Notable upgrades for Digital Audio G4

Howdy fellow PPC users!

Last year I did a whole bunch of upgrading on my Digital Audio G4, and though I had a lot of fun doing it, the time and expense yielded much lesser results than I had hoped for. My goal was to cut down rendering time in FCE4, and it worked, but only to a small degree. It wasn't worth what was put into it, so I won't go into great detail about all that was done, but one upgrade of note that was especially fun was with the GPU. I wanted something more powerful than what I had (GE Force FX5200 IIRC), but options were limited until I came across info about taping pins 3 and 11 for modding a G5 AGP GPU to work in a G4.

I'd like to mention two guys who were a tremendous help to me with this upgrade: Christian at old-macs.com and Donald at atxg4.com. Christian helped me pick out a GPU that would work in my machine (the EVGA nVidia 7800 GS CO), and helped me with a LOT of other stuff like info for picking the right PSU to power the GPU. We emailed back and forth constantly for more than two weeks. He was extremely helpful (and patient), before, during and after my purchase. The man really went WELL above and beyond to help. If you're in the market for a GPU, I highly recommend him.

Which brings me to Donald, who was kind enough to answer my plea for help when I realized (I was on a learning curve with all of this) that standard PC ATX power connectors and Apple's version of ATX were two different animals. At the time the info on his website said that he wasn't taking orders for his G4 ATX adapters at all, but when I contacted him through the website he replied and hooked me up with one! He was very easy to work with and expedited the shipment. Just FYI, he has since updated his info and is taking orders again (but note he's not using Paypal). So for anyone who needs to or is otherwise interested in changing out their PSU, he can help with that if your G4 model is one of the ones he's developed an adapter for. My adapter works great, by the way.

Incidentally, for my PSU I chose the OCZ Fatal1ty 750-watt unit. It sits on top of the computer (too big to go inside). I went this route because the 7800 GS needed molex power on a dedicated +12V rail with a particular amperage rating (18A IIRC). As a side benefit, the PSU itself is modular.

Some other notable upgrades I did at the time (yes, these were worth it too): I also bought an OWC SSD for my boot drive (installed in the Zip drive bay), and 3 sticks of OWC "high-performance" CL2 RAM to replace my CL3 RAM (in case anyone has ever wondered, that CL2 RAM does make a difference - not a significant difference mind you, but I noticed an immediate difference after booting up and everything was just zippier overall). I also overclocked my aftermarket processor upgrade, for which I mounted 3 extra fans on the heatsink to get more air on and underneath so that I could run steady at dual 1.8 gHz (instead of dual 1.6 which is how it was configured when I bought it back in '06). I switched it at one point to dual 2.0 gHz, but afterward About This Mac reported only dual 1.93 gHz, and it was unstable when playing movies - it would freeze every few minutes. So, I switched it back to dual 1.8. It runs great at 1.8 - rock solid actually.

As an aside, one thing I tried to do (unsuccessfully) earlier this year was to replace my ATA optical drive with SATA optical drive hooked up via a PCI SATA card like my other drives. Alas, after much investigation, the only real possibility I found was the FirmTek SeriTek/1V4, which is advertised on FirmTek's website as having ATAPI device support. But, the most I could get out of it when I tried to use it with a SATA optical drive (this one and one other LG model) was that it responded to the eject key on my keyboard, one time. It did not respond that way again, nor did it mount a CD or DVD on the desktop. It seems that PCI SATA cards for Mac that fully support SATA ATAPI devices don't exist.

Cox Orange said:

Hey,

thanks for your report. Since you have a SATA-PCI card, may I ask you to copy a bigger file (like 5 to 8GB big) from one SATA-Drive to another via the SATA-PCI card and stop the time with a watch?

Please do not use a benchmark tool, the reason of my question is to find out, what you get, when you actually stop the time it takes for copying the file in reality. A DV-file (from iMovie) would be great, but you can take other files or a folder that contains big files that make up 5-8GB.

