Intel Matrix..is it a marketing scam?

[manbush]

Sure Intel Matrix gives you phenomal burst rates but burst rate is the rate the INTERFACE can transfer data...not the harddrive itself.

The average reads I seen in the Intel Matrix thread can easily be accomplished with a normal RAID 0. My 80gigs on raid 0 (non intel matrix) were scoring like 100mb/s average read.

So why bother doing Intel Matrix raid at all? Is it because you can Raid 0 and Raid 1 with only 2 drives?

[tuskenraider]

So why bother doing Intel Matrix raid at all? Is it because you can Raid 0 and Raid 1 with only 2 drives? Yep. Or you want to create a small RAID0 array to make pretty benchmarks.

[RangerXLT8]

Sure Intel Matrix gives you phenomal burst rates but burst rate is the rate the INTERFACE can transfer data...not the harddrive itself.

The average reads I seen in the Intel Matrix thread can easily be accomplished with a normal RAID 0. My 80gigs on raid 0 (non intel matrix) were scoring like 100mb/s average read.

So why bother doing Intel Matrix raid at all? Is it because you can Raid 0 and Raid 1 with only 2 drives?
The only use I see in matrix RAID is if the array needs to be split into partitions. Otherwise the plain Intel RAID is fast, probably some of the fastest onboard RAID.

[manbush]

You can make pretty benchmarks with a normal raid 0!!

And you can make partitions in normal raid!

In all honesty, it's just another marketing gimmick like hyperthreading right? (dont get me wrong, i love intel..i own core 2)

I need to know the HARDCORE truth...cause I'm undecided between a new gen 74gig raptor and 2x 250gig segate perpins.

[RangerXLT8]

You can make pretty benchmarks with a normal raid 0!!

And you can make partitions in normal raid!

In all honesty, it's just another marketing gimmick like hyperthreading right? (dont get me wrong, i love intel..i own core 2)

I need to know the HARDCORE truth...cause I'm undecided between a new gen 74gig raptor and 2x 250gig segate perpins.

Well if it's Raptor VS 2xperp drives, then get the 2 perps and RAID0 them, seek time will still be slower though.. BUT, 2xRAID0 Raptors will beat the 2xperps everytime.

[manbush]

If seek time is slower...

then for an example:
1x raptor = avg read = 80 mb PER SECOND seek 8ms
2x perp = avg read = say 120 PER SECOND seek 13ms

the raptor gets 80MB BUT (13ms of perps-8ms of 1x raptor) = within 5 MS, it can gain another additional 80MB therefore getting 160 MB theoritical speed over the 120 MB of 2x perp BECAUSE OF seek time.

Am I right?

[MadMan007]

What you can't do with other on-board RAID (or any RAID controller for that matter?) is mix RAID types by making different partitions on the same disk. If you were to make small RAID partitions on the fastest part of the disks with any other controller I beleive you wouldn't be able to use the rest of the disk. Also there's nothing wrong with doing that, this is actually the way I looked at it as a 'hype thing' at first but then I realized you can still use the rest of the disks too - sort of a have your cake and eat it too situation. If anything it's a much more better solution for that reason than other RAID controllers.

Finally, the real-world 'feel' in actual use is said to be phenomenal by everyone who did the small RAID0+Raid1 so it's not just for benchmarks, although only looking at one benchmark like HDtach is never a good idea. So no it's not just a marketing-hype thing, and yes to the questions at the end of the OP.

[manbush]

Is it really pheonomal as in will I noticed a difference from going from a old 74 gig raptor to 2x 250gig perps? My credit card is near my keyboard tee heehe

[MadMan007]

Read the long Matrix RAID thread to get a definitive answer, I'm pretty sure there are people in there who went from a single Raptor to matrix RAID.

