raid or raptor

[regan1985]

yeh a riad setup is cheaper which is wot i want as i can use the spare money to buy a 400 sat2 drive for my other free port!

however im finding it hard to find a 16mb 80gig drive?anybody?idea if the is in in the uk!

 

[t_ski]

smallest I find is this one

http://www.newegg.com/Product/Product.asp?Item=N82E16822136062

 

[Alec§taar]

We shouldn't be agreeing on HD topics

Well, here & there is "ok" I suppose!

I don't think the warranty is really relevant. The price of the disk is nothing compared to the data on it. If it dies a new drive won't make you happy, you still lsot your data.

So if you go RAID, at least be sure to back up everything often.

That's the KEY point, your last one: Backup, always... we all know this. Question is - how many folks here (primarily gamers mostly from what I have observed) have data that is 'irreplaceable'? Still, again, BACKUP like mad!

A raptor is a expensive paperweight, a perpendictular recording drive is faster on reads, as well as a single 300Gb Maxtor-Seagate drive. There is a thread here to prove it.

On reads, yes... on access time, no. Raptors outright WHIP other disks of LIKE class (EIDE/SATA/PATA) here due to their 10,000 rpm speeds... AND, the way I have it setup on a caching RAID 0 controller? On CPU usage as well (0%)!

That same thread (here):

http://forums.techpowerup.com/showthread.php?t=16252&page=13&highlight=Tach

Proves that much... (that was the last day I charted the HD Tach test, & the proof is 2 posts down of my statement).

Raptors are for itty bitty files that are all over the harddrive. Not contingious large files like movies, games, any sort of anything other than a small system file.

Which is EXACTLY what I am doing here, while writing code... largely TINY files used by compilers in program generation

It always "boils down to/distills to" HOW you use your system mostly.

But access rates are everthing when it comes to games.

Yes, especially for saves (init. loads too to an extent) & swap of data (which games do, especially in low memory situations) during gameplay, & level changes...

To answer the cache question, yes more is better. Always. And for Raid 0, I have two 80GB Hitachi drives that are nearly as fast (97MB/s reads) as two 36gb raptors in raid (99MB/s on same controller, same setup).

Absolutely, & note: After WD started putting 16mb of cache onto their ENTIRE Raptor series (yes, even the 36gb & 74gb latest have this over their init. predecessors which only carried 8mb)? Everyone else followed suit...

raptors are the best choice imo. I would go with 2 36G raptors with 16mb cache wich are quite cheap or if you want more space,

Price is right on them (bit higher than others, but you get 10k rpm access/seek + 16mb of cache on them, & you're proof of my statement on cachesize above on smaller raptors currently also) too...

16mb is better than 8mb and also the 16mb version has ncq and native sata support. Sure it doesn't have sata2 but sata2 is useless as no drive can max out the 150mb/s sataI capacity. also the 8mb is noisier.

Good point... burst reads MIGHT take advantage of SATA2, but NOT CONSTANTLY!

I have yet to see anybody hit 300mb bursts constantly in use as well.

APK

 

[Steevo]

I hate to disagree Alec but the hard drive hos no bearing on CPU useage. My Silicone controller and many others use the CPU to computer the disk algorythim.

 

[strick94u]

Cheetah 15k.4 ST3146854LW
Features Description
15K RPM spindle speed Delivers more I/O's per second than 10K, and faster sustained transfer rates, enabling the same high performance to be met with fewer drives.
Ultra320 SCSI Up to 320 Mbyte-per-second throughput and faster sustained transfer rate
Background Media Scan Proactively scans the media for potential defects during drive idle time without using system bandwidth. It enables incipient errors to be corrected before data is lost.
Enhanced Error Correction Code Better maintains performance throughout the life of the drive and reduces the probability of lost data.
Fluid Dynamic Bearing Motor No ball bearings means reduced acoustics and improved reliability.
Full Duty Cycle Reliability The Cheetah 15K.4 is designed to have the highest reliability, even when in used in demanding applications around the clock, 365 days per year.
Capacity Size: 146 GB
Speed: 15K rpm
Seek Time: 3.5 ms avg
Configuration: Discs: 4
Heads: 8
Nonrecoverable Read Errors per Bits Read: 1 sector per 1015
MTBF (hours), AFR: 1,400,000/0.62%

And its a steal at 1,100.00 us$ And yes I cut-n-paste this

 

[Steevo]

SAS is faster.

