• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

SCSI RAID results: Some thoughts on them and other musings *UPDATED w/New BMarks!*

MichaelD

MichaelD

Lifer
15 March update:

At Shuttleteam's suggestion, I changed the Cache Flush setting to 10 seconds, and gained significant improvements!


New ATTO.

Here's Drivebench...I didn't have this one before.


Drivebench results.

COSBI scores still suck though...I dunno, I believe ATTO and Drivebench, personally.

Minimum creation time (s), 39.5112
Maximum file creation rate (MB/s), 25.309
Average creation time (s), 39.9364
Average file creation rate (MB/s), 25.043
Minimum copy time (s), 90.0057
Maximum copy rate (MB/s), 11.110
Average copy time (s), 90.5404
Average copy rate (MB/s), 11.045

******** SCORES ********
File creation score, 693
File copy score, 1593
COSBI score, 1143

I'm pretty happy with the new results. :beer:🙂:beer:

Two-fisted, even!!!






____________________________________________________________________________________________________________-
Hi guys,

Well, I finally got the SCSI RAID going. There's definitely a learning curve involved with a real SCSI HBA (Megaraid Elite 1600...settings are listed in the pics) as the cards' bios as almost as extensive as my motherboards' bios! I must say that the depth and breadth of the bios/menus/utilities just blows away anything available for IDE.

I have a 3Ware7000-2 IDE RAID card; it was what I was running before I purchased the SCSI stuff. It's very nice but just doesn't compare. Two drives is all you can put on the 3Ware; I can do a 10-drive stripe on the Megaraid. (Oh..can I borrow about $2K to purchase the 10 drives? 😉)

Big thanks to Shuttleteam for holding my hand thru the process. If not for his advice and patience, I'd still be lost in the bios somewhere, randomly enabling things and wondering "why the hell isn't this stupid thing working?"
rolleye.gif


First off, some good old ATTO results.

ATTO results for RAID 0

The write speed isn't that great, but I'm very pleased with the read speeds!

Here's what a single MAS3184 did on a LSIU160 controller.

ATTO results for single MAS3184

Note the excellent write speeds...no comparison on the reads, though. 😉

I do not understand why the write speeds went down so much; you would think with a two-drive stripe happening, on a real controller that they would be out of this world. *shrug*

I do believe it has something to do with the fact I'm not using the controller's full capabilities by having it running in a 32-bit, 33MHz slot instead of the 64-bit, 66MHz slot it would normally be in.

For comparison's sake, here's a single WD-SE 120GB IDE drive attached right to the motherboard. (Epox 8RDA+)

ATTO results for single WDSE 120GB.

Here's the results with the latest big thing, the COSBI benchmark. I'm not at all impressed with my scores, but hey, I'm honest and will throw them up anyway. MY IDE RAID array did much better than this (almost 2,000 points) but hey, benchmarks only tell half the story.

Minimum creation time (s), 39.8424
Maximum file creation rate (MB/s), 25.099
Average creation time (s), 40.0252
Average file creation rate (MB/s), 24.985
Minimum copy time (s), 91.4863
Maximum copy rate (MB/s), 10.931
Average copy time (s), 92.1574
Average copy rate (MB/s), 10.851

******** SCORES ********
File creation score, 691
File copy score, 1565
COSBI score, 1128
************************



My quest for speed may just push me to purchase
this motherboard for my next build instead of an A64 rig simply b/c of the 64-bit PCI slots. That's a big maybe though...we'll see. Now, if I can find a modern 64-bit, 8xAGP equipped mobo that takes dual P4's (NOT Xeons!) then I'm all over it...as long as it's not like, $1K in cost!

It is very important to note that the benchmarks most definitely do not tell the whole story about SCSI! Before I started using SCSI, I was ignorant of it's benefits...I was strictly a benchmark kind of guy. Wrong.

Everything involving the disk subsystem is faster, way faster than IDE.

For example, I created a 10GB partition on the array and installed W2K on it.

The ENTIRE process from right after the intial format where it starts copying files THRU the SECOND reboot (after it does the "detecting devices" thing) was 14 minutes and change. It took LONGER to "detect devices" than it did to actually copy all the files from the CD!

SP4 installed in under a minute. No, I still haven't figured out how to Slipstream a SP into the CD. 😱 Bite me, OK? 😛

Big games like UT2003 that would take five minutes per CD to install now install in half the time. Things like that.

I can do anything I want while NAV runs in the background...almost no detectable load-time difference. 🙂

Well, that's the story. I figured with as many "SCSI threads" as I started, I'd owe you guys a review of sorts.

