Does partitioning produces performance decrease ??

[gustavo]

I wanted to know if partitioning your HDD results in any performance decrease.

I have one Seagate Barracuda IV 40 GB 7200 RPM and made
a primary partition of 30 GB and left the rest as a kind of backup just in case i have to reinstall windows 2000.

This is very convenient anyway for my purposes but want
to know if I should expect any performance decrease.

Running Sisoft Sandra Pro benchmarks the benchs dropped
from 24500 to 21000 but as benchmarks are not fully reliable
I wanted to know if there is any theoretical reason to expect this.

Thanks Gustavo.-

[brad]

In general, performance degrades as the free space declines because the outer tracks on the platter(s) are the ones being used. The amount of degradation varies a bit from one make and model to the next.

To see how much degradation there is, partition and format the drive as one. Then run HDTach in extended mode and look at the speed drop (STR) as the outer tracks are used. The degradation is generally only important to those running dual-boot on a single drive. It isn't an important factor if logical drives in an extended partition are used for archival storage, pics, mp3s, backups, etc.

I would post a pic of a typical HDTach run here but this BBS software doesn't support uploding an image into a post, only links to URLs. I can e-mail one to you if you prefer.

Brad

[gustavo]

Thanks Brad I just sent you a private message with my e-mail address if you would take the time to send me that image.

I wanted to ask the original question because I did not just partition the HDD but also upgraded the VIA busmaster driver (they released the version 5.0.2195.120 with the new 4in1 4.36 saying there is a performance improvement).

Also my primary 30GB partition is 98% free space.

Could the upgrade of VIA PCI IDE Bus Driver be the cause of the performance degradation if any (I mean benchmarks are just benchmarks you cannot tell any diference if you do not run them)

Thanks Gustavo.-

[brad]

The HDTach test pictures are in the mail now. A picture is worth a thousand words and you will readily see what is happening, and where.

I have no win2k experience so I really can't comment on the effect different IDE drivers have on performance, other than to say the standard MS drivers work better in win98 and 98-se.

Brad

[gilles lussier]

Originally posted by gustavo
Thanks Brad I just sent you a private message with my e-mail address if you would take the time to send me that image.

I wanted to ask the original question because I did not just partition the HDD but also upgraded the VIA busmaster driver (they released the version 5.0.2195.120 with the new 4in1 4.36 saying there is a performance improvement).

Also my primary 30GB partition is 98% free space.

Could the upgrade of VIA PCI IDE Bus Driver be the cause of the performance degradation if any (I mean benchmarks are just benchmarks you cannot tell any diference if you do not run them)

Thanks Gustavo.-

It could if udma is not activated. Go in device manager, double click on the hard-drive controller and make sur dma is enabled. Also, still in device manager, double-click on the hard-drive, select the property sheet and make sure that the write back cache is enabled. In win2k there can be a 50% performance increase in benchmarks when these 2 items are enabled.

If you can't enable dma, just reinstall the previous ide driver.

[brad]

Gilles,

Gustavo's Sandra benchmarks are about right for his drive in win98 so I'm a bit surprised that he got them in win2k. According to what I read on usenet, the win2k file system in general is slower than in win98. Is this true or a bunch of malarky from folks that don't have their systems configured right?

Brad

[gilles lussier]

Brad, when a system is well configured, you normally get 3000 to 4000 points more in win2k than you get in win98. win2k maybe somewhat slower in games but is much more efficient than win98 in using all hardware in general. For example my memory benchmarks are about 50% higher in win2k than they are in win98.

[gustavo]

I can see looking at Brad pictures that performance decreases at certain sectors of the drives located (i just guess) at the center of the drive because more spins are needed and therefore more time to get the same amount of data, BUT the question is if is there a pattern when you partition a HDD that produces one partition to keep the better sectors and the other the worsts or the partition is made at random ?

I run my own HD Tach and got a random acces time of 14,4 ms I think it should be much lower as the specifications of the disc

The average read speed was 36810 kps, (from 43212 max
to 23952 min), CPU utilization 12% and the graphic shows performance stays above 40.000 kps up to 14 gb, then at aprox 40.000 kps up to 21 gb, 38.000 kps up to 24gb and then linearly down until the end (37.13 gb).

Brad: Why your discs have so much lower random access times
(8.7 and 12.4 ms) and cpu utilization (5 to 7%) but mine has higher average read speed ?? Cannot understand it.

Gilles: UDMA is active as is the cache.

Thank both Gustavo.-

[brad]

One must be carefull when reading and/or comparing drive specs because not all manufacturers use the same terminology, or use them in the same way. Access time and seek time are often used terms with very different meanings and it helps to understand what they are.

SEEK time is the time it takes the head stack to move to the required track/cylinder. Nothing more.

