PDA

View Full Version : PowerMac G4 SATA-PCI-Card Owners, post your real life benchmarks!




Cox Orange
May 27, 2013, 12:47 PM
Since benchmarks come up recently again and are referred to every now and then, I thought I search for some insight to a question I asked myself.

What is real life speed and are benchmark utilities anything useful? My point is no.

If I connect two SATA-drives to a SATA-PCI-card in a PowerMac G4 (AGP), I get between 33-39MB/s, when copying a file from one HDD to another. I found someone that found out the same.

So, if someone be so kind and help me and copy a file from one drive to another on his/her SATA-PCI-cards and stop the time.

IMPORTANT! Don't use utilities, just stop the time and divide the file size through the time elapsed (in seconds).

I am really risking proving to be a fool, but then again I want to know, if anyone has the speeds they always report from their xbench, quickbench and AJA-tests.

Please forgive me!

PS: I used 5400rpm Drives, with 2MB and 10k rpm Drives and others, but all are in the range of 33-39MB/s. I used a 5GB, 20GB DV-file (iMovie) and several files gathered to one copying action of 800GB (some DV-files up to 40GB under them).
What makes it even worse is, I have Sonnet and Newertech CPU Upgrades, so the evildoer is not the slow stock CPU.



Cox Orange
May 29, 2013, 04:25 PM
Hm, no one?

Is there suddenly no one with a PowerMac G4 anymore? Is it you find my question stupid anyway? Or am I right and the truth is so embarrassing and shocking?

Antonius
May 29, 2013, 04:55 PM
In a few months I'll be buying myself an SATA PCI card for my MDD and I'll add my benchmarks here

philz4life
May 29, 2013, 07:39 PM
I recently got a SATA PCI card for my Powermac G4.. I can't do any tests the way you want because my SSD is the only drive on the card.

philz4life
May 29, 2013, 07:47 PM
Here's my benchmarks, the SATA SSD has a disappointingly small performance lead.

MisterKeeks
May 29, 2013, 07:52 PM
IMPORTANT! Don't use utilities, just stop the time and divide the file size through the time elapsed (in seconds).


Here's my benchmarks, the SATA SSD has a disappointingly small performance lead.

The goal is to find out whether the utilities produce anything useful. Though you could use them and compare.


HOWEVER, read and write speeds can vary based on the size of the file. Hard disks excel at large files.

skinniezinho
May 30, 2013, 10:44 AM
Here are the benchmarks for my Vertex 2 fw1.37
Sonnet Tempo Sata PCI , fw5.1.3

Powermac G4 GE fw4.2.8
Sonnet 1.2GHz CPU
2GB PC133


http://img14.imageshack.us/img14/6476/ssdx.png (http://imageshack.us/photo/my-images/14/ssdx.png/)

http://img221.imageshack.us/img221/2158/ssdaja.png (http://imageshack.us/photo/my-images/221/ssdaja.png/)

Guess it is the max of controller/pci bus.


There are some threads with more benchmarks like this (http://forums.macrumors.com/showthread.php?t=1093080&highlight=study) or search for SSD.

seveej
May 30, 2013, 12:09 PM
Since benchmarks come up recently again and are referred to every now and then, I thought I search for some insight to a question I asked myself.

I have a Sonnet Tempo two port SATA card in my MDD, and can not provide you with a "copy from one disk to another" speed, as they are in (soft)RAID.

What I can say, is that when transferring files over the (gigabit) network to and from that RAID stack, the transfer speeds are within the 60-80 MBps bracket.

The disks are 2 TB Seagate Barracuda Green's (those weird drives with an rpm of 5900 - that's not a typo).

RGDS,

666sheep
May 31, 2013, 12:15 AM
Hm, no one?

Is it you find my question stupid anyway?

Kinda. To get comparable results one should have your exact files. That's why we use benchmarks.

Doward
May 31, 2013, 09:24 AM
Please keep in mind that the PCI bus max is usually 133MB/s, and is shared with other subsystems.

Copying from one drive to another, I'm unsurprised by that 33MB/s average - after all, you are both reading AND writing at the same time on the PCI bus.