If there's anything you want me to run for giggles, LMK and I'll do it. 🙂
 
Interesting results, did you try having write cache disable ? Only because on a server if the battery on the controller dies & the cache gets corrupt then you loose your array.. :0

I noticed on my customers systems that when I used a small stripe I got good overall system response but horrible response when reconfiguring their database's .

I also thought of moving to a scsi raid controller, but figured that the extra overhead + having only a 32bit pci slot precluded any benefits I might see. I'm sure if I did any video editing on a professional level I'd really go w/ a raid controller. I really hope more 64bit mobo's become available like the one you listed. I don't think eve pci-express w/ be an improvement over current Xeons mobo's w/ 64bit 133mhz pci slots.

How is your overall system response w/ the raid controller vs the scsi card ?

Regards,
Jose
 
Thanks Adul. 🙂

Jose,

W/the write cache disabled, my write scores were MISERABLE. Down in the 20MB/s range.

My system response is about the same with the SCSi vs. IDE controller. The loads (I mainly game on this box) are slighty faster...it's noticeable, but not the same jump as say single IDE drive vs. IDE RAID array.

What is MUCH faster is the multitasking. I run a NAV scan daily and scan anything I DL from the net before opening. I am paranoid. 😱

Anyway, while scanning, I can open proggys, execute stuff w/almost no noticeable slowdown, delay or whathaveyou. That was not the story w/the IDE array.

Also, the W2K install was about 100% faster. 😀
 
Well Michael, My results are very similar with my 5 10k drives VS your 2 15k drives, except my write speeds went up to 73k, and my read was 114k. Cosbi was almost the same if I read and write to the same drive (SCSI), but to different drives (SATA read or write, SCSI other) it jumps to 3200 ! I haven;t tried multitaking yet, but it took 4 minutes to install XP, including the device detection (that doesn;t count the 3 minute first boot file copy from CD). I don;t think these banchmarks are showing up what these things are capable of yet. Time will tell !. The only thing I don;t like is the amount of heat that 5 10k drives puts out. They were running all day, and even though it only got up to 64F outside today, it was 77F inside, and I had to turn the air conditioning on to be confortable.
 
Hi Mark, my partner in SCSI Addiction. 😀

I wonder how much, if any different my scores would be w/a better mobo/CPU platform? True, you are still on a 33MHz PCI bus, but you have a much more robust CPU/memory subsystem than I do. Hmm. That Supermicro board is calling me....caaaaaalllllling meeeeeeee. 😉

Your reads are 73MB/s? :Q Dayum. That's some juice, man! 😎 If I had the change, I'd love to make a FOUR-drive stripe and check that out. The motherboard would probably explode. 😀

I agree on the benchmarks being ignorant comment. This thing is darn fast...gaming is sweet. I used to be able to go crack another beer b/t levels loading. Now I'm lucky to be able to take a good long swig before the fragging starts again. 😀
 
Originally posted by: MichaelD
Hi Mark, my partner in SCSI Addiction. 😀

I wonder how much, if any different my scores would be w/a better mobo/CPU platform? True, you are still on a 33MHz PCI bus, but you have a much more robust CPU/memory subsystem than I do. Hmm. That Supermicro board is calling me....caaaaaalllllling meeeeeeee. 😉

Your reads are 73MB/s? :Q Dayum. That's some juice, man! 😎 If I had the change, I'd love to make a FOUR-drive stripe and check that out. The motherboard would probably explode. 😀

I agree on the benchmarks being ignorant comment. This thing is darn fast...gaming is sweet. I used to be able to go crack another beer b/t levels loading. Now I'm lucky to be able to take a good long swig before the fragging starts again. 😀
Click to expand...

Very interesting results Mike, although I'm pretty sure that something configured sub-optimally, even if the RAID array did add overhead, the performance drop is a bit much.

BTW, slipstreaming Win2K or XP is really easy, just google it. There's a command-line switch you can use after downloading the service pack that does the job for you 🙂
 
Originally posted by: MichaelD
Hi Mark, my partner in SCSI Addiction. 😀

I wonder how much, if any different my scores would be w/a better mobo/CPU platform? True, you are still on a 33MHz PCI bus, but you have a much more robust CPU/memory subsystem than I do. Hmm. That Supermicro board is calling me....caaaaaalllllling meeeeeeee. 😉

Your reads are 73MB/s? :Q Dayum. That's some juice, man! 😎 If I had the change, I'd love to make a FOUR-drive stripe and check that out. The motherboard would probably explode. 😀

I agree on the benchmarks being ignorant comment. This thing is darn fast...gaming is sweet. I used to be able to go crack another beer b/t levels loading. Now I'm lucky to be able to take a good long swig before the fragging starts again. 😀
Click to expand...
No writes are 73k (due to 5 drives ??) , reads are 114k. i need to try COD, everything else runs too fast. You got a good suggestion for a gome that loads slow ? I'm not normally a fragger, but I do strategy, but C&C generals loads a little faster.....

