10,000 RPM vs. RAID

[SickBeast]

Originally posted by: Fallen Kell

Originally posted by: Aikouka

Originally posted by: SickBeast
The Raptors are overrated IMO. If you get a top performing 7,200RPM SATA drive, you will barely notice a difference between it and the Raptor. Not only that, but the faster 7,200 drives outperform the Raptors in quite a few areas.

Wrong, and I'm willing to bet that you've never actually had a personal system with a new 7200rpm drive and a raptor. My new machine that I built loads everything slower than my old machine (both are fairly speedy machines, but the newer one definitely has the better components (sans the HDD)). The HDDs being used on the two machines are a Seagate Barracuda 7200.10 750GB (OS is partitioned off) and a 74GB Raptor.

The differences are why a 150GB Raptor is being delivered to me today from NewEgg .

I agree completely on this. It is also why I have a 150 GB Raptor and 4x 500 GB Seagates going into my new rig. The Seagates will be RAID 5 on a dedicated RAID card (Promise SuperTrak EX8350).

Perhaps the Raptors have gotten alot better. I owned a 36GB Raptor, and found that my 250GB WD drive was just as quick when it came to just about everything, so I sold it.

The Raptors are too expensive. How long does it take to load windows with the 150GB version?

 

[Blain]

Load times... Burst speed... Yadda Yadda Yadda...
It's all about Average read and Average write. :laugh:

 

[docinthebox]

Those who say a good 7200RPM drive performs about the same as a 10,000RPM drive do not distinguish between access time and throughput in terms of drive performance. A 10,000RPM drive will always have better access time because of the lower rotational latency. The throughputs can be similar for example if you use denser platters or if you do RAID-0, but nothing you can do is going to make the access time of a 7200RPM comparable to that of a 10,000RPM drive. Access time affects response time, affects performance in random access, affects basically how "snappy" a system feels.

 

[SickBeast]

Originally posted by: Blain
Load times... Burst speed... Yadda Yadda Yadda...
It's all about Average read and Average write. :laugh:

So then the Raptos are +/- 10% faster using that logic. Not worth it for double the cost IMO, especially considering the low capacity.

AT ran articles that discredited both the Raptors and RAID.

 

[StopSign]

If you want to compare performance, docinthebox is right. The Raptors will definitely perform better.

But if you want to compare numbers, compare sustained transfer speeds, not burst. And to another extreme, don't compare SATA 150 with SATA 300...

 

[Blain]

Originally posted by: SickBeast

Originally posted by: Blain
Load times... Burst speed... Yadda Yadda Yadda...
It's all about Average read and Average write. :laugh:

So then the Raptos are +/- 10% faster using that logic.

It's handy to toss out "10%", but where did you get that number?
Did you calculate the difference of only +/- 10% from the two links above? :roll:

 

[Aikouka]

Originally posted by: SickBeast
Perhaps the Raptors have gotten alot better. I owned a 36GB Raptor, and found that my 250GB WD drive was just as quick when it came to just about everything, so I sold it.

The Raptors are too expensive. How long does it take to load windows with the 150GB version?

I believe the 36GB models were only 8MB cache. I know my old roommate from college has one in his computer and he's happy enough with it.

As for the 150GB model, I haven't been home, so I don't have the chance to work with it. It's also getting Vista on it, so I'd rather not attempt to make a comparison between XP load times and Vista load times on two different hard drives when you're looking for the hard drives loading the same thing. I could've went with the tried and true 74GB like I already have, but the 150 was only marginally more ($40 more I believe it was) and to avoid ever having to uninstall games... I just got the bigger one. Now I can move the two 750s to my server PC and avoid having this horrid situation I have now where I don't remember what exists on which PC.

EDIT:

Originally posted by: SickBeast
So then the Raptos are +/- 10% faster using that logic. Not worth it for double the cost IMO, especially considering the low capacity.

AT ran articles that discredited both the Raptors and RAID.

