Link
Savvio 15K drive, The Fastest, and it's 2.5"
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
Bluefront
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Savvio 15K drive, The Fastest, and it's 2.5"
Wow....I don't know how quiet it will be, but it's fast, really fast. Get out your check books.....
Link
Link
Already discussed in:
viewtopic.php?t=37444
I'm quoting myself from that thread:
"I don't believe there to be any benefit in noise either. Going laptop sized for 15krpm drives is definitely not a good thing noisewise. Why laptop drives are quieter than desktop HDDs is due to two things:
- lower rpm
- smaller platters.
The fact that Savvio is 2.5" drive doesn't make it a laptop drive. Laptops don't use SCSI interface so it's useless for laptops. It's a very specialized server hard disk. There's no reason to use it in desktop enclosures either as the main benefit (smaller size than 3.5" SCSI drives) is no benefit at all if the enclosure only has 3.5" and 5.25" slots and one would have to use adapters to mount a 2.5" drive.
But why isn't 2.5" 15krpm Savvio necessarily silent?
Very high rpm. But how about the platter size? Well, it's no different from the platter size used in 3.5" 15krpm drives. 3.5" SCSIs use smaller platters that what could be physically fitted into that size of HDD casing. The extra space is just used to make the HDD a big heatsink. Take "the heatsink" off and you can make the 15krpm drive to 2.5" sized HDD. With almost the same sized platters and the same rpm, I doubt there's anything positive for silencer enthusiast to notice. Or say: the only positive potential I could see is that fitting the mechanics inside a smaller enclosure forces the manufacturer to miniaturize the actuator and that could (but not necessarily would) reduce seek noise. Using a miniaturized actuator in 3.5" SCSI drives is possible but probably not high on their goals as these drives are designed for enterprise use only. We should not forget that, whether it's discussion about noise or about performance."
Thus I wouldn't consider there's any possibility of Savvio being regarded as a silent drive. (Surprizingly quiet, perhaps, but certainly not quiet.) Savvio's smaller size has nothing to do with making it more quiet.
The factor that makes it quiet (or less noisy) is smaller platter size, not smaller form factor of the base cast. All 15krpm drives use very small platters even though most of them appear normal sized 3.5" drives! Savvio has no benefit over those others. (And higher rpm is likely going to cause more noise than smaller platters are going to prevent. Not to mention the super-aggressive drive current to coil driving the actuator arm. Expect quite a bit of noise when seeking.)
I'm not against Savvio - I'm only against buying Savvios for the wrong reason (silence or real-life desktop performance). Savvios are intended for servers. You would not benefit from Savvio unless you need to have several of these drives. Savvio is all about saving space. If you want a single 15krpm SCSI, a one with 3.5" form factor offers everything a 2.5" SCSI can offer.
This is when 2.5" SCSIs comes into the picture:
http://en.wikipedia.org/wiki/Blade_server
I'm not against reviewing these drives but I think the time is better spent with drives that are more likely to be of real consideration for silent systems. Of course reviewing a drive that will prove to be noisy is good too... to eliminate some common misconceptions around the issue. Even my opinions are based on reasoning instead of careful evaluation of real noise profiles, thus my opinion could be a misconception too.
(Would prefer to see drives like WD5000AAKS and T7K500 reviewed instead, though.)
viewtopic.php?t=37444
I'm quoting myself from that thread:
"I don't believe there to be any benefit in noise either. Going laptop sized for 15krpm drives is definitely not a good thing noisewise. Why laptop drives are quieter than desktop HDDs is due to two things:
- lower rpm
- smaller platters.
The fact that Savvio is 2.5" drive doesn't make it a laptop drive. Laptops don't use SCSI interface so it's useless for laptops. It's a very specialized server hard disk. There's no reason to use it in desktop enclosures either as the main benefit (smaller size than 3.5" SCSI drives) is no benefit at all if the enclosure only has 3.5" and 5.25" slots and one would have to use adapters to mount a 2.5" drive.
