[NEWS] - Raptors in RAID articles

[CougTek]

Some interesting anomalies - the drives tested (purchased retail) perform on par with the BETA drives tested by SR and X-Bit - which raises the question, did some BETA models make it into the wild, or has WD supplied SR and other review sites "optimised" drives (a la NVIDIA)?

Hmmm...

I love a good conspiracy theory. Now where did I put my aluminium foil hat...

News Source

Submitted by LiamC

 

[honold]

this is all over sr

i, too, have a 'beta performance' raptor

 

[CougTek]

Honold, now you impress me. You posted your reply not even fifteen seconds after I posted the thread. Maybe not even ten. Must have a HyperTransport bus between brain and fingers...

 

[honold]

just lucky timing i guess

i'm 25 and i've been in chatrooms/bbses/forums since i was about 12, so my touch typing skills (never learned proper typing) have gotten pretty extreme.

 

[LiamC]

Honold, not wanting to read the entire SR thread, can you give a brief synopsis? What's going on with the Raptors? Anyone?

 

[honold]

the performance of all known shipping drives is identical to that of the beta firmware tested by sr, not the final firmware tested by sr. the allegation is that sr received a 'golden' firmware rev.

http://www.storagereview.com/articles/200303/20030320WD360GD_2.html demonstrates the biggest difference

 

[LiamC]

MWAAAHHHAAAAHHAAAAHAAAA! SR got owned! :rofl:

There needs to be checks and balances lest you (the one publishing a review) become a puppet. Needless to say, I will be very wary purchasing a Raptor.

 

[zx]

That's why i did not buy a raptor and went scsi instead. It's IOMeter scores were too low for 10k. And you can't fool IOMeter. Something was wrong with this drive and i decided not to go with SATA 10k drives before more models are released.

I'm about to recieve a Fujutsu MAN drive I bought at a very low price via e-bay. Yes, it's desktop benchmark scores are weak, but that will give me the chance to compare it with a WD400JB to analyse performance in real life, both for general usage and for more special cases.

 

[LiamC]

zx,

want to post your thoughts on it when you've run your tests?

 

[zx]

Sure...next week.

 

[honold]

sr doesn't use iometer because they don't consider it a consistent metric of use...

 

[CougTek]

SR does use Iometer for their server-oriented benchmark (file and web server). They use IPEAK to monitor the results given by Iometer.

 

[honold]

right, and i don't suspect he was buying it for a server...

 

[CityK]

honold,

1) How would you characterise the performance behaviour of your Raptor? ... especially in comparison to that provided by your JBs

2) Outside of the fact that a specific benchmark performance anomonaly has been observed, are you impressed by the drive, or do you find it falls short of all the hype.

CK

 

[honold]

well my jbs were in raid1 on a 3ware controller. 3ware controllers are treated as scsi, so i had the whole 2k/xp scsi mess... their new firmware/drivers improved things a bit. to answer both questions at once it feels decidedly better than the jbs, and i'm not disappointed with the performance even though i'm 'beta'.

whether or not my tune would be the same with the jb in a single drive on the motherboard's ata controller, dunno. i don't recall being that blown away by it during the brief time i used it before the mirror.

i've also gone from 2.4ghz to 3.1ghz, pc800 to dual-channel low latency ddr416, 875 chipset, etc so those are factorings. i can generally 'feel' the storage though, just in stuff like browser cache lookups and file browsing.

 

[CityK]

Just noticed that Eugene wrote that he is going to formally address the Raptor enigma shortly.

I'm looking forward to see if there are performance difference outside of the STR....although by the sound of Eugene's tone, it doesn't look like any serious differences exist.

CK

 

[flagreen]

I've never been able to duplicate SR's test results with any drive I've bought and tested. I can come close but always fall a bit short. I have assumed it is because of the "clean" way in which Eugene runs his tests.

 

[Santilli]

With Eugene and David, any ide drive misinformation is inconsequential.

Seriously, isn't the raptor a huge jump in seek time for ide? I don't care about no stinking sustained transfer rates...

:wink:

gs

 

[blakerwry]

I've never been able to duplicate SR's test results with any drive I've bought and tested. I can come close but always fall a bit short. I have assumed it is because of the "clean" way in which Eugene runs his tests.

my tests are always very close, within a few percent on the seek and STR. I have never ran the specific aplication performance WinBench tests.

 

[CityK]

Well I'm not sure about the other tests, but I know for a fact that SR's Winbench STR measurements have been almost identical to my own....of course, I've only drawn my testing from a limited pool of drives

CK

 

[Santilli]

I used their tests as a a mild indicator of a general performance for drives.

My own experience showed many other factors lead to a persons ability to
notice performance...


GS

 

[blakerwry]

very true, in both server and desktop roles.

the 1st thing that comes to mind is someone with alot of junk installed... no matter how many I/O's their disk can provide it will still be sluggish compared to a clean system. Using faster drives will only help reduce the symptoms of the junk.

 

[Santilli]

This may sound really weird, but I just put in an adaptec 2940, with 4 drives on it, external, and my system seems snappier at start up then before. Everytime I've connected the raid box, I've been amazed at how fast my first couple programs open, and function.


That said, I do do regular mat. delete junk stuff, spyware, etc. get rid of unused programs, watch my reg with Reg Cleaner to make sure nothing new gets added, and keep it lean, and clean.

Perhaps each time you add a pci card that has an efficent processor on it, you help your machine function, because the processor takes over I/o transfer that would otherwise be done by a cheaper mobo chip?

In other words, you have chips on the mobo that are responsible for PCI bus speed transfer. Each time you add a processor that affects that bus, you can increase the bus speed. In particular, when you add a card, in this case the 2940, that really doesn't do much. The drives are connected, but other then that, it doesn't transfer much data.

This is all theory on my part, or hypothesis. Anyone that knows if I'm even close, jump in anytime...

s
:jumpin:

 

[Pradeep]

I don't think that any PCI card would increase the bus speed. But perhaps it could lessen the impact.

That's why I won't buy any mobo with less than 64 bit PCI. It's insane to cripple a healthy modern processor with 133MBps 32/33 PCI. Yeah I know the on board crap can have it's own channels to the cpu but I still like the potential of slipping in a nice 64bit card and going crazy with data

I've got a 64bit FW800 PCI card and matching external 3.5" enclosure on it's way. Will post results when up and running.

 

[Santilli]

It's a weird time. I've always wanted to setup a 64bit/33/66/mhz bus board.

Problem was at the time I wanted to do it, it was WAY too expensive.

Other problem is the raid card I bought is 32 bit/33mhz, not the 64 bit, backwards compatible that I thought I had spent 700 bucks on.

I'm really wondering what the next generation of scsi will bring, what sata is going to work like, and, I would, if the prices become reasonable, like a 64 bit, dual processor setup. Last time I priced it, I ended up with a single processor. Just wasn't worth it.

We will see what the future brings...

I'm waiting for something like 5 ghz to upgrade, and, I would like a system capable of carrying something like 4-8 gig of ram, like the new macs should be able to do.

s
PS What I meant by increase bus speed, was actually make it function closer to optimum, without any slow down. I suspect the mobo I bought wasn't designed for server type data speed, though it does appear to go up to 110 mb/sec.

 

[blakerwry]

my guess is that the drivers are having an effect globally on all I/O adapters.....