Hi I'm planning to change from IDE to SCSI. Currently I have 7200RPM IBM 40GB IDE. I have been looking into a mid-priced solution. Is there a real speed increase with SCSI? Could someone recommend some adapter manufacturer / model that has been proven to work? Jaska.
I'm planning to change from IDE to SCSI.
Currently I have 7200RPM IBM 40GB IDE.
I have been looking into a mid-priced solution.
There's not really any such thing with SCSI. It's still insanely expensive! I have a SCSI based system and wanted to add a drive about a month ago. I looked at the reviews at the time and a 40GB Fujitsu drive was winning all the reviews. The IDE version was 85UKP, the SCSI version was 270UKP. I got the IDE one!
Is there a real speed increase with SCSI?
On one spindle, no. Many drives, including the Fujitsu I got, have the same rotating mechanism in both the SCSI and IDE versions. It's just the interface that's different. If you want to connect several drives to one machine, or you want lots of parallel access (for RAID arrays or LVM trickery) then SCSI will give significantly better performance because the data can be moved through the bus much faster. With just one or two drives on an largely unstressed machine you'll be hard pushed to tell the difference.
Could someone recommend some adapter manufacturer / model that has been proven to work?
Adaptec are the safest bet. Decent products, supported Linux for years.
On Thu, Jan 03, 2002 at 01:34:25PM +0200, Jaakko Tamminen wrote: -> I'm planning to change from IDE to SCSI. Dust off your wallet. -> Is there a real speed increase with SCSI? There certainly can be, but very fast scsi stuff is expensive. I have reasonably fast scsi stuff on this computer: One Adaptec 29160 Ultra 160 host adapter Two IBM UltraStar 36GB Ultra160 10,000RPM drives One Seagate Cheetah X15 18GB Ultra160 15,000RPM drive ...and the disk i/o is phenomenally fast. Multitasking work is also far superior to IDE/ATA. But if you total up the costs of all that stuff, it's a LOT of money. Also, these 10 and 15K RPM drives run pretty hot, so factor in the cost of supplementary cooling mechanisms for your computer. Mine has *9*, count 'em NINE fans. Gets a little noisy in here. 1 in the power supply blowing out 1 below ps on case wall blowing out 1 on cpu 1 in slot blowing out 1 on video card 1 on drive cooler 2 on front of case sucking cool air in 1 on side blowing on motherboard Personally, I think it's all worth the money. YMMV. Michael -- "# chmod a+x /bin/laden" Michael Nelson San Francisco, CA
On 03 Jan 2002 13:34:25 +0200, Jaakko Tamminen wrote:
Hi
I'm planning to change from IDE to SCSI.
Currently I have 7200RPM IBM 40GB IDE.
I have been looking into a mid-priced solution.
Is there a real speed increase with SCSI?
Could someone recommend some adapter manufacturer / model that has been proven to work?
Hello, Shifting to SCSI eh ?...... Welcome to temptation island. My unsolicited advice on SCSI in personal / private box(es) 1. NEVER buy New ,except when tax-deductable. 2. Cross Reference http://www.linhardware.com and http://www.storagereview.com/ 3. Join this religious cult by starting at http://scsifaq.org:9080/scsi_faq/ 4. Dont be easily intimidated by so-called experts. Ask whatever's on your troubled mind. 5. Pray , that you know of a big computer-fair stall that supplies bits and bobs. Dont know of speed increases when compared to modern IDE HDs.... Just check #2. Holy Smoke.!!!!! And finally, dont bother with bleeding-edge like U160 or U320 for now until you can make it worthwhile ,in terms of money. Let the big shops and server farms unwittingly iron out any known quirks. Enjoy the stability of the SCSI bus. I pray your Power Supply Unit dont get swamped by a multi-drive SCSI system. -- Kemdi IN_SuSE_d Since 5.2 ICQ:112290572 kemdi@btopenworld.com +++++++++++++++++++++
Il 12:34, giovedì 3 gennaio 2002, Jaakko Tamminen ha scritto:
Hi
I'm planning to change from IDE to SCSI.
Currently I have 7200RPM IBM 40GB IDE.
I have been looking into a mid-priced solution.
Is there a real speed increase with SCSI?
Could someone recommend some adapter manufacturer / model that has been proven to work?
Jaska.
Even if new EIDE disks are quite fasts, SCSI are much more faster not only because of the bus, but because they have lower seek time. Seek time is what make your HD slow. I have been noticing more increase of performance switching to SCSI then switching from 128Mbyte to 512Mbyte of RAM. I am using Adaptec 29160 and it's pretty fast and well supported. Praise
I agree. SCSI is much faster than IDE. Even if the drives have equal output (say 30MB/sec), the SCSI drive will have a much smaller seek time (say 10ms for IDE and 4 for SCSI), and when copying from drive to drive you get the max output of each drive (it seems anyway . . .). On IDE it seems you only get max output when going from disk to CPU (say when loading a program). I strongly recommend the Tekram DC-3903W over the adaptec cards, it has all the features you need and the retail kit (with cables) is cheaper than the bare adaptec card. I paid around $170 for mine. There are reviews comparing the Tekram to the Adaptec 29160, and the conclusion is that the cards are pretty much equal in performance. Check on www.storagereview.com for reviews . . . _Nick On Thu, Jan 03, 2002 at 05:47:12PM +0100, Praise wrote:
Il 12:34, giovedì 3 gennaio 2002, Jaakko Tamminen ha scritto:
Hi
I'm planning to change from IDE to SCSI.
