[t_hah]

I am curious how the G5 performs...and now that XBench has been optimized for the G5, the resuls might look much better.


t

[krassy]

http://www.xbench.com/

[Shaktai]

It would also be helpful if people ran the latest version of X-Bench on their current machines and submitted those results to the X-Bench site as well.

The differences between v1.0 and v1.1 are significant enough to warrant it in order to have accurate comparisions. I have already done my iBook 600 and PowerMac800. (DWR)

[BZ]

Sawtooth Upgraded to 1 GHz G4/2MB L3, 2 GB Ram, Radeon 8500, WD 120 GB 7200 RPM 8 MB cache Drive:

Results 84.69
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 2048 MB
Model PowerMac3,1
Processor PowerPC G4 @ 1.00 GHz
Version 7455 (Apollo) v2.1
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 500 MHz
L3 Cache 2048K
Bus Frequency 100 MHz
Video Card ATY,R200
Drive Type WDC WD1200JB-75CRA0
CPU Test 60.96
GCD Recursion 115.27 4.50 Mops/sec
Floating Point Basic 125.54 446.48 Mflop/sec
AltiVec Basic 124.70 1.81 Gflop/sec
vecLib FFT 20.40 314.29 Mflop/sec
Floating Point Library 120.10 4.81 Mops/sec
Thread Test 61.89
Computation 61.83 498.13 Kops/sec, 4 threads
Lock Contention 61.94 777.57 Klocks/sec, 4 threads
Memory Test 85.15
System 104.63
Allocate 118.55 39.97 Kalloc/sec
Fill 175.71 1398.67 MB/sec
Copy 68.74 343.71 MB/sec
Stream 71.78
Copy 72.71 296.54 MB/sec [altivec]
Scale 71.60 297.42 MB/sec [altivec]
Add 72.02 303.90 MB/sec [altivec]
Triad 70.82 291.49 MB/sec [altivec]
Quartz Graphics Test 117.73
Line 116.27 2.96 Klines/sec [50% alpha]
Rectangle 112.30 7.90 Krects/sec [50% alpha]
Circle 121.47 2.80 Kcircles/sec [50% alpha]
Bezier 116.25 1.26 Kbeziers/sec [50% alpha]
Text 123.01 2.01 Kchars/sec
OpenGL Graphics Test 95.39
Spinning Squares 95.39 66.75 frames/sec
User Interface Test 107.71
Elements 107.71 36.65 refresh/sec
Disk Test 99.12
Sequential 102.71
Uncached Write 106.65 42.46 MB/sec [4K blocks]
Uncached Write 104.87 40.93 MB/sec [256K blocks]
Uncached Read 100.48 15.91 MB/sec [4K blocks]
Uncached Read 99.20 40.08 MB/sec [256K blocks]
Random 95.77
Uncached Write 101.94 1.46 MB/sec [4K blocks]
Uncached Write 102.54 23.13 MB/sec [256K blocks]
Uncached Read 99.53 0.66 MB/sec [4K blocks]
Uncached Read 82.26 16.93 MB/sec [256K blocks]

[BZ]

And another...

TiG4 500Mhz, 1GB Ram, New 5,200 RPM Drive.

