[Grrr]

Just been messing about with Xbench on my Macs, and comparing figures with other users.. And I notice G5 iMacs are pretty consistently returning an over all Disk test score in Xbench of 91, or there abouts. Not just mine, but everyones. Even my old eMac beats that by a noticeable margin. Is SATA not all that? Is the drive being mounted upright really having an impact on performance?
Whats all that about then??

PS, And no "But Xbench is meaningless!" posts, cos I disagree! You only have to compare peoples results with with others to get a good comparative idea of whats going on.

[macaddict0001]

With the Speed of the disk in the imac you couldn't overload a sata bus anyway so that is irrelevant. It could very well be that the drive is mounted vertically, because the read/write head has to move up and down.

[Randman]

More info would help.
What are your XBench scores? Is your energy saver setting set to maximum performance? Is bluetooth turned off? Are you using other apps at the same time? What about resource hogs such as Konfabulator.

[Simon]

Originally posted by macaddict0001:
With the Speed of the disk in the imac you couldn't overload a sata bus anyway.

That's what I had in mind too.

IIRC SATA1 goes up to about 3Gbps so I highly doubt any available 7200rpm disk could saturate the bus.

[Strix]

The MacFormat magazine in the UK recently reviewed the 1.8Ghz 17" iMac, and found that overall performance was slightly better than a 1.6Ghz G5.

However the big surprise was the disc performance, which was not as expected. I don't have the figures with me, and will post them when I do, but the disc performance was poor.

[Simon]

Is it possible that in order to keep costs low Apple chose to use a cheaper and slower disk on the iMac? I guess the PowerMac's higher margins certainly would allow for a more expensive and better disk. Does anybody know what disk types are used in the 1.6GHz PowerMac and in the iMac? Just because both have 7200rpm disks, does not necessarily mean they perform equally.

[Grrr]

Originally posted by Randman:
More info would help.
What are your XBench scores? Is your energy saver setting set to maximum performance? Is bluetooth turned off? Are you using other apps at the same time? What about resource hogs such as Konfabulator.

Is that relevant? Im only talking about disk scores here, which I did already give the score for. Plus this iMac has no bluetooth and no other apps were running, And yes, CPU set to highest etc.

Anyway, it just seems like the supplied iMac drives are a bit lame (In my case, its a 160gig drive). Looking at the figures for a few 1ghz eMacs, including my own with a 60gig whatever in it, the G5 iMac SATA drive is completely spanked.
Looking at the Xbench site where you can view other peoples performance figures and compare, its clear a few folk have chucked out the standard clockwork gramophone style HD within their shiney new G5 iMacs in favour of something a bit more speedy. And the gains appear to be pretty big too. Which would indicate the upright mounting orientation probably isn't a factor.
See for yourself, Heres one, much the same spec as mine, but with an upgraded 300gb Maxtor SATA drive (Plus 2gb vs my 1gb ram).
http://ladd.dyndns.org/xbench/merge.xhtml?doc2=84696
Compared to mine with factory issue drive..
http://ladd.dyndns.org/xbench/merge.xhtml?doc2=87330
Again, compare disk scores with something like an eMac or 3, and you will see just how weak it really is.

Standard drive is poopy, no?

[Simon]

Originally posted by Grrr:
Standard drive is poopy, no?

Well, before you jump to conclusions, why don't you first tell us what exact model of drive your iMac has.

System Profiler -> Hardware -> ATA -> Vendor, Model, Revision, etc.

That way we'll know what kind of drive we're dealing with and we can compare with what Apple builds into the PowerMacs.

[Grrr]

Originally posted by Strix:
The MacFormat magazine in the UK recently reviewed the 1.8Ghz 17" iMac, and found that overall performance was slightly better than a 1.6Ghz G5.

However the big surprise was the disc performance, which was not as expected. I don't have the figures with me, and will post them when I do, but the disc performance was poor.

Exactly, I've run the test maybe half a dozen or so times now, and each time, the results are consistently poor. And as far as I can tell, its the same for all G5 iMacs, until the drive is swapped out for a better one.
Just makes me curious why the standard offerings are not too hot..

[Grrr]

Originally posted by Simon:
Well, before you jump to conclusions, why don't you first tell us what exact model of drive your iMac has.

System Profiler -> Hardware -> ATA -> Vendor, Model, Revision, etc.

That way we'll know what kind of drive we're dealing with and we can compare with what Apple builds into the PowerMacs.

Did you not look at the Xbench links I provided? It says on those pages what drive is fitted.

However, I checked the system profiler anyway, and a brand name does not appear to be given, like I see for most other Macs, instead of brand and type, it says "ST3160023AS" Which is the same as it says in Xbench.
Capacity, 149.05gb
Revision, 5.06

[Simon]

ST3160023AS is a Seagate Barracuda SATA disk with 7200 rpm and an 8MB cache as you can see here.

AFAIK this is the exact same type of disk that Apple puts in the PowerMac G5 1.8GHz. And xBench scores for those machines look similar. So at least Apple wasn't just cutting corners on the iMac.