Currently I have 7200RPM IBM 40GB IDE.
I have been looking into a mid-priced solution.
Is there a real speed increase with SCSI?
Could someone recommend some adapter manufacturer / model that has been proven to work?
Jaska.
Even if new EIDE disks are quite fasts, SCSI are much more faster not only because of the bus, but because they have lower seek time. Seek time is what make your HD slow. I have been noticing more increase of performance switching to SCSI then switching from 128Mbyte to 512Mbyte of RAM. I am using Adaptec 29160 and it's pretty fast and well supported.
Praise
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
-- Nick Webb http://www.uidaho.edu/~nickw/
Il 18:26, giovedì 3 gennaio 2002, Nick Webb ha scritto:
I agree. SCSI is much faster than IDE. Even if the drives have equal output (say 30MB/sec), the SCSI drive will have a much smaller seek time (say 10ms for IDE and 4 for SCSI), and when copying from drive to drive you get the max output of each drive (it seems anyway . . .). On IDE it seems you only get max output when going from disk to CPU (say when loading a program).
The main problems with IDE bus are: 1) CPU utilization grows up when tranferring data. There is no Bus Mastering in EIDE, AFAIK. So your output drops when you use the CPU very much. 2) An IDE controller can only read or write from all disks it serves. I mean, it can only write on all disks or read from all disks at once. Some times ago, I used a little program to measure the speed of a cdrom. Asus 50x EIDE = 24/25x when I was lucky. Once it reached only 8x, because CPU had something better to do. Plextor 40X SCSI = 41,5x was ALWAYS my best... The performance was always the same. A friend of mine has got a Yamaha eide cd-burner. He cannot burn 16x cd-r because of buffer underruns, and he has got a p3 800 with 256mbyte. I got a Waitec scsi cd-burner. I can copy cd on the fly at 16x while surfing the web, acting as a file server for my home network and listening to some mp3. And I only got an athlon 600... Ok, I also got Linux while he is using Windows... :-) Last story: I installed Win 98 in 8 minutes in my scsi machine. When it was EIDE, the installation took 25mins... Praise
Yes. I have a 16x Yamaha CD-R and a 32x TEAC CD-ROM (both SCSI) and have never had a buffer underun, nor does copying CDs seem to affect CPU utilization (I'm sure it does a LITTLE, but it's not noticeable). I used to have a 48x Toshiba IDE CD-ROM, and the 32x SCSI is certainly faster. If I was in the market for a new PC I would get SCSI drives and drop down a few notches on the CPU. I heard in a Linux journal article that 1ms less seek time on your hard disks used to be equivalent to 30MHz of CPU power on a Unix system, and that today it is probably close to 300MHz. I don't know if it's that drastic, but close. The moral of the story is that a fast disk subsystem is usually more important to Unix systems than fast CPU(s) . . . The Linux Journal Article: "Building the Ultimate Linux Box" http://www.linuxjournal.com/article.php?sid=5420 It's a pretty good read even if you (like most of us) are on a strict budget. They stress that you should get the best (fastest) hard drives you can get your hands on, as that has most to do with overall performance. On Thu, Jan 03, 2002 at 07:53:14PM +0100, Praise wrote:
Il 18:26, giovedì 3 gennaio 2002, Nick Webb ha scritto:
I agree. SCSI is much faster than IDE. Even if the drives have equal output (say 30MB/sec), the SCSI drive will have a much smaller seek time (say 10ms for IDE and 4 for SCSI), and when copying from drive to drive you get the max output of each drive (it seems anyway . . .). On IDE it seems you only get max output when going from disk to CPU (say when loading a program).
The main problems with IDE bus are:
1) CPU utilization grows up when tranferring data. There is no Bus Mastering in EIDE, AFAIK. So your output drops when you use the CPU very much.
2) An IDE controller can only read or write from all disks it serves. I mean, it can only write on all disks or read from all disks at once.
Some times ago, I used a little program to measure the speed of a cdrom. Asus 50x EIDE = 24/25x when I was lucky. Once it reached only 8x, because CPU had something better to do. Plextor 40X SCSI = 41,5x was ALWAYS my best... The performance was always the same.
A friend of mine has got a Yamaha eide cd-burner. He cannot burn 16x cd-r because of buffer underruns, and he has got a p3 800 with 256mbyte. I got a Waitec scsi cd-burner. I can copy cd on the fly at 16x while surfing the web, acting as a file server for my home network and listening to some mp3. And I only got an athlon 600... Ok, I also got Linux while he is using Windows... :-)
Last story: I installed Win 98 in 8 minutes in my scsi machine. When it was EIDE, the installation took 25mins...