But why isn't 2.5" 15krpm Savvio necessarily silent?
Very high rpm. But how about the platter size? Well, it's no different from the platter size used in 3.5" 15krpm drives. 3.5" SCSIs use smaller platters that what could be physically fitted into that size of HDD casing. The extra space is just used to make the HDD a big heatsink. Take "the heatsink" off and you can make the 15krpm drive to 2.5" sized HDD. With almost the same sized platters and the same rpm, I doubt there's anything positive for silencer enthusiast to notice. Or say: the only positive potential I could see is that fitting the mechanics inside a smaller enclosure forces the manufacturer to miniaturize the actuator and that could (but not necessarily would) reduce seek noise. Using a miniaturized actuator in 3.5" SCSI drives is possible but probably not high on their goals as these drives are designed for enterprise use only. We should not forget that, whether it's discussion about noise or about performance."
Thus I wouldn't consider there's any possibility of Savvio being regarded as a silent drive. (Surprizingly quiet, perhaps, but certainly not quiet.) Savvio's smaller size has nothing to do with making it more quiet.
The factor that makes it quiet (or less noisy) is smaller platter size, not smaller form factor of the base cast. All 15krpm drives use very small platters even though most of them appear normal sized 3.5" drives! Savvio has no benefit over those others. (And higher rpm is likely going to cause more noise than smaller platters are going to prevent. Not to mention the super-aggressive drive current to coil driving the actuator arm. Expect quite a bit of noise when seeking.)
I'm not against Savvio - I'm only against buying Savvios for the wrong reason (silence or real-life desktop performance). Savvios are intended for servers. You would not benefit from Savvio unless you need to have several of these drives. Savvio is all about saving space. If you want a single 15krpm SCSI, a one with 3.5" form factor offers everything a 2.5" SCSI can offer.
This is when 2.5" SCSIs comes into the picture:
http://en.wikipedia.org/wiki/Blade_server
I'm not against reviewing these drives but I think the time is better spent with drives that are more likely to be of real consideration for silent systems. Of course reviewing a drive that will prove to be noisy is good too... to eliminate some common misconceptions around the issue. Even my opinions are based on reasoning instead of careful evaluation of real noise profiles, thus my opinion could be a misconception too.
(Would prefer to see drives like WD5000AAKS and T7K500 reviewed instead, though.)
Not true. It's true that 3.5" 15k rpm drives use platters with roughly 2.5" diameter, however, if you look at the pictures of the savvio 15k (for instance http://www.seagate.com/www/en-us/produc ... avvio_15k/), you can clearly see that it does in fact not (unlike the savvio 10k) use 2.5" platters, but much smaller ones (I'd guess something like 1.8"?).whiic wrote:Very high rpm. But how about the platter size? Well, it's no different from the platter size used in 3.5" 15krpm drives. 3.5" SCSIs use smaller platters that what could be physically fitted into that size of HDD casing.
(But otherwise I quite agree, there is absolutely no reason why this hd would be quiet, the small platter size (lower mass) might provide some benefits, but the high rpm will certainly negate that - not to mention actuator noise, and that low noise was certainly no consideration in designing this unit)
"Not true. It's true that 3.5" 15k rpm drives use platters with roughly 2.5" diameter, however, if you look at the pictures of the savvio 15k (for instance http://www.seagate.com/www/en-us/produc ... avvio_15k/), you can clearly see that it does in fact not (unlike the savvio 10k) use 2.5" platters, but much smaller ones (I'd guess something like 1.8"?)."
Ah, my mistake. I had not seen pictures of 15krpm Savvio before. To me, that appears the platter size is noticeably smaller than typical platter used in 2.5" 7200rpm and 2.5" 10000rpm units. It doesn't appear to be 1.8" either but probably something in between... The area used by platters is roughly the same as with most 10000rpm 3.5" SCSIs. 15000rpm 3.5" SCSIs use even smaller platters when compared to the area usable.
