G5 2.7GHz

[kaye]

Has anybody run Motion 2 on a G5? I was apprehensive after reading all the tales of horror on the Apple Support forums but most of those seemed to be related to not having religiously repaired permissions, or not having done probably ever an Open Firmware reset, or upgrading from Motion to motion 2, or finally that Apple still has some work to do. So I bit the bullet and purchased it. And I am impressed with its capabilities as well as, so far, it works and works fast.

At Rob-ART's suggestion, I went to Motion Mark Benchmarks and followed the instructions for both tests. The RAM Preview Test average was 18.88 seconds.

The 100 Frame Export Render Test is puzzling. All of the listed results are Motion 1 results in Panther and the best is about 18 minutes. My results with Motion 2 and Tiger is 2 minutes 14 seconds averaged for 6 runs with a reboot between each. I have emailed the author but it looks like Motion 2 is way faster than Motion 1 for this test. Anybody with Motion 2, a G5, running 10.4.2 see if you see the same? k

[kaye]

Finally dug out my 4x Cheetah X15.3 Burly to go with a 4x Raptor74 Burly and a 4x Maxtor Atlas 15K II Burly.

4x X15.3 on UL4D, 100MHz slot.
4x Raptor74 on SeriTek/1VE4, 100MHz slot.
4x Atlas 15K II on UL4D, 133MHz slot.
Striped all 12 drives, 8 channels, with SR 3.2.1, 128KB SU.

QB 2.1 report:
QuickBenchª 2.1 Test Results File
©2000-2005 Intech Software Corp.
Test file created on Thursday, October 20, 2005 at 10:25:23 AM
Test Volume name: 12xSCSIsataSCSIsr

Xfer Size Sequential Read Sequential Write Random Read Random Write

1 KByte 1.089 MB/sec 156.592 KB/sec 235.878 KB/sec 277.044 KB/sec
2 KBytes 6.689 MB/sec 330.394 KB/sec 2.155 MB/sec 475.494 KB/sec
4 KBytes 4.624 MB/sec 1.131 MB/sec 5.885 MB/sec 1.158 MB/sec
8 KBytes 8.830 MB/sec 2.156 MB/sec 19.195 MB/sec 2.307 MB/sec
16 KBytes 17.845 MB/sec 4.302 MB/sec 23.328 MB/sec 4.610 MB/sec
32 KBytes 30.290 MB/sec 7.072 MB/sec 10.648 MB/sec 9.253 MB/sec
64 KBytes 22.483 MB/sec 14.240 MB/sec 49.544 MB/sec 18.576 MB/sec
128 KBytes 63.789 MB/sec 32.178 MB/sec 66.906 MB/sec 33.221 MB/sec
256 KBytes 78.832 MB/sec 58.597 MB/sec 53.481 MB/sec 56.365 MB/sec
512 KBytes 225.225 MB/sec 107.172 MB/sec 98.082 MB/sec 120.799 MB/sec
1024 KBytes 291.834 MB/sec 170.710 MB/sec 213.424 MB/sec 174.557 MB/sec

Extended Test Size: 20 MB Read: 461.159 MB/sec Write: 740.384 MB/sec
Extended Test Size: 30 MB Read: 606.772 MB/sec Write: 748.951 MB/sec
Extended Test Size: 40 MB Read: 635.950 MB/sec Write: 759.388 MB/sec
Extended Test Size: 50 MB Read: 672.938 MB/sec Write: 749.142 MB/sec
Extended Test Size: 60 MB Read: 665.816 MB/sec Write: 750.826 MB/sec
Extended Test Size: 70 MB Read: 669.357 MB/sec Write: 759.137 MB/sec
Extended Test Size: 80 MB Read: 672.959 MB/sec Write: 753.267 MB/sec
Extended Test Size: 90 MB Read: 645.295 MB/sec Write: 765.013 MB/sec
Extended Test Size: 100 MB Read: 689.803 MB/sec Write: 762.125 MB/sec

[Nicolas]

over 750MB/s


How much more than the Mac does this setup cost.

Seriously, great to see such speeds on a Mac.

Regards

Nicolas

[kaye]

How much more than the Mac does this setup cost.