Praise
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
-- Nick Webb http://www.uidaho.edu/~nickw/
On 3 Jan 2002, Praise wrote:
Some times ago, I used a little program to measure the speed of a cdrom. Asus 50x EIDE = 24/25x when I was lucky. Once it reached only 8x, because CPU had something better to do. Plextor 40X SCSI = 41,5x was ALWAYS my best... The performance was always the same.
The issue is more than just SCSI vs. IDE; it's Asus vs. Plextor, and Plextor is almost guaranteed to make better CDR(W) / ROM drives. That's why I have a Plextor 12/10/32S CDRW drive, with which I am extremely satisfied (25X-30X ripping w/ Linux 2.4.16). I went for SCSI over IDE because I doubt my machine (233 Pentium I MMX w/ 32MB of RAM) could handle the load on its IDE bus during burning at 12X. -- Karol Pietrzak PGP KeyID: 3A1446A0
Il 03:40, venerdì 4 gennaio 2002, Karol Pietrzak ha scritto:
On 3 Jan 2002, Praise wrote:
Some times ago, I used a little program to measure the speed of a cdrom. Asus 50x EIDE = 24/25x when I was lucky. Once it reached only 8x, because CPU had something better to do. Plextor 40X SCSI = 41,5x was ALWAYS my best... The performance was always the same.
The issue is more than just SCSI vs. IDE; it's Asus vs. Plextor, and Plextor is almost guaranteed to make better CDR(W) / ROM drives.
Not completely. Asus is one of the best CD-ROM EIDE unit I have seen around. And my tests were done with the same CD, always: why should the cdrom be faster or slower on a CD, if not because of the CPU getting stressed?
That's why I have a Plextor 12/10/32S CDRW drive, with which I am extremely satisfied (25X-30X ripping w/ Linux 2.4.16). I went for SCSI over IDE because I doubt my machine (233 Pentium I MMX w/ 32MB of RAM) could handle the load on its IDE bus during burning at 12X.
Yes you are right.... I got a friend of mine with an 8x Cd-rw and he can only go at 4x without risking uverruns... Praise
On Friday 04 January 2002 05:44, Praise wrote:
Il 03:40, venerdì 4 gennaio 2002, Karol Pietrzak ha scritto:
That's why I have a Plextor 12/10/32S CDRW drive, with which I am extremely satisfied (25X-30X ripping w/ Linux 2.4.16). I went for SCSI over IDE because I doubt my machine (233 Pentium I MMX w/ 32MB of RAM) could handle the load on its IDE bus during burning at 12X.
Yes you are right.... I got a friend of mine with an 8x Cd-rw and he can only go at 4x without risking uverruns...
That's what you get for running Intel CPU's. I have a K6-233 on a FIC PA-2007, 64 megs, 19.2 gig Bigfoot, and an HP 9500 12/8/32 burner. It's a headless file-server that sits next to my desk. I burn CDRs at 12x and RW's at 8x while it's also serving ogg files over NFS to my workstation. I've had exactly one coaster in ~6 months. I use RWs for daily incremental backups. On average, probably 6 burns a day. The single coaster was a 12x CDR burn while the machine was also trying to serve a very large mpeg to my workstation. IOW, it was my fault. -- Robert S. McMillan rmcmilla@centurytel.net "The right to privacy... is the most comprehensive of rights and the right most valued by civilized man" - Justice Louis Brandeis, US Supreme Court, 1928
Il 21:39, venerdì 4 gennaio 2002, Robert S. McMillan ha scritto:
On Friday 04 January 2002 05:44, Praise wrote:
Il 03:40, venerdì 4 gennaio 2002, Karol Pietrzak ha scritto:
That's why I have a Plextor 12/10/32S CDRW drive, with which I am extremely satisfied (25X-30X ripping w/ Linux 2.4.16). I went for SCSI over IDE because I doubt my machine (233 Pentium I MMX w/ 32MB of RAM) could handle the load on its IDE bus during burning at 12X.
Yes you are right.... I got a friend of mine with an 8x Cd-rw and he can only go at 4x without risking uverruns...
That's what you get for running Intel CPU's.
I have a K6-233 on a FIC PA-2007, 64 megs, 19.2 gig Bigfoot, and an HP 9500 12/8/32 burner. It's a headless file-server that sits next to my desk. I burn CDRs at 12x and RW's at 8x while it's also serving ogg files over NFS to my workstation. I've had exactly one coaster in ~6 months. I use RWs for daily incremental backups. On average, probably 6 burns a day. The single coaster was a 12x CDR burn while the machine was also trying to serve a very large mpeg to my workstation. IOW, it was my fault.
