Here is a small benchmark test, testing various filesystems. Elsewhere, 10.3 Panther is tested on an iMac G3 600 Mhz and –except the reduced OpenGL performance (possibly because of changes in the QuartzExtreme that this iMac does not support with its ATi Rage Pro card)– it is showing to be overall faster than 10.2 according to the XBench test suite. Update: More benchmarks. HP released new server speed-test results Wednesday that for the first time compare its version of Unix with Windows on the company’s top-end Itanium server–and Unix came out ahead.
Benchmarks: Test of 6 Linux Fs; Panther Faster even on G3s
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
It is good to hear that Mac OS X’s speed is improving. Does anyone yet know what changes that Apple made to improve the over all speed?
Everytime I see another x-bench benchmark I think my confidence in that tool goes down a little more. Why are the CPU results different between operating systems? I’m assuming that it’s the only other application being run so why should they be different? They should be very close.
This is going to be great for all the old G3 iMacs and iBooks. Its good that Apple is still making some investment in their older hardware. Unfortunately I do not think that Beige G3s will be supported.
Also, if you take a look at the benchmarks for the Quartz system, you’ll notice that while the ‘overall’ score hasn’t improved much, you’ll notice that the number of characters being rendered has almost doubled. Yet the commentary doesn’t note this.
This of course assumes that Xbench measures the text rendering in a way that will help the rest of us.
SteveTooth is right. There is something very wrong with these benchmarks. How can be CPU tests so much higher? Does 10.3 upload any magic microcode to the CPU or is there something running in the background (like indexing service or something)?
To properly test ONLY the CPU, you need to use the SPEC benchmark which is a whole process. ANY other “normal” benchmark is just using the OS’ API to test a few things, and so it is kinda OS-specific. There is nothing wrong with XBench here IMHO.
The cpu improvements will probably be because of an improved scheduler.
“Unfortunately I do not think that Beige G3s will be supported.”
They are not supported by Jaguar, either. Unfortunately, Panther is even choosier than Jaguar about which machines it will allow itself to be installed upon. It is one upgrade too far for my machine, unfortunately
iMac G3 600mhz are cheap too! I could get one for under $5,000 right?
Since my Mac happens to be a 600 MHz G3 iMac, this seems to be good news.
what machine is that?
an original iMac?
that is still supported. are you bitching because your hardware is no longer supported by Apple in software? well, it is old. they did say that OS X would cut a lot of backward compatibility with hardware.
I mean, Mac OS 9 can still run on a 10 year old mac, but when they say they are cutting legacy, what did you expect?
get a new computer.
you are such a troll.
> Since my Mac happens to be a 600 MHz G3 iMac, this seems to be good news.
Except the fact that OpenGL support is worse than before and in order to fully have a good experience, you will need a QE-enabled card in order to have faster UI.
> what machine is that? an original iMac? that is still supported. are you bitching because your hardware is no longer supported by Apple in software? well, it is old.
I am sorry, but you are in error. The Classic iMac 600 Mhz, was still sold by Apple up to April 2003. Therefore, it can NOT be conceived as “old”. Even if it is, they were sold up to now. Therefore, it is just not correct to say that it should not be supported.
I was not talking about that, I was talking about the beige mac…the ones sold just before steve got back and during his first few months back.
There are a lot of Beige G3s still out there and Sonnet offers a 1 GHZ upgrade for a machine that was never intended to run past 366.
I am sure Sonnet and Ryan Rempel will offer a solution or hack around this. Regardless going this route may cost you more in the long run and it might be better to sell your rig and get an eMac.
“iMac G3 600mhz are cheap too! I could get one for under $5,000 right?”
Yes…. the same way that all Macs are a fraction of that figure… as they are the same price as comperable PCs.
I’m surprised by the speed of XFS. I didn’t knew it was that fast with big files. I’m also surprised by the abyssmal speed of VFAT. I suspect there’s a problem with the Linux driver. I also wonder how ext2/ext3 managed to beat ReiserFS…
…then again, these tests were made with the 2.4.5 kernel. It’s more than two years old, and ReiserFS wasn’t excellent in these days. I wonder if there are benchmarks with a more recent kernel?