Incorrect, the Anandtech article only discredited people who screamed and raved about running two Raptor 74GB in RAID 0 and how it dramatically increased loads. The Anandtech times showed loads to be practically the same. I do remember some article that said the MaxLine III drives from Maxtor were supposedly near the speeds of the Raptor. I do own a MaxLine III but I've never used it as a system drive.

 

[SickBeast]

Originally posted by: Aikouka

Originally posted by: SickBeast
Perhaps the Raptors have gotten alot better. I owned a 36GB Raptor, and found that my 250GB WD drive was just as quick when it came to just about everything, so I sold it.

The Raptors are too expensive. How long does it take to load windows with the 150GB version?

I believe the 36GB models were only 8MB cache. I know my old roommate from college has one in his computer and he's happy enough with it.

As for the 150GB model, I haven't been home, so I don't have the chance to work with it. It's also getting Vista on it, so I'd rather not attempt to make a comparison between XP load times and Vista load times on two different hard drives when you're looking for the hard drives loading the same thing. I could've went with the tried and true 74GB like I already have, but the 150 was only marginally more ($40 more I believe it was) and to avoid ever having to uninstall games... I just got the bigger one. Now I can move the two 750s to my server PC and avoid having this horrid situation I have now where I don't remember what exists on which PC.

EDIT:

Originally posted by: SickBeast
So then the Raptos are +/- 10% faster using that logic. Not worth it for double the cost IMO, especially considering the low capacity.

AT ran articles that discredited both the Raptors and RAID.

Incorrect, the Anandtech article only discredited people who screamed and raved about running two Raptor 74GB in RAID 0 and how it dramatically increased loads. The Anandtech times showed loads to be practically the same. I do remember some article that said the MaxLine III drives from Maxtor were supposedly near the speeds of the Raptor. I do own a MaxLine III but I've never used it as a system drive.

The AT article showed that the most a Raptor could beat a good 7,200RPM drive by was a few seconds loading a game like Far Cry.

Just because your roommate is happy with his 36GB Raptor it doesn't say much. Has he used a modern 7,200RPM drive? I wouldn't have sold my Raptor if it were significanly quicker than my new drive at loading windows! I actually installed Far Cry on both my Raptor and the new 250GB WD drive, and timed level loads. The WD drive was just as fast!

I'm not denying that the new 150GB Raptor will beat the top 7,200 drives in most performance areas, seek times especially. What I'm saying is that the older raptors (36GB and 74GB drives) have been outclassed by the newest 7,200RPM drives. The increased data density on the newer drives helps to offset the faster rotational speed of the Raptor.

 

[SickBeast]

Originally posted by: Blain

Originally posted by: SickBeast

Originally posted by: Blain
Load times... Burst speed... Yadda Yadda Yadda...
It's all about Average read and Average write. :laugh:

So then the Raptos are +/- 10% faster using that logic.

It's handy to toss out "10%", but where did you get that number?
Did you calculate the difference of only +/- 10% from the two links above? :roll:

Hey, they're your links.

Why not post something showing your beloved Raptor blowing other drives out of the water if you want us to believe otherwise?

 

[docinthebox]

Originally posted by: SickBeast
The increased data density on the newer drives helps to offset the faster rotational speed of the Raptor.

only if your drive is very well defragged, and then only if you're reading from a very big file. If your files are fragmented and all over the drive, the access time component is going to play a more important role than the transfer rate component in determining the overall performance.

 

[Fallen Kell]

Originally posted by: SickBeast

Originally posted by: Blain

Originally posted by: SickBeast

Originally posted by: Blain
Load times... Burst speed... Yadda Yadda Yadda...
It's all about Average read and Average write. :laugh:

So then the Raptos are +/- 10% faster using that logic.

It's handy to toss out "10%", but where did you get that number?
Did you calculate the difference of only +/- 10% from the two links above? :roll:

Hey, they're your links.

