View Full Version : **Vio1's SCSI Benchmarks and Observations**
Finally got my new SCSI setup going!
Background info:
A few months back I sold my 36gb raptors I had in raid-0 and bought myself a 74gb raptor. A few weeks ago I sold that raptor and went SCSI for the first time.
SCSI Drive:
Fujitsu MAS3735 73GB 15000rpm hdd (one of, if not THE fastest SCSI drive available)
http://www.phillylan.com/vio1/photos/November%2022%20001.jpg
SCSI Controller:
LSI u160
http://www.phillylan.com/vio1/photos/November%2022%20002.jpg
Here are some benchmark results:
HDtach 2.7:
http://www.phillylan.com/vio1/photos/HDTach.jpg
PCmark2004:
http://www.phillylan.com/vio1/photos/PCMark04.jpg
SiSoftsandra:
http://www.phillylan.com/vio1/photos/SiSoftware%20Sandra.jpg
Observations:
Computer feels way more "zippy" then even the 74gb raptor. Everything pops up real quick. I love it.
Questions:
Is this normal?
http://www.phillylan.com/vio1/photos/Bios.jpg
-Everytime I boot up the scsi controller scans the computer for devices... any way of having it not scan for new devices and just recognize the drive?
-My computer doesnt boot up faster, actually it takes longer then it did on the raptor... it takes 7 bars before windows opens. Ive noticed, since this harddrive has a led on it (showing harddrive activity) that it doesnt start to load up windows until the 4th bar has passed in windows load up.... before the 4th bar the harddrive is just sitting there idling. Any way of speeding up so that it starts reading the hdd faster?
-Anything I can do/should do to increase SCSI performance on a desktop computer? Looking for somekind of software or something.
Ad Rock
11-25-04, 09:00 PM
So have you noticed a "seat of your pants" difference? It would be interesting to see what your old 74 gb raptor would do on these same benches. Maybe you have some saved?
Congrats on the sweet SCSI set-up man :D.
So have you noticed a "seat of your pants" difference? It would be interesting to see what your old 74 gb raptor would do on these same benches. Maybe you have some saved?
Congrats on the sweet SCSI set-up man :D.
There isnt a HUGE difference from the raptors, but there is something. Something noticable.
As to benchmarks of my older drives... i dont have screenshots, but I do have some benchmark results:
74gb raptor:
HD Tach 2.70............7.6ms seek time
max 73.4mb/s
min 38.1mb/s
avg 65.3mb/s
36gb raptor:
HD Tach 2.70............8.5ms seek time
max 60mb/s
min 36.4mb/s
avg 50.1mb/s
If you have some benchmark program you want me to try out, let me know.
Daewood
11-25-04, 09:09 PM
nice vio but i think that i will stick with buying the 74 gig raptor...
theELVISCERATOR
11-26-04, 02:43 AM
and still slower than two 36s raided but access is quicker....
burst speed less than half
man_utd
11-26-04, 09:35 AM
vio: I notice the same forever load time, when it is looking for stuff. I haven't really messed around with the controller, but have you pressed ctrl+c before it loads all the stuff up?
Ya i did, and I dont understand the options in there.
theELVISCERATOR: 172.5mb burst speed is impossible ... the most your computer could do is 150... or is it 120mb? I dont remember.
Anyone know how to set these jumpers on the harddrive for optimal performance (see first pic up top for jumper pin diagram)?
http://www.phillylan.com/vio1/photos/Jumpers%20001.jpg
man_utd
11-26-04, 10:02 AM
86.25mb/s per drive does seem a bit high...
Also vio, it looks like the maxtor atlas 15k II is the new performance leader, but it doesn't seem like there is much of a difference between it and the rest of the competition, if any at all.
man_utd
11-26-04, 10:04 AM
As far as I can tell (I'm a noob to scsi as well), they are in all the right places
i havent seen any reviews of the maxtor atlas 2... where did you find one? From storagereviews.com they tested the new Fujitsu and its a bit faster then my drive... but I doubt it would be even noticable.
86.25mb/s maybe is not, but since the motherboard is the bottleneck there is no way that you could have it run at 2x 86.25mb/s
Vio,
nice setup. I have been running scsi on my computer for a few years and it is definatley fast as hell, it is the only thing keeping my old 2.8 533 fsb seem fast. ONe thing that i have noticed is that Benchmarks do not give SCSI justice. A raptor or even IDE will almost always beat my setup when it comes to benchmarks, but when it comes to realworld use my scsi's in raid-0 smoke them.
another thing, that screen on bootup will always be there, you get used to it, many SATA and IDE raid controlers have that as well. As far as bootup time, yours seems a bit to slow. when i first redo my computer before i get it on the network the little bar will not even make it across once, the fastest i seen was less then half. After i get the network on it takes about 3 bars across. I am thinking that much of the delay in bootup has to do with a network. I have done some tests when it comes to strip size for my setup similar to what you did with your raptors, except i could only do it for 2 strip sizes. I'll post the results file when i get home for your viewing pleasure.
edit, one thing i would recommend is putting a HD cooler on that drive. They tend to get hot.
theELVISCERATOR
11-26-04, 05:31 PM
lol @ impossible....are you saying my screenshot is a forgery?
hmm sata 150x2 should do 300 burst right?