2.4.5? I’m running 2.4.6 on my freakin PDA. Why did they use such an old and unstable kernel for their tests?
Looks like they made a concerted effort, but when did they do this? Two years ago?
Don’t know about kernel, but the author must live in a cave…
ext3 still in test ????? it’s default in RedHat distribution. It’s considered as stable everywhere I read about it. I use it on a critical server for at least a year.
So I can be wrong, but I am curious : Who is doing those tests on ext3 that make it sounds like if it was an experimental fs ??
Does ext3 support files larger than 2GB?
Maybe vfat is faster under windows, but it’s still a really slow and outdated filesystem. NTFS I believe, is even slower (under windows I mean).
Is it possible that the filesystem benchmark is really old?
Why else test on 2.4.5?
I remember seeing filesystem article around six months ago when I was searching for information about which filesystem was the fastest and most robust. It looks like it has been around for a while and is probably pretty outdated.
uh?
Yes, ext3 supports files larger than 2GB.
The nice thing about ext3 is that it’s backwards compatible. Meaning, if you for some reason need to mount it on a system that only supports ext2 it’s not a problem, and only the journal features of the FS will be missing. Then, when you remount it as ext3 it will rebuild the journal and be all happy again.
Because the post says “Here is a small benchmark test, testing various filesystems” and in the title it says “Benchmarks: Test of 6 Linux Fs”?
Great news since I have an older iMac 400dv that I love.
“Unfortunately I do not think that Beige G3s will be supported.”
I think you could find a “hack” that would solve that problem.
Mac OS X keeps getting better and better 😛
The linux tests are for 2.4.5.
Running 2.6.0-test2 here however.
Well, this is quite an interesting development, especially in the face of so many lambasting Sun’s price/performance.
The SunFire V480 has the best price/performance of any system on the chart, at $44/QphH, edging out the PowerEdge 6650 at $45/QphH.
Combine this with the lower TCO of a Sun system and the higher reliability, and the choice becomes obvious.
wish the author of that article would come in for round 2 with either 2.4.21 or one of the test 2.6 kernels
i think reiser is *alot* faster nowdays than it was in 2.4.5, and im sure xfs has had some improvements
— Ophidian
In regards to CPU benchmark differences; I would have to say that it is very possible they differ. Scheduling and microkernal and algorithm improvements may have made code that works better with the processor–it may be a result of GCC 3.3 or someother API improvement.
In regards to the SPEC (200?) benchmark being the only one that matters–that is a PC derived benchmark (if I’m not mistaken), and any process test will be compiler dependent. All software has to be compiled by something to run. So a Pentium can be a piece of junk as a processor, but if it has great compiling libraries, then it can do OK. There was an arguement that using the GCC compiler to test P4 versus G5 was unfair, because it wasn’t as fast as Intel compilers–which is true. But it was a fair test of a decent compiler that gave a slight advantage anyway to Intel (as the GCC compiler is more mature on Intel). So the test was relatively accurate in predicting that a G5 is faster than a P4 of any kind. What it might not say is how good the G5 with OS X is compared to a P4 with XP–which is what users care about. The end results matter more to the user.
The Mac platform still has some room to grow with better compilers, and we may be seeing that with Panther. Whether it is just recompiling with GCC 3.3 or something else–well only the Apple programmers will know that.
It would be nice to see application benchmarks (please, only use MS products for reference or humour).
The OpenGL test might be slower because the benchmarking utility might be using its own OpenGL calls. Perhaps their is a new library based on OpenGL 1.5 coming out that will improve things, but works poorly on legacy apps (but most all apps should be making generic calls–so is this true about the benchmarking?). I don’t have the answer on that.