2x UL4Ds at the peak of their price
4x Raptor74s, three at the peak
4x Cheetah X15.3s at the peak
4x Atlas 15K IIs at the peak
k

[kaye]

I ran the 12x RAID in ZoneBench this morning

Transfer Size: 511 Megabytes is the largest size ZoneBench will test.

[kaye]

Ran the OpenGL Extensions Viewer Test mentioned today at xlr8yourmac today on my G5 with ATI X800XT. The Viewer says I have 85 OpenGL extensions.

First test default:

Multisample added:

Now all:

k

[Nicolas]

2x UL4Ds at the peak of their price
4x Raptor74s, three at the peak
4x Cheetah X15.3s at the peak
4x Atlas 15K IIs at the peak
k

Kaye,

bit scary isn't it?

Man, we are truly "storage speed junkies"

Best regards

Nicolas

[kaye]

Ain't it the truth Nicolas. k

[rwm]

Big laugh here.

[TZ]

Maxtor sold off all their older models, and the price on the 15K II really did plunge, not just trickle down, to $189 each now. Doubt it will improve. The original USA price was $280 ($259 @ Zipzoomfly). So if you had waited three months you saved, and had you waited another three months (mid-October) you saved another $30 or not quite enough to buy one "spare" drive.

But, while worried at first over the "uh oh, Quad G5" I think having the last PCI-X 2.7DP was godsend, now "only" $2799.

Around the next corner? Nah.

The Hitachi 7K500 is fast, big, 16MB cache, and equal of the Raptor in most ways, especially the first 100GB. Just $380!

[kaye]

Now that 10.4.3 is loaded I reran the previous tests and all the same, Cinebench being 1-2 points lower/higher or the same for its various tests.

Also ran Xbench 1.2 and it worked just fine 162.41 with a RAM Disk. k

[TZ]

When trying to decide what to get, a good look at the options.
http://www.barefeats.com/dc20.html

Benchmarks Include information on video, RAM, cpu intensive tasks.

Here is Dual 2.7 Cinebench score:

CINEBENCH 2003 v1
************************************************** **
Processor : PowerMac G5 Dual
MHz : 2.7
Number of CPUs : 2
Operating System : 10.4.2

Graphics Card : ATI 9650 256MB
Resolution : 1680x1050
Color Depth : millions
************************************************** **
Rendering (Single CPU): 385 CB-CPU
Rendering (Multiple CPU): 663 CB-CPU

Multiprocessor Speedup: 1.72

Shading (CINEMA 4D) : 359 CB-GFX
Shading (OpenGL Software Lighting) : 1028 CB-GFX
Shading (OpenGL Hardware Lighting) : 1736 CB-GFX

OpenGL Speedup: 4.83
************************************************** **

Here is Quad G5 Cinebench score:

CINEBENCH 2003 v1
************************************************** **

Processor : Quad G5
MHz : 2.5 GHZ
Number of CPUs : 4
Operating System : 10.4.2

Graphics Card : GeForce 6600
Resolution :
Color Depth :
************************************************** **
Rendering (Single CPU): 359 CB-CPU
Rendering (Multiple CPU): 1016 CB-CPU

Multiprocessor Speedup: 2.83

Shading (CINEMA 4D) : 353 CB-GFX
Shading (OpenGL Software Lighting) : 1051 CB-GFX
Shading (OpenGL Hardware Lighting) : 1871 CB-GFX

OpenGL Speedup: 5.29
************************************************** **

Would have bee nice to see with X850 used though.
Also worth mention: Barefeats: Dual-Core vs. Quad-Core G5

[kaye]

I didn't keep my notes for my last Cinebench 10.4.3 tests. Notes went to the landfill yesterday. But here are my earlier test results to compare to the 2.5-Quad.