 

[Alec§taar]

Promise SuperTrak Ex8350 has onboard CPU for HDD I/O Processing, less SYSTEM CPU USE

See my subject line-title above for this reply, & these specs from the Promise SuperTrak Ex8350:

# Intel IOP333 XScale I/O processor (500 MHz)
# 128MB DDR ECC on-board controller cache

That first one? Makes me offload my mobo CPU, to 0% CPU use in fact, & our tests here showed it in fact:

http://forums.techpowerup.com/showthread.php?t=16252&page=13&highlight=Tach

CPU USAGE CATEGORY & MY SYSTEM'S PLACEMENT IN IT of 20 total testers: avg - 3.46% & highest cpu usage was 11%:

1. Alec§taar=0% - (2xWD "Raptor 'X'" RAID 0 (Promise Caching 128mb ECC RAM + onboard CPU Controller))

Our own test here w/ HD Tach, also proves that much... (that was the last day I charted the HD Tach test, & the proof is 2 posts down of my statement - 300% to 1100% variances in fact, in MY favor!).

I hate to disagree Alec but the hard drive hos no bearing on CPU useage.

Ah, but it does in mine (as I specifically stated above in fact "how I have it set up") - ALL via the controller I use...

E.G.-> It not only has 128mb of ECC cache onboard it, but it has its own onboard "CPU" to handle any & all disk related request activity... lessening the system CPU usage to drive it, via its filtering driver taking interrupt requests!

Interrupt request for DISK I/O that MOST systems have their mobo CPU drive - this one directs them @ the HDD Controller CPU for disk I/O by INTEL onboard it!

(0% System CPU use was shown in the URL above in our tests WE DID HERE no less, & this is HOW it was done: A good driver redirecting Interrupt requests to the HDD Caching controller card's onboard HDD I/O processor by INTEL, instead of the mobo System CPU)



The type of controller I use? Server-Oriented more I would say because it is like high-end UltraScSi controllers this way, & using its OWN dedicated CPU onboard for disk I/O processing!

It is a LOT diff. than most RAID controllers (cache & onboard CPU dedicated ASICS for it) folks here use...

APK

P.S.=>

My Silicone controller and many others use the CPU to computer the disk algorythim.

IN yours yes (Silicon Image variant, lol, "SILICONE")...

Not mine (look up the model # @ Promise Technology if you wish & see its specs)... if the test results above don't prove that to you that is... mine's pretty "advanced" over most by how it works is why.

(It's WHY I picked it up: TO OFFLOAD THE SYSTEM CPU, because it has a dedicated "brain" of its own for disk I/O processing!)

By lessening SYSTEM CPU usage? Other tasks get the interrupts they need, MORE, from the SYSTEM CPU (for stuff like gaming or other tasks)... apk

 

[t_ski]

I believe what Steevo is saying is correct, that the hard drive has no bearing on CPU usage. It is the controller, not the hard drive, that steals CPU cycles. That is why a good controller like Alec§taar's does not use the CPU - it has it's own built-in processor to do the work.

So again, it is not the HDD, but the controller that needs the CPU.

 

[Deleted member 3]

SAS is faster.

Ah yes, SAS is the new goal. Wait two generations and buy a cheap used SAS controller and hang cheap SATA drives to it. It'll be great.

 

[Alec§taar]

I believe what Steevo is saying is correct, that the hard drive has no bearing on CPU usage.

I never said it did DIRECTLY (the actual HDD case mechanism)... which is why I wondered WHY he stated what he did in reply.

That machine (all the HDD really is) is driven largely by driver programs & the OS HAL layer only talking to/with/from it.

(HOWEVER- Some internal to disk buffer logic in PROMs on them may have SOME to do w/ performance though, keep that in mind).

AND, even on systems w/out their own "HDD I/O Brain" onboard (as my system has on its disk controller)?

The disks themselves STILL have nothing to do with CPU usage!

(However, their drivers (in cases like mine too even) are run off the system cpu)

HOWEVER: Drivers that do not let another cpu cycle offloading controller do the disk I/O related work (as mine does) not only use the CPU to drive that disk I/O's via running their drivers, but use more system cpu also for running other I/O processes via interrupt requests directed to the mobo system CPU, rather than one that offloads the system mobo CPU).

It is the controller, not the hard drive, that steals CPU cycles.

And, their drivers also, & this is ALWAYS the case... even on Silicon Image setups for RAID built into the mobo as an ASIC/firmware... moreso than mine (has own disk I/O cpu by INTEL), but even mine has to run a driver on the mobo system CPU, & steals SOME cycles for that from the mobo system CPU, but negligible by comparison to most RAID 0 onboard mobo firmware ASICS.

That is why a good controller like Alec§taar's does not use the CPU - it has it's own built-in processor to do the work.

UltraScSi higher end controllers tend to do this as well today & in the past (had ones from Adaptec that did this as far back as ScSi II, had their own onboard CPU for disk I/O) + noted it during my ScSi/UltraScSi-III run circa 1996-1999!