[greenmaji]

First bench in the matrix raid thread here (just two drives)

http://i65.photobucket.com/albums/h201/vbin/HDTachIntelRaid0.jpg

Ignore the burst

look at random access, average read and sequential read.

btw.. I can only assume Matrix Raid has been around since the ICH5R chipset, since its compatable.. And I only heard tell of this when it was first posted here. That would be what a 5 year gap and no advertizing hmmm.. If they were marketing this, they must have had a diffent target market then the ones that include me.

btw.. hyperious went from 4-75gig raptors in raid 10 to 4-perps in matrix raid, and won't go back.
btw.. you might want to ask if those ATTO runs were not with the full drives rather then Matrix raid, he was in a hurry to do the benches, he builds computers for people, he doesnt have time to play with his stuff sometimes.

[tuskenraider]

btw.. hyperious went from 4-75gig raptors in raid 10 to 4-perps in matrix raid, and won't go back. Did he try 4 Raptors in Matirx RAID for a fair comparison?

[greenmaji]

Did he try 4 Raptors in Matirx RAID for a fair comparison?

yes he did.. he thinks the limitation might be the lack of SATA 3gb/s for the raptors.
IE matrix raid's taking advantage of the bandwith somehow were that wasn't much of a consideration before.

[RangerXLT8]

EDIT. Drive for drive Raptors are the fastest SATA HDD you can buy.

[dominick32]

Like a few previous posters stated. No this is not a marketing scam. The incredible burst rates mean absolutely nothing to me. It is obviously a software optimization (enabled by write cachingi) utilizing cache in a different way and it very well could be something we may benefit from in the future, or not. The amazing thing about the Matrix is the ability to split up a hard drive into "almost physical" drive slices, not partitions. That is where you have a misunderstanding. Splicing up partitions for a raid 0 array is much different than using the matrix to slice up physical drives.

Read my Raptor150 review in my signature: http://www.ocforums.com/showthread.php?t=476776

By slicing up the array and creating a physical 20GB slice (not standard partition) I was able to drop random access times to 6.1 ms from 8.5ms and increase sustained reads over 30 MB/s more than using a standard raid 0/. (non matrix)

[dominick32]

And as far as my stance on the raptor/perps issue that has been beaten to death I usually copy and paste this statement:

This is my response for this issue(taken from another one of my posts in a previous thread):

Quote by: Dominick32
When dealing with 74 Gig Raptors, a lot of people tend to forget to specify that there is an older 74G 8mb Cache model and a new model with 16mb of cache and Raptor X technology. Everyone seems to overlook this.

After reading some of the previous posts in this thread, lets clarify a few things:

One single 74 gig Raptor (older 8mb model) = 65 MB/s sustained read 7.7ms access time.
One single 74 gig Raptor (brand new 16mb model) = 84 MB/s sustained read 7.7ms access time.

One single 7200.10 Seagate Perp = 66 MB/s sustained read 13.2ms access time.

I would say that is quite a substantial difference^^^^^^^^^^^

So the answer to some of the previous statements is both yes and no. A single perpendicular will perform slightly better than an older 8mb raptor. Raid 0 will produce very similar performance figures when using the Matrix compared to Perps.

However, when dealing with the newer revision raptor. In both sustained reads and random access time, the drive will perform significantly stronger both in a single drive configuration, and a Raid 0 setup VS. the perps.

Price to performance, I still say hands down that the perps are the winners. However, if you want maximum performance and money is not part of the equation, the Raptors will definitely shine over the perps.

Regards,

[dominick32]

btw.. hyperious went from 4-75gig raptors in raid 10 to 4-perps in matrix raid, and won't go back.
btw.. you might want to ask if those ATTO runs were not with the full drives rather then Matrix raid, he was in a hurry to do the benches, he builds computers for people, he doesnt have time to play with his stuff sometimes.

Greenmaj,
Just let me clarify that hyperius went from 4X74gig (original 8mb) raptors not the newer 16mb revision. Thats why he never went back after buying the perps.

Read my previous response.

Regards,

[manbush]

I already have a WD raptor 8mb cache....should i buy 1x new 16mb cache raptor and raid them?

[tuskenraider]

The only way you'll benefit from the new drives performance is to get two new ones which will be some cash. I'd get a another used 8MB cache drive for cheap and you'll still have a flyin' setup.