BTW heres Hd tach 3.0 shows a 194.4 burst....
pity...
take it easy dude... im not saying you forged it... i had something like that too when i raid-0 my 36gb raptors... All im saying is that realistically, your pci bus is bottlenecking your harddrives, and therefore a speed of 190+mb /s is impossible.... Im saying that HDtach isnt correct.... its value is higher then your motherboard can support.
99'erForever
11-26-04, 07:09 PM
I have a couple of Seagate 15ks in raid 0. I'm not much of a benchmark man but they do seem awful quick. A windows installation take no time at all. Scsi drives are like an old muscle car. They are loud, hot, and not always the fastest thing on the block but when you fire them up that roar makes you feel like a real man.
Vio1: theELVISCERATOR is running from the ICH5R southbridge, if it's on the system in his sig. 200MB/s bursts are easily possible, since the ICH5R is not PCI 2.0 linked. It does not share bandwidth with the PCI bus. Also, these are burst speed readings that we are referring to. They consist of bursts of data from cache on the drive and are often much higher than disk throughput can be.
Vio1: theELVISCERATOR is running from the ICH5R southbridge, if it's on the system in his sig. 200MB/s bursts are easily possible, since the ICH5R is not PCI 2.0 linked. It does not share bandwidth with the PCI bus. Also, these are burst speed readings that we are referring to. They consist of bursts of data from cache on the drive and are often much higher than disk throughput can be.
ah. okay. Stand corrected.
theELVISCERATOR
11-26-04, 10:57 PM
Vio1: theELVISCERATOR is running from the ICH5R southbridge, if it's on the system in his sig. 200MB/s bursts are easily possible, since the ICH5R is not PCI 2.0 linked. It does not share bandwidth with the PCI bus. Also, these are burst speed readings that we are referring to. They consist of bursts of data from cache on the drive and are often much higher than disk throughput can be.
This is why vios results with raid have been less successful, he was limited by pci bus speeds.
I take it the long initializing of the scsi card on bootup is not to your liking?
i was hoping that boot up would be even faster... but im content with the operational speed in windows. Its darn fast....
DEZMOND
11-27-04, 02:34 AM
Nice benches, I'm kinda new to scsi myself. I'm running a lsi21320r. My load times also seem a bit long,about 6 bars across, but I'm also running 3 controllers.I'm running 2 internal drives and one external with a ide-scsi adapter. So far I really like it.
man_utd
11-27-04, 12:33 PM
Your new to it and running 3 controllers? When you get your feet wet, does that mean you dive into lakes?
DEZMOND
11-27-04, 01:17 PM
Your new to it and running 3 controllers? When you get your feet wet, does that mean you dive into lakes?
Actually only the one is a scsi controller,the other 2 are ide.I love the high speed external,feels like a regular drive in both benches and real world use.
man_utd
11-28-04, 07:31 PM
Ah, still very impressive though. Hopefully I will get my drive in on tuesday and make my forray into scsi.
i stuck my 15k fugitsu in a vantec vortex, totally quiet with the rubber mounts plus the digital readout attached to drive ~~~ i wish they would have designed it to have the fan turn on with the system...duh ~~~ these 15k drives need some type of cooling.... forget to turn the fan on, then click and watch it cool down 10C in a minute
mounted mine in the top case slot across the intake psu fan, keeping that hot air out of the mix of the rest of the case airflow
man_utd
11-28-04, 09:04 PM
Probably will keep mine down next to normal hdd, and let it burn up, might just reverse air flow to pull, not push.
donny_paycheck
11-29-04, 01:56 PM
Very nice setup. Congratulations on finally getting it up and running.
As you know, I have the same thing with the exception of my drive being the Maxtor Atlas 15k, which is just a hair faster than the MAS in a few benchmarks, but about the same in general.
The LSI controller has two virtual controllers that you can see when it scans devices during boot up. You can disable one of these and use only the one with the drive on it if you want, but that doesn't do much for boot time. You can do it by entering the SCSI BIOS setup and messing around with the menus.
As has been said, bootup time is determined largely by how much crap the system has to load/process before you get to the desktop. Network interfaces, antivirus software, and other services will all affect your bootup time drastically.
