Small Dog is giving away a G5 Power Mac

BARE FEATS - real world Mac speed test lab

Xserve Solo vs XServe Dual vs Power Mac Dual

Posted 8/10/02 by rob-ART morgan, mad scientist.
Updated 8/21/02 with commentary on the XINET tests.
Mahalo to my fellow mad scientists, Brendan and Greg, for the use of their Xserve's for testing. Thanks also to
TransIntl.com for shipping DDR memory to me overnight.

I was curious how the Xserve with its DDR RAM sporting 2.1GB/s memory throughput would fair running non-server applications. Like, what if I wanted to use the Xserve as a number cruncher or several rack mounted Xserve's as a rendering farm? Is it worth the extra money?

I tested both the dual and single CPU Xserve models. Check it out...

 

 

 

 

 

 

 

 

 

CONCLUSION

After all the hype touting the Xserve's "phenomenal processing power," I was disappoined in the performance. The more expensive Double Data Rate Memory gives it no measureable advantage. I guess memory with 2GB/s throughput won't help you if the system bus is limited to 1GB/s throughput... just like the Power Mac.

Where's the bottleneck? Here's some theories proposed by readers:

1. The two processors share a 1GB/s pipe to Apple's custom AGP/Memory controller. Since the processors can only be fed data at 1GB/s, having memory with a 2GB/s data rate is of no use.

2. The advantages of the Xserve with DDR memory will show only if you load up all the buses (CPU, Memory, PCI, FireWire, AGP, etc.). (Though all the tests apps we used exercise the most of the PCI and AGP buses to some degree, they are primarily CPU and Memory intensive. So we are running several new tests in an attempt to address this concern. -- rob-ART)

What will it take to fix the bottleneck?

1. A CPU than can handle the full speed of DDR memory (like the fabled PPC 7470).
2. A redesigned motherboard with separate bus for each CPU.

Some readers point to the XINET tests showing the Xerver significantly faster than the Power Mac "simularly" configured. However, after careful study of their results, I have concluded that there were two signficant differences in the configurations:

1. The RAID array on the Power Mac topped out at 30MB/sec WRITE speed while the RAID on the Xserve topped out at 55Mb/sec. That doesn't make sense because I have a dual drive ATA RAID in the lab that consistantly measures 90MB/sec. So the RAID array used on the Power Mac is suspect.

2. The number of concurrent print jobs supported by the Xserve was double that of the Power Mac. But keep in mind that the printers were connected to Ethernet ports. The Xserve has two Gigabit Ethernet channels/ports. The Power Mac has one. You could always add a second Gigabit Ethernet card to the Power Mac if you want to match the Xserve's configuration.

At any rate, if you plan to use the Xserve for any purpose other than as a server, it's not worth the extra $1000 to $1500.

As for the 100MB/sec ATA drive bus, the 120GXP that comes standard ran no faster than the same drive in a Power Mac with a 66MB/sec drive bus. That's because no single Utra ATA drive will go faster than 50MB/sec. Duh.

As you can see from this table below, the Xserve does have its advantages:

Power Mac
Xserve

Max CPU Speed

1GHz *2
1GHz *2

System Bus Speed

1GB/s
1GB/s

Memory Bus Speed

1GB/s
2GB/s (DDR)

Maximum Memory

1.5GB
2.0GB

L3 Cache

4MB
4MB

L3 Cache Throughput

4GB/s
4GB/s

Standard Graphics Card

GeForce4 MX AGP (64MB DDR)
ATI (Radeon 7000?) PCI (32MB DDR) *

Optional Graphics

GeForce4 Titanium AGP
Radeon 8500 AGP (Apple) or any AGP card you can obtain

ATA drive bus speed

66MB/s
100MB/s

ATA drive connections

2
4
w/ SMART support

full length PCI slots

4
2

PCI bus speed

33MHz
66MHz

FireWire Ports

2
3

Gigabit Ethernet Ports

1
2 *

Price for Dual 1GHz Base Model

$2999
$3999

(* Combo PCI/AGP short slot can be used for AGP graphics card or Gigabit Ethernet card but for both.)

Xserve can be rack mounted. If not , it takes up a lot of desk space. The dual "horn" fans are very noisy, if you plan to use it in a normal office environment. The four drive bays are easier to access than those of the G4 tower.

 

RELATED XSERVE LINKS

Apple has their own Xserve performance page.

Apple's Xserve specs

According to the testing done by Xinet, the Xserve is faster than a similarly configured Dual 1GHz G4 Power Mac doing server duties. But I take issue with the "matching" configurations. (See comments above.)

Xserve compared to the new DDR Power Mac G4/1GHz MP.

Article on why Apple will NOT likely switch to Intel or AMD processors.

 

TEST NOTES

Each system was running OS X (10.1.5).
The Dual G4 Power Mac and Dual G4 Xserve had 1.5GB of memory.
The Single G4 Xserve had 768MB of memory.
(Extra PC2100 DDR memory ordered from
Trans International.)

For details on each real world test, read "HOW I TEST."










© 2002 Rob Art Morgan
"BARE facts on Macintosh speed FEATS"
Email webmaster at
rob-art@barefeats.com