Rotational latency is the time it takes for the disk to spin around so that the sector on the track being accessed is under the head stack. In a 7200 rpm disk, it is 4.2 ms for 1/2 revolution. 8.3 ms for 1 full revolution. The former is used in most access time tests and specs because it affords a 50/50 chance of the sector being more than 1/2 turn, or less than 1/2 turn away from the head stack.

That means Access time = seek time + rotational latency

Since all 7200 rpm drives have equal rotational latencies, a difference in access time is caused by different seek times.

If your drive is a ST340016A, the seek time is 9ms average. Add 4.2 ms rotational latency and we get 13 ms. The difference between that and the 14.4 ms you measured is caused by a combination of things that aren't terribly important here but here is a link to a wealth of related info that will explain those factors and a lot more:

http://www.storagereview.com/guide2000/ref/hdd/perf/perf/spec/pos.html

CPU utilization is a function of how the drive's on-board controller is interfacing with the rest of the computer. Most drives have different results for this as well.

Please note that there is nothing wrong with the hdtach numbers you provided. They are excellent and the Baracuda IV is among the quickest IDE drives available now.

Brad

[gustavo]

Thanks Brad.

Learning is a pleasure, more enjoyable with professors like you.-

Gustavo.-

[brad]

My pleasure and thanks for the kind words. The HDTach graphic results you sent are excellent, better than the IBM 75GXP drives I'm using. On going back to the review at storagereview.com, I now see that it is indeed. You would have to spend quite a chunk of money to do much better.

Brad

[Propellerhead]

Brad, It doesn't cost too much for faster IDE HDD access. Raid is quite afordable nowadays, especially if u get a mobo with it onboard.
I run 2x 40gb IBM GXP60 series drives in a RAID 0 array, & man is it qwik! I could never go back to a single HDD.
I've sold 3 Promise TX2 Controller cards to people in the last few months, & they all say the same thing, "Could never go bak to 1".

So Gustavo, If u do a lot of HDD intensive work like analysing Database data, & u want speed, give RAID a go.

As for the newest offerings from seagate, not only r they just as fast as IBMs GXP series drives, they're now also just as quiet!

That's my 2 cents worth.
CU.

PS. This is only my first visit to eyos' forum so it may be a new feature, but u can post pictures & other files here.

Example, here is the picture that represents me over at the Atomic forums:

[brad]

Try a pair of WD1000SE drives in a raid setup!! This drive is the current king of the IDE hill by far, rivaling SCSI performance in a single drive IDE environment, so imagine what a pair of them would do in RAID. Not sure if RAID will take a pair of 100 gig drives though. I think I read somewhere that either 60 or 80 gigs per drive is the limit on some RAID controllers.

Raid is nice but I'm not a fan of using raid for speed alone. I'd rather make use of the better data safety other RAID modes provide.

Brad

[jarred]

hey Propellerhead


what site can i download pictures 2 put next to my name

[brad]

Don't know of any particular site with a mess of these for d/l. I rather guess a lot of people just do a right-click save as to grab them as they encounter them on the web. That's what I do anyway. Just make sure they are 64x64 max size, or resize them if they're larger. I have an animated one I want to upload to the forum also but it is larger than 64x64 and when I resize it, the animation disappears.

Brad

[alan norman]

Hi Brad

Nice to see you back.

You could resize the image but you would have to resize every frame and then re-animate it. I have resized similar images using Paint Shop Pro.

Bye

Alan

[brad]

Thanks Alan. EYO helps maintain my sanity. I never realized how big this house is until I'm the only one left in it. Sure sounds like I have some learning to do with animated GIFs. Don't know if I have any tools to do what you said or not. Maybee I'll send it to ya and you can advise..... Gotta keep busy ya know.

Brad

[Propellerhead]

The "Little Pictures" are called "Avatars".
I make my own using Paint Shop Pro 7, & like Brad says, must be 64x64 pixels in size.
There R sites where u can download AVs (Avatars) aimed at the VP Chat progy, but most of them r lame, & VP uses a diferent size.

Best way is to find a picture u like, & manipulate it yourself using somthing like PSP7.

BTW, I love your AV brad, v'cool. (This is the 1st site I'v seen/used that supports animated AVs !)

PS, to those 13 people that downloaded My AV from the post above, please don't use it as an AV for yourselves. I made that AV, & it is my regular AV (besides which, if u wear it to certain websites, u may get yelled at. lol)

Cop-U-L8r

[brad]

Thanks. Thats the only animated AVATAR I've ever seen that fits a hard core anti-Intel fanatic. My conscience is bugging me though because I lifted it from another forum and Alan re-sized it for me! Guess I better look for something else. Like to find one that is anti-Microsoft AND anti-Intel. Can someone point me in the right direction?

Brad

[Propellerhead]

Hey Brad,

I think that u will find my anonymous email address amusing. It is:

wintel_sux@hotmail.com

U just gotta love the irony of it being a hotmail address.
lol.