I'd guess while 15krpm drives can use 2.5" platters (as they use that size in 3.5 inch form factor SCSIs) the use of ~2" platters would decrease seek length an thus improve performance. It would also make the actuator arm shorter, thus both making it lighter and having significantly lower rotational inertia (both due to being short and due to being light). 2.5" inch form factor does limit the physical size of actuator coil, thus the force used to move the actuator is limited and making a smaller actuator might prevent the performance loss of using a smaller 2.5" form factor instead of typical 3.5".
So maybe there would be some benefit from these 2.5" 15krpm when compared to 3.5" 15krpm drives. Both due to smaller platter and due to smaller actuator. That'd make less idle noise and less seek noise. Still: why would you compare Savvio to 15krpm 3.5" drive? Is 15krpm 3.5" quiet? This is SPCR, so I assume being quiet is somewhat high on the list of priorities.
Surely, silence isn't the only priority. There's things such as performance, capacity and power consumption. Performance of a 15krpm drive is optimized for random access. They use small platters (and 2.5" 15krpm appears to use even smaller) which basically mean they sacrifice sequential read speed (STR) to lower access time. Desktop use (office, photo editing, video recording, gaming, whatevah) is typically either
- sequential or
- consisting of multiple sequential streams.
First is defined by STR. Latter is defined by careful optimization of cache, STR and seek time.
Sure, 15krpm will allow quite high STR too, despite sacrificing some of it to obtain very low access times. If you make the platters spin twice as fast but make the platters half the size, STR is unaffected. (If the media uses same linear data density.)
There are "good" and "bad" performing desktop drives. STR, seek time and amount of cache alone doesn't tell which will perform best: it's the firmware that optimizes these three to make the best use of them. There's no drive that's good for every imaginable environment. Best performing 7200rpm drive is usually pretty close to average 15krpm SCSI drives in some certain desktop use environment.
In server environment it's obvious 15krpm SCSI is the best, followed by 10krpm SCSI, 10krpm SATA, 7.2krpm SATA and PATA. But in desktop environment 10krpm SATA (=Raptor) tend to come first, followed by 15krpm SCSIs and best 7.2krpm SATAs, then trailed by 10krpm SCSIs and rest 7.2krpm SATAs and PATAs.
SCSI is waste of money if you don't need a real server. Gaming & multi-tasking applications may be intensive HDD use but they are still localized access patterns with high STR requirements or highly dependent on caching algorithms due to multiple sequential I/O streams.
Ah, my mistake. I had not seen pictures of 15krpm Savvio before. To me, that appears the platter size is noticeably smaller than typical platter used in 2.5" 7200rpm and 2.5" 10000rpm units. It doesn't appear to be 1.8" either but probably something in between... The area used by platters is roughly the same as with most 10000rpm 3.5" SCSIs. 15000rpm 3.5" SCSIs use even smaller platters when compared to the area usable.
I'd guess while 15krpm drives can use 2.5" platters (as they use that size in 3.5 inch form factor SCSIs) the use of ~2" platters would decrease seek length an thus improve performance. It would also make the actuator arm shorter, thus both making it lighter and having significantly lower rotational inertia (both due to being short and due to being light). 2.5" inch form factor does limit the physical size of actuator coil, thus the force used to move the actuator is limited and making a smaller actuator might prevent the performance loss of using a smaller 2.5" form factor instead of typical 3.5".
So maybe there would be some benefit from these 2.5" 15krpm when compared to 3.5" 15krpm drives. Both due to smaller platter and due to smaller actuator. That'd make less idle noise and less seek noise. Still: why would you compare Savvio to 15krpm 3.5" drive? Is 15krpm 3.5" quiet? This is SPCR, so I assume being quiet is somewhat high on the list of priorities.
Surely, silence isn't the only priority. There's things such as performance, capacity and power consumption. Performance of a 15krpm drive is optimized for random access. They use small platters (and 2.5" 15krpm appears to use even smaller) which basically mean they sacrifice sequential read speed (STR) to lower access time. Desktop use (office, photo editing, video recording, gaming, whatevah) is typically either
- sequential or
- consisting of multiple sequential streams.