[manbush]

but they are both raptors except one has 8mb cache instead of 16??

[tuskenraider]

The difference between the 8 and 16MB drive is about 15% as it has greater average read/write times and revised algorithems in the electronics for desktop use. So if you by a new Raptor it will slow down to the performance of the older model, so why spend the extra dollars on one drive unless you plan to get a pair of new drives. The older ones aren't slouches either.

[bing]

Sure Intel Matrix gives you phenomal burst rates but burst rate is the rate the INTERFACE can transfer data...not the harddrive itself.


That burst is nothing, its just another "nice to have" side effect of enabling the Intel Raid Write Back Cache feature.

Yeah, spotted a few non Intel campers was looking jealously and keep shouting "That burst is nothing !" ... LOL..., actually it is the average and linear read speed that improved significantly. See this pic below for the difference :

53015


Now, the real question , if you were me, which one you prefer ? Especially looking that significant approx. 20% increase in average and linear read speed and also a trivial CPU increase ! This picture speaks for it self and it is pretty obvious isn't it ? :D

Again, to non Intel campers, god damn it, just close your eyes when seeing that burst number ! We are not talking about that silly bust here!

HDTach is a "bad" program to see this difference, its just I'm too lazy to launch my IOMeter or other high level benchies proggies to show you, but almost all of them as showing similiar pattern.

Fyi, by enabling it that WB Cache, it won't even consume the main memory, isn't that nice or not ? :)


The average reads I seen in the Intel Matrix thread can easily be accomplished with a normal RAID 0. My 80gigs on raid 0 (non intel matrix) were scoring like 100mb/s average read.

Currently, you simply "can not" beat Intel plain Raid 0 with other non Intel on board Raid 0, period.

Few points that make it superior which are :

1. The linear and average read speed, Intel is alot better, been there and tested it out my self.
2. Significantly better CPU utilisation, compared with the result it delivered
3. Low OS or raid driver interrupt latencies
4. Low Mobo PCI backplane utilization

Why ?

Intel has been diligently doing their "home work" on improving ICHxR pretty well for the past "few years" since the early ICHxR generations, let me show you proof that they sort of "integrate" this Raid feature in their ICHxR Southbridge.

53016

See that ? That quote was captured in their ICH8 datasheet and there are a lot of other entries that pointed ICH8R has something to do with Intel Raid, so it is not just pure software tricks, and also I have a sort of connection with a hardcore low level device driver programmer that confirmed that they had improved their ICH chips significantly in recent generation. Just be prepare to expect another nice surprises on upcoming ICH9R next year with bearlake nb ! :D



So why bother doing Intel Matrix raid at all? Is it because you can Raid 0 and Raid 1 with only 2 drives?

Yep, simple, until now and today, you could not find other non Intel raid that capable of "Matrix Raid Kungfu". Maybe you'll have to wait until next year for their competitor to mimic that, but by that time they will be too late to beat Intel since they're just a "follower" while Intel had advanced in few steps ahead.

See the bigger picture now on how Intel is defining their grand scale war strategy ? They're not just improving their CPU by releasing C2D only, they're doing it for other as well ..hint...hint.... :)

[MadMan007]

I seriously doubt you would see seek times that low I'm sure it doesn't go down in a linear fashion but in a 'diminishing returns' fashion.

[greenmaji]

I would have to ask him..

I just noticed SR had benched Segates new perp. 15K Cheetah, did a comparo of it, the Raptor 150 and the 750gb perp.

http://www.storagereview.com/php/benchmark/suite_v4.php?typeID=10&testbedID=4&osID=6&raidconfigID=1&numDrives=1&devID_0=321&devID_1=309&devID_2=323&devCnt=3

The 150gb raptors transfer rate is better, but they dont have benches for the 16mb 75gb raptors or the 250 or 320gb perps :-/

btw.. I concure if you break 7ms seeks with a 4 drive setup, that would be phonminal in itself.