Why not post something showing your beloved Raptor blowing other drives out of the water if you want us to believe otherwise?

You asked:
Raptor VS 10K and 15K SCSI and other SATA random seek, average transfer/max transfers
http://www.storagereview.com/articles/200601/WD1500ADFD_3.html

Same disks in singl user performance (even beats the 15k SCSI's here let alone the SATA disks)
http://www.storagereview.com/articles/200601/WD1500ADFD_4.html

Same disks in gaming (blows away many 15k SCSI's as well here)
http://www.storagereview.com/articles/200601/WD1500ADFD_5.html

 

[Blain]

Originally posted by: SickBeast
I owned a 36GB Raptor, and found that my 250GB WD drive was just as quick when it came to just about everything, so I sold it.

Originally posted by: SickBeast
Hey, they're your links.

Why not post something showing your beloved Raptor blowing other drives out of the water if you want us to believe otherwise?

Those links show...
74GB Raptor "Average read transfer", 49% faster than the 250GB WD SE16.
WD score > 50.5 + 49% = 75.245 ...Raptor score > 75.3
74GB Raptor "Average write tranfer", 48% faster than the 250GB WD SE16. WD score > 50.5 + 48% = 74.74 ...Raptor score > 74.7

How are those scores anywhere close to only +/- 10%?

 

[Dutchmaster420]

i have a 74gb raptor for my boot and a 500gb 3.0 samsung spinpoint for everything else...could not be happier

just go for the raptor...their a little loud though when accessing them

 

[VTrider]

Don't you just love these 'RAID' posts? Well, as this thread is going down in flames i'll throw in my .02c

I have a lot of experience with RAID 0 and 10,000K raptors in all sorts of configurations, and you know what...bottom line, don't worry about - everybody gets their panties all tied up in an uproar over insignificant technicalities, benchmarks, etc. When I read the OP question, the key word to me was 'noticeable difference'

You have good, fast equipment and all the other components other people recommended are fast also - crap, this is not like apples -n- oranges - my advice, you don't need to spend a lot of money, lot of good options listed, pick one - it's not going to make a significant noticeable difference.

 

[bob4432]

as you can see from my sig i run both 10K and 15K scsi and have for years as my main boot drive. you have to understand that a pci bus has a theoretical max speed of i think 133MB/s, but real world is more like 100-125MB/s when you add in all the other stuff on that channel, so for scsi you can pick up a u160 card as it has more speed capability than the 32bit pci slot it is more than likely going into.

next, you need to remember that only 1 drive has a str of over 100MB/s - the u320 seagate 15k.5 - perpendicular recording and they are awesome drives for str but i would prefer the fujitsu mau, mas or max series for a boot/system/app/game drive. these mau and mas fujitsu drives can usually be picked up pretty cheap "reconditioned", the u160 card is old skool so they are pretty cheap and then all you need is a u320 or u160 cable, again can be picked up pretty cheap - scsi is not this too expensive beast people always say it is, sure if you go out and buy full retail it is crazy expensive, but with scsi you can get 1 or 2 gen old and still be very well off, especially with the 15k line because their seek times and rotational latencies haven't changed. usually a 15k drive is going to have a str of ~70-85MB/s with the exception of the 15k.5 seagate, mine bursts and has a str of the same speed - ~97MB/s (can we way 32 bit pci bus limitation???)

also, remember that buying a reconditioned scsi drive is, imho not like buying a reconditioned pata/sata drive - a scsi drive is a true enterprise level drive meant to be on 24/7/365/multiple years so they are built like tanks

i would say you could go a 15k 36GB fujitsu mas setup for ~$100-$120 (u160 card, cable, adapter, drive) (remember you can get the 80pin sca drives and use an adapter on them, you don't always need to go a 68pin drive). also remember that that the 15k drives do run warm so active cooling is required

this information is from somebody who uses this on their main rig everyday, not somebody looking at benchmarks or reading articles.