Cox Orange
May 31, 2013, 10:23 AM
In a few months I'll be buying myself an SATA PCI card for my MDD and I'll add my benchmarks here
Thanks!
Here's my benchmarks, the SATA SSD has a disappointingly small performance lead.
Can you redoo the test like I had described?
(...)
HOWEVER, read and write speeds can vary based on the size of the file. Hard disks excel at large files.
You are right, but I tested 5GB, 20GB and 800GB not much of a difference.
Here are the benchmarks for my Vertex 2 fw1.37
Sonnet Tempo Sata PCI , fw5.1.3

Powermac G4 GE fw4.2.8
Sonnet 1.2GHz CPU
2GB PC133
(...)
Do you have a second drive, to do the test like I described in my first post? If not, you could copy a file one that one Drive, just out of curiosity for me, perhaps?
I have a Sonnet Tempo two port SATA card in my MDD, and can not provide you with a "copy from one disk to another" speed, as they are in (soft)RAID.

What I can say, is that when transferring files over the (gigabit) network to and from that RAID stack, the transfer speeds are within the 60-80 MBps bracket.

The disks are 2 TB Seagate Barracuda Green's (those weird drives with an rpm of 5900 - that's not a typo).

RGDS,
OK, that must be with a really good NAS and a Gigabit-ethernet-PCI card, I guess? I heard, that the on board Gigabit ethernet port of the G4s is slow and that the consumer NASes are slow, too.
Do you have energy saving options disabled? I heard that one can have problems with green drives. If they go to sleep, they can not reactivated or they disappear from Finder. It may depend on the SATA-card used... Did you experience something like that?
Kinda. To get comparable results one should have your exact files. That's why we use benchmarks.
I would upload it, if I had the possibility to. But what does a benchmark tell you, if daily tasks, real life tasks, are different from what the benchmark does when testing.
I mean either way, in a G4, one will copy files on one disk or he will copy files from one disk to another, if it is about bigger files. He will do this on a) onboard ATA, b) SATA-PCI-card, c) via ethernet, d) Firewire. Since a + d make no sense, if you want to know something about a fast Drive and its transfer speeds, it leaves us with b + c.

With small files you probably will not notice anything anyway, so a test which wants to test something that is perceivable to the human eye, will have bigger files involved (or a big number of small files, if you want to test access time).

Please keep in mind that the PCI bus max is usually 133MB/s, and is shared with other subsystems.

Copying from one drive to another, I'm unsurprised by that 33MB/s average - after all, you are both reading AND writing at the same time on the PCI bus.
Yes, but I guess, that is what we do anyway, if we write a file, we must read and write it from somewhere at the same time. (Reading+writing on only one drive should be very slow, too. But the Benchmark-utility gets higher benchmarks, as if you do that with a small file on you own, manually).
(Someone in another forum mailed me, he had an SATA-SSD in a Quicksilver and got 30-40MB, when copying a movie file on that same disk, not to another.)

Since it would make no sense to read/write between ATA-Bus and SATA-PCI-Bus, I was argumenting testing it with two SATA-drives on the presumably faster PCI-Bus.
I also have two SATA-PCI-cards, but since all PCI-buses share the total bandwidth, I guess I would not be better of, doing the test with, too SATA-cards. (I did, as far as I remember, but since there was no difference, I proceeded only with one card and different drives).

-----------------------------------------------------------------------------
Reorganizing my mind:

presumption A: you will want to test something that is relevant to your everyday experience.
presumption B: you will choose the fastest interface available for your test (which leaves us to SATA-PCI or Gigabit-ethernet, when using an ATA-Bus-PowerMac.)

presumption C: there are three scenarios of using files.
a) file read/written between two drives on same controller (e.g. 2xSATA-PCI-CARD, or - "or" in excluding meaning - ATA-Master and Slave.
b) file read/written on one drive, from one part of its surface to a neighbouring part. (one drive!)
c) file read/written between two drives, which are on separate BUSes. (e.g. from onboard ATA to SATA-PCI-Card)

conclusion:
C-c) makes no sense, since it involves a known slow Bus
if A and B then C-a or C-b with SATA-PCI-Card or Gigabit-ethernet.

Letting PCI and Gigabit-ethernet out of sight, what now limits it, is the FSB involved, right? (Real question).

What I admit:
- what I elaborated above does not come into account, when just accessing files (like opening a folder). Either way you will not notice a difference or you will have an SSD (or a SATA-Drive hooked up via SATA-IDE adapter) and see, that e.g. browsing your folders has become quicker.