[dominick32]

Well, currently my ultimate dream "desktop" storage is using 4 X 750GB 7200.10 perp drives, and slicing out just 25GB from each drive to make a 100GB Raid 0 plus RAID 10 for the left over space. Or even crazy, slice up 10GB each !! :mad:

Expecting a 2-3 ms average seek time :eek: since they're on a very-very tight area and with > 400MB/s average transfer speed since its on the highest dense disk platter (sata disk of course) that money can buy now. :drool:

With that monster, trust me, the current Raptor "fanatic" campers will simply knocked out with bleeding nose on the floor, humbly praising for forgiveness and admitting who is the master ....lol... :D
Bing,
I agree with everything you said except for this. ^^^^ You are most definitely wrong bro and there is no way in heck you would have 2-3ms average seek time. Put 4X150 Raptors against 4X750perps and you are telling me the perps would win? Not only in sustained read, but average seek would be better on the raptors. Its simple physics bro. I do not see seek time dropping any lower than 8.0 ms on that setup you mention. Its just not possible on those drives. You could give me 2gb slices. I do admit, price to performance you cannot beat the perps, but if you want the fastest sata raid setup on the market the raptors win hands down even though most may not be willing to pay the price.

[greenmaji]

bing.. the perp drives have good numbers for low level test for 7200rpm drives.. but for single user preformance and a SATA of considerable size. The Hitachi 500Gb Deskstar 7K500 whips the 750Gb perp drives soundly.

http://www.storagereview.com/php/benchmark/suite_v4.php?typeID=10&testbedID=4&osID=6&raidconfigID=1&numDrives=1&devID_0=297&devID_1=323&devCnt=2

And for lower capasity (less Gb's to shave off for your Raid0) the 150Gb raptor wips it soundly in all respects.

http://www.storagereview.com/php/benchmark/suite_v4.php?typeID=10&testbedID=4&osID=6&raidconfigID=1&numDrives=1&devID_0=309&devID_1=297&devCnt=2

Now I have to deside if four 250Gb perps are worth it after seeing how poor the single user marks of the 750Gb perp :(

I could get two 150Gb raptors or two 500Gb Deskstars instead..

[Max0r]

I think BING just owned you all with his description of Intel raid.

NOW, he is being pretty speculative about the last part of his PhD (pimpin hard drives) dissertation... so I think we should wait for someone to give us some hard numbers :) Now the argument makes sense... due to the extremely high data density.. using only a small slice of the fastest portion would mean that every revolution of the data discs = MORE data flies past the heads, so I think the potential for extreme bandwidth/seektimes is there... BUT can the read/write implement actually keep up and seek things out fast enough, or does it have to wait extra revolutions to seek something? Is this an issue of actual limitations due to speed of electrical impulses too?

I'll be getting a DFI expert soon and I want to do something like 4x320gb Seagate Perp Raid 0 on the NF4 controller, or some other similar drives that are cheap and have decent storage. Maybe Hitachi's, or whatever seems to have the best performance in that class. Comments? How will it compare :)

[bing]

Ok..ok..., I surrender, have to admit that 2-3 ms claim was wayyyy too low ! Apologize for that trolling statement, was in the "naughty" mood when I typed that last part ! :)

Actually I was trying to provoke for somebody to shout out loud challenging me with sort of calculation to proof I was wrong ! :D Just realised it is not nice to sort of making that "trolling" tone, again sorry for that. :bang head

Now, after intensely reading the 7200.10 datasheet and also reading those benchmarks pointed by Greenmaji out there, I'm quite confident with 4 X 750GB 7200.10 with 5GB slices from each drive, they could deliver lower than 7 ms, actually it will be around 6 to 7, not much though ! Again just a rough calculation, this needs to be proven by a real benchmark. Btw, that Hitachi looks hot too, thanks for the link.


You are most definitely wrong bro and there is no way in heck you would have 2-3ms average seek time. Put 4X150 Raptors against 4X750perps and you are telling me the perps would win? Not only in sustained read, but average seek would be better on the raptors. Its simple physics bro. I do not see seek time dropping any lower than 8.0 ms on that setup you mention. Its just not possible on those drives. You could give me 2gb slices. I do admit, price to performance you cannot beat the perps, but if you want the fastest sata raid setup on the market the raptors win hands down even though most may not be willing to pay the price.