 
I tell ya; I read multiple how-to guides on slipstreaming. I can follow directions, darnit! Well....maybe not, since the handful of times I tried, it didn't work! 😱 I dunno what went wrong.
rolleye.gif


Anyway, when it takes under a minute to load a SP, who cares, right? 😀

I have the HBA configured exactly how Shuttleteam told me to. He plays with servers that cost more than my truck on a daily basis...I definitely trust his instructions!

I honestly think it's that I'm running a Ferrari on Joe's Superstop & Go 82 Octane pump gas. 😱

I have a nice gaming rig, but it's not what this HBA should be run on. I wish I could see some benchies on that Supermicro P4SCT mobo.

Not only do a small handful of resellers have it (Newegg does not) but it's not exactly a gaming-rig board. Not an OCers board either, from what I can tell. More of a CAD/CAM, video editing Workstation board. Hell, stick a 3.2GHz P4 in there and it's a gaming rig. 😀

You know...you just can't win this Hardware Game....you just can't. You get the Super-Mega-Maxi Videocard and then you get the Fry-Your-Pants clean off CPU...but there's no 64-bit PCI slots. Then you find the 64-bit PCI-equipped mobo, and it's got the Super-Mega-Maxi 8x AGP slot...but no OCing options.

you just can't win.
 
on my 15k drive i was getting lousy writes (benchmarking)
converted the disks to dynamic and the benchmarks returned
to where they should have been...but i didn't see any increase in speed
so i quit worrying about the benchmarks.
now that i lost the drive (anybody know how to fix a fubared partition?)
the return to an ide drive has left an awful taste in my mouth.

load times that were never there are now rampant.
wait a few seconds everytime i let the system rest,surfing the web is
a pain as it constantly spins up the drive

running seti and dpad was easy before now if i am running both the lag is horrid
game loads are also horrid compared to the speed i was used to

but the thing i miss the most is the responsiveness of my system even under full loads
with scsi it just went on and on...ide seems like it has to stop and catch its breath
way too much.

and since my bellyaching isn't helping i'll just mosy back to the dark ages and tinker a bit.

weird thing is i can format the drive just when i partition it won't start stays at 0%...

mike
 
Here are benches I got with my ATTO UL3D with 4 second generation 18GB 15K last year on an Athlon MP motherboard. Bear in mind the controller is a 32bit controller, not 64bit. WinXP using dynamic disk.

Im glad to see you got yours working well now.. 🙂
 
Those ATTO controllers are host based RAID and those results will reflect pure file R/W performance only. The i/o performance will still be the same is a single drive. (like IDE raid0 vs. single drive) I've used LSi and Adaptec host based RAID HBA's and bascially get drive str n times number of drives on samples of 64kb and higher. The system does NOT feel any faster than a single drive connected to a $40 U160 HBA unfortunately.

These pure hardware controllers are a different animal altogether. These benches are not telling the story obviously. In the case with the 1600's running at 32/33, you have a six speed gearbox but you aren't getting out of third gear! :Q (Still beats the crap out of IDE...that's like being in first.)

On the thought of a PCI-X based system...

You cannot get a dual P4. There is no such animal. You will have to get either a P4 based uniprocessor system with a workstation/entry level server chipset or Xeon. The Xeon is *the* way to go. If you're hardcore AMD, the only choice is Opteron and they get rather expensive especially if you want the middle end cpu's. I haven't used them on the desktop so I have no idea how they would perform in that realm. Windows XP 64bit Edition and 8+GB of RAM sounds like a nice setup however. 🙂

Cheers!
 
MichaelD:

Your running w2ksp4 right ? Have you or anyone tried XPsp1 w/ a scsi raid controller ? I've read that XP doesn't work great w/ scsi. I'm about to re-install , and I'm thinking to switching to XP to take adavantage of HT. I have 2- 15K 36g Atlas & 2 - 10k 36g atlas w/ a Plextor 708a ide drive.

Regards,
Jose
 
mastertech01
Man, those are some kickass scores! :Q You have all four drives striped, two per channel, right? Zooom! Your write scores are outrageous too...why do mine suck so badly? Only two drives? "Neutered" controller? *sigh* I"ll get it right one of these days.