CINEBENCH 2003 v1
************************************************** **
Tester : Dual kyum/Quad Barefeats

Processor : G5-2.7 DP/Quad G5
MHz : 2700/2500
Number of CPUs : 2/4
Operating System : 10.4.1/10.4.2

Graphics Card : ATI X800XT ME/GeForce 6600
Resolution : <1920x1200>/unknown
Color Depth : < Millions >/unknown
************************************************** **
Rendering (Single CPU): 386 CB-CPU/359 CB-CPU
Rendering (Multiple CPU): 701 CB-CPU/1016 CB-CPU

Multiprocessor Speedup: 1.82/2.83

Shading (CINEMA 4D) : 364 CB-GFX/353 CB-GFX
Shading (OpenGL Software Lighting) : 1071 CB-GFX/1051 CB-GFX
Shading (OpenGL Hardware Lighting) : 1833 CB-GFX/1871 CB-GFX

OpenGL Speedup: 5.04/5.29
************************************************** **
Maybe the amount of memory in my 2.7 G5 and the video card are causing better scores than Barefeats 2.7 G5 results. Certainly the video card. k

[kaye]

The 12x FrankenRAID is faster now.
4x X15.3 on UL4D, 100MHz slot.
4x Raptor74 on SeriTek/1VE4, 100MHz slot.
4x Atlas 15K II on UL4D, 133MHz slot.
Striped all 12 drives, 8 channels, with SR 3.3, Workstation, 128KB SU, 10.4.3.
Current test 10.4.3/SoftRAID 3.3:
Extended Test Size: 20 MB Read: 480.561 MB/sec Write: 760.774 MB/sec
Extended Test Size: 30 MB Read: 632.498 MB/sec Write: 751.371 MB/sec
Extended Test Size: 40 MB Read: 669.154 MB/sec Write: 746.575 MB/sec
Extended Test Size: 50 MB Read: 683.359 MB/sec Write: 749.389 MB/sec
Extended Test Size: 60 MB Read: 689.394 MB/sec Write: 757.060 MB/sec
Extended Test Size: 70 MB Read: 690.465 MB/sec Write: 751.759 MB/sec
Extended Test Size: 80 MB Read: 679.937 MB/sec Write: 746.986 MB/sec
Extended Test Size: 90 MB Read: 673.063 MB/sec Write: 766.016 MB/sec
Extended Test Size: 100 MB Read: 697.175 MB/sec Write: 767.790 MB/sec

Previous test 10.4.2/SoftRAID 3.2.1:
Extended Test Size: 20 MB Read: 461.159 MB/sec Write: 740.384 MB/sec
Extended Test Size: 30 MB Read: 606.772 MB/sec Write: 748.951 MB/sec
Extended Test Size: 40 MB Read: 635.950 MB/sec Write: 759.388 MB/sec
Extended Test Size: 50 MB Read: 672.938 MB/sec Write: 749.142 MB/sec
Extended Test Size: 60 MB Read: 665.816 MB/sec Write: 750.826 MB/sec
Extended Test Size: 70 MB Read: 669.357 MB/sec Write: 759.137 MB/sec
Extended Test Size: 80 MB Read: 672.959 MB/sec Write: 753.267 MB/sec
Extended Test Size: 90 MB Read: 645.295 MB/sec Write: 765.013 MB/sec
Extended Test Size: 100 MB Read: 689.803 MB/sec Write: 762.125 MB/sec

Current ZoneBench:

Previous ZoneBench:

Not only is the current slightly faster but more important is the near elimination of the yellow portion of the read/write bars which indicates that the Ave Read/Ave Write are much higher. k

[kaye]

When I purchased the iSkin keyboard protector for my iBook, I also purchased one for my G5 keyboard http://www.iskin.com/protouch_xt.html

I never liked the G5 white keyboard because it looked as if it would look dirty very quickly. The iSkin solves that problem completely. And, as the iSkin page sez, it does not work on earlier keyboards because of different spacing of the F keys. So My MDD and QS keyboards will not work with it. k

[kaye]

Finally got around to testing my DSL via the groundwork done by CB, Nicolas, and TZ here http://macgurus.com/forums/showthread.php?t=20856

1. Checked and wrote down the IP Address of my G5 in Tiger (10.4.3) Prefs>Network>Configure>TCP/IP
2. Then clicked Ethernet and changed Configure: Automaticallly to Manually (Advanced).
3. Left Speed: autoselect
4. Duplex: auto (grayed out)
5. Maximum Packet Size (MTU): Standard (1500) button on but I saw another button Custom: with a window set at 1500 which would be where I would later set 1492.
6. Clicked Apply Now to exit.

Went to http://www.dslreports.com/stest
and selected MegaPath Test speed using our server at SF, CA, USA
traffic: was very low.

