Velociraptor vs. Raptor - RAID Benches

canislupy

[H]ard|Gawd
Joined
Jan 31, 2003
Messages
1,030
Finally got my system up and time to run some HD Tune benches.

System:
E3110 @ stock
P5E3 Premium Wi-Fi
4x Cellshock 1800 1GB @ 1600
Areca ARC-1222 PCIe 8x
OS is Vista 64 Ultimate.

All disk drives on the Areca controller, including OS drive, WD 1TB GP. Default block size used in OS, block size for all tests set to 64k on controller unless specified otherwise.

And now for the pretty pictures.

Raptor 150
HDTune_R150.png


Velociraptor 300
HDTune_VR300.png


Raptor 150 x2 RAID-0
HDTune_2R150R0.png


Velociraptor 300 x2 RAID-0
HDTune_2VR300R0.png


Velociraptor 300 x4 RAID-0
HDTune_4VR300R0.png


Velociraptor 300 x4 RAID-5
HDTune_4VR300R5.png


Velociraptor 300 x4 RAID-10
HDTune_4VR300R10.png


And for comparison sake, here are some 4x RAID-0 benches with differing stripe sizes.

Velociraptor 300 x4 RAID-0 16k
HDTune_4VR300R0_16.png


Velociraptor 300 x4 RAID-0 32k
HDTune_4VR300R0_32.png


Velociraptor 300 x4 RAID-0 128k
HDTune_4VR300R0_128.png


I'd really like to see the write benches, but can't quite bring myself to pay $35 just to see that. RAID-5 read times are great, while giving me some fault protection, but I don't trust it to be so hot in writes. So, currently based on these numbers, I've been leaning towards RAID-0 with 32k stripe, then using Acronis to image... as well as my 1TB GP RAID-1 serving to save the critical data.
 
Awesome work. That actually has provided me some information I was really wanting on the Velociraptor. I'm quite amazed by the performance of the new drives.
 
Thanks for posting these, the raid0 was what I wanted to see, 200MB/s average looks very tempting.
 
Really nice post.. so much useful & focused data in such a short amount of space/time to review.

I wish HW review sites were this effective... don't get me wrong they are useful, but typically filled with 5-7 pages of fluff, background, repetitiveness, and links to MFG & ads sites :p
 
This is exactly the information I was hoping to see. I appreciate it!
 
Thanks for the review... I have one that I just ordered on the counter, looking at ordering a 2nd one now before I reload :)
 
Cool, nice to see an increase over the Raptor X, which I love and still have not found anything that matches it for a single drive for normal consumers (aka not SAS or the likes)
 
These are my benches using the Intel ICH9R. Does a PCI-Xpress add in Raid card really make a difference?

(2x Raid 0 300GB Velociraptor Abit ix38 Intel ICH9R)
Minimum: 145.0MBs
Max: 197.0MBs
Avg: 170.7MBs
Access Time: 7.1ms
Burst: 167.8MBs
CPU: 9.7%
 
These are my benches using the Intel ICH9R. Does a PCI-Xpress add in Raid card really make a difference?

(2x Raid 0 300GB Velociraptor Abit ix38 Intel ICH9R)
Minimum: 145.0MBs
Max: 197.0MBs
Avg: 170.7MBs
Access Time: 7.1ms
Burst: 167.8MBs
CPU: 9.7%

A good hardware raid add in card, in an appropriate slot does make a difference. The cheap ones are no better than your typical on board solutions however. On board does not truly suck, but is typically a very low end raid implementation. If they used a higher end one it would add $100, $200, or more to current mobo prices.


Nice benches OP, been thinking of picking up an Areca and 4 or 5 of those. But, I think I may just end up picking up 2 new Velocirapters and a new vid card instead. Speed is nice, though.
 
question - what exactly are you guys doing that you need a raid setup for?
 
Just wanted to thank you for now making me feel like my 150 is slow and need an upgrade.
 
Damnit...Just bought a new gun. Looks like the Raptor 150's will have to soldier on a few more months...


 
For me, heavy VMWare use - which is very disk intensive. Other than that, mostly games and coding.
 
Is gaming load times really going to benefit that much to justify the costs? I think I saw a review where the load time was only 1.5 to 2 seconds faster than the old raptors.
 
Is gaming load times really going to benefit that much to justify the costs? I think I saw a review where the load time was only 1.5 to 2 seconds faster than the old raptors.


That is prolly close enough to right. A second or three faster level load times over the old raptor seems probable.
Whether the cost to benefit ratio is justifiable or not, is a matter of personal opinion.
 
Very nice work, OP.

I hadn't seen any data like this until now, and that's very useful to me... I have been contemplating just such a RAID 0.
 
Really nice post.. so much useful & focused data in such a short amount of space/time to review.

I wish HW review sites were this effective... don't get me wrong they are useful, but typically filled with 5-7 pages of fluff, background, repetitiveness, and links to MFG & ads sites :p

That is just so they can count each page you load as a hit to there site, if the review was one page its only a single hit, if it s 5 pages thats 5 hits to there webpage and it looks alot better when there trying to sell adspace, also it means you load 5 times as many adds!

I hate the review sites that only have 1 to 2 paragraphs per page, at least when the [H] does reviews they only switch pages for differant sections of the review.

Back on topic, its a shame these drives cost so much, I am happy with the size of my 250gig HD but would love to get some increased speed.
 
3 bills is a lot of $ but the 74GB costs about 135-150 so not so bad.

These should be 250 at most.
 
3 bills is a lot of $ but the 74GB costs about 135-150 so not so bad.

These should be 250 at most.