Those server benchmarks don’t apply too well to the G5. Even though the power4 is a lot more expensive and built for a server, it is supposed to be a lot slower. It is built with more insulation on a larger process die (adding to its costs), but this makes it more reliable for servers. Supposedly, the G5 (970) is a lot faster than the Power 4. So we will have to see a QphH score based on the actual processor we want to compare. The Power 4 here is a sibling of the G5–not the same processor. So why are we even looking at it?
Aye, probably. I don’t even know if ReiserFS was marked as “stable” (and not “experimental”). I also wonder the performance of FreeBSD’s UFS/UFS2 and its VFAT driver.
Hmm… Perhaps I’ll do a benchie like that when I’ll have the time.
…what happens when you have slow hardware. Apple really optimised their OS. I’m running Panther DP1 on my iBook 500 and it really shines on it. I really don’t think of upgrading anymore. I’m just sad that they released G5 and with that speed they will probably go the Win route, adding crap into their OS because fast hardware can handle it.
I believe that this will be the last OSX revision that would be good on older hardware. OSX has really matured and it’s only two years old OS.
Check out the price tag on some of those SunFires…
$50,000 for a SunFire v240 @ 2 CPUs @ 1 Ghz
$100,000 for the SunFire v480 @ 8 CPUs @ 1 Ghz
$230,000 for the SunFire v880 @ 4 CPUs @ 1 Ghz
$5,000 for a Linux server @ 2 CPUs @ 2 Ghz
Are you sure you want to do a comparison here? You can get 10 Linux PCs for the price of one SunFire v240. I sure hope your SunFires last 10 times as long as your competition’s PCs, because they can afford to be upgrading their PCs every year.. Something to think about.
Here’s a quote from the Gentoo x86 Installation page, Filesystems section, which is here:
http://www.gentoo.org/doc/en/gentoo-x86-install.xml#doc_chap6
The quote is:
<quote>XFS is a filesystem with metadata journaling that is fully supported under Gentoo Linux’s xfs-sources kernel. It comes with a robust feature-set and is optimized for scalability. We only recommend using this filesystem on Linux systems with high-end SCSI and/or fibre channel storage and a uninterruptible power supply. Because XFS aggressively caches in-transit data in RAM, improperly designed programs (those that don’t take proper precautions when writing files to disk, and there are quite a few of them) can lose a good deal of data if the system goes down unexpectedly.</quote>
I’m using this information (and the other information on that page — there’s a lot) to help me decide which file system to use.
Since it’s a desktop, I am mostly gravitated toward ReiserFS. But, wow, with those benchmarks, I am seriously rethinking it.
Does anyone know which uses more CPU time, for example? That might be a determining factor, as my desktop is fairly slow. (1Ghz Nehemiah C3-2). Also, I’ve heard that the ReiserFS does a “better job” (than what? I don’t know) of handing priority elevation so that no process gets starved for data. I am suspicious of this because I didn’t know that the filesystem even handled that, but thought that was done by another module in the kernel. Any insights on this?
Thanks.
Erik
ummm…his argument was price/performance ratio.
so all the dell fanboys who trounce around about P/P ratio can now run out and get a 50 k computer since it is better P/P.
I use XFS on two machines — a mail server with two IDE drives striped in a RAID-1, and in another machine with a RAID-5 array on SCSI disks. Both are ups’d. I’ve only had problems on the SCSI machine, while the IDE machine works fine. When I did have problems, xfs_repair fixed things perfectly. XFS, for me, has been as stable as Ext2/3, and seems about as stable as FFS/UFS (which would make sense).
Reiser, otoh…..I will never use Reiser3 again. I’ve had two machines where reiserfs crashed hard. Even booting from a CD, and trying to run repair utilities didn’t fix the problems. I have never experienced this with another FS, save FAT (but it’s not fair to compare windoze here). I may try Reiser4 on a testing machine when it’s released — it’s got some good ideas, but it certainly isn’t stable from what I’ve seen.
As an aside, I’ve never had a problem with an NTFS volume, other than fragmentation. They still haven’t solved that problem over at MS.