I did not mention they use windows... :-) Praise
Hi to all who reside here! (or lurk) :o) I have been following this thread since it started and could see there were several knowledgable responses to the original question. I thought adding a little of my experiences to the discussion might help also. SCSI and Unix/Linux have always been best friends and if budget would allow, should be anyone's choice when building/buying a system. It is just better all around, plus for a true multitasking OS such as Linux, it's the best answer. Since moving from the Amiga computer/w scsi to Linux on a very fast PC system/w IDE, I have had several opportunities to make comparisons. Now realize, that the Amiga is only a 060/50mhz cpu, while the Linux machine is an AthlonXP 1700+ setup! I mention this point because some have made reference to the cpu playing a big part in the speed differences. I don't think this is the case. Multitasking is much more easily accomplished on the Amiga with scsi versus the PC with IDE. CDRW write times, with very nearly equal drive specs, are easily one half to one quarter the time on the Amiga compared to the IDE! Loading programs, reading, writing on the scsi is considerably faster on a 4gb drive than the 13gb IDE drive in the PC. Bottom line is that IDE is cheaper, slower version of scsi and will never give you the advantages of scsi. Plus, one has to add in the still older and clunkier design of the PC hardware versus the more updated design of the Amiga and scsi. Just some observations, your mileage may vary. My point is if budget will allow and you need or want the added capabilities of scsi, then get it, you won't want to go back to IDE again! Patrick -------------------------- On Friday 04 January 2002 08:44 am, Praise, went on about:
Il 03:40, venerdì 4 gennaio 2002, Karol Pietrzak ha scritto:
On 3 Jan 2002, Praise wrote:
Some times ago, I used a little program to measure the speed of a cdrom. Asus 50x EIDE = 24/25x when I was lucky. Once it reached only 8x, because CPU had something better to do. Plextor 40X SCSI = 41,5x was ALWAYS my best... The performance was always the same.
The issue is more than just SCSI vs. IDE; it's Asus vs. Plextor, and Plextor is almost guaranteed to make better CDR(W) / ROM drives.
Not completely. Asus is one of the best CD-ROM EIDE unit I have seen around. And my tests were done with the same CD, always: why should the cdrom be faster or slower on a CD, if not because of the CPU getting stressed?
That's why I have a Plextor 12/10/32S CDRW drive, with which I am extremely satisfied (25X-30X ripping w/ Linux 2.4.16). I went for SCSI over IDE because I doubt my machine (233 Pentium I MMX w/ 32MB of RAM) could handle the load on its IDE bus during burning at 12X.
Yes you are right.... I got a friend of mine with an 8x Cd-rw and he can only go at 4x without risking uverruns...
Praise
-- ---KMail 1.3.2--- SuSE Linux v7.2 Pro--- Registered Linux User #225206 /tracerb@sprintmail.com/ *Magic Page Products* Amiga-Sales & Service-http://home.sprintmail.com/~tracerb
On Thu, 2002-01-03 at 05:34, Jaakko Tamminen wrote:
I'm planning to change from IDE to SCSI. Currently I have 7200RPM IBM 40GB IDE.
Why?
I have been looking into a mid-priced solution.
Fast, cheap, good; pick any two. You cannot have all three, though.
Is there a real speed increase with SCSI?
No. Not with a single disk on a single controller with a single user. At this point, we're up to hardware throughput, not wire throughput. Though ATA133 and Ultra160 boast incredible burst throughput, you'll have a hard time finding a disk that can actually feed more than 30MB/S in a stream, and your current IBM ATA disk isn't slow enough for you to notice a difference.
Could someone recommend some adapter manufacturer / model that has been proven to work?
http://cdb.suse.de -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
On Thu, Jan 03, 2002 at 02:48:32PM -0600, Jon Pennington wrote:
On Thu, 2002-01-03 at 05:34, Jaakko Tamminen wrote:
I'm planning to change from IDE to SCSI. Currently I have 7200RPM IBM 40GB IDE.
Why?
I have been looking into a mid-priced solution.
Fast, cheap, good; pick any two. You cannot have all three, though.
Is there a real speed increase with SCSI?
No. Not with a single disk on a single controller with a single user. At this point, we're up to hardware throughput, not wire throughput. Though ATA133 and Ultra160 boast incredible burst throughput, you'll have a hard time finding a disk that can actually feed more than 30MB/S
My Quantum altlas III (10,000RPM) is old and slow by SCSI standards and it can push over 40MB/sec sustained . . . I'm sure the newer SCSI models are over 50MB/sec. Although I had an IBM IDE model that could push 35MB/sec sustained, the SCSI drive noticibly faster by far. I also have an old IBM SCSI drive (7200RPM) that is comparable to the IBM IDE drive I had; the old IBM SCSI drive can only push 30MB/sec sustained, but it feels much faster than the IDE drive that can push 35. I think the much smaller seek time is the cause . . . When copying from drive to drive (or CD-ROM to drive) there is a very huge difference in performance, IDE can't even touch it.
in a stream, and your current IBM ATA disk isn't slow enough for you to notice a difference.
Could someone recommend some adapter manufacturer / model that has been proven to work?