Results 50.09
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 1024 MB
Model PowerBook3,2
Processor PowerPC G4 @ 500 MHz
Version 7410 (Nitro) v1.3
L1 Cache 32K (instruction), 32K (data)
L2 Cache 1024K @ 250 MHz
Bus Frequency 100 MHz
Video Card ATY,RageM3
Drive Type TOSHIBA MK6022GAX
CPU Test 43.21
GCD Recursion 48.83 1.91 Mops/sec
Floating Point Basic 76.40 271.69 Mflop/sec
AltiVec Basic 62.68 910.37 Mflop/sec
vecLib FFT 17.98 277.07 Mflop/sec
Floating Point Library 94.58 3.79 Mops/sec
Thread Test 36.92
Computation 32.14 258.90 Kops/sec, 4 threads
Lock Contention 43.37 544.44 Klocks/sec, 4 threads
Memory Test 80.08
System 74.08
Allocate 182.02 61.37 Kalloc/sec
Fill 72.29 575.44 MB/sec
Copy 47.23 236.16 MB/sec
Stream 87.15
Copy 85.96 350.55 MB/sec [altivec]
Scale 85.19 353.90 MB/sec [altivec]
Add 88.24 372.38 MB/sec [altivec]
Triad 89.33 367.68 MB/sec [altivec]
Quartz Graphics Test 62.00
Line 57.62 1.47 Klines/sec [50% alpha]
Rectangle 60.38 4.25 Krects/sec [50% alpha]
Circle 69.96 1.61 Kcircles/sec [50% alpha]
Bezier 67.38 732.10 beziers/sec [50% alpha]
Text 56.85 926.71 chars/sec
OpenGL Graphics Test 72.90
Spinning Squares 72.90 51.02 frames/sec
User Interface Test 58.31
Elements 58.31 19.84 refresh/sec
Disk Test 65.41
Sequential 72.91
Uncached Write 72.44 28.84 MB/sec [4K blocks]
Uncached Write 67.97 26.53 MB/sec [256K blocks]
Uncached Read 94.33 14.93 MB/sec [4K blocks]
Uncached Read 63.51 25.66 MB/sec [256K blocks]
Random 59.31
Uncached Write 50.52 0.72 MB/sec [4K blocks]
Uncached Write 58.51 13.20 MB/sec [256K blocks]
Uncached Read 65.14 0.43 MB/sec [4K blocks]
Uncached Read 65.76 13.53 MB/sec [256K blocks]

[Jupeman]

PowerMac 1.42:

Results 139.95
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 1536 MB
Model PowerMac3,6
Processor PowerPC G4x2 @ 1.42 GHz
Version 7455 (Apollo) v3.3
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 1417 MHz
L3 Cache 2048K @ 237 MHz
Bus Frequency 167 MHz
Video Card ATY,RV250
Drive Type ST3120024A
CPU Test 177.41
GCD Recursion 176.72 6.90 Mops/sec
Floating Point Basic 176.85 628.95 Mflop/sec
AltiVec Basic 174.71 2.54 Gflop/sec
vecLib FFT 181.09 2.79 Gflop/sec
Floating Point Library 177.81 7.12 Mops/sec
Thread Test 171.29
Computation 174.00 1.40 Mops/sec, 4 threads
Lock Contention 168.67 2.12 Mlocks/sec, 4 threads
Memory Test 158.89
System 190.52
Allocate 172.19 58.05 Kalloc/sec
Fill 335.46 2670.26 MB/sec
Copy 143.72 718.62 MB/sec
Stream 136.26
Copy 138.21 563.66 MB/sec [altivec]
Scale 140.62 584.15 MB/sec [altivec]
Add 137.36 579.65 MB/sec [altivec]
Triad 129.39 532.53 MB/sec [altivec]
Quartz Graphics Test 156.29
Line 159.68 4.06 Klines/sec [50% alpha]
Rectangle 136.87 9.63 Krects/sec [50% alpha]
Circle 176.39 4.07 Kcircles/sec [50% alpha]
Bezier 173.84 1.89 Kbeziers/sec [50% alpha]
Text 142.84 2.33 Kchars/sec
OpenGL Graphics Test 140.71
Spinning Squares 140.71 98.47 frames/sec
User Interface Test 151.59
Elements 151.59 51.58 refresh/sec
Disk Test 82.33
Sequential 91.08
Uncached Write 84.03 33.45 MB/sec [4K blocks]
Uncached Write 83.34 32.53 MB/sec [256K blocks]
Uncached Read 130.38 20.64 MB/sec [4K blocks]
Uncached Read 81.00 32.73 MB/sec [256K blocks]
Random 75.11
Uncached Write 52.66 0.75 MB/sec [4K blocks]
Uncached Write 80.84 18.23 MB/sec [256K blocks]
Uncached Read 93.47 0.62 MB/sec [4K blocks]
Uncached Read 89.30 18.38 MB/sec [256K blocks]

[Busemann]

None of these are G5's are they

I dunno why people are so obsessed with Xbench results, what matter is the real world apps. Where are the UT 2003, PS 7.0.1, iTunes, FPC 4, etc results!?

btw. I got over 140 on my 1.25Ghz with only 512MB RAM, which goes some way of explaining why a 1.42 got 139..

[Jack The Dog]

Something's fishy. A lot of people are reporting very low vecLib FFT scores in xbench 1.1. I had a very low score my first time testing (<30) but every other time i've run it, it's been 150+. This really affects the CPU and overall scores.