But, and this again shows how much xBench sucks, take a look at the dual 2.5GHz PowerMac disk scores. With the same type of drive the scores are up to 20% higher. So actually the so-called disk test doesn't only test the disk, but also the system's general performance.

[discotronic]

Originally posted by macaddict0001:
It could very well be that the drive is mounted vertically, because the read/write head has to move up and down.

The drive being mounted vertically has absolutely nothing to do with the performance. I'm not sure where people are getting that idea. I've seen this statement on numerous occasions. Am I missing something?

[Randman]

Originally posted by Simon:
But, and this again shows how much xBench sucks.

Well-said.

[deboerjo]

Originally posted by Simon:
IIRC SATA1 goes up to about 3Gbps so I highly doubt any available 7200rpm disk could saturate the bus.

There's a lot of misinformation about SATA speeds floating around. SATA1 is 150MBps (note the capital B, that's 1.2Gbps). SATA2 is 300MBps (2.4Gbps). I don't know where the oft-quoted 1.5Gbps and 3Gbps comes from.

Either way, while it's not SATA itself that's the problem here (150MBps is more than enough bandwidth) I suspect the SATA controller they're using has some kind of performance problem. It's fairly common, since SATA is so new. I was just reading last night how Linux is having an issue where Silicon Image SATA controllers + Seagate (and others) hard drives = poor performance, due to some kind of packet size issue. A firmware update from Silicon Image that fixes the issue is supposedly forthcoming. Maybe a firmware update from Apple in the future will speed things up?

(Edited for terrible spelling)

[Simon]

Originally posted by deboerjo:
There's a lot of misinformation about SATA speeds floating around. SATA1 is 150MBps (note the capitol B, that's 1.2Gbps). SATA2 is 300MBps (2.4Gbps). I don't know where the oft-quoted 1.5Gbps and 3Gbps comes from.

I got the numbers from James Wiebe's white paper on the evolution of high-speed connections. He claims SATA goes up to 1.5GHz (which I translated to Gbps since it's a serial bus).

But, as you correctly point out, this is baloney.

According to Wikipedia SATA is 150MBps and SATA II should go to 300MBps and 600MBps.

[BTW I think it makes more sense to give speeds of serial busses in bps than in Bps, but that's just a side remark.]

[Grrr]

Originally posted by Randman:
Well-said.

Ahh, I wondered when the Xbench doubters would preach this.. From what I have seen, all Xbench performance figures are very comparable between Macs of the same type, and are quite repeatable on the same Mac too. Xbench easily shows one Macs strengths and weaknesses compared to another. So I say Xbench is doing its job just fine. I'm yet to see any proof to suggest otherwise.
Only issue I see, is that unlike the Norton tests of old, Xbench does little to prompt the user to use suggested default system settings during the test to ensure comparable scores.

Comparing with a Dual G5 2.5 with the same drive is interesting though. But just because the disk score is different, does not mean its an Xbench issue. I'd be far more inclined to believe that something else causes the differing scores.
And for those Xbench doubters, I'd bet any other disk test software would show very similar differences.

So, Back to the matter in hand.. Whats with the G5 iMacs poor drive performance? All I could think of was, that performance was deliberately strangled somehow in order to reduce heat/noise perhaps? In much the same way as CPU performance was strangled in the energy saver control panel. Difference being, I don't see any way to uncap the drive, if indeed it is capped in some way.

[Strix]

The MacFormat magazine in the UK recently reviewed the 1.8Ghz 17" iMac, and found that overall performance was slightly better than a 1.6Ghz G5.

However the big surprise was the disc performance, which was not as expected. I don't have the figures with me, and will post them when I do, but the disc performance was poor.

------

The MacFormat results were based on CocoaBench 1.2:

1.6 Ghz G5:
Number Crunching: 23.01
Image Processing: 31.72
File I/O: 24.53
Total Score: 25.57

1.8Ghz iMac
Number Crunching: 31.18
Image Processing: 39.56
File I/O: 7.95
Total Score: 27.47

The test machine had 512MB RAM, 80GB drive.

[macaddict0001]

Originally posted by discotronic:
The drive being mounted vertically has absolutely nothing to do with the performance. I'm not sure where people are getting that idea. I've seen this statement on numerous occasions. Am I missing something?

The read write head may have to mover vertically slowing it down slightly, of course speeding it up when it comes down. This affects access time.

[deboerjo]

Originally posted by macaddict0001:
The read write head may have to mover vertically slowing it down slightly, of course speeding it up when it comes down. This affects access time.

The weight of the drive head relative to the motor driving it is negligible, gravity has no effect. Remember, this thing is designed to accelerate that head and bring it to a stop again in 9ms. That's like saying it's harder for you to flick a pine needle in your hand up and down than it is to flick it sideways.

[P]

Originally posted by Simon:
I got the numbers from James Wiebe's white paper on the evolution of high-speed connections. He claims SATA goes up to 1.5GHz (which I translated to Gbps since it's a serial bus).