EDIT: For the drive jumpers, don't bother. They set stuff like on-drive termination for the SCSI bus, spin-up delay, and things that you don't need to change, and don't affect performance unless they're set wrong and keep the drive from working at all.
man_utd
11-29-04, 03:51 PM
Ah, thanks for the clearing up. I will probably notice a speed up in boot up, as I am do for a format soon anyway, just got way too much junk on this drive. (I have 4gb left on a 200gb drive)
Here are some benchmarks that i did with some SCSI drives in Raid0 and also some IDE drives in both Raid and Single drive setup. The time is in seconds, I did all the tests from a fresh install, no drive images. I did this for real world data, every time I reinstalled I did sp2 most recent drivers and i even did the tests in the same order. Below I have a ZIP file with all the benchmark results. If you have any questions just ask. I am posting this just to add the the SCSI topic.
All of the Benchmark files (Excel, screenshots, and benchmark saved Files) (http://www.mcd.gotdns.com/RAID0Tests.zip)
http://www.mcd.gotdns.com/IDE.jpg
http://www.mcd.gotdns.com/SCSI.jpg
DEZMOND
12-01-04, 01:37 AM
Nice work cooter :thup:
awesome... do you have any comparisons between 1 single scsi drive versus raid-0?
I currently do not have any comparison on the single SCSI drive verse raid0. I originally had 1 of hte scsi drives in my rig but for the last year it has been in raid. If i get time ill try to run all the tests in single drive configureation to compare, but it will pry not be for about a week. I was also wondering if any of you have load times of these same programs and maps that I could compare with. The maps I tested on are in the excel file. A few things to remember are that on the SCSI test the internet explorer load times are pretty inaccurate cuz i could not start and stop the stopwatch fast enough. Also remember I used adobe acrobat pro, not the reader for the tests.
Vio I am glad to hear that you have converted to SCSI.
im happy with scsi.... however boot ups and shut downs are slower then id like.
Id love to see a comparison of 1 scsi versus 2 in raid-0... cause im thinking of getting a second fujitsu mas3735.
Please do try it out. Im dieing to see the difference.
I'll post later some of my times (using the same tests you did...) however they will be very different cause we have different systems.
man_utd
12-01-04, 06:48 PM
Well, I got my drive in today. I ended up buying a fujitsu MAM (thought it was a MAS, late at night) But I am not complaining. Haven't had a chance to mess around with it yet, as I just haven't had a chance to work on it.
eduncan911
12-01-04, 10:16 PM
and still slower than two 36s raided but access is quicker....
burst speed less than halfNot slower then my 36 gigabyters:
http://www.ocforums.com/showthread.php?t=346936
theELVISCERATOR: 172.5mb burst speed is impossible ... the most your computer could do is 150... or is it 120mb? I dont remember.Correct if using the standard 32bit PCI bus we all know and love (it's around 133MB/s, and I got 129MB in link above).
Incorrect if using 64bit PCI buses (i.e., my previous Ultra320 mobo I had), it goes far beyond that. Servers I've setup with RAID0 spanned across 4 drives in Ultra320 is blazingly fast!
My previous motherboard had dual Ultra320 scsi controllers (dual P4 Xeon mobo). Even though I was using only Ultra160 drives, I wanted more. ;)
Since you're new to scsi, the rule of thumb is this:
SCSI/SCSI2: 20MB/s, SCSI2 allowed bursting to 40MB/s on the bus
U/W: 40MB/s substained, possible burst to 80MB/s (I've personally seen 148MB/s on these, back in 1998!)
Ultra160: 80MB/s substained, possible burst to 160MB/s (I saw ~140MB/s bursting on my previous 7,200rpm drive)
Ultra320: 160MB/s substained, possible burst to 320MB/s
And they are coming out with Serial SCSI later next year, or might be out now for Ultra640 or something like that (haven't kept up with the news lately).
And then there's Optical drives... Do I need to keep going up in burst speeds that are possible? ;)
Looking at your speeds, I'd say yours is rated right about normal. SCSI really shines on RAID in PCI64 bit slots.
Across two channels (each PCI 64bit path to the process/memory is individual, to allow max through-put PER slot!), I've setup several RAID 0, and 0/1 combinations. Putting a drive on one channel, and a 2nd drive on another channel splits the data packet across two true 64 bit paths to each controller. I've seen throughput in Unix across my servers using 4 Ultra320 drives (two on each controller, two controllers) around 420MB/s if I can recall. It was insane! Instant write and read.
Btw, if you went through the expense of Utlra160 why didn't you just get an Utlra320 card? Humm, now that I think about it I don't think they make Utlra320 cards for standard PCI 32bit slots, only 64bit slots.
Should have gotten a controller with write-behind cache. ;) Mucho faster!