-- -=|JP|=- Need a good geek? I'm unemployed!
'01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com |
_________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
-- Nick Webb http://www.uidaho.edu/~nickw/
On Thu, 2002-01-03 at 15:02, Nick Webb wrote:
My Quantum altlas III (10,000RPM) is old and slow by SCSI standards and it can push over 40MB/sec sustained . . . I'm sure the newer SCSI models are over 50MB/sec. Although I had an IBM IDE model that could push 35MB/sec sustained, the SCSI drive noticibly faster by far. I also have an old IBM SCSI drive (7200RPM) that is comparable to the IBM IDE drive I had; the old IBM SCSI drive can only push 30MB/sec sustained, but it feels much faster than the IDE drive that can push 35. I think the much smaller seek time is the cause . . .
50MB/S, eh? I'm a little out of the loop in the SCSI arena, but 50MB/S clearly isn't a "mid-priced" solution. We're talking about, what, $300+ for the disk by itself? Tack on another $200 or so for a controller, and let's assume that somebody *gave* you a cable? Can you show me proof of this seek time phenomena? Seek is determined by how fast the arm inside the disk can move the head; given the same internal hardware (IBM, WD, Fujitsu do this), why is a SCSI drive quicker?
When copying from drive to drive (or CD-ROM to drive) there is a very huge difference in performance, IDE can't even touch it.
Well, I'll humbly say, "Not in my experience at the workstation level," but I've been out for a while. You know me, I love being *proven* wrong, but hate all this rhetoric about "My bike's faster than yours because it has chrome handlebars." ;) -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
On Thu, Jan 03, 2002 at 03:17:01PM -0600, Jon Pennington wrote:
On Thu, 2002-01-03 at 15:02, Nick Webb wrote:
My Quantum altlas III (10,000RPM) is old and slow by SCSI standards and it can push over 40MB/sec sustained . . . I'm sure the newer SCSI models are over 50MB/sec. Although I had an IBM IDE model that could push 35MB/sec sustained, the SCSI drive noticibly faster by far. I also have an old IBM SCSI drive (7200RPM) that is comparable to the IBM IDE drive I had; the old IBM SCSI drive can only push 30MB/sec sustained, but it feels much faster than the IDE drive that can push 35. I think the much smaller seek time is the cause . . .
50MB/S, eh? I'm a little out of the loop in the SCSI arena, but 50MB/S clearly isn't a "mid-priced" solution. We're talking about, what, $300+ for the disk by itself? Tack on another $200 or so for a controller, Probably. I paid $200 for my drive and $175 for the controller (with 3 cables: LVD, WIDE, and NARROW). I agree that it isn't the cheapest . . .
and let's assume that somebody *gave* you a cable? Can you show me proof of this seek time phenomena? Seek is determined by how fast the arm inside the disk can move the head; given the same internal hardware (IBM, WD, Fujitsu do this), why is a SCSI drive quicker?
If the internal hardware was the same, then so should the seek time. I heard someone on this thread (you?) say that they had two drives, one IDE and one SCSI, that had the same internals. I would like to know the model numbers for these (so I could look up the specs), I would be surprised (pleasely perhaps) if the seek times were the same.
When copying from drive to drive (or CD-ROM to drive) there is a very huge difference in performance, IDE can't even touch it.
Well, I'll humbly say, "Not in my experience at the workstation level," but I've been out for a while. You know me, I love being *proven* wrong, but hate all this rhetoric about "My bike's faster than yours because it has chrome handlebars." ;)
I'll look when I get back from work on some comparisons.
-- -=|JP|=- Need a good geek? I'm unemployed!
'01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com |
_________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
-- Nick Webb http://www.uidaho.edu/~nickw/
On Thu, 2002-01-03 at 15:38, Nick Webb wrote:
If the internal hardware was the same, then so should the seek time. I heard someone on this thread (you?) say that they had two drives, one IDE and one SCSI, that had the same internals. I would like to know the model numbers for these (so I could look up the specs), I would be surprised (pleasely perhaps) if the seek times were the same.
I wasn't the one who said that I had one of each, but it's very common practice these days for the internal hardware to be the same. The chips and the interfaces are the most significant differences from mid-high-end SCSI and top-end ATA. I don't know why Seagate doesn't make 15kRPM ATA drives, but it's probably due to the fact that nobody wants to *live* with a 15kRPM disk on their desktop. Those who do want to live with it buy SCSI, anyway. :) -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
On Thu, Jan 03, 2002 at 03:58:37PM -0600, Jon Pennington wrote:
On Thu, 2002-01-03 at 15:38, Nick Webb wrote:
If the internal hardware was the same, then so should the seek time. I heard someone on this thread (you?) say that they had two drives, one IDE and one SCSI, that had the same internals. I would like to know the model numbers for these (so I could look up the specs), I would be surprised (pleasely perhaps) if the seek times were the same.