First is defined by STR. Latter is defined by careful optimization of cache, STR and seek time.
Sure, 15krpm will allow quite high STR too, despite sacrificing some of it to obtain very low access times. If you make the platters spin twice as fast but make the platters half the size, STR is unaffected. (If the media uses same linear data density.)
There are "good" and "bad" performing desktop drives. STR, seek time and amount of cache alone doesn't tell which will perform best: it's the firmware that optimizes these three to make the best use of them. There's no drive that's good for every imaginable environment. Best performing 7200rpm drive is usually pretty close to average 15krpm SCSI drives in some certain desktop use environment.
In server environment it's obvious 15krpm SCSI is the best, followed by 10krpm SCSI, 10krpm SATA, 7.2krpm SATA and PATA. But in desktop environment 10krpm SATA (=Raptor) tend to come first, followed by 15krpm SCSIs and best 7.2krpm SATAs, then trailed by 10krpm SCSIs and rest 7.2krpm SATAs and PATAs.
SCSI is waste of money if you don't need a real server. Gaming & multi-tasking applications may be intensive HDD use but they are still localized access patterns with high STR requirements or highly dependent on caching algorithms due to multiple sequential I/O streams.
A longe since not posting.
What I read in this thread made me return.
Savvio are the best HDs, noise, power comsuption, and performance wise.
http://www.storagereview.com/php/benchm ... 8&devCnt=2
Do you think they are just a "sized down SCSI"? What about these 9W vs 19W?
http://www.storagereview.com/php/benchm ... 7&devCnt=2
15K Series have smaller plattes because they use perpendicular technology. Since they are sized as 10K Series, 36GB and 73GB, the platters have to be smaller.
And that's why they can do 15K rpm in a 2,5" form: smaller platters.
SCSI may be a waste of money, access time is not.
Did you see that 2.9ms average random access time on the .pdf? Samsung Spinpoint SP120 250GB has 14.5ms. 'nuff said.
The only problem here is the money. That makes the Savvios a no-go.
What I read in this thread made me return.
Savvio are the best HDs, noise, power comsuption, and performance wise.
http://www.storagereview.com/php/benchm ... 8&devCnt=2
Do you think they are just a "sized down SCSI"? What about these 9W vs 19W?
http://www.storagereview.com/php/benchm ... 7&devCnt=2
15K Series have smaller plattes because they use perpendicular technology. Since they are sized as 10K Series, 36GB and 73GB, the platters have to be smaller.
And that's why they can do 15K rpm in a 2,5" form: smaller platters.
SCSI may be a waste of money, access time is not.
Did you see that 2.9ms average random access time on the .pdf? Samsung Spinpoint SP120 250GB has 14.5ms. 'nuff said.
The only problem here is the money. That makes the Savvios a no-go.
"Savvio are the best HDs, noise, power comsuption, and performance wise."
All of them? I strongly disagree.
In your first link you compared Savvio (desktop mode) to Raptor 150 (NCQ enabled = server mode). Even though NCQ can hurt desktop performance, Raptor still beat Savvio 10krpm. And it beat Savvio into pulp, hands down.
Or did you skip that "Single-User Suite 4.0" section of the benchmarks? if you did, then why? "Single-User Suite 4.0" would be the only section I'd be interested in if I was building a top-end gaming computer. How on earth can you ignore the results compeletely?
"Do you think they are just a "sized down SCSI"? What about these 9W vs 19W?"
Sure, that's because 10krpm Savvio uses smaller platters than 10krpm 3.5" SCSIs. When comparing 15krpm Savvio and 15krpm 3.5" SCSIs the difference isn't probably that big. But like mczak corrected, the platters are still somewhat smalled and would give some positive effect.