With MTU set 1500, test results were:
Your download speed : 1255 kbps or 156.9 KB/sec.
Your upload speed : 244 kbps or 30.6 KB/sec.

Going back to Step 5 above and resetting 1500 to 1492 and rebooting, the test results were:
Your download speed : 1256 kbps or 157 KB/sec.
Your upload speed : 271 kbps or 33.8 KB/sec.

Not much change in download speed but a nice boost in upload speed. There is always some loss from your rated connection such as how far your DSL signal has to go from your local ISP, I can throw a stone and just about hit the building, but the connection goes from the building to downtown Concord and back to me. That trick put me outside of their max allowable distance because they came out and checked that. That is until they installed a booster on a telephone pole somewhere and the technician rerouted my line underground thru an apple orchard where they had rights and then above ground to that booster. Finally, in the box under my street he put a note on my connection to not install a splitter on this line. The ISP would not do this work but the technician was more than happy to do so and worked many long hours into the night, about 12 hours on my job. So line loss, noise, TCP overhead, and on and on.

On to the Nicolas post on pinging an address with different MTU values from the Terminal. I chose to ping 10 times with a yield of averages. Some revealing info comparing G5 with 7 foot shielded CAT 5 enhanced (CAT5e) 568B 350MHz cable to router. Router shows 100Base-TX connection LED and Full Duplex LED. With G5 shutdown, iBook showed the same info on the router, just a different port.

with iBook shutdown, G5 in Terminal, just start typing in either
ping -c 10 -s 1492 xxx.xxx.xxx.xxx
where the x's are your IP Address
or
ping -c 10 -s 1500 xxx.xxx.xxx.xxx
Other computers on the same LAN will have a different IP address to type in. With my G5 sutdown, did the iBook. Comparing results:
MTU 1500= 10 packets transmitted, 10 packets received, 0% packet loss
G5=round-trip min/avg/max/stddev = 0.065/0.079/0.114/0.012 ms, iBook= 0.126/0.130/0.141/0.004 ms
MTU 1492= 10 packets transmitted, 10 packets received, 0% packet loss
G5=round-trip min/avg/max/stddev = 0.062/0.077/0.092/0.008 ms, iBook= 0.123/0.128/0.134/0.004 ms
Comparing G5/1500 to G5/1492 (and G4/1500 to G4/1492), both see a nice boost. But comparing G5 to G4, what's with that? As long as the pipe is big enough, why does the G4 underperform when electricity travels at the speed of light or about 186,000 miles per second? The answer is that it does but the G4 goes thru a 114 foot shielded CAT 5 enhanced (CAT5e) 568B 350MHz cable to router and a 568B jack in the kitchen, 100 feet router to jack and 14 foot jack to iBook. That 114 feet and the fact that milliseconds (One thousandth (10-3) of a second) are being measured may be part of the difference. My son plugged in another 100 foot cable to the kitchen jack and measured significantly worse ping results in ms but no errors. So the CAT 5 spec of max 100 meters or 300 feet plus will just about guarantee no errors but not speed, tho I can't tell the difference when running the dslreports test on the iBook.

Both G5 and iBook yield a Shields UP first 1056 ports (All Service Ports) perfect "TruStealth" rating. https://www.grc.com/x/ne.dll?bh0bkyd2 k

[TZ]

kaye,

searching for the latest information on G5 with CS2, gobs of RAM, and fast scratch disk

[kaye]

TZ, I've got the data with 8GB RAM and 12x FrankenRAID sort of outlined in the Tribal but will need to put it all together for posting on the Sample PshopTest Results. May be a day or so to get it all together. k

[Nicolas]

Hello Kaye,

lower MTU size can improve web browsing while larger MTU is better for networking like copying files from one to another Mac on your network.
But, if you are using a router 1500 or 1492 should be OK.
You can test 1472 and 1424 too if you like.

On thing you can do is tuning the send/receive buffers a little using SystemOptimizer or Cocktail.

Set in and out to 65536 this will help you to get more throughput.

EDIT: Kaye, turn off IPv6 in systemprefs>network if it is on.

Best regards

Nicolas

[TZ]

The one place I didn't look, and that is where you posted results...