Yes, now I stand corrected, no way on earth this thing could get that low ! Sorry ! :)

But, you've launch an interesting challenge. Since its weekend now, give me few hours ahead, even though it is too small for realistic usage, I'll try that 2GB slices on my 2 X 7200.10 250GB disks to see what will happened.

Will get back here with the result, time to refresh my drive image on my C: raid 0 boot drive ! :D

[dominick32]

Yes, now I stand corrected, no way on earth this thing could get that low ! Sorry ! :)

But, you've launch an interesting challenge. Since its weekend now, give me few hours ahead, even though it is too small for realistic usage, I'll try that 2GB slices on my 2 X 7200.10 250GB disks to see what will happened.

Will get back here with the result, time to refresh my drive image on my C: raid 0 boot drive ! :D

Bing,
I respect you a lot on here bro. :beer: You are the reason why I stepped up to the Matrix and I appreciate your early insight. Just for comparison sake, my two raptor 150's using 10GB slices net me a 6.1ms seek time and 175.5 MB/s transfer rate. What seek time do you think 4 of my drives with 2gb slices would yield? :santa:

Anyway, no disrespect intended bro. I truly appreciate you bringing these results to us with the whole Intel Matrix setup.

Cheers.

[bing]

Bing,
I respect you a lot on here bro. :beer: You are the reason why I stepped up to the Matrix and I appreciate your early insight. Just for comparison sake, my two raptor 150's using 10GB slices net me a 6.1ms seek time and 175.5 MB/s transfer rate. What seek time do you think 4 of my drives with 2gb slices would yield? :santa:

Anyway, no disrespect intended bro. I truly appreciate you bringing these results to us with the whole Intel Matrix setup.

Cheers.

Hey ! None taken, infact thanks for the kind word, really appreciate it ! Btw, OOT, hows Jinu's progress ? :D :bday:

Bout that 2gb slice, actually it was trickling myself on why the heck I didn't try it in the 1st place ?

Decided not to use HDTach anymore since its not good enough to represent the head's seek performance mapping and worst it created lot of pessimistic reactions ... lol. Actually HDTach also has similiar figures.

Result of raid 0 with 2 GB slices each from 2 X 7200.10 250GB perp drives firmware 3.AAC, nothing fancy though. :)

53073

Since it is too small (4GB) for normal usage, the only thing that I could think of is improving your PCMark scores ! LOL :cool: :D Anything else anyone ?

Ok, bout using 2GB slices on 4 of your WD drives ? Just post a HDTune screen shot of your current running raid 0, it will be easy to "guestimate" it, prolly 4-5 ms ? :drool:

Also found a few old screen shots that were taken during testing diff. strip sizes in Matrix Raid, I hope this will be helpfull enough for new adopter to stop the arguments about 16K vs 128K strip size at Intel Raid, hey.. this is different beast we're talking bout here ! :D

The pic should speak for it self, ignore the burst since its not consistent from every run and considered as equal. Just watch the CPU usage.

53074

[Max0r]

No comments on my proposed setup?

[fritzman]

No comments on my proposed setup?

There's been some discussion on the differing platter sizes between the 250's & 320's but I'm uncertain as to what the final word is on it. I think it is that the 250's use the same size/density platter as the 500, and that was thought to be an advantage, but someone else will have the definitive answer.

Anyways... even if the DFI doesn't do it for you, you at least have a killer set of drives to move ahead with. Speed is the curious one... the matrix allows you to slice a particular piece of the pie out, which seems to have advantages in real-world performance, but you'll be running straight Raid0... and everyone I've spoken to says that any raid0 rig with 4 x drives is really quick, a greater leap ahead in both performance and feel than the jump from 2 x drive raid0 to 3 x drive raid0.