I would thank you a lot!

Cox.

Click to expand...

Sure thing, but best I can do is a 28GB file. It's a .mov file that I saved without compression. I don't have anything in the 5-8GB range - all my video-related files are much smaller than that. So I'll do a test copy of the 28GB file and post the result and let you do the math. Sounds good?

Results

In Finder (that is, on my desktop), I copied a 28.07GB file from one standard HDD to another (also across two PCI SATA cards). It took 10 min 37 sec which is 10.62 min.

28GB / 10.62 = 2.64GB per min.

8GB / 2.64GB = 3.03 min.

So, slightly over 3 min. to copy 8GB on PCI SATA card (Sonnet Tempo SATA I, 1.5 Gb/s)

Hope that helps!

----------

Forgot to say I used the stopwatch feature on my phone to time the copy.

Cox Orange said:

Thank you very much! That confirms, what I have seen!
So it is about 45MB/s.

Many people on the net tell how fast their drives are and especially how much they profit from SSDs or modern HDDs. They use Benchmark Tools and get 80-100MB/s transfer rate via SATA-PCI in a G4 PowerMac.
Since I allready learned that Xbench is crap, I began to doubt these numbers of other Benchmark Tools as well and tested my drives (I have a big collection of different Drives from different ages and spindle speeds etc.).
On my drives I had 33-37MB/s. I tested 5400rpm drives as well as 10k Velociraptors.

I once started a thread and hoped that more people would do, what I asked you, but the result was that everybody said, they do not see the reason, because they think their benchmark Tools are precise enough. Yes, I admit we would have had used the same file and so on and we need more data, but I think this gives us a hint, that Benchmark Tools are useless or at least doubtful and that even puts purchasing a SSD or faster SATA-Drives into a new perspective. It seems our SATA-PCI-cards can't utilize the whole technically given transfer speed.

I, btw, have a Sonnet 2-port-SATA-PCI card (which has the Firmtek Firmware) and a Macally SUA-100e (2 SATA, 1 external SATA, 1 IDE, to PCI), in an AGP-Sawtooth (with 100MHz Bus). I once had a ACARD SATA-PCI 2port RAID card, too, but since I wanted to have money to spend on other Mac parts and didn't see the need for 6 SATA-drives in my G4, I somehow made the decision to sell the ACARD, not really based on information or knowledge (just thinking Sonnet "might" be better and the Macally one has the IDE-133 port as an extra, and I didn't need RAID abilities).

Click to expand...

No problem, glad to help! I have the Sonnet (Tempo) PCI SATA card also - 2 of them, and they work great. As for SSDs, the one I use for my boot drive really made a difference in overall performance (zippiness of things opening/loading/boot time etc.).

With PCI I reckon the limiting factor of transfer speed would be the PCI bus itself, especially since a SATA II SSD isn't the one "holding up the show" at that point, so to speak. It does seem odd though that the transfer rate didn't get above 45MB/s, even with standard HDDs. Maybe because I copied the file across cards...? I don't see it making THAT big of a difference.

Cox Orange said:

it could have something to do with the fact, that your G4 has a 133MHz BUS and I have 100MHz, and hence you get 45MB/s vs my 37MB/s.

Click to expand...

Sounds logical to me.

Overall, and from a troubleshooting/process of elimination standpoint, what variables are there? The PCI card, the HDDs and the PCI bus, right?

So, if the PCI SATA cards support up to 1.5 Gb/s (or about 150 MB/s), and if the standard HDDs themselves can transfer up to about 150 MB/s, then it seems that the only limiting factor would have to be the PCI bus itself. Specifically what, I'm not sure.

Hmm. That being the case, it seems that either the HDDs or the PCI cards, or something as yet unknown, is the cause of the slower transfer rate.

----------

Cox Orange said:

Thank you very much! That confirms, what I have seen!
So it is about 45MB/s.