I wasn't the one who said that I had one of each, but it's very common practice these days for the internal hardware to be the same. The chips and the interfaces are the most significant differences from mid-high-end SCSI and top-end ATA. I don't know why Seagate doesn't make 15kRPM ATA drives, but it's probably due to the fact that nobody wants to *live* with a 15kRPM disk on their desktop. Those who do want to live with it buy SCSI, anyway. :)
Yep, but then why aren't there 10,000 RPM IDE drives? Or are there . . . my 10,000 RPM SCSI drive isn't any louder than IDE drives I've had. I would like to agree with you, but I haven't seen any recent drives with anywhere similar specs on both IDE and SCSI. All recent SCSI drives are either 10K or 15K RPM (and have been for quite a while), for example, and you can't get that in IDE. Which brings it down to another 'fact': The newest (fastest) technology is deployed on SCSI. But, as you pointed out, that comes at a price. -- Nick Webb http://www.uidaho.edu/~nickw/
Il 22:17, giovedì 3 gennaio 2002, Jon Pennington ha scritto:
On Thu, 2002-01-03 at 15:02, Nick Webb wrote:
My Quantum altlas III (10,000RPM) is old and slow by SCSI standards and it can push over 40MB/sec sustained . . . I'm sure the newer SCSI models are over 50MB/sec. Although I had an IBM IDE model that could push 35MB/sec sustained, the SCSI drive noticibly faster by far. I also have an old IBM SCSI drive (7200RPM) that is comparable to the IBM IDE drive I had; the old IBM SCSI drive can only push 30MB/sec sustained, but it feels much faster than the IDE drive that can push 35. I think the much smaller seek time is the cause . . .
50MB/S, eh? I'm a little out of the loop in the SCSI arena, but 50MB/S clearly isn't a "mid-priced" solution. We're talking about, what, $300+ for the disk by itself? Tack on another $200 or so for a controller, and let's assume that somebody *gave* you a cable? Can you show me proof of this seek time phenomena? Seek is determined by how fast the arm inside the disk can move the head; given the same internal hardware (IBM, WD, Fujitsu do this), why is a SCSI drive quicker?
Because SCSI driver are usually better built than EIDE ones. Look at the techincal details.. the Bus is not the reason here.
When copying from drive to drive (or CD-ROM to drive) there is a very huge difference in performance, IDE can't even touch it.
Well, I'll humbly say, "Not in my experience at the workstation level," but I've been out for a while. You know me, I love being *proven* wrong, but hate all this rhetoric about "My bike's faster than yours because it has chrome handlebars." ;)
You can prove yourself wrong by yourself:-) Make the test! As I already told to this list: try to install a win98 in 8 minutes like I did. Usually EIDE guys does not even believe it possible! And try to benchmark a scsi cdrom vs a faster eide one.. you will find what I have found. The scsi cdrom if it's 40x, will get 40x. The eide one wont reply 2 times the same performance, and will use lots of CPU more. Try also to use a cdrom as e primary slave and the hd as primary master. In this configuration (the secondary eide was broken) 6 hours ran away installing a basic suse system!! Praise
On Thu, 2002-01-03 at 16:06, Praise wrote:
Because SCSI driver are usually better built than EIDE ones. Look at the techincal details.. the Bus is not the reason here.
No, drivers have nothing to do with it. We're well within the disk at this point, and *very* far away from both the bus and the software controlling the bus.
As I already told to this list: try to install a win98 in 8 minutes like I did. Usually EIDE guys does not even believe it possible!
You're right. I don't believe you.
Try also to use a cdrom as e primary slave and the hd as primary master.
Far from optimal. Ah, but we're getting into symantecs. ATA is fast enough, and a helluvalot cheaper than new SCSI. It's fast enough for amatuer video production, it's fast enough for ripping DVDs, and it's cheap enough to allow me to eat reasonable food. The decision is yours as a consumer, but nothing's perfect for everything. :) -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
Il 23:24, giovedì 3 gennaio 2002, hai scritto:
On Thu, 2002-01-03 at 16:06, Praise wrote:
Because SCSI driver are usually better built than EIDE ones. Look at the techincal details.. the Bus is not the reason here.
No, drivers have nothing to do with it. We're well within the disk at this point, and *very* far away from both the bus and the software controlling the bus.
Sorry! I am not english.. I wrote driver, I wanted to write hard disk mechanical part:-) We in Italy we call hard disk and storage stuff "driver" sometimes.-)
As I already told to this list: try to install a win98 in 8 minutes like I did. Usually EIDE guys does not even believe it possible!
You're right. I don't believe you.
Come here and try:-) I wont pay your airplane ticket though...
Try also to use a cdrom as e primary slave and the hd as primary master.
Far from optimal.
Obviously.