Prior to that? I have equipment that does this from disk controllers as far back as 1994 in the ISA/VLB days (& still have them no less in a 486 Dx/4 133mhz rig running a dual WD "Caviar" 424mb disk raid 0 stripe running Windows NT 3.51 fully patched here) - the TekRam units I have (VLB) & had (ISA) had their own I/O brain cpu onboard them to offload the system CPU.

So again, it is not the HDD, but the controller that needs the CPU.

Always, & in ALL cases to some extent (to @ least run their drivers, always, & in the case of offloading the system CPU? More "CPU relief" is possible on 3rd party SATA controllers like mine, or UltraScSi controllers today/nowadays)...

APK

 

[G.T]

RAID 0 with 2 x Raptors for your OS drive, duh. It's a no-brainer.

You can even partition those 2 drives into more to keep the drive area small so fast as you like for different areas such as OS load times, another bit for your games.

End of the day Raptors have their uses and what they have uses for they are good drives.

 

[Alec§taar]

TOM'S HARDWARE HAS A REVIEW ON THE SAME QUESTION POST STARTER HAD (w/ data in tests)

Tom's Hardware: RAID or RAPTOR? Article:

http://www.tomshardware.com/2006/02/06/wd1500ad_raptor_xtends_performance_lead/

* Something to read, on the VERY SAME TOPIC AS THIS THREAD, no less...

Conclusion: The Raptor-X Is Unrivaled

"Finally, we have a new high-performance enthusiast hard drive that is capable of beating its competition in the blink of an eye. Of course it's the high rotation speed of 10,000 RPM that lets the Raptor perform noticeably better than any 7,200 RPM drive in everyday uses. This applies to both access time and data transfer performance, and it helps to further reduce the annoying little delays inherent in everyday PC operation.

The Raptor-X's performance is even good enough to beat a RAID 0 array consisting of two modern 7,200 RPM drives, except in terms of pure throughput, of course.... Speaking of safety, we should refer to the five year warranty, which should give you a good feeling."

APK

P.S.=> Might be of some use/reference to the post thread starter here... apk

 

[G.T]

& something else in the Raptor debate.

Face it, the name alone adds extra points onto their performance cos it's so cool. >:}

 

[regan1985]

sorry only people with raptors would say that!!!!!

some would say you just have no much money to play with as you only have a raptor for benchmarks

i am just trying to find out with its worth becoming one of those people!

 

[Alec§taar]

sorry only people with raptors would say that!!!!!

Yea: They are like muscle cars - especially, in the way I have them setup: UNRIVALLED @ TECHPOWERUP.COM in the arena of HDD performance in CPU USAGE & SEEK/ACCESS COMBINATION... writes? We ought to test THAT now, lol...

some would say you just have no much money to play with as you only have a raptor for benchmarks

No, for me?

Small files are what I hit TONS of!

(Hence, why I use an SSD for a temp ops + webbrowser cache faster seek/access than my RAID 0 Raptor "X" setup even: FOUND FAST, 0ns speed!)

With smaller files, the overall throughput's NOT so important w/ tiny files, & seeks, TRULY ARE... seek/access> "burst reads, or writes" (rare in & of themselves by way of comparison of overall I/O performed)...

My diskdrive setup, really was unrivalled here as a combination in CPU use + Access/Seek performance @ techpowerup.com in HD Tach 3.x

I hit TONS of relatively smaller files... compiling code, & using the internet... being able to find them fast? PARAMOUNT - for performance!

Writing them? THE SAME!

This is exactly right for how I use a system... hi power diskdrive performance is a required parameter, in the areas specified.

i am just trying to find out with its worth becoming one of those people!

Oh, I'm happy in the area of diskdrives, this is certain/I can assure you... "join us, won't you?" lol...

(Strictly speaking from the 10,000rpm range though... on Solid-State? There is NO SUCH LIMITATION!)



APK (w/ the hi-tech/torque diskdrive arrangement)...

 

[Deleted member 3]

My diskdrive setup, really was unrivalled here as a combination in CPU use + Access/Seek performance @ techpowerup.com in HD Tach 3.x

When I receieve my new board I'll be setting my ancient disks up as RAID again, I'll run HDtach for you then. I think my disk will be able to keep up decently. Then again my disks
are of a different class, older generation though.

Untill then.. I WANT MY BOARD

 

[Alec§taar]

When I receieve my new board I'll be setting my ancient disks up as RAID again, I'll run HDtach for you then. I think my disk will be able to keep up decently. Then again my disks are of a different class, older generation though.

They'll also have an advantage though because you'll have set them up "fresh" again:

Their advantage will be that they'll be "emptier" than usual, because it will be a new setup again done on them, & thus, those disks (older or not) will leverage their outermost tracks speed advantage better (most likely)...

APK

 

[Deleted member 3]

They'll also have an advantage though because you'll have set them up "fresh" again

Nah I'm already coping with a week latency cause of of the board, probably those damn customs.