BZ - have you tried retesting?

[Eug Wanker]

Xbench 1.1 Version History [2003-08-23]

* Optimized for the PowerMac G5, compiled with GCC 3.3
* Avoid use of vec_dst on PowerPC 970 for better performance
* Changes to the AltiVec test to reduce dependence on memory bandwidth
* Removed use of vec_dstt
* Changes to floating point test
* Added vecLib FFT test
* Added automatic machine type detection
* Fixed an issue where the results submission sheet caused the results window to become catatonic
* Prettier submit button
* Recalibration of CPU tests, Disk tests
* Fixed spinning rainbow cursor during OpenGL test

Note that overall, my TiBook scores about the same as it used to. Dunno about each individual test though.

TiBook SuperDrive 1 GHz, 768 MB PC133, Radeon 9000, Fujitsu MHS2060AT 4200 rpm drive:

Code:
Results	91.58	
	System Info		
		Xbench Version		1.1
		System Version		10.2.6
		Physical RAM		768 MB
		Model		PowerBook3,5
		Processor		PowerPC G4 @ 1.00 GHz
			Version		7455 (Apollo) v3.2
			L1 Cache		32K (instruction), 32K (data)
			L2 Cache		256K @ 667 MHz
			L3 Cache		1024K @ 134 MHz
			Bus Frequency		134 MHz
		Video Card		ATY,RV250M9
		Drive Type		FUJITSU MHS2060AT
	CPU Test	121.73	
		GCD Recursion	116.80	4.56 Mops/sec
		Floating Point Basic	122.82	436.78 Mflop/sec
		AltiVec Basic	124.98	1.82 Gflop/sec
		vecLib FFT	123.39	1.90 Gflop/sec
		Floating Point Library	121.00	4.84 Mops/sec
	Thread Test	63.89	
		Computation	63.12	508.50 Kops/sec, 4 threads
		Lock Contention	64.68	811.94 Klocks/sec, 4 threads
	Memory Test	104.41	
		System	100.34	
			Allocate	118.34	39.90 Kalloc/sec
			Fill	162.66	1294.78 MB/sec
			Copy	65.35	326.76 MB/sec
		Stream	108.83	
			Copy	107.65	439.03 MB/sec [altivec]
			Scale	107.34	445.92 MB/sec [altivec]
			Add	112.16	473.28 MB/sec [altivec]
			Triad	108.31	445.77 MB/sec [altivec]
	Quartz Graphics Test	125.97	
		Line	128.61	3.27 Klines/sec [50% alpha]
		Rectangle	120.31	8.46 Krects/sec [50% alpha]
		Circle	135.65	3.13 Kcircles/sec [50% alpha]
		Bezier	119.76	1.30 Kbeziers/sec [50% alpha]
		Text	126.88	2.07 Kchars/sec
	OpenGL Graphics Test	96.95	
		Spinning Squares	96.95	67.85 frames/sec
	User Interface Test	109.79	
		Elements	109.79	37.36 refresh/sec
	Disk Test	63.97	
		Sequential	67.11	
			Uncached Write	66.34	26.41 MB/sec [4K blocks]
			Uncached Write	58.66	22.90 MB/sec [256K blocks]
			Uncached Read	93.63	14.82 MB/sec [4K blocks]
			Uncached Read	59.52	24.05 MB/sec [256K blocks]
		Random	61.11	
			Uncached Write	73.96	1.06 MB/sec [4K blocks]
			Uncached Write	55.26	12.46 MB/sec [256K blocks]
			Uncached Read	58.57	0.39 MB/sec [4K blocks]
			Uncached Read	59.67	12.28 MB/sec [256K blocks]

Xbench author's dual G4 800, 1152 MB PC133, Geforce2 MX, WD 1200JB 7200 rpm drive:

Code:
Results	99.68	
	System Info		
		Xbench Version		1.1
		System Version		10.2.6
		Physical RAM		1152 MB
		Model		PowerMac3,5
		Processor		PowerPC G4x2 @ 800 MHz
			Version		7450 (V'ger) v2.1
			L1 Cache		32K (instruction), 32K (data)
			L2 Cache		256K @ 800 MHz
			L3 Cache		2048K @ 200 MHz
			Bus Frequency		134 MHz
		Video Card		GeForce2 MX
		Drive Type		WDC WD1200JB-75CRA0
	CPU Test	100.16	
		GCD Recursion	99.46	3.88 Mops/sec
		Floating Point Basic	100.79	358.46 Mflop/sec
		AltiVec Basic	100.36	1.46 Gflop/sec
		vecLib FFT	100.06	1.54 Gflop/sec
		Floating Point Library	100.12	4.01 Mops/sec
	Thread Test	100.55	
		Computation	100.32	808.22 Kops/sec, 4 threads
		Lock Contention	100.78	1.27 Mlocks/sec, 4 threads
	Memory Test	100.90	
		System	98.71	
			Allocate	98.31	33.15 Kalloc/sec
			Fill	97.82	778.68 MB/sec
			Copy	100.02	500.10 MB/sec
		Stream	103.20	
			Copy	102.34	417.37 MB/sec [altivec]
			Scale	103.23	428.83 MB/sec [altivec]
			Add	103.45	436.53 MB/sec [altivec]
			Triad	103.79	427.18 MB/sec [altivec]
	Quartz Graphics Test	101.66	
		Line	101.94	2.60 Klines/sec [50% alpha]
		Rectangle	100.69	7.08 Krects/sec [50% alpha]
		Circle	102.97	2.37 Kcircles/sec [50% alpha]
		Bezier	100.34	1.09 Kbeziers/sec [50% alpha]
		Text	102.43	1.67 Kchars/sec
	OpenGL Graphics Test	100.16	
		Spinning Squares	100.16	70.09 frames/sec
	User Interface Test	94.07	
		Elements	94.07	32.01 refresh/sec
	Disk Test	100.64	
		Sequential	101.62	
			Uncached Write	103.02	41.01 MB/sec [4K blocks]
			Uncached Write	103.35	40.34 MB/sec [256K blocks]
			Uncached Read	100.24	15.87 MB/sec [4K blocks]
			Uncached Read	99.98	40.40 MB/sec [256K blocks]
		Random	99.67	
			Uncached Write	99.74	1.43 MB/sec [4K blocks]
			Uncached Write	99.83	22.51 MB/sec [256K blocks]
			Uncached Read	100.14	0.66 MB/sec [4K blocks]
			Uncached Read	98.99	20.37 MB/sec [256K blocks]

[osiris]

no laughing pleeeze, at least until a G5 user posts something here:

Results 46.02
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 896 MB
Model PowerMac3,1
Processor PowerPC G4 @ 400 MHz
Version 7400 (Max) v2.6
L1 Cache 32K (instruction), 32K (data)
L2 Cache 1024K @ 200 MHz
Bus Frequency 100 MHz
Video Card ATY,RADEON
Drive Type WDC WD205BA
CPU Test 38.77
GCD Recursion 39.71 1.55 Mops/sec
Floating Point Basic 57.22 203.48 Mflop/sec
AltiVec Basic 49.83 723.71 Mflop/sec
vecLib FFT 18.83 290.20 Mflop/sec
Floating Point Library 76.10 3.05 Mops/sec
Thread Test 29.85
Computation 25.51 205.54 Kops/sec, 4 threads
Lock Contention 35.96 451.34 Klocks/sec, 4 threads
Memory Test 51.50
System 51.47
Allocate 84.97 28.65 Kalloc/sec
Fill 57.44 457.25 MB/sec
Copy 34.36 171.79 MB/sec
Stream 51.52
Copy 57.43 234.21 MB/sec [altivec]
Scale 55.53 230.67 MB/sec [altivec]
Add 44.61 188.26 MB/sec [altivec]
Triad 50.51 207.89 MB/sec [altivec]
Quartz Graphics Test 58.09
Line 56.05 1.43 Klines/sec [50% alpha]
Rectangle 65.68 4.62 Krects/sec [50% alpha]
Circle 65.47 1.51 Kcircles/sec [50% alpha]
Bezier 56.61 615.11 beziers/sec [50% alpha]
Text 49.85 812.49 chars/sec
OpenGL Graphics Test 54.47
Spinning Squares 54.47 38.12 frames/sec
User Interface Test 49.80
Elements 49.80 16.94 refresh/sec
Disk Test 56.32
Sequential 60.03
Uncached Write 56.60 22.53 MB/sec [4K blocks]
Uncached Write 57.39 22.40 MB/sec [256K blocks]
Uncached Read 74.68 11.82 MB/sec [4K blocks]
Uncached Read 55.09 22.26 MB/sec [256K blocks]
Random 53.05
Uncached Write 33.80 0.48 MB/sec [4K blocks]
Uncached Write 59.31 13.38 MB/sec [256K blocks]
Uncached Read 75.47 0.50 MB/sec [4K blocks]
Uncached Read 63.68 13.11 MB/sec [256K blocks]