Ah, but we're getting into symantecs. ATA is fast enough,
Maybe for you:-) Not for me:-)
and a helluvalot cheaper than new SCSI. It's fast enough for amatuer video production, it's fast enough for ripping DVDs, and it's cheap enough to allow me to eat reasonable food. The decision is yours as a consumer, but nothing's perfect for everything. :)
SCSI stresses my nervous system much less then EIDE, btw:-) With EIDE I had some slowdown of my system for I/O reasons I think. Praise
Il 23:24, giovedì 3 gennaio 2002, Jon Pennington ha scritto:
On Thu, 2002-01-03 at 16:06, Praise wrote:
Because SCSI driver are usually better built than EIDE ones. Look at the techincal details.. the Bus is not the reason here.
No, drivers have nothing to do with it. We're well within the disk at this point, and *very* far away from both the bus and the software controlling the bus.
Just for information, look at seek time of seagate disks (both EIDE and SCSI). See why I prefer scsi. Here it is the URL: http://www.seagate.com/cda/products/discsales/index Praise
On Thu, 2002-01-03 at 17:53, Praise wrote:
Just for information, look at seek time of seagate disks (both EIDE and SCSI).
I'm not saying that it's impossible. Looking at seek times vs. spindle speed, it's a practically linear relationship, though. The older SCSI disks with lower spindle speeds are right inline with the ATA disks, speaking for my hardware similarity data. We've already discussed the fact that high-spindle-speed ATA disks simply aren't available. Be it the fact that engineers can't package it or that salesmen can't market it, they simply don't exist for consumers.
See why I prefer scsi. Here it is the URL:
My stomache prefers being full. :) -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
On Friday 04 January 2002 01.16, Jon Pennington wrote:
I'm not saying that it's impossible. Looking at seek times vs. spindle speed, it's a practically linear relationship, though. The older SCSI disks with lower spindle speeds are right inline with the ATA disks,
I thought disk size played in on the seek times, or is that wrong (I'm no hardware wiz, to say the least). The 181 GB SCSI drive at 7200 rpm has a lower seek time than the 10 GB IDE drive at 7200 rpm. And at about $4/GB lower price. //Anders
On Thu, 2002-01-03 at 18:38, Anders Johansson wrote:
I thought disk size played in on the seek times, or is that wrong (I'm no hardware wiz, to say the least).
Some of each. The higher the areal density, the (statistically) quicker the seek time is going to be; the same for spindle speed. If you crank up the RPM fast enough, and cram enough data per square mm of platter space, you could theoretically have a single side of a single platter with <0.1ms of seek time. We don't have that kind of technology yet, though. :) Conversely, the faster the platter is spinning, the less likely (statistically) the head is to actually hit the track it was looking for on the first try.
The 181 GB SCSI drive at 7200 rpm has a lower seek time than the 10 GB IDE drive at 7200 rpm.
Probably due in part to data density per platter. Seek time refers to how long it takes for the head to go from "idle" to "reading" mode, including *seeking* to the location on the platter. The 10GB disk is probably 1.5 sides; about 3.3GB per side on two platters. The 180GB is how many platters? Then divide the storage by the number of used sides on those platters; the areal densitiy of the platter is very different. Also much more expensive per unit, but...
And at about $4/GB lower price.
Interesting point. I personally don't need 181GB of disk space at the moment, but that's a good example of when it *is* economical to buy SCSI. -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
I'm hoping to make the switch to SCSI as well for several reasons. I _hate_ having to deal with master / slave combinations. I hate having to deal with 7 different (U)DMA modes. I'd hate having to have my CDRW and hard drive use separate buses. SCSI, IMO, is just _simpler_. Of course, it's also more expensive, but prices are dropping. SCSI drives that cost 600USD just a year and half ago are now in the 100USD-200USD. Speaking of real buys, how are these hard drives? Fujitsu 36.4GB 10,000 RPM, 3.5" Wide Ultra2 LVD SCA SCSI-2 Hard Drive : $149.99 [harddriveoutlet.com] IBM:07N3140, 18GB ULTRA 160 SCSI LP ULTRASTAR7200 RPM 36LP 80PIN 4MB : $107.95 [axiontech.com] -- Karol Pietrzak PGP KeyID: 3A1446A0