Daniel Robbins (Mr Gentoo) seems to have a problem with XFS, I’m not sure why. I’ve used it for awhile (since it came out for Linux, and before that on SGI IRIX) and never had a problem with it. It’d be nice if the real-time XFS stuff was in the LInux version, for the data starving stuff. (actaully, it says on SGI’s page “This hasn’t been tested”, so maybe it’s actually in there already)
I’m running Panther (7A202) on my 400 mHz iMac. For what it’s worth, it seems to be running significantly faster than Jaguar. Programs launch faster, and the big thing I’ve noticed is that Get Info is no longer embarassingly slow. I’m really liking it.
A couple of points, first is the fact that this article appears to have been done in July of 2001 (based on modified date of conlcusion page). This test was done before ext3 had even been considered stable, and ReiserFS has undergone multiple fixes/changes since then. Second, as has been mentioned before, he uses 2.4.5 which is a fairly old kernel. Third the drive tested is a UDMA66 and he does not even qualify his tests with IDE chipset and level of support for it in the kernel he used. DMA settings make large differences in performance.
As a comparison of the various file systems against each other at the time it is fine. But as a comparison to any other system, or even modern versions of the various file systems I do not think the comparison is valid.
$44 per transaction is still too high when you compare it to Power4 + AIX which is around $7 per transaction. Yes, that is unclustered.
What sort of MORON buy an iMac “Classic”? I went in to my local Apple Store on friday last week (where I bought my eMac + Office X) and they have NONE on display, what does that tell me? it tells me there are a bunch of IDIOTS going out of their way to purchase OBSOLETE equipment. That does not make sense!
I’ve recently bought an eMac, I certainly don’t expect it to last 6 years, but I do expect to be able to upgrade and keep using it for another 3 years. This stupid expectation by some people that some how the whole world should remain static because they happen to be a Chucky Cheapskate is just rediculous.
As for the benchmarks, again it proves what I said on another forum, Mac’s problem totally AREN’T related to the CPU but the lack of a properly optimised compiler. It is nice to see they’ve adopted 3.3 for the compiler and it will be interesting to see what they are going to add to harness the full power of the hardware.
my mistake, I was thinking of the TPC-C benchmark.
I don’t know, someone who wants a cheaper Apple?
What a MORON, huh?
CooCooCaChoo,
There is nothing wrong with an iMac and if one can be had reasonably then fine, although the eMacs are very inexpensive and a better value overall.
I hope you like the eMac you should write an article on what your experience has been like so far. We have one at work that I setup for a PR person and its easily better than any iMac and much more durable than a iMacG4.
I have some concerns about the quality of the display when displaying a pure white background but the refresh rate goes up pretty high. The case feels durable and solid. I wish they made the HD and CDROM easier to upgrade. It might then be the perfect office machine. Perhaps that is what Apple needs to do. I plan on getting more in the office. The eMacs initially had a lot of quality control issues which seem to have been fixed.
Everytime I see another x-bench benchmark I think my confidence in that tool goes down a little more. Why are the CPU results different between operating systems? I’m assuming that it’s the only other application being run so why should they be different? They should be very close.
Maybe they used the Intel compiler
I will write a review if ELQ wants me to, however, until I hear, I am certainly not going to slog my guts out only to find it thrown back into my face.
My email address has been made available on this post, just remove the extra oo, it should be:
coocoocachoonz at yahoo dot co dot uk
ELQ or who ever else is in charge, I can write an article regarding a *BSD/Linux to eMac migration. Email me and I’ll start writing one.
RE: Anonymous (IP: —.red.retevision.es)
You can’t honestly tell me that AUS1899 is too much. I am a student, yet, I was able to save up and pay for it. If I can do it, considering I am the worlds worst money manager, I am sure ANYONE can do it.
Classic iMacs are OLD and crusty. If the average user is too stupid not to read up the product they are about to buy, I say, stiff-bikkie.