You also provided a second link to back up that power use argument. I have to ask you: are you blind? First you claim 2.5" 10krpm Savvio is the BEST, then you compare it to Fujitsu MAU and, surprize, Fujitsu beats it in every single performance benchmark (both desktop and server benchmarks)!
Certainly, MAU is a 15krpm 3.5" SCSI. It's unfair to compare it to 10krpm 2.5" Savvio. But it wasn't me who made the comparison. What were you trying to prove by showing us 15krpm 3.5" drive uses more power than 10krpm 2.5" Savvio? Who would have doubted that if you just said it instead of showing it? I wouldn't. It's quite obvious.
A bit more fair comparison would be to compare 10krpm Savvio to 10krpm (instead of 15krpm) 3.5" SCSIs. The results are here:
http://www.storagereview.com/php/benchm ... 4&devCnt=5
As you may see (if you are not blind) Savvio isn't the best performer in ANY of the performance benchmarks. Maxtor Atlas 5th generation and Fujitsu MAT compete with each other in both desktop and server performance. (That is, if we compare only SCSI drives. Include Raptor into comparison with 10krpm SCSIs and all SCSIs are beaten. Except Fujitsu which manages to tie.)
Compare 15krpm SCSIs to Raptor and only Maxtor and Fujitsu have a chance. Seagate's and Hitachi's 15krpm offerings aren't even close to 10krpm Raptor.
http://www.storagereview.com/php/benchm ... 9&devCnt=4
A good-performing 7.2krpm drive can beat a 15krpm drive. Don't belive it? Here:
http://www.storagereview.com/php/benchm ... 7&devCnt=2
The newest generation of Seagate's 15krpm SCSI. I didn't even have to find an old one to make the comparison. 7K500 is older. How could an "old" 7200rpm drive like 7K500 beat the newest perpendicular recording Seagate 15krpm drive with 135.0 MB/s STR and a low 5.8 ms access time? (Answer: firmware)
Did you see that 2.9ms average random access time on the .pdf? Samsung Spinpoint SP120 250GB has 14.5ms. 'nuff said.
Bullsh*t! 15krpm drive has 2.9ms average access time! Average latency alone is 2.0ms so that'd leave 0.9ms for the average seek time. (Average access time = average seek time + avearage latency.)
You probably meant 15krpm drives have approximately 2.9ms seek time (= 4.9 ms access time)? Is it really fair to compare 15krpm drives seek time to 7.2krpm drive's access time?
Even if you compared access time to access time (instead of access time to seek time) what indication of real-world performance would that really give? None.
"15K Series have smaller plattes because they use perpendicular technology."
Seagate use perpendicular recording on 3.5 inchers too.
"Since they are sized as 10K Series, 36GB and 73GB, the platters have to be smaller."
Funny. That probably means newer 160GB Seagates with perpendicular recording use smaller platters than the older (7200.9) 160GB Seagates with longitudal recording. Both have only one platter and both sides are used. Because perpendicular allow higher density, the platters have to be smaller. No... not really. I don't think the way you do.
All of them? I strongly disagree.
In your first link you compared Savvio (desktop mode) to Raptor 150 (NCQ enabled = server mode). Even though NCQ can hurt desktop performance, Raptor still beat Savvio 10krpm. And it beat Savvio into pulp, hands down.
Or did you skip that "Single-User Suite 4.0" section of the benchmarks? if you did, then why? "Single-User Suite 4.0" would be the only section I'd be interested in if I was building a top-end gaming computer. How on earth can you ignore the results compeletely?
"Do you think they are just a "sized down SCSI"? What about these 9W vs 19W?"
Sure, that's because 10krpm Savvio uses smaller platters than 10krpm 3.5" SCSIs. When comparing 15krpm Savvio and 15krpm 3.5" SCSIs the difference isn't probably that big. But like mczak corrected, the platters are still somewhat smalled and would give some positive effect.
You also provided a second link to back up that power use argument. I have to ask you: are you blind? First you claim 2.5" 10krpm Savvio is the BEST, then you compare it to Fujitsu MAU and, surprize, Fujitsu beats it in every single performance benchmark (both desktop and server benchmarks)!