I really like my 29160 with <4> Vendor: QUANTUM Model: ATLAS 10K 18WLS Rev: UCJP <4> Type: Direct-Access ANSI SCSI revision: 03 <4>(scsi0:A:3): 40.000MB/s transfers (20.000MHz, offset 31, 16bit) <4> Vendor: PLEXTOR Model: CD-R PX-R820T Rev: 1.08 <4> Type: CD-ROM ANSI SCSI revision: 02 <4>(scsi0:A:4): 10.000MB/s transfers (10.000MHz, offset 8) <4> Vendor: PLEXTOR Model: CD-ROM PX-40TW Rev: 1.01 <4> Type: CD-ROM ANSI SCSI revision: 02 <4>(scsi0:A:8): 40.000MB/s transfers (20.000MHz, offset 15, 16bit) <4> Vendor: QUANTUM Model: ATLAS IV 18 WLS Rev: 0909 <4> Type: Direct-Access ANSI SCSI revision: 03 <4>(scsi0:A:14): 40.000MB/s transfers (20.000MHz, offset 31, 16bit) <4>hdb: DVD-ROM BDV212B, ATAPI CD/DVD-ROM drive <4>ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 I have all 3 CD drives on different busses. Very nice :-) -- W.D.McKinney (Dee) (907)349-4308 (Office) (907)349-2226 (Fax) http://3519098920
Here's an interesting article somewhat related to SCSI throughput on a VIA chipset. http://www.tecchannel.de/hardware/817/index.html Chris Geske -----Original Message----- From: W.D.McKinney [mailto:deem@wdm.com] Sent: Thursday, January 03, 2002 9:01 PM To: suse-linux-e@suse.com Subject: Re: [SLE] Changing to SCSI I really like my 29160 with <4> Vendor: QUANTUM Model: ATLAS 10K 18WLS Rev: UCJP <4> Type: Direct-Access ANSI SCSI revision: 03 <4>(scsi0:A:3): 40.000MB/s transfers (20.000MHz, offset 31, 16bit) <4> Vendor: PLEXTOR Model: CD-R PX-R820T Rev: 1.08 <4> Type: CD-ROM ANSI SCSI revision: 02 <4>(scsi0:A:4): 10.000MB/s transfers (10.000MHz, offset 8) <4> Vendor: PLEXTOR Model: CD-ROM PX-40TW Rev: 1.01 <4> Type: CD-ROM ANSI SCSI revision: 02 <4>(scsi0:A:8): 40.000MB/s transfers (20.000MHz, offset 15, 16bit) <4> Vendor: QUANTUM Model: ATLAS IV 18 WLS Rev: 0909 <4> Type: Direct-Access ANSI SCSI revision: 03 <4>(scsi0:A:14): 40.000MB/s transfers (20.000MHz, offset 31, 16bit) <4>hdb: DVD-ROM BDV212B, ATAPI CD/DVD-ROM drive <4>ide0 at 0x1f0-0x1f7,0x3f6 on irq 14 I have all 3 CD drives on different busses. Very nice :-) -- W.D.McKinney (Dee) (907)349-4308 (Office) (907)349-2226 (Fax) http://3519098920 -- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
On Fri, 2002-01-04 at 09:12, Chris Geske wrote:
Thanks for the article. We've known for a very long time that VIA chipsets were not acceptable if you needed low PCI latency, but now that it's clear that their throughput sucks, it may be time for VIA to change their ways... :) -- -=|JP|=- Need a good geek? I'm unemployed! '01 B15 SE/PP | http://www.xanga.com/cowboydren/ | < '95 SL2 Auto | cowboydren @ yahoo . com | _________________________________________________________ Do You Yahoo!? Get your free @yahoo.com address at http://mail.yahoo.com
On Thu, Jan 03, 2002 at 09:40:54PM -0500, Karol Pietrzak wrote:
I'm hoping to make the switch to SCSI as well for several reasons. I _hate_ having to deal with master / slave combinations. I hate having to deal with 7 different (U)DMA modes. I'd hate having to have my CDRW and hard drive use separate buses.
SCSI, IMO, is just _simpler_. Of course, it's also more expensive, but prices are dropping. SCSI drives that cost 600USD just a year and half ago are now in the 100USD-200USD.
I agree here . . .
Speaking of real buys, how are these hard drives?
Fujitsu 36.4GB 10,000 RPM, 3.5" Wide Ultra2 LVD SCA SCSI-2 Hard Drive : $149.99 [harddriveoutlet.com]
IBM:07N3140, 18GB ULTRA 160 SCSI LP ULTRASTAR7200 RPM 36LP 80PIN 4MB : $107.95 [axiontech.com]
I have one of these and it is a good, solid, drive. Not the fastest, but good for the price. I bought two drives, one IBM (like the above) and a 10k RPM quantum. I use the fast drive for the main partitions (/, /usr, etc.) and the slower IBM for archival use. Note that you will probably need a 80pin to 68pin conversion, as the 80pin drive is for hot-swap bays, they are very cheep. Here is the output of hdparm for you on the IBM drive: nickw@gizmo:~ > sudo /sbin/hdparm -tT /dev/sdb Password: /dev/sdb: Timing buffer-cache reads: 128 MB in 0.87 seconds =147.13 MB/sec Timing buffered disk reads: 64 MB in 2.13 seconds = 30.05 MB/sec nickw@gizmo:~ >
-- Karol Pietrzak PGP KeyID: 3A1446A0
-- To unsubscribe send e-mail to suse-linux-e-unsubscribe@suse.com For additional commands send e-mail to suse-linux-e-help@suse.com Also check the FAQ at http://www.suse.com/support/faq and the archives at http://lists.suse.com
-- Nick Webb http://www.uidaho.edu/~nickw/
participants (13)
-
Anders Johansson -
Chris Geske -
Derek Fountain -
Jaakko Tamminen -
Jon Pennington -
Karol Pietrzak -
Michael Nelson -
Nick Webb -
Patrick -
Praise -
Robert S. McMillan -
Uzo Kemdi Anyamele -
W.D.McKinney