Transform 'Cuda 7200.11 1.5tb into Velociraptor Killer
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- Site Admin
- Posts: 12285
- Joined: Sun Aug 11, 2002 3:26 pm
- Location: Vancouver, BC, Canada
- Contact:
Transform 'Cuda 7200.11 1.5tb into Velociraptor Killer
Transform 'Cuda 7200.11 1.5tb into Velociraptor Killer
TechWare Labs posted a mod article by Artiom that transforms a $120 Seagate Barracuda 7200.11 1.5TB into a speed demon that outperforms a $230 VelociRaptor 300GB in almost every way. The mod only requires the use of Seagate's own Sea Tools utility... and turns the 1.5TB drive into a 300GB drive. (This is reversible.) Apparently, setting the size to 20% of max capacity makes use of only "the inside 20% of the disc, thus bringing superior performance."
TechWare Labs posted a mod article by Artiom that transforms a $120 Seagate Barracuda 7200.11 1.5TB into a speed demon that outperforms a $230 VelociRaptor 300GB in almost every way. The mod only requires the use of Seagate's own Sea Tools utility... and turns the 1.5TB drive into a 300GB drive. (This is reversible.) Apparently, setting the size to 20% of max capacity makes use of only "the inside 20% of the disc, thus bringing superior performance."
Last edited by MikeC on Thu Jul 16, 2009 6:37 am, edited 2 times in total.
Or what's the "breakeven" size - Can you get 500 GB and EQUAL a Velociraptor? 800 GB? I wonder how it picks it's cylinders that it uses? Does it prefer a single platter (doubt it), or start from the middle and work it's way out of all platters equally?chahahc wrote:Wow....That is impressive. But I don't really get how downsizing the drive increases performance....
Wonder if there would be even more of a performance boost if it was downsized more....150gb or even 100gb for example....
-Dan
-
- Posts: 286
- Joined: Sat Dec 09, 2006 9:45 pm
- Location: Montréal, Canada
I didn't read the article but i imagine i know what it's about. It's called short stroking. It's old like the planet, but with each generation it gets better and better. The same can be achieved with partitionning. You can always put the data partitions in the inner disc, and if it doesn't access it often, the stuff on the outer rim (OS most notable, and programs) will be as fast as this article would claim. Except that if you benchmark it that way, it wont give you the results you get from doing it with seatools because it will bench the whole disc.
By the way, i would imagine you mean the 20% OUTER disc, where the linear velocity is highest for an equal angular velocity.
By the way, i would imagine you mean the 20% OUTER disc, where the linear velocity is highest for an equal angular velocity.
-
- Friend of SPCR
- Posts: 356
- Joined: Wed Jan 16, 2008 8:56 pm
- Location: Council Bluffs, Iowa
- Contact:
I'm not sure I see the point of the article. The Velociraptor isn't noted for its sequential performance, which can be approached, if not matched, by a great many high-capacity drives. It's noted for ridiculously short seek and random access times, which make the computer more responsive in day-to-day use. Short-stroking a high-capacity drive doesn't dramatically improve its seek time, because the drive still has to wait for the platters to complete an average of half a rotation to bring data to bear. All this achieves is to hold data in the outer portion of the disk where the sequential rates are higher. There are certain situations in rendering, data serving, and heavy number-crunching where this can produce a significant benefit. On a personal desktop computer, the only tangible effects are a smaller drive with more impressive synthetic benchmark results.
Why go into the seatools program? Ever heard of partitioning? That way you can still get all of the space you paid for too! Admittedly software like hddscan will see past that and benchmark across the entire physical hdd so the random access times in the report will still be quite high.
Another way of doing it would be to get a couple of drives (I'd choose 7200.12 myself) and using Intel Matrix Raid or similar and use that to partition off the first 100gb or whatever you feel like and Raid 0 them or something. Hddscan will only 'see' the first 100gb of each hdd and will report extremely quick results including random access. The ultimate result will be that you win the benchmark tests In real life there probably won't be much of an advantage.