Certainly, MAU is a 15krpm 3.5" SCSI. It's unfair to compare it to 10krpm 2.5" Savvio. But it wasn't me who made the comparison. What were you trying to prove by showing us 15krpm 3.5" drive uses more power than 10krpm 2.5" Savvio? Who would have doubted that if you just said it instead of showing it? I wouldn't. It's quite obvious.
A bit more fair comparison would be to compare 10krpm Savvio to 10krpm (instead of 15krpm) 3.5" SCSIs. The results are here:
http://www.storagereview.com/php/benchm ... 4&devCnt=5
As you may see (if you are not blind) Savvio isn't the best performer in ANY of the performance benchmarks. Maxtor Atlas 5th generation and Fujitsu MAT compete with each other in both desktop and server performance. (That is, if we compare only SCSI drives. Include Raptor into comparison with 10krpm SCSIs and all SCSIs are beaten. Except Fujitsu which manages to tie.)
Compare 15krpm SCSIs to Raptor and only Maxtor and Fujitsu have a chance. Seagate's and Hitachi's 15krpm offerings aren't even close to 10krpm Raptor.
http://www.storagereview.com/php/benchm ... 9&devCnt=4
A good-performing 7.2krpm drive can beat a 15krpm drive. Don't belive it? Here:
http://www.storagereview.com/php/benchm ... 7&devCnt=2
The newest generation of Seagate's 15krpm SCSI. I didn't even have to find an old one to make the comparison. 7K500 is older. How could an "old" 7200rpm drive like 7K500 beat the newest perpendicular recording Seagate 15krpm drive with 135.0 MB/s STR and a low 5.8 ms access time? (Answer: firmware)
Did you see that 2.9ms average random access time on the .pdf? Samsung Spinpoint SP120 250GB has 14.5ms. 'nuff said.
Bullsh*t! 15krpm drive has 2.9ms average access time! Average latency alone is 2.0ms so that'd leave 0.9ms for the average seek time. (Average access time = average seek time + avearage latency.)
You probably meant 15krpm drives have approximately 2.9ms seek time (= 4.9 ms access time)? Is it really fair to compare 15krpm drives seek time to 7.2krpm drive's access time?
Even if you compared access time to access time (instead of access time to seek time) what indication of real-world performance would that really give? None.
"15K Series have smaller plattes because they use perpendicular technology."
Seagate use perpendicular recording on 3.5 inchers too.
"Since they are sized as 10K Series, 36GB and 73GB, the platters have to be smaller."
Funny. That probably means newer 160GB Seagates with perpendicular recording use smaller platters than the older (7200.9) 160GB Seagates with longitudal recording. Both have only one platter and both sides are used. Because perpendicular allow higher density, the platters have to be smaller. No... not really. I don't think the way you do.
-
Bluefront
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
Look.....no sense arguing about the thing before anyone actually gets one and tries it in a low-noise setup. The small size is a benefit, as is the speed. But the cost is a negative for the time being, and if it proves to be noisy, it would be a no-no, no matter what else. I'll wait for some reviews that focus on sound.
"Look.....no sense arguing about the thing before anyone actually gets one and tries it in a low-noise setup. The small size is a benefit, as is the speed."
You make the assumption it has "the speed". Sure, it has rpm, but what use is rpm? Lower access time? Yes. So what?
The fact is, 7200rpm drive has more of "the speed" than Savvio. Even if Savvio was quiet, why would you recommend a poorly performing and expensive drive?
You make the assumption it has "the speed". Sure, it has rpm, but what use is rpm? Lower access time? Yes. So what?
The fact is, 7200rpm drive has more of "the speed" than Savvio. Even if Savvio was quiet, why would you recommend a poorly performing and expensive drive?