[krassy]

Originally posted by osiris:
no laughing pleeeze, at least until a G5 user posts something here:

no problem - this is the 400MHz G4 - its speed is ok i think ... at least you will have the pleasure to feel that buying a G5 will be really worth it

[Horsepoo!!!]

Dual 800:

Results 107.60
System Info
Xbench Version 1.1
System Version 10.3
Physical RAM 640 MB
Model PowerMac3,5
Processor PowerPC G4x2 @ 800 MHz
Version 7450 (V'ger) v2.1
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 800 MHz
L3 Cache 2048K @ 200 MHz
Bus Frequency 134 MHz
Video Card GeForce2 TwinView
Drive Type IBM-IC35L060AVER07-0
CPU Test 100.15
GCD Recursion 99.62 3.89 Mops/sec
Floating Point Basic 99.31 353.20 Mflop/sec
AltiVec Basic 99.38 1.44 Gflop/sec
vecLib FFT 101.56 1.57 Gflop/sec
Floating Point Library 100.90 4.04 Mops/sec
Thread Test 112.44
Computation 101.83 820.37 Kops/sec, 4 threads
Lock Contention 125.53 1.58 Mlocks/sec, 4 threads
Memory Test 110.38
System 176.13
Allocate 492.79 166.15 Kalloc/sec
Fill 190.87 1519.29 MB/sec
Copy 102.41 512.07 MB/sec
Stream 80.38
Copy 81.80 333.61 MB/sec [altivec]
Scale 81.61 339.01 MB/sec [altivec]
Add 79.50 335.46 MB/sec [altivec]
Triad 78.70 323.93 MB/sec [altivec]
Quartz Graphics Test 115.49
Line 102.31 2.60 Klines/sec [50% alpha]
Rectangle 89.66 6.31 Krects/sec [50% alpha]
Circle 107.32 2.47 Kcircles/sec [50% alpha]
Bezier 112.70 1.22 Kbeziers/sec [50% alpha]
Text 239.45 3.90 Kchars/sec
OpenGL Graphics Test 97.48
Spinning Squares 97.48 68.22 frames/sec
User Interface Test 158.03
Elements 158.03 53.77 refresh/sec
Disk Test 84.23
Sequential 83.41
Uncached Write 59.97 23.87 MB/sec [4K blocks]
Uncached Write 88.64 34.60 MB/sec [256K blocks]
Uncached Read 114.80 18.17 MB/sec [4K blocks]
Uncached Read 88.56 35.78 MB/sec [256K blocks]
Random 85.07
Uncached Write 66.32 0.95 MB/sec [4K blocks]
Uncached Write 90.38 20.38 MB/sec [256K blocks]
Uncached Read 94.29 0.62 MB/sec [4K blocks]
Uncached Read 97.35 20.04 MB/sec [256K blocks]

[BrunoBruin]

Here's my dual-1.25 (FW800):