Does raise some interesting questions about a pair of 320gb or 500gb 7200rpm 2.5" hard drives though doesn't it? Quiet, vibration free, and reasonably quick?
Another way of doing it would be to get a couple of drives (I'd choose 7200.12 myself) and using Intel Matrix Raid or similar and use that to partition off the first 100gb or whatever you feel like and Raid 0 them or something. Hddscan will only 'see' the first 100gb of each hdd and will report extremely quick results including random access. The ultimate result will be that you win the benchmark tests In real life there probably won't be much of an advantage.
Does raise some interesting questions about a pair of 320gb or 500gb 7200rpm 2.5" hard drives though doesn't it? Quiet, vibration free, and reasonably quick?
People buy the Velociraptor for the lower access time due to the higher rotational speed. A random access time of 10.3 ms does not outperform the velociraptor's 7.2 ms.
AFAIK Seagate drives tend to be slower so I'm not sure why they used those. The 640GB WD blue has a seek time of ~12 ms with no short stroking and short stroking it to 200-300 GB should give you ~10 ms.
AFAIK Seagate drives tend to be slower so I'm not sure why they used those. The 640GB WD blue has a seek time of ~12 ms with no short stroking and short stroking it to 200-300 GB should give you ~10 ms.
"that outperforms a $230 VelociRaptor 300TB in almost every way."
Every way but the most important one: seek time.
Moreover, in a real world desktop condition using a reasonable filesystem, you will frequently benefit more from having the extra space so that the filesystem can avoid fragmentation than you would from the guaranteed sequential speed performance that you get from short stroking.
In nearly every situation that effects the desktop user, faster sequential read speed will basically only make fast operations faster. Anything that is taking a significant amount of time is either a very special case, like large file copy from another local disk, or almost certainly seek-limited.
Every way but the most important one: seek time.
Moreover, in a real world desktop condition using a reasonable filesystem, you will frequently benefit more from having the extra space so that the filesystem can avoid fragmentation than you would from the guaranteed sequential speed performance that you get from short stroking.
In nearly every situation that effects the desktop user, faster sequential read speed will basically only make fast operations faster. Anything that is taking a significant amount of time is either a very special case, like large file copy from another local disk, or almost certainly seek-limited.
-
- Posts: 434
- Joined: Tue Oct 04, 2005 6:10 am
The author doesn't really seem to understand what he's talking about. The fastest part of the disks is the outside, not the inside.
As someone else already mentioned, partitioning is the better way to go.
'Virtual' partitioning would be even better, where you only have one partition but could tell the OS to store certain directories / files (like movies) at the end of the disk. Unfortunately I don't know of any OS / FS that can do that.
As someone else already mentioned, partitioning is the better way to go.
'Virtual' partitioning would be even better, where you only have one partition but could tell the OS to store certain directories / files (like movies) at the end of the disk. Unfortunately I don't know of any OS / FS that can do that.
-
- Posts: 2198
- Joined: Thu Feb 10, 2005 11:20 am
- Location: TN, USA
1. Partitioning is much simpler to do
2. What is up with the Seagate 7200.11 love from this guy? You could do the same short stroking by one or more methods with a Samsung HD502HI or Western Digital WD6400AAKS. There is no magic to buying the specific model or brand he mentioned or magic to buying any drive I mentioned. Buy a drive format it for less than the full capacity and enjoy the speed benefit.
I've found on the drives I've tested recently the first 20% of the drive capacity has a pretty flat transfer rate graph. If you want max speed just format the drive to 20% capacity. If you want more space pick the trade off you want, something in the 20% to 50% range will still give you noticeable improvements but the higher you go the slower those access times and sustained transfer rates will get.
2. What is up with the Seagate 7200.11 love from this guy? You could do the same short stroking by one or more methods with a Samsung HD502HI or Western Digital WD6400AAKS. There is no magic to buying the specific model or brand he mentioned or magic to buying any drive I mentioned. Buy a drive format it for less than the full capacity and enjoy the speed benefit.