You want speed you have to pay for it. ;)
I get your point though. You can buy two 750gig drives for less than the $300 one of the new 300gig raptors cost. 5 times the storage space for the same or less money kinda puts the new raptor in a different light.
 
so does a single velociraptor beat the 150 raptors raid 0? I want to see if I can just replace them for one.
 
so does a single velociraptor beat the 150 raptors raid 0? I want to see if I can just replace them for one.

The second and third HD tune screen caps show 2x150 raptors are faster than a single new 300 raptor. The new raptor seems to split the difference between a single 150 raptor and two in raid0.
 
wouldnt it bet better to get 4 x 640GB WD6400AAKS which achieve around 400MB/s STR in RAID-0 on a standard Intel ICH10 ?

I mean the cost of 4 x 300GB raptors + controller is like .... 16 !!!! of the above drives ... thats bit crazy ...

even for enthusiast ... I reckon the drive is way too overpriced ... I was gonna grab 2 or 4 .. but ... I refuse now ... I was expencting more ... and pls ... dont give me the story of ... "but it has better access time" ... its been proven many times that faster access time gives u minimal advantage in daily desktop (even gaming) environment ...
 
Expecting more? WD put out a drive in a form factor of enterprise server class, of higher density and performance than currently available in that class for a fraction of the price that it will cost enterprise customers whenever the other manufacturers get around to catching up.

There is also something to be said for the lower power, heat and noise of these fantastic little drives.

Worth is subjective, if for your applications, these drives are not worth the cost, and the 4x640's are, then by all means, go with what works for you. But don't sit here and complain about the drives under performing your baseless expectations when they are arguably the best performing magnetic disk drives available today. Look at what people are paying for SSD, you could buy crap loads of 640's for the same price as the arrays of SSD's some are making... but so what? What would you actually do with all those drives, and how would you even raid them all together without going to a megabuck enterprise solution?
 
A few more data points now that I've got mine up and running. Using WinXP-64, e8400, and 8G of RAM with the RAID built into the 780i mainboard, with my old Raptors as data points.

2x300G, 780i based RAID-0
r300%20-%20raid0.jpg


150G, single drive
r150%20-%20single.jpg


74G, single drive
r74G%20-%20single.jpg
 
OS installed to the 2x300r0 set when you benchmarked? Might explain the performance, unless that's just the controller. Even running my tests, since I was using Vista, I had to wait for Vista to become idle (truly idle, no indexing, etc.) before I ran the tests in order to get those "clean" numbers. This was because my OS drive was on the same controller, so when Vista was doing things and I ran a test, it would drop my numbers by up to 20mb/s.
 
OS installed to the 2x300r0 set when you benchmarked? Might explain the performance, unless that's just the controller. Even running my tests, since I was using Vista, I had to wait for Vista to become idle (truly idle, no indexing, etc.) before I ran the tests in order to get those "clean" numbers. This was because my OS drive was on the same controller, so when Vista was doing things and I ran a test, it would drop my numbers by up to 20mb/s.

Nope. Booted from my old 150G raptor, mounted the new 300x2 disk array, and benched it with XP-64. Raptor 150 numbers may have been lower due to it running the OS... but there is quite a gap between the Areca ARC-1222 PCIe 8x controller and the fakeraid you get on the mainboard.
 
Yeah, I wasn't expecting it to be that much, but I guess its reinforcement of the advantage to spending the money on the controller that I did. ;)
 
I missed originally that HD Tune Pro had a 15 day trial period, so I did some write tests late into last night. Unfortunately I have to work this weekend, so I won't get a chance to post them until tonight, but there were kind of interesting. The RAID-5 chart looks funny... not sure if its normal or my controller... well, you'll see for yourself soon enough (or maybe not soon enough for ya, but I was too tired to post 'em last night, and had to run into work this morning, sigh...).
 
If you get a chance, love to see what your software based RAID benches at, as well has the normal SATA ports. I assume the single drive specs were using the Areca. (If not, I'll need to try things with the NVidia drivers and compare)
 
Yeah, single drive runs were on the Areca as well. I have considered testing on the ICH9R, but... a lot more work. :p
 
even for enthusiast ... I reckon the drive is way too overpriced ... I was gonna grab 2 or 4 .. but ... I refuse now ... I was expencting more ... and pls ... dont give me the story of ... "but it has better access time" ... its been proven many times that faster access time gives u minimal advantage in daily desktop (even gaming) environment ...

The dairy's over that way-----------> just in case you wanted a little cheese for that whine. :p
 
Here they are. All tests this time at 32k block size. Including read and write.

4x VR300 RAID-0 32k - Read
HDTune_4VR300R0_R_32.png

4x VR300 RAID-0 32k - Write
HDTune_4VR300R0_WR_32.png


4x VR300 RAID-5 32k - Read
HDTune_4VR300R5_R_32.png

4x VR300 RAID-5 32k - Write
HDTune_4VR300R5_WR_32.png


4x VR300 RAID-10 32k - Read
HDTune_4VR300R10_R_32.png

4x VR300 RAID-10 32k - Write
HDTune_4VR300R10_WR_32.png


So is it just me or does that RAID-5 write test look kind of ugly?
 
Very nice results, that Areca controller is a hardware raid standalone controller right?
If so how much is it?

I'd still like to see the difference with the onboard controller, could you post those results as well?

What does the striping size mean, if a harddrive has less capacity, does it mean that the striping size has to be smaller to be more optimal??
 
The Areca ARC-1222 is about $550.

The stripe size is the block size of data written to each drive. If the file size is less than this number, it will be written to only 1 drive, if larger, will be split accordingly. Wiki has a much better explanation of all of this than I can give you. ;)

Yeah, maybe one of these days I will get some time to do the on-board comparison.
 
Back
Top