Results 138.32
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 1536 MB
Model PowerMac3,6
Processor PowerPC G4x2 @ 1.25 GHz
Version 7455 (Apollo) v3.3
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 1250 MHz
L3 Cache 1024K @ 250 MHz
Bus Frequency 167 MHz
Video Card ATY,RV250
Drive Type IBM-IC35L090AVV207-1
CPU Test 146.47
GCD Recursion 156.18 6.10 Mops/sec
Floating Point Basic 158.11 562.30 Mflop/sec
AltiVec Basic 157.12 2.28 Gflop/sec
vecLib FFT 115.36 1.78 Gflop/sec
Floating Point Library 156.83 6.28 Mops/sec
Thread Test 154.27
Computation 154.70 1.25 Mops/sec, 4 threads
Lock Contention 153.85 1.93 Mlocks/sec, 4 threads
Memory Test 144.18
System 147.02
Allocate 155.63 52.47 Kalloc/sec
Fill 248.28 1976.32 MB/sec
Copy 100.48 502.42 MB/sec
Stream 141.45
Copy 141.79 578.23 MB/sec [altivec]
Scale 143.07 594.31 MB/sec [altivec]
Add 144.18 608.42 MB/sec [altivec]
Triad 136.97 563.76 MB/sec [altivec]
Quartz Graphics Test 144.92
Line 150.42 3.83 Klines/sec [50% alpha]
Rectangle 129.83 9.13 Krects/sec [50% alpha]
Circle 165.54 3.82 Kcircles/sec [50% alpha]
Bezier 152.87 1.66 Kbeziers/sec [50% alpha]
Text 132.11 2.15 Kchars/sec
OpenGL Graphics Test 121.25
Spinning Squares 121.25 84.85 frames/sec
User Interface Test 139.56
Elements 139.56 47.48 refresh/sec
Disk Test 124.23
Sequential 137.62
Uncached Write 142.79 56.84 MB/sec [4K blocks]
Uncached Write 127.84 49.90 MB/sec [256K blocks]
Uncached Read 167.18 26.46 MB/sec [4K blocks]
Uncached Read 121.09 48.93 MB/sec [256K blocks]
Random 113.22
Uncached Write 111.03 1.59 MB/sec [4K blocks]
Uncached Write 130.56 29.45 MB/sec [256K blocks]
Uncached Read 103.92 0.69 MB/sec [4K blocks]
Uncached Read 110.62 22.77 MB/sec [256K blocks]

[Jack The Dog]

Originally posted by Eug Wanker:
Results 90.56
System Info
Xbench Version 1.0
System Version 10.2.6
Physical RAM 768 MB
Model PowerBook3,5
Processor PowerPC G4 @ 1.00 GHz

Can you try using x-bench 1.1? ;p

[The Placid Casual]

867 TiBook. 256MB RAM (soon to be more!).

Results 80.54

System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 256 MB
Model PowerBook3,5
Processor PowerPC G4 @ 867 MHz
Version 7455 (Apollo) v3.2
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 667 MHz
L3 Cache 1024K @ 134 MHz
Bus Frequency 134 MHz
Video Card ATY,RV250M9
Drive Type TOSHIBA MK4021GAS

CPU Test 103.63
GCD Recursion 99.59 3.89 Mops/sec
Floating Point Basic 108.91 387.31 Mflop/sec
AltiVec Basic 108.76 1.58 Gflop/sec
vecLib FFT 108.83 1.68 Gflop/sec
Floating Point Library 93.98 3.76 Mops/sec

Thread Test 55.45
Computation 54.48 438.92 Kops/sec, 4 threads
Lock Contention 56.45 708.54 Klocks/sec, 4 threads

Memory Test 97.08
System 87.05
Allocate 104.57 35.26 Kalloc/sec
Fill 111.28 885.81 MB/sec
Copy 62.84 314.20 MB/sec
Stream 109.72
Copy 108.34 441.84 MB/sec [altivec]
Scale 108.18 449.41 MB/sec [altivec]
Add 113.47 478.85 MB/sec [altivec]
Triad 109.05 448.83 MB/sec [altivec]

Quartz Graphics Test 108.46
Line 111.35 2.83 Klines/sec [50% alpha]
Rectangle 113.98 8.02 Krects/sec [50% alpha]
Circle 119.61 2.76 Kcircles/sec [50% alpha]
Bezier 105.43 1.15 Kbeziers/sec [50% alpha]
Text 95.22 1.55 Kchars/sec

OpenGL Graphics Test 83.39
Spinning Squares 83.39 58.35 frames/sec

User Interface Test 94.40
Elements 94.40 32.12 refresh/sec

Disk Test 52.18
Sequential 66.67
Uncached Write 57.32 22.82 MB/sec [4K blocks]
Uncached Write 57.03 22.26 MB/sec [256K blocks]
Uncached Read 137.53 21.77 MB/sec [4K blocks]
Uncached Read 56.34 22.77 MB/sec [256K blocks]
Random 42.87
Uncached Write 33.29 0.48 MB/sec [4K blocks]
Uncached Write 47.33 10.67 MB/sec [256K blocks]
Uncached Read 44.88 0.30 MB/sec [4K blocks]
Uncached Read 50.34 10.36 MB/sec [256K blocks]

[Eug Wanker]

Originally posted by Jack The Dog:
Can you try using x-bench 1.1? ;p

The weird part is I did. I downloaded Xbench 1.1 to my desktop and then ran it. The score I posted (complete with 1.0 business) is what I came up with. Anyways, I reran it just now, and it's posted above.