The two SCSI partitions on the array were both upgraded to Dynamic Disks right after I finished installing SP4. I've done everything I can...I think.

AFA the P4 vs. A64 upgrade goes...I would really have to think on that one. The A64 is a better overall choice, I know, but man...64-bit slots! Yummy.

Shuttleteam
In your opinion, if I stayed w/this same 2-drive stripe, what performance increase, if any, would I see moving the card to a 64-bit, 66MHz platform? With W2K, of course. Unless a free, legal copy of XP Pro falls into my lap, I'm waiting for Longhorn...SP3. 😉
 
I have tried the ATTO on XP pro and win2k. The scores on mine seem almost identical. I didn't want to bother with dynamic drives, since I have the nifty SCSI hardware controller with 128 meg cache, and 5 drives in raid0. 73k write and 114k read. I would post a link showing my score, but don't have a picture hosting site right now. (maybe I can find one) I need to find a better bench to show the results, installing is allways waiting on the DVD drive. But game loading is fast.
 
Shuttleteam
In your opinion, if I stayed w/this same 2-drive stripe, what performance increase, if any, would I see moving the card to a 64-bit, 66MHz platform? With W2K, of course. Unless a free, legal copy of XP Pro falls into my lap, I'm waiting for Longhorn...SP3.
Click to expand...

If you don't have any BIOS conflicts, the 1600 series typically will see writes in the 90MB/S range and reads in the 145MB/S range with caching enabled.

ATTO scores are much higher with the cache off especially with more disks, but the system will feel faster with the cache ON.

Cheers!

 
Mark,

Host your pics right here! It's free and it's what most ATers use. You cannot beat the price!!! It's run by guys from right here at AT. 🙂

Shuttleteam,
That's what I was afraid you were gonna say. 😉

*grumbling*
/flipping coin
*mumbling*


😀
 
Originally posted by: shuttleteam
32/33


64/66

Questions?

Cheers!
Click to expand...

:Q

You always just "happen to have a complete set of benchmarks handy." How do you do that? 😎

MichaelD: "Hmm, I wonder what the airspeed velocity of an unladen swallow is?"
Shuttleteam: "Well, I just did the research last week..." *links to PowerPoint presentation complete w/videos and CAD/CAM drawings*

Ahahaha!!
*ahem*

A moment of levity there. *straightens tie*


I notice that the first graph is using three drives and the second, four drives. How much of a diff does that make? I'm curious b/c my current results are better than the first set you posted...but dismal compared to the second.

Also, how much of a diff is the OS making here? I"m not XP-equipped...not that I want to be, to be honest.
 
You always just "happen to have a complete set of benchmarks handy." How do you do that?
Click to expand...

We have everything! 😉

MichaelD: "Hmm, I wonder what the airspeed velocity of an unladen swallow is?"
Shuttleteam: "Well, I just did the research last week..." *links to PowerPoint presentation complete w/videos and CAD/CAM drawings*

Ahahaha!!
*ahem*

A moment of levity there. *straightens tie*
Click to expand...

Don't forget the 808 nM goggles and NEVER, EVER Stare into the light! :Q

I notice that the first graph is using three drives and the second, four drives. How much of a diff does that make? I'm curious b/c my current results are better than the first set you posted...but dismal compared to the second.
Click to expand...

If I used 4 drives in the first it would not made any difference. The card *works* at 32/33 but just for compatibility reasons. It's just teasing you; giving you enough taste of its capabilities while you really want to see what it can do!

Also, how much of a diff is the OS making here? I"m not XP-equipped...not that I want to be, to be honest.
Click to expand...

Probably not much difference. XP loads faster and launches applications faster due to prefetch. This goes a long way in the IDE realm. SCSI actually hurts XP startup time (any OS for that matter) just due to the fact that the SCSI bus must be scanned. This is why you see a fairly long period of no disk activity on a SCSI based system booting up. If you do a load trace (bootvis) you will see this stand out clearly. If it could be eliminated, SCSI systems would boot up lickety split. I'm not worried about boot time as the OS is plenty stable for me. The only time I'm restarting is if I have to.

Cheers!

 
Thanks for all the info. 🙂 I know what I must do...I must find an A64 board with 64-bit slots. And then I woke up. 😉

BTW...what the hell is that picture? It's got like...eyeball thingees at the top and bottom. It almost looks like a CPU substrate....
 
Oversimplified explanation: 808 nM pump bar. It produces light in the invisible spectrum that can burn stuff at great distance. It's not a toy but then again a lot of stuff technically isn't to be played with!

Cheers!
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top