 

[chizow]

With the quality and speed of the large SATA drives available today, I don't see as much of a need for RAID 0 as before. Just run a Raptor 150GB for your boot partition/apps and a large 300GB+ SATA drive or two for your storage in JBOD. If you don't need that much storage and want a little bit more performance for stuff where you transfer data between disks, 2x Raptor 150GB in JBOD offers a nice balance of speed and storage. Not only will your case be cleaner, but your temps will be lower as well.

I plan on picking up a 2nd Raptor 150GB (@$160 for the retail kit this week at BB) and maybe a 300GB+ SATA down the line. I scrapped the idea of porting over my RAID 0 array from my old rig (2x DM9+ on a 3ware 7000-2) after seeing how fast the Raptor was. It just wasn't worth losing a PCI slot and cluttering my case with drives and cables for 120GB of decent performance compared to today's drives.

 

[SuperNaruto]

If you got money to burn, 2 4gb iram in raid 0 - 1 8gb partition

otherwise just 10k raptor

 

[Enlightenment]

Originally posted by: SickBeast
RAID is also overrated and only really helps with transferring very large files, which does not reflect normal usage. Most files are less than 1MB, which negates most of the benefit that RAID provides.

Totally untrue!

RAID0 does not only yield higher STR (Sequential Transfer Rate), but also speeds up realistic workloads due to parallism. With an outstanding I/O queue the I/O requests will be divided over the available disks and the aggregate performance is much higher than a single disk. I hear so many people say RAID0 only yields higher STR which is simply not true.

Observe this HIGHLY random I/O benchmark (random transfers ranging from 16KB to 128KB MAX; 50% read, 50% write):

Single drive (ad8)
concurrency Performance in I/O's per sec. average
1 106 106 107 106
4 106 106 106 106
16 116 116 116 116
32 127 125 126 126
128 151 151 150 150
256 156 156 157 156

RAID0 with 4 drives: gstripe 4xad - 128KB stripe - FM off
concurrency Performance in I/O's per sec. average
1 173 173 173 173
4 270 270 270 270
16 338 338 338 338
32 370 370 370 370
128 444 434 434 437
256 465 465 465 465

Performance increase ranging from 63% to 300%.

 

[bob4432]

Originally posted by: Enlightenment

Originally posted by: SickBeast
RAID is also overrated and only really helps with transferring very large files, which does not reflect normal usage. Most files are less than 1MB, which negates most of the benefit that RAID provides.

Totally untrue!

RAID0 does not only yield higher STR (Sequential Transfer Rate), but also speeds up realistic workloads due to parallism. With an outstanding I/O queue the I/O requests will be divided over the available disks and the aggregate performance is much higher than a single disk. I hear so many people say RAID0 only yields higher STR which is simply not true.

Observe this HIGHLY random I/O benchmark (random transfers ranging from 16KB to 128KB MAX; 50% read, 50% write):

Single drive (ad8)
concurrency Performance in I/O's per sec. average
1 106 106 107 106
4 106 106 106 106
16 116 116 116 116
32 127 125 126 126
128 151 151 150 150
256 156 156 157 156

RAID0 with 4 drives: gstripe 4xad - 128KB stripe - FM off
concurrency Performance in I/O's per sec. average
1 173 173 173 173
4 270 270 270 270
16 338 338 338 338
32 370 370 370 370
128 444 434 434 437
256 465 465 465 465

Performance increase ranging from 63% to 300%.

please give information regarding what hardware was used to get your results? also, is this server data or desktop? is there any info from whatever source you are using using 2hdds? plus don't forget of the higher chances of drive failure when you make a 4drive striped array....

 

[Enlightenment]

Originally posted by: bob4432
please give information regarding what hardware was used to get your results? also, is this server data or desktop? is there any info from whatever source you are using using 2hdds? plus don't forget of the higher chances of drive failure when you make a 4drive striped array....