I will point out however, that Xbench 1.0 was buried in my utilities folder, so it was on my system. I did NOT launch that one though. Strange.

[blueshift]

These results for a 1.6Mhz G5 were posted on spymac's forum..

As you'll see the processor tests arn't as fast as a dual 1.4, and i wouldn't expect them to be...but the system tests show the G5's subsystems advantage over the older systems. The Dual G5 is going to scream because of this...

Results 126.13

System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 1280 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 91.85
GCD Recursion 72.97 2.85 Mops/sec
Floating Point Basic 213.26�> 758.44 Mflop/sec
AltiVec Basic 78.43 1.14 Gflop/sec
vecLib FFT 51.25 789.76 Mflop/sec
Floating Point Library 264.65�> 10.59 Mops/sec
Thread Test 80.68
Computation 59.27 477.48 Kops/sec, 4 threads
Lock Contention 126.33 1.59 Mlocks/sec, 4 threads
Memory Test 251.00�>
System 199.11
Allocate 327.59�> 110.45 Kalloc/sec
Fill 155.75 1239.80 MB/sec
Copy 178.75 893.76 MB/sec
Stream 339.49�>
Copy 335.86�> 1369.69 MB/sec [G5]
Scale 319.23�> 1326.10 MB/sec [G5]
Add 347.79�> 1467.62 MB/sec [G5]
Triad 357.52�> 1471.49 MB/sec [G5]
Quartz Graphics Test 156.39
Line 174.55 4.44 Klines/sec [50% alpha]
Rectangle 174.20 12.26 Krects/sec [50% alpha]
Circle 175.11 4.04 Kcircles/sec [50% alpha]
Bezier 148.88 1.62 Kbeziers/sec [50% alpha]
Text 123.84 2.02 Kchars/sec
OpenGL Graphics Test 188.62
Spinning Squares 188.62 132.00 frames/sec
User Interface Test 141.20
Elements 141.20 48.04 refresh/sec
Disk Test 105.75
Sequential 114.68
Uncached Write 146.61 58.36 MB/sec [4K blocks]
Uncached Write 138.07 53.89 MB/sec [256K blocks]
Uncached Read 75.75 11.99 MB/sec [4K blocks]
Uncached Read 131.30 53.05 MB/sec [256K blocks]
Random 98.12
Uncached Write 86.38 1.24 MB/sec [4K blocks]
Uncached Write 102.33 23.08 MB/sec [256K blocks]
Uncached Read 95.15 0.63 MB/sec [4K blocks]
Uncached Read 112.26 23.10 MB/sec [256K blocks]

[:XI:]

Powermac G4 933 1280mb RAM, everything else is stock.

Results 87.17
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 1280 MB
Model PowerMac3,5
Processor PowerPC G4 @ 934 MHz
Version 7455 (Apollo) v2.1
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 934 MHz
L3 Cache 2048K @ 234 MHz
Bus Frequency 134 MHz
Video Card GeForce4 MX
Drive Type ST360021A
CPU Test 102.33
GCD Recursion 97.40 3.80 Mops/sec
Floating Point Basic 98.41 349.99 Mflop/sec
AltiVec Basic 106.90 1.55 Gflop/sec
vecLib FFT 109.47 1.69 Gflop/sec
Floating Point Library 100.58 4.03 Mops/sec
Thread Test 54.86
Computation 53.61 431.90 Kops/sec, 4 threads
Lock Contention 56.17 705.08 Klocks/sec, 4 threads
Memory Test 97.11
System 101.09
Allocate 101.43 34.20 Kalloc/sec
Fill 141.84 1129.04 MB/sec
Copy 78.33 391.64 MB/sec
Stream 93.43
Copy 95.46 389.32 MB/sec [altivec]
Scale 94.78 393.73 MB/sec [altivec]
Add 88.96 375.41 MB/sec [altivec]
Triad 94.84 390.34 MB/sec [altivec]
Quartz Graphics Test 126.55
Line 126.19 3.21 Klines/sec [50% alpha]
Rectangle 183.49 12.91 Krects/sec [50% alpha]
Circle 156.83 3.61 Kcircles/sec [50% alpha]
Bezier 97.49 1.06 Kbeziers/sec [50% alpha]
Text 105.25 1.72 Kchars/sec
OpenGL Graphics Test 108.45
Spinning Squares 108.45 75.89 frames/sec
User Interface Test 92.96
Elements 92.96 31.63 refresh/sec
Disk Test 71.35
Sequential 67.41
Uncached Write 61.96 24.67 MB/sec [4K blocks]
Uncached Write 60.53 23.63 MB/sec [256K blocks]
Uncached Read 100.31 15.88 MB/sec [4K blocks]
Uncached Read 59.83 24.18 MB/sec [256K blocks]
Random 75.78
Uncached Write 85.51 1.22 MB/sec [4K blocks]
Uncached Write 66.10 14.91 MB/sec [256K blocks]
Uncached Read 85.29 0.56 MB/sec [4K blocks]
Uncached Read 70.25 14.46 MB/sec [256K blocks]