Many people on the net tell how fast their drives are and especially how much they profit from SSDs or modern HDDs. They use Benchmark Tools and get 80-100MB/s transfer rate via SATA-PCI in a G4 PowerMac.
Since I allready learned that Xbench is crap, I began to doubt these numbers of other Benchmark Tools as well and tested my drives (I have a big collection of different Drives from different ages and spindle speeds etc.).
On my drives I had 33-37MB/s. I tested 5400rpm drives as well as 10k Velociraptors.

I once started a thread and hoped that more people would do, what I asked you, but the result was that everybody said, they do not see the reason, because they think their benchmark Tools are precise enough. Yes, I admit we would have had used the same file and so on and we need more data, but I think this gives us a hint, that Benchmark Tools are useless or at least doubtful and that even puts purchasing a SSD or faster SATA-Drives into a new perspective. It seems our SATA-PCI-cards can't utilize the whole technically given transfer speed.

I, btw, have a Sonnet 2-port-SATA-PCI card (which has the Firmtek Firmware) and a Macally SUA-100e (2 SATA, 1 external SATA, 1 IDE, to PCI), in an AGP-Sawtooth (with 100MHz Bus). I once had a ACARD SATA-PCI 2port RAID card, too, but since I wanted to have money to spend on other Mac parts and didn't see the need for 6 SATA-drives in my G4, I somehow made the decision to sell the ACARD, not really based on information or knowledge (just thinking Sonnet "might" be better and the Macally one has the IDE-133 port as an extra, and I didn't need RAID abilities).

Click to expand...

In your testing, did you at any point use SSDs to time the copy? I no longer have a second SSD to copy to/from, but I could copy a large file to my SSD, duplicate the file, and time the duplication. Just trying to eliminate variables (in this case, whether or not standard HDDs are a significant factor).

If that doesn't yield better results, maybe the Firmtek firmware on the Sonnet cards is a contributing factor...? Both of mine have ROM version 5.3.1.1S2, and Firmtek Tech Support has said that version is not ideal for Leopard (which I am running), though I haven't had any known issues with it.

Anyway, I can try timing the duplication of a file if you're interested.

Cox Orange said:

If you don't mind dublicating a file on your SSD, do it. Would be nice!

Click to expand...

Okay, I created a test TIFF file in Photoshop and saved to my SSD. File size in Finder is 3.35GB. Then duplicated the file and it took 1 min 8 sec which is 1.13 minutes, so:

3.35GB / 1.13 min = 2.965GB/min

2.965 / 60 sec = 49.42 MB/s

I'm beginning to wonder what's going on too! Why is the real time data transfer rate so much slower than spec if the PCI SATA card, the PCI bus and the HDDs (and SSDs) can all move faster than 49.42 MB/s?

And, like you mentioned, I opened up a previous test I did in xbench. Sure enough, I see numbers that for the most part are more than 49.42 MB/s - as high as about 114 MB/s.

I wonder if the PCI slot itself matters. For instance, I'm pretty sure my SSD is plugged into the card that occupies PCI slot #2. I wonder if slot #1 would be any faster? I did some testing with this last year and didn't see any significant differences in speed, so maybe that's not it.

Cox Orange said:

Btw., if you like to make some further testing, test Xbench. Do 10-20 tests with the same SSD, etc. just run the test several times, over and over again and save them. I would be interested, how much the results in Xbench themselves differ from one another (find lowest and highest result out of 10 tries for one parameter. Examine all of the 8 test scenarios (with scenarios I mean 4k-blocks write, 256k-blocks write, 4k-blocks read etc.). Thank you!

Click to expand...

Sure I can do that. Will post ASAP (but not tonight!).

Cox Orange said:

Any news?

Click to expand...

Ooooooooo, my sincere apologies. I completely forgot about doing this. Tonight is the first time I've logged in to MR in about 2 months so I just now saw your post. I very recently bought a MP and am using that instead of my DA G4. Does that matter to you for purposes of xBench tests?