If money weren't an issue... 4 x 74Gb 16Mb Raptors would have to be amazing, but then I'd be wanting a large external for backin up every night.

[greenmaji]

@Max0r.. Im debating two of these instead of four perps.
http://www.newegg.com/Product/Product.asp?Item=N82E16822147007&ATT=22-147-007&CMP=OTC-Froogle

[Max0r]

Fritz: My friend has a DFI Expert and two 74GB 16MB raptors in Raid 0 and is planning on expanding that to 4! And yes, absolutely, backup will be integral.

greenmaji: Why only 2? =p I want 4 drives for maximum performance. I see Hitachi has 250's for very cheap with Sata 3.0... but only 8 megs of cache.

[fritzman]

My friend has a DFI Expert and two 74GB 16MB raptors in Raid 0 and is planning on expanding that to 4! And yes, absolutely, backup will be integral.

Wowser:eek: ... I think just about anything with 4 x 16Mb Raptors will be killer! 250Gb is a nice size for a drive and that's about what the sum will be in Raid0 and having an inexpensive backup system will be a breeze. Sounds great.

[greenmaji]

greenmaji: Why only 2? =p I want 4 drives for maximum performance. I see Hitachi has 250's for very cheap with Sata 3.0... but only 8 megs of cache.

Its a cost thing.. I kinda planned on $400 in HD's. Those just happen to be second to the 150Gb raptor for SATA single user performance but have 500Gb of capasity (more capasity is more of a matrix raid consideration).
But if you can swing $800 for HD's.. four of those or 150GB raptors would be the ticket :thup:

[Max0r]

Ahhhhhhhhhhhhhhhhhhhhh I see. What comes out as best for around $100 or less per drive?

No way I'm swingin that much :D

[fritzman]

Here's a question for ya...

How am I going to transition from a 3 x drive array to a 4 x drive array on a P5W-DH board?

Once I get the 4 x drives in place, my understanding is, that 'that's it'... I can no longer use the external SATA connector.

It's an interesting one, and the only way I can think, is to have a large PATA rive, and put the two DriveXL images on their, then plug that into the PATA slot instead of the DVD drive, once the array is all constructed.

Value any other ideas that might work.

Only other thing I thought of, was to grab a P5B Deluxe board, that has 6 ports... I assume I could array 4 x of them, and use another single SATA2 drive just to get the data across.

[Max0r]

****, maybe a long-ass PATA cable that takes 2 devices?

Additional PCI controller if you want to use more SATA drive?

[greenmaji]

Ahhhhhhhhhhhhhhhhhhhhh I see. What comes out as best for around $100 or less per drive?

No way I'm swingin that much :D

36Gig 16mb cache raptors, right about $100 each right now.

more capasity and that would be a consideration of mine.
moving up to four 75Gig would bump up the price another two hundred :-/
And if I was going low on capasity I would really want to bump it up to four drives.
8Mb raptors are kinda out of the quesiton for me, again the 500gigers would likely put a hurting on them.

[Max0r]

36Gig 16mb cache raptors, right about $100 each right now.

Need more capacity! :D What's next in line?

[greenmaji]

my guess.. one of these
http://www.storagereview.com/php/benchmark/suite_v4.php?typeID=10&testbedID=4&osID=6&raidconfigID=1&numDrives=1&devID_0=285&devID_1=300&devID_2=299&devCnt=3

depends on what you are going to use the setup most for, gaming, video encoding

gaming, better seeks and big map loading perf.. Western Digital Caviar WD2500KS (250 GB SATA) modle# WD2500KS

video encoding looks good on this one, better transfer speeds.
Samsung SpinPoint P120 no NCQ (250 GB SATA) modle# SP2504C

*edit* missed one
http://www.storagereview.com/php/benchmark/suite_v4.php?typeID=10&testbedID=4&osID=6&raidconfigID=1&numDrives=1&devID_0=285&devID_1=300&devID_2=299&devID_3=301&devID_4=303&devCnt=5

you get the idea.. ;)

[Max0r]

OMG I'm so foolish. You know I was all "performance performance performance!!!!!1111one" but now I have another important consideration: reliability!