But, as you correctly point out, this is baloney.

According to Wikipedia SATA is 150MBps and SATA II should go to 300MBps and 600MBps.

[BTW I think it makes more sense to give speeds of serial busses in bps than in Bps, but that's just a side remark.]

Not only serial busses... bps is the correct unit for data transfer speeds. The problem with using bytes is that 1KB = 1024B, but one 1Kb = 1000b. Make it easy on yourself, stick with bps.

[macaddict0001]

Originally posted by deboerjo:
The weight of the drive head relative to the motor driving it is negligible, gravity has no effect. Remember, this thing is designed to accelerate that head and bring it to a stop again in 9ms. That's like saying it's harder for you to flick a pine needle in your hand up and down than it is to flick it sideways.

it still makes a difference, I didn't mean that it was the sole factor.

[discotronic]

Originally posted by macaddict0001:
it still makes a difference, I didn't mean that it was the sole factor.

I couldn't find anywhere in my research where having a hard drive mounted vertically will cause a performance hit. I decided to perform my own test.

First I ran Xbench with my iMac sitting normal two times. The first test gave the overall disk test score of 91.37 and the second test a score of 91.71.

Second I positioned my iMac to where the hard drive would be working exactly parallel to my desk. I even used a level to make sure. I repeated the test and the first score was 91.68 and the second score was 90.89.

Basically there isn't a difference one way or the other.

[Grrr]

Originally posted by discotronic:
I couldn't find anywhere in my research where having a hard drive mounted vertically will cause a performance hit. I decided to perform my own test.

First I ran Xbench with my iMac sitting normal two times. The first test gave the overall disk test score of 91.37 and the second test a score of 91.71.

Second I positioned my iMac to where the hard drive would be working exactly parallel to my desk. I even used a level to make sure. I repeated the test and the first score was 91.68 and the second score was 90.89.

Basically there isn't a difference one way or the other.

I thought of trying that myself, but you beat me to it. That certainly settled that argument. Good work!

While im here, I remember reading on this page, something about hard drive performance management, I'll quote the specific part..

Modern hard drives have an internal feature called AAM - Automatic Acoustic Management. It allows to trade few percents of drive's performance for nearly silent operation by eliminating head movement noise. Apple supplies Macs with drives set for silent operation but most OEM hard drives are sold set for maximum performance mode. This utility (see actual site this came from for the links) allows you to control the drive's AAM feature. It works only on internal drives in OS 9 but not in Classic. However if you can not boot into OS 9 there is a way out - AAM feature is "sticky", so you can put the new drive in another Mac or a PC and set its AAM parameters there. There are quite a few AAM utilities for PC.

Might this be the culprit? It would explain why 3rd party drive installs on G5 iMacs show much better performance numbers, even tho the manufacturers stats are similar from drive to drive. However, seems like in the case of the iMac, the performance difference is more than just a 'few percent'

[P]

Originally posted by Grrr:
<AAM>

Might this be the culprit? It would explain why 3rd party drive installs on G5 iMacs show much better performance numbers, even tho the manufacturers stats are similar from drive to drive. However, seems like in the case of the iMac, the performance difference is more than just a 'few percent'

It may be part of it, but hardly all of the reason. No, I think we need a Firmware update - the new mobo design is not perfect yet, and there are likely timing issues to be resolved.

[Unregistered Us]

Originally posted by macaddict0001:
it still makes a difference, I didn't mean that it was the sole factor.

Man, give it up. There is nothing wrong with being wrong, acknowledge your mistake, and then move on.

I mean, according to your argument, humidity could make a difference. How about altitude? If I live on a mountain where the air is less dense, should the HD be faster then? Sheesh, talk about a sore loser. HOwever, I can't believe discotronic took on the challenge and actually tested it though.

[deboerjo]

Originally posted by macaddict0001:
it still makes a difference, I didn't mean that it was the sole factor.

Apparently I was not clear. Maybe this will help:

IT....DOES...NOT...MAKE...ANY...DIFFERENCE...WHATS OEVER

[macaddict0001]

Humidity and altitude do make a difference, even infinitesimal differences add up, I also didn't say it was the reason for the score difference. That being said, I also have noticed that the imac's hard drive speed is slower than the powermac's, and i would also like to see the reason.

[Pierre B.]

Originally posted by Grrr:
Is SATA not all that? Is the drive being mounted upright really having an impact on performance?
Whats all that about then??

I cannot answer directly your question. However, you can try iozone. It performs a much more detailed set of I/O tests than Xbench, offering many options in the command line. You will need the developer tools in order to install it (it is also available through Fink).

[Grrr]

Originally posted by Pierre B.:
I cannot answer directly your question. However, you can try iozone. It performs a much more detailed set of I/O tests than Xbench, offering many options in the command line. You will need the developer tools in order to install it (it is also available through Fink).

Thanks for that. But i'll pass on it. As I think we have already established that the G5 iMac drive is not performing as well as expected. Just a question of finding out why, and what can be done about it.