All kinds of clutter. And it was crunching UD proteins.
It may be interesting for me to run a test on a barebones OS.
Still, can't deny that with the latest drives, the Raptor's advantages seem to be diminishing.
But I still love mine
I don't trick out a empty system to bench. I load an OS and all my apps with driver updates and stuff. The bottom right hand tray is full and anti virus progs and crap is all loaded but I don't test with folding running or anything. You should test a system in the same state it will be used in or whats the point? We won't be giving out prizes here or anything. If we were I would post crap like this...
Tex
That's OK, cos I wasn't expecting one.
And your point is?
I just contributing to a thread here, is all.
You're obviously far more clued up than I.
On the other hand.
Thats all.
Tex
P.S. The last benchmark was unrealistic really. In way it is and a way it isnt I guess since it does reflect real life performance. Thats from a single scsi drive. But the test isnt fair really because of the controller used. Its got a really fast cpu onboard with 512mb pc3200 DDR cache onboard the controller. So ATTO fits into the cache and probably never hit the disk at all. We tested the controller but not the disk.
The new are supposed to be not only bigger but faster also. Sweet. I have raid onboard the MSI neo with the Athlon 64. I'm thinking about getting a couple sata's to play with but have not decided if I want to drop that kind of bread on a raptor or just get two big sata maxtors like the one I posted the bench of.
Tex
The Maxtors, going by what you said also, are probably quieter, obviously cooler and much more of an economical buy because of the additional capacity.
Pretty weighted towards the Maxtors when you look at it like that isn't it?
So why would anyone buy a Raptor. Well...
They are faster, even if not by much. They are made by Western Digital (IMO, WD are a better manufacturer with regard to quality control). They sound much more hip.
So three points, the last seemingly irrellevant, but really not in today's age of poncy show offs. I know, I'm one of them.
So the bottom line here for me is...
Economy just ain't cool. That's why the Raptors sell even though they are very pricey. Oh yeah, and of course, because they are the fastest. It doesn't matter to people by how much, it's just whether it is, or it isn't.
Make sense?
And I have seen as many people grumbling in the forums over failed WD drives as maxtors. Thats more of a personal preferance and I understand that.
My view is if your going to try and top the drives like the maxtor I posted which atto's as fast as a raptor but has a slower access time, is quieter and cooler then you might as well bite the bullet and go scsi. scsi has so many advanatges over sata you can't begin to compare them.
Cheers Spinner... and tell floppy I wasn't picking on him again. I was thrilled to see him posting here at all. I really liked him back at the old site. I considered him a friend.
Tex
Oh yeah, personal preference all the way. One can only speak by experience.
I think you've just told him yourself.:)
Thanks mate.
What did you pay for 36gb Raptors? more then a hundred bucks?
maybe it's time to rethink things.
Most nice scsi drives have a five year warranty. Most OEM ide/sata's have 12 months and are not intended to run 24/7. Some retail boxed one do have a 3 year warranty. I shut the system here down when heavy lightning comes through for storms. I have run servers here for up to 3 or 4 months and never rebooted, and never powered it off. This just isnt the strong suit of ide/sata. They are designed to be inexpensive drives for desktops primarily as thats what their firmware is tweaked for. Thats why they hang close to many scsi drives on some benchmarks. Because the benchmarks are geared to desktop activity. And I don't mean running three heavy background process's but rather running one process at a time. The more you put a heavy load onto scsi the better it looks and the worse raptors do because they were carefully optimized them for the desktop not multi-tasking or multi-user disk usage. So if you adhere to their projected usage patterns they perform very well. I have scsi tools to tweak my scsi firmware for differant environments. Can't do that with sata. The thing I don't like about the raptors I have touched is that they don't all seem to perform the same even with a carefully tweaked environment. You can see atto's ranging from 48,000 to 60,000 even after you mess with them a lot. Every one of the maxtors I HAVE TOUCHED will hit 57,000 to 61,500 on the outside edge. Thats a substantial differance. Would you be thrilled with a raptor hitting 50,000? Didnt think so.
tex