After a decade of only using SCSI drives at home, and now I wanted to get into the overclocking world, I decided to dump SCSI and just go SATA. Yes, I can notice a little difference (especially the CPU usage now). But for something less then 1/2 of the cost for RAID0 I'll for these days. Now that SATA RAID is faster then U/W that I was on for many of years... And swore if things are faster then U/W, I'll switch. :)
So, I went from SCSI after 8 to 10 years of use to SATA. Hey, my pocket book was very happy with this upgrade. ;)
eduncan911
12-01-04, 10:31 PM
Since you went through the expense of an add-on card, and HD, let's compare our setups. This is a great thread for comparision!!! I think we have near identicle performance, at almost the same cost. SATA cables are much easier to route and mucho cheaper.
Your setup:
32bit PCI SCSI Controller card: [unknown] most likely around $75
Fujisu 73GB 15,000rpm HD: $343
Ultra160 "round" cable: ~$30 (given, the card may have a cheap flat cable with it)
Total: ~$450 (+ s&h)
SATA "True" RAID (my setup):
Promise SX4 SATA Raid Controller card w/Cache: $130
Samsung PC166 256MB ECC: $22 :D (you can use any old PC100 SDRAM laying around)
2xWD 36GB Raptors: $100ea (oem)
Total: ~350 (+s&h)
Now that I total things up, I'm liking my setup more and more. :)
Not sure if it makes any dif but here's mine.
2 74gb rappys in raid0
Asus P5GDC-v
Intel onboard 82801FR
Burst Speed= 226.6 MB/s
Random Access= 8.0 ms
CPU utilization= 4%
Avg Read= 119.8 MB/s
My scsi drive is a 68pin scsi. If i wanted to raid-0 it, could i use another drive like it, but an 80 pin version instead?
Yes, you'd need a SCA adapter to convert the 80 pin SCA drive to 68 pin. For single to a couple of drives, they're not bad. Adding more than 3 adapters to a channel is usually bad news. Capacitative impedance can cause signalling issues.
I seem to remember hearing that that controller may support RAID, but it may require some research and modification. I never tried when I had one, due to having SCSI HW RAID controllers with BBUs and cache memory. If not, then you'd need a RAID controller. The LSI 21320-R retail kit is one of the less expensive cards, around 130 on newegg. It's PCI-X, but backwards compatbile to PCI 2.0. Unless you are getting a planar with at least PCI 64/66, I probably wouldn't recommend the expense. The PCI bandwidth is a wall that gets hit by 2 drives in RAID 0 very easily. You coud easily add another drive on the existing controller and install apps and other frequently accessed files there. That would allow more efficient use of the controller, since there is very little degradation in single drive access rates when accessing multiple drives on a channel.
eduncan911
12-02-04, 09:27 AM
I've used these "converters" in the past to go from 80 to 64 and from 64 to 80. I have yet to find one that let me use Ultra160 (they all registered the drive as 80 drives, not 160 as you see on start up). I evetually broke down and bought an 80pin card for the 80 pin drives.
There is no speed difference between 64 and 80, just stick to what you got. The advantage of 80pin is the built-in power on the same connector, allowing you to hot-swap the 80pin drive in a drive array.
Not needed for home.
But you asked if you could use an 80pin version for RAID0, the answer is yes... But just like any other RAID, you'll want exactly matched drive specs (seek, throughput, cache, etc). Else it could (and will) cause more damage then good in adding delays in accessing data.
man_utd
12-02-04, 03:44 PM
Small correction: 68 pin, not 64, though on all else, looks right.
eduncan911
12-02-04, 04:28 PM
Bleh. 68. lol 64bit... lol
Ok,
Vio i know that you requested this and for anyone else that is interested I have done all the tests above on a single SCSI drive. The results are in that original post. One thing to keep in mind is that I have 2 different SCSI drives in my raid array and I used the faster of the 2. they are virtually identical except one is U320 and one is U160. i may do a test with the other drive in single drive configureation as that will pry be better to compare against the rest of my raid results but I want to get my computer up and running again. I listed the model number of the 2 drives as well in the results.
cooter
there doesnt seem to be any improvement with having raid-0.... does it feel slower at all with 1 drive versus 2?
One of the big differences is boot time. Raid makes the computer boot much faster. As far as the rest of the speed I dont think it makes alot of difference otehrwise. I am mainly using it now for the increased HD space, 1 - 36gb hard drive is not enough for me for my OS and apps and games. My results may have been different if i would have checked the slower drive but the difference in speed between the two is pry very minimal.
how many bars does it take to load up windows with 1 scsi? with 2 scsi?
With one SCSI it was taking about 5 or 6 bars, with 2 SCSI in RAID0 it takes anywhere from 1.5 - 2 bars.
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.