I've found on the drives I've tested recently the first 20% of the drive capacity has a pretty flat transfer rate graph. If you want max speed just format the drive to 20% capacity. If you want more space pick the trade off you want, something in the 20% to 50% range will still give you noticeable improvements but the higher you go the slower those access times and sustained transfer rates will get.
Last edited by dhanson865 on Thu Jul 16, 2009 5:35 am, edited 1 time in total.
-
- Friend of SPCR
- Posts: 356
- Joined: Wed Jan 16, 2008 8:56 pm
- Location: Council Bluffs, Iowa
- Contact:
It certainly would be. Unfortunately, the modification described in the article actually doesn't achieve a worthwhile performance improvement at all - read back through the responses in this thread discussing seek time.yacoub wrote:[It's arguably more impressive to get the same/better performance without needing to increase the rotational velocity.
-
- Posts: 64
- Joined: Thu Sep 18, 2008 7:05 am
- Location: Rhode Island
Way back when I really cared about these things I used to partition my drive
- partition for OS + programs
- partition for swap
- partition for everything else
the idea being that in effect the OS and swap live on the outer sectors of a short-stroked drive. All my other stuffs were on the lower peformance part of the disc, but who needs tip-top performance for playing MP3s?
- partition for OS + programs
- partition for swap
- partition for everything else
the idea being that in effect the OS and swap live on the outer sectors of a short-stroked drive. All my other stuffs were on the lower peformance part of the disc, but who needs tip-top performance for playing MP3s?
If you're really interested in short stroking, there was a much better article (with a different emphasis but still) here: http://www.tomshardware.com/reviews/sho ... ,2157.html
Well I am wondering from a different perspective.
Since we are here at the SPCR website, it would be interesting to see if this alteration has any benefits from audio and power side.
Is SPCR thinking of doing their own analysis?
Also last comment. Short Stroking the Velociraptor.. That would be impressive to see.
Since we are here at the SPCR website, it would be interesting to see if this alteration has any benefits from audio and power side.
Is SPCR thinking of doing their own analysis?
Also last comment. Short Stroking the Velociraptor.. That would be impressive to see.
-
- Posts: 1406
- Joined: Tue Feb 13, 2007 7:28 pm
- Location: USA
There is one simple benefit to both. Faster transfer means certain tasks get accomplished faster and the drive can be returned to its idle state (which will be quieter and consume less power). The real question is how much time are we actually talking about in real-world scenarios? Most people aren't continuously writing/reading giant files from their primary drive (and if they were the loss of capacity from short-stroking would be too big a drawback). Also, the Windows OS like to pretty much continuously keep the drive in use with the random writing and reading of OS files, so completing our large tasks faster still won't get us back to idle state for long.prodeous wrote: Since we are here at the SPCR website, it would be interesting to see if this alteration has any benefits from audio and power side.
As many previous posters have noted, short-stroking is fairly pointless. The real benefit of higher RPM drives is fast random operations. Faster sustained operations are certainly nice, but not at the expense of drive capacity, which is going to be very important on drives that do a lot of sustained operations.
The primary benefit of short stroking a drive is to reduce seek times. For example, short stroking their 450GB 15k SAS drive to 45GB cut seek times from 6ms to 3.8ms. For the 250GB 7.2k SATA drive short stroking to 34GB cut seek times from 17ms to 10ms.jessekopelman wrote:so completing our large tasks faster still won't get us back to idle state for long.
Those are substantial improvements in seek times.
There are also substantial improvements in raw through put, too (the SATA drive had a minimum of 55MB/s at full capacity but stayed over 90MB/s when short stroked, but for most people, raw throughput isn't what limits system performance, random IO is.
Last edited by drees on Fri Jul 17, 2009 12:44 pm, edited 1 time in total.