The 250gb WDC RE's seem to be a good choice for that, WDC's have always been good to me, but so have Maxtors, Seagates, and Quantums. I'm finding it hard to judge the long term reliability of a drive. But I've had the most WDCs. How are the Samsungs in this category? Seag perps? Hitachi deskstars?

Say, why didn't you include the Seagate perps in your consideration?

Help me make sense of all this! XD!

BTW I already have 2 of the WD2500KS drives... not bad so far :) As you can see in the sig.

[greenmaji]

they only did a review of the 750gig perp, and the single drive info is lacking for the 250.
I wasnt all that impressed with the 750's single user performance :rolleyes:

If you already have 2 of the WD's, why not get 2 more?

[fritzman]

Anyone using a P5B-Deluxe board with more than 4 drives?

Wondering about the viability of having a 4 x drive matrix (split per the Bing formula) and the ability to attach another Sata drive for data transfer later should I ever need to break up the array/take a drivexml image of it, etc.

[Max0r]

they only did a review of the 750gig perp, and the single drive info is lacking for the 250.
I wasnt all that impressed with the 750's single user performance :rolleyes:

If you already have 2 of the WD's, why not get 2 more?

What about Bing's results and such? Also, the Seagates as well as certain WD's have a 5 year warranty, which means the company is confident in their reliability. I have two seagate drives, one of which I had for over 4 years and heavily abused, and another which I've had a couple years as well, and used as a 24/7 download drive (torrents and such). And they're still runing without a hitch. The WD's I have also are running without a hitch (although 1 has encountered a couple bad sectors (my oldest one that I've had a couple years, but it is an early revision, not the black colored final rev).

So what I'm saying is... it seems to me that the perps have good sustained read performance (not good seek times tho), and that they should be just as reliable as the WD's, which to me makes them a serious candidate.

As for the WD's... do you mean get 2 more and do a quad-raid? :) Or do you mean based on how much I'm already liking them, it makes sense to get a couple more? ;) I'm seriously considering the RE versions tho. Those seem to have an even higher prospect of reliability than the KS. I haven't had the KS's more than a couple months... so I can't tell just how reliable they are, but I have maximum confidence in Western Digital drives.

Then we have the Hitachis and Samsungs, brands I've never messed with before. I wonder if anyone here has had those brands a long time and can speak for their reliability.

I will probably end up only doing 2 drive raid... I have been uh.. disheartened to pay twice as much and get vastly diminishing returns :( Plus life is biting me on the ass financially.

[Seven]

Honestly, I've never heard of Hitachis or Samsungs being bad, people generally seem to have the most problems with Maxtor, although I did have a personal friend have 2 WD drives come in DOA.

Maxtor is just big on the "oh ****, your data is gone!" approach.

[greenmaji]

What about Bing's results and such?

Exellent low level performance, but no single user benches because thats a bit beyond my knologe or I would have instructed him on how to do so when he started posting here ;)

I'm just going by SR's single user benches, he developed them for his testbed with different applications and macro's to perform the same set of tests to every drive he puts the image on.

Besides, any set of HD's in matrix raid = :drool:

If your really worried about reliabity, sell them near the end of the warrentee period and buy new ones.

[bing]

Exellent low level performance, but no single user benches because thats a bit beyond my knologe or I would have instructed him on how to do so when he started posting here ;)


Green, what benches ?

[Max0r]

Yeah green... what benches? Huh? huh? >=p

[greenmaji]

Green, what benches ?
Yeah green... what benches? Huh? huh? >=p

http://www.storagereview.com/articles/200510/Testbed4_3.html
http://www.storagereview.com/articles/200510/Testbed4_5.html

[Max0r]

Thanks:)

Hmmm KS is performing nicely. I bet the YS performs similarly... don't they perform better for games and other single user stuff when native command qeueing is turned off (I mean vs NCQ on, not vs KS)?

I think I'll go either for 2x2500YS or 2x3200YS. About same $/gb so might as well go with the big one.