Fragged/Defragged benches don't make sense...
So, after reading Guyute's thread on defragging, I figured I'd test something. The results were bizarre. Below is what my harddrive looked like before a defrag (I think the crap on the right is mostly music and such) and an ATTO bench. The bench, by the way, was the best I've EVER seen with this drive. In my second post (2 attachment limit kinda sucks) is after a defrag and a bench. Those scores are approximately average.
0
Comments
EDIT
BTW, where do you get that program?
maybe tex knows what's going on with that ...it's just illogical.
Is that a RAID array or what?
This is my drive partitioned to C:\ and D:\.
//edit: Are reads usually that much slower than writes?
http://www.attotech.com/software/files/eptscsi.exe
While I'm here I guess I'll post mine too hehe. Sorry I don't wanna crap up your thread.
ATTO is maybe the worst disk test to reflect defragging as its so very small. I didn't spend any time looking at the differant tests after I saw the purpose of the thread so PM or email me if you want further input
Tex
http://www.short-media.com/download.php?dc=54
EDIT by Admin
Download fixed.
Gigabyte Socket 940 (K8NNSXP-940)
FX-53 Processor
after defragging
NOTE: middle of the drive (it's partitioned) tests slower than C: at beginning...about a 7k difference.
Does that small read speed rate increase make that big a difference?
EDIT
Actually I'm sure there are better benchmarks out there that could better show the advanteges... Nevermind me!
Trash away...lol.
The problem is that SATA isn't really SATA. It's most commonly a PATA drive with a SATA connector. Should be called SPATA.
SATA indeed isn't faster...it's just a more convenient connection space wise at the moment. I remember my 40 GB drives PATA Maxtor usually clocked in around 42-44 K peak...but those were with 2 MB cache.
Tex has a whole bunch of tweaks that I haven't applied or can remember but remember this is NON raid.
But is this more wow for you? 2 x 10k 36 GB SCSI drives in RAID 0 on a 1600 Elite MegaRaid controller with 2 x 2800+ MP processors.
It's a damn shame the Socket 940 board doesn't have 64-bit wide PCI slot....
I have to admit that I do like the "feel" of RAID 0 drives over single. Dual Processors helps too but in the dragstrip of benchmarks...th FX-53 will beat my dual MPs.
So I have both PCs running on a KVM and switch back and forth so neither get's lonely.
Great now I forgot what I came in here wanting to say.
I'll defraggle my motherdisk and post back tomorrow with new scores...
nf7-sv2
barton@2.4g/400 1.825v
my sys temp went up 2 degrees while running this!
Tex
Linkage to main page here.
I couldn't get it to run in the 2 minutes I played with it but StorageReview says it's much better at assessing hard drive speed.
Whatya think Tex?
PS: I think you should post an ATTO of your 200+k drives and put everyone to shame.
There are good and bad points about every disk benchmark. ATTO is a quick and dirty look but only shows str. Which isnt the best indication for most desktop users of the performance as a whole. It doesnt begin to take into consideration disk latency and access times. The old winbench ones are still as relevant as anything we have IMO because they tested the whole disk subsystem including the windows cache. Which is also why I still have uses for Sandra though many poopoo on it. Most of their bitchs are because they don't understand what it's doing. Sandra uses a test size 1.5 times the size of your ram so varying the amount of ram you have in the system varies the test size and can alter the results. they also start at the very beginning of the disk and use every teeny weeny bit of free space which is why defragging for sandra is key. Even using DIFFERANT defraggers varies the sandra score if its a disk in use because they vary how much free disk space they leave at the front of the disk and thats where I-O is fastest. Thats why running sandra often tells you more about your real windows disk speed because it really shows your disk getting fragmented even hours after your original test if its a system disk and you can also test both with and without the windows cache. Most raid-0 users are aghast when they run sandra with and without the cache as they bench higher WITHOUT the cache. Which means their disk subsystem needs tweaked. In real life when you get away from benchmarks we always HAVE to use the windows disk cache. So running sandra without the cache is just a baseline number of what the disk/controller can do and you want the windows cache score to be at least 50 percent HIGHER. Or your just shooting yourself in the foot and paying to much attention to benchamrks at the expense of real life performance.
Which is again why you have to take EVERY disk benchmark with a grain of salt IF you tune a disk subsystem as a whole. Because no benchmark can really show a score for A DISK SUBSYSTEM, as they compare single disks, so if you have more then one disk and its not in raid then no benchmark shows the results. To tune a disk subsystem you want to balance the I-O across multiple drives. And no disk bench shows the advantages of doing that. Which is why it still takes someone that knows what a disk subsytem including the windows cache is doing to really wring full benefits out of it. Its really not as hard as it sounds to get good performance though from multiple drives.
Tex
Here you go. Here is some benchs to play with. Look close at the average access time on the sandra bench. 2ms. You will never realize the differance that makes in real life compared to your ide/sata 6 to 9ms average access. few ide/sata controllers can shorten the access time like that no matter how many drives you had.
Oh yeah... That ATTO ? Thats one scsi drive not a array. Which shows a weakness in the atto test on high end controlelrs as we obviously are not testing the disk now but rather the cpu/cache on the controller as the entire test fits in the 512mb ddr cache on the controller right? So if I wanted to play games with the sandra bench mark I would sticka small amount of system ram in forcing sandra to use a test size smaller then the 512mb cache on the controller. Benchmarks can be tweaked every which way if thats your intent. You need to test real systems loaded just like you use them evey day. With all the AV software and MBM and anything else you run in real life running. My lower right hand tray in Windows was jammed with a dozen running process's for both those benchmarks and I had a GB or ram in the box. Real life performance. thats what we are searching for
Everyone else: I told you Tex would post something astronomical!
but thats a awesome raptor score. The 2nd gen raptors are a good deal faster. My nice scsi's only hit 5,000 higher for atto. Nothing on that atto not to be tickled pink about. Congrats!
tex
And you're definitely right (once again) with using different test sizes (for ATTO, even if it isn't entirely accurate). I used 4 mb and got around 130,000 on my new harddrive and using 32 got around 60. Now my question is this: Is my new harddrive (Samsumg 1213c i think...) faster than the one I'm currently on? I know it was benched totally empty, and my os drive was benched 80% full, but is there anyway I can know if it's worth trying to move everything over to it, in hopes of better performance (it's also quieter)? Sandra gave the new drive a 50 MB/s and ATTO gave it this:
Thanks Tex ...I realized the performance and quietness of this hd and decided not to go raid this time ...I just don't need it.
Besides ...I honestly think that the raid array I originally had that I was booting too met its demise because I was in fact booting to it. Those quantums should have lasted way longer than only a year or two. I know I know I should have listened and done the soft array. :banghead:
Anyway I am curious to see scores on RAID-0 with dual 2nd gen raptors. I think spinner and maybe leo have that setup. Just curious.
Tex