“Check out the price tag on some of those SunFires…
$50,000 for a SunFire v240 @ 2 CPUs @ 1 Ghz
$100,000 for the SunFire v480 @ 8 CPUs @ 1 Ghz
$230,000 for the SunFire v880 @ 4 CPUs @ 1 Ghz
$5,000 for a Linux server @ 2 CPUs @ 2 Ghz
Are you sure you want to do a comparison here? You can get 10 Linux PCs for the price of one SunFire v240. I sure hope your SunFires last 10 times as long as your competition’s PCs, because they can afford to be upgrading their PCs every year.. Something to think about.”
The price tag includes the database license. A 2 Cpu 2 GB RAM v240 is $6495. And this a 64-bit cpu. And the Sunfire v65x with Dual 3.06 GHz Xeon 1 GB RAM running linux is $3,995.00.
The v65x is cheaper than a similarly configured dell poweredge.
How could you say that? you’ve just crushed the Dell fanboys last line of defence, aka, “we’re cheap, we’re cheap even though Dell can’t innovate themselves out of a paper bag”.
If Dell spent the same amount on R&D (as a percentage) as SUN or IBM then maybe we would see something a little more substantial than a cheap machine.
1.2 kchars/s? Am I the only one who thinks Quartz’ text engine is ridiculously slow? “x11perf -f24itext16” on my iBook draws 80kchars/s, should Quartz’ AA make text rendering 80 times slower than necessary?
These benchmarks aren’t even close to new – I think I saw them at least a year ago. Not sure how this qualifies as “news”.
Not to mention that the 2.4.5 kernel doesn’t give you half a clue how 2.4.21 or 2.6.0 will do.
-Erwos
These benchmarks are absolutely garbage. I dont understand why it was even posted here to OSNews. 2.4.5 kernel…..come on now. Nearly all current benchmarks that Ive seen have XFS and Reiser nearly neck and neck and both trouncing Ext3. XFS was designed for larger files by SGI to be used for video editing etc. Its really not designed to be an everyday FS. Ext3 is supposed to protect your data best, but its also slow. I personally prefer ReiserFS since its still got good journaled protection of your data and has excellent performance. Ive been using Reiser personally on tons of different systems at home and at work on servers and on desktops, and Ive never had a single issue with it. I also hear Reiser4 will be merged into 2.6.x shortly. Bring it on
yeah, and when you talk about cheap, your not just talking inexpensive.
you get crap from Dell in the way of consumer level laptops!!!
900 gets you the low end, and it comes with a CD-ROM…ok, that is fine, 950 gets you the same in an iBook.
1200 gets you a DVD-ROM…..umm…ibook gives you a combo drive.
1300-ish gets you a dvd-rom and a larger screen with dell
1350-ish gets you a larger screen and a combo drive with apple
1500 gets you a 15 in screen and a dvd-rom….ibook, 14.1 in screen and a combo drive
now, that is not all you get with either model, but screen size and writable drives are important to laptop users. is dell on crack?
oh, and not to mention an ibook is put together much better…..if you have ever held an iBook and then hold a Dell laptop, you can feel the difference.
I think it is because the test was run on a machine that does not support QE, and they made some changes to the Q engine, but I seriously doubt that they are THAT slow, there was a mistake somewhere.
XBench’s OpenGL test absolutely needs a rewrite, as I have noticed major discrepancies between two separate XBench results on the same machine, with the same Jaguar version… other tests are so-so, but I guess XBench remains a good general performance metering utility.
Surely OpenGL also needs more optimisation on Panther, but it ain’t outta beta yet, so lets hope for the better!
As for iMacs, not everyone wants to upgrade their multi-hue gumdrop machine… In fact, judging by preliminary benches, older Macs will gain value instantly (well, whatever instant it takes to install Panther) and that is a big deal for a lot of schools, organisation, small businesses etc…
Actually, I’m considering buying a second-hand Graphite iMac DV sometime in the future, to complement my PowerBook G3, for MP3 serving, surfing and DVD-viewing purposes. The original iMac series is still of great value, the design is very good, and the DV iteration of those machines have virtually silent operation.
Someone mentioned that Jaguar/Panther will not run on an old G3. You need to check the install README…it will indeed.
You need to base your assumptions on facts rather than pop culture on such subjects.