[RooneyX]

Alubook 17, default specs. Had to quit several apps which might have lowered the score if resources weren't quite perfectly allocated to Xbench after that.

Result: 85.61

[:XI:]

I ran it again without iTunes running in the background:

Results 92.90
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 1280 MB
Model PowerMac3,5
Processor PowerPC G4 @ 934 MHz
Version 7455 (Apollo) v2.1
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 934 MHz
L3 Cache 2048K @ 234 MHz
Bus Frequency 134 MHz
Video Card GeForce4 MX
Drive Type ST360021A
CPU Test 113.58
GCD Recursion 109.22 4.27 Mops/sec
Floating Point Basic 117.97 419.54 Mflop/sec
AltiVec Basic 116.47 1.69 Gflop/sec
vecLib FFT 114.54 1.77 Gflop/sec
Floating Point Library 110.22 4.41 Mops/sec
Thread Test 59.72
Computation 58.44 470.81 Kops/sec, 4 threads
Lock Contention 61.05 766.27 Klocks/sec, 4 threads
Memory Test 106.70
System 115.17
Allocate 111.59 37.62 Kalloc/sec
Fill 156.69 1247.22 MB/sec
Copy 93.42 467.09 MB/sec
Stream 99.39
Copy 98.84 403.09 MB/sec [altivec]
Scale 99.56 413.58 MB/sec [altivec]
Add 101.08 426.56 MB/sec [altivec]
Triad 98.11 403.80 MB/sec [altivec]
Quartz Graphics Test 125.13
Line 123.03 3.13 Klines/sec [50% alpha]
Rectangle 129.41 9.10 Krects/sec [50% alpha]
Circle 134.52 3.10 Kcircles/sec [50% alpha]
Bezier 117.75 1.28 Kbeziers/sec [50% alpha]
Text 122.29 1.99 Kchars/sec
OpenGL Graphics Test 109.54
Spinning Squares 109.54 76.65 frames/sec
User Interface Test 107.51
Elements 107.51 36.58 refresh/sec
Disk Test 71.42
Sequential 67.60
Uncached Write 61.02 24.29 MB/sec [4K blocks]
Uncached Write 60.69 23.69 MB/sec [256K blocks]
Uncached Read 106.25 16.82 MB/sec [4K blocks]
Uncached Read 59.17 23.91 MB/sec [256K blocks]
Random 75.70
Uncached Write 84.88 1.21 MB/sec [4K blocks]
Uncached Write 63.27 14.27 MB/sec [256K blocks]
Uncached Read 88.55 0.58 MB/sec [4K blocks]
Uncached Read 71.63 14.74 MB/sec [256K blocks]

[Eug Wanker]

Xbench doesn't like the G5:

Stock G5 1.6: Processor PowerPC 970 @ 1.60 GHz - CPU Test 118.6
Eug's TiBook: Processor PowerPC G4 @ 1.00 GHz - CPU Test 121.73

[timmerk]

Here comes the 1.6 Ghz G5 Xbench 1.1 results:

Results 115.30
System Info
Xbench Version 1.1
System Version 10.2.7
Physical RAM 256 MB
Model PowerMac7,2
Processor PowerPC 970 @ 1.60 GHz
L1 Cache 64K (instruction), 32K (data)
L2 Cache 512K @ 1600 MHz
Bus Frequency