Hardware system is:
AMD Athlon 64 X2 3800+ (irrelevant; disks are bottleneck)
2x1GB memory (irrelevant; RAIDTEST operates outside file caching architecture)
Maxtor MaxLine III 250GB SATA/150 disks (irrelevant; numbers will vary but performance concept remains unchanged)
Operating System: FreeBSD 6.1-REL amd64
System role: test system (so neither server nor workstation -- concept applies to both system roles)
RAID subsystem: geom stripe (gstripe) software RAID0

Higher chance of drive failure? RAID0 in itself does not speedup failure though 4 disks ofcourse have more risk to fail than a single disk -- but that is unfair comparison since with 4 disks in RAID0 you also have four times the capacity. And if you do not utilize redundancy you need a proper backup anyway if your data has any value to you.

 

[Fullmetal Chocobo]

Originally posted by: Enlightenment

Originally posted by: bob4432
please give information regarding what hardware was used to get your results? also, is this server data or desktop? is there any info from whatever source you are using using 2hdds? plus don't forget of the higher chances of drive failure when you make a 4drive striped array....

Hardware system is:
AMD Athlon 64 X2 3800+ (irrelevant; disks are bottleneck)
2x1GB memory (irrelevant; RAIDTEST operates outside file caching architecture)
Maxtor MaxLine III 250GB SATA/150 disks (irrelevant; numbers will vary but performance concept remains unchanged)
Operating System: FreeBSD 6.1-REL amd64
System role: test system (so neither server nor workstation -- concept applies to both system roles)
RAID subsystem: geom stripe (gstripe) software RAID0

Higher chance of drive failure? RAID0 in itself does not speedup failure though 4 disks ofcourse have more risk to fail than a single disk -- but that is unfair comparison since with 4 disks in RAID0 you also have four times the capacity. And if you do not utilize redundancy you need a proper backup anyway if your data has any value to you.

4 disk in RAID 0 -- 1 disk failure, all data lost.
4 disks, not in RAID -- same capacity, 1 disk failure, 1 disk's data lost.

How is that an unfair comparison referencing capacity? I don't follow.

 

[Enlightenment]

If your data is important, you do not want to loose any. That implies the usage of either RAID or proper backups. If you obey this you can happily use any RAID0 or single disks you want.

 

[Blain]

Originally posted by: Blain

Originally posted by: SickBeast
I owned a 36GB Raptor, and found that my 250GB WD drive was just as quick when it came to just about everything, so I sold it.

Originally posted by: Blain
Load times... Burst speed... Yadda Yadda Yadda...
It's all about Average read and Average write. :laugh:

Originally posted by: SickBeast
So then the Raptos are +/- 10% faster using that logic.

Originally posted by: SickBeast
Hey, they're your links.

Why not post something showing your beloved Raptor blowing other drives out of the water if you want us to believe otherwise?

Those links show...
74GB Raptor "Average read transfer", 49% faster than the 250GB WD SE16.
WD score > 50.5 + 49% = 75.245 ...Raptor score > 75.3
74GB Raptor "Average write tranfer", 48% faster than the 250GB WD SE16. WD score > 50.5 + 48% = 74.74 ...Raptor score > 74.7

How are those scores anywhere close to only +/- 10%?

"How are those scores anywhere close to only +/- 10%?"

 

[CloE]

The raptors not only fast, but durable too. While all other WD drives offer 1 or 3 yrs warranty, the raptors come with 5 and solid running 24/7 for me like a rock. I won't hesitates to setup raid 0 with 2 raptors, most people use them for boot drive anyway.

In gaming performance, since i got the raptors raid-0, I have 90% of time being the fastest person arrive to the server when map changes. I don't know what other people is running, but 2 x 36GB raptors have enough room for all my games and programs. Save your BIG FILES on the other HDs

 

[BDawg]

Storage Review Article

I'd rather have 1 better drive than 2 drives in RAID 0.