-
Bluefront
- *Lifetime Patron*
- Posts: 5316
- Joined: Sat Jan 18, 2003 2:19 pm
- Location: St Louis (county) Missouri USA
'The Savvio 15K drive's unique combination of features including its 70 percent smaller size, lower power consumption (30 percent lower than any other 15K drive), industry's fastest seek time and the industry's highest reliability, make it the ideal storage solution for all server platforms."
It's not the rpm of the drive.....it's the low seek times that make it so fast.
It's not the rpm of the drive.....it's the low seek times that make it so fast.
Seek time alone has no meaning at all. Random access time consists of rotational latency + seek time (+ settling time + command overhead). 15krpm drives typically (but not necessarily) have lower seek times. The only thing they have certainly is higher rpm. Even though you dismiss rpm as a meaningful factor, it does have a meaning by itself: rotational latency is dependent on rpm... and rpm alone.
Rotational latency (in milliseconds) = 30000/rpm.
Thus for 7200rpm drives rotational latency is always 4.17 ms. If the seek time is for example 8.5 ms, total random access time would be 12.67 ms (this is the access time we see when running benchmarks!).
For 15krpm drives rotational latency is always 2.00 ms.
For 10krpm drives it's 3.00 ms.
Even though low seek times make drive faster, and low latency makes it even more faster, there is still something else to drive performance and it has nothing to do with something being randomly accessed. I never claimed Raptors and 7.2krpm SATA drives were good performers when there's a lot of random access. On the contrary, I said Savvio sucks in a gaming freaks environment because running games has nothing to do with random seeks.
While it may appear impossible for some to accept the fact that there's other attributors to real-life performance than just STR, seek time or rpm, there really is. Other factors include cache, interface speed and last (but definitely not least) firmware. (And by firmware I mean read and write cache algorithms, native/tagged command queuing, error correction, background SMART routines, ... everything.)
SCSI drives are not intended for gaming, thus their firmware is not optimized for it. They are intended for highly random access and thus cannot do much read-ahead caching to improve performance. In typical gaming (or other desktop) environment access patterns are not random. Even when the drive appears to be seeking constantly, it's typically seeking between 2 or 3 locations. Read-ahead caching reduces number of seeks and increases performance noticeably in this type of operation and SCSI drives are not optimized for it. (I believe their cache is mainly optimized for write caching as write caching can be used for random writes better than read cache used for random reads.)
Outside of a server, SCSI drives suck. Inside a server, they kick ass and chew bubble gum.
Rotational latency (in milliseconds) = 30000/rpm.
Thus for 7200rpm drives rotational latency is always 4.17 ms. If the seek time is for example 8.5 ms, total random access time would be 12.67 ms (this is the access time we see when running benchmarks!).
For 15krpm drives rotational latency is always 2.00 ms.
For 10krpm drives it's 3.00 ms.
Even though low seek times make drive faster, and low latency makes it even more faster, there is still something else to drive performance and it has nothing to do with something being randomly accessed. I never claimed Raptors and 7.2krpm SATA drives were good performers when there's a lot of random access. On the contrary, I said Savvio sucks in a gaming freaks environment because running games has nothing to do with random seeks.
While it may appear impossible for some to accept the fact that there's other attributors to real-life performance than just STR, seek time or rpm, there really is. Other factors include cache, interface speed and last (but definitely not least) firmware. (And by firmware I mean read and write cache algorithms, native/tagged command queuing, error correction, background SMART routines, ... everything.)
SCSI drives are not intended for gaming, thus their firmware is not optimized for it. They are intended for highly random access and thus cannot do much read-ahead caching to improve performance. In typical gaming (or other desktop) environment access patterns are not random. Even when the drive appears to be seeking constantly, it's typically seeking between 2 or 3 locations. Read-ahead caching reduces number of seeks and increases performance noticeably in this type of operation and SCSI drives are not optimized for it. (I believe their cache is mainly optimized for write caching as write caching can be used for random writes better than read cache used for random reads.)
Outside of a server, SCSI drives suck. Inside a server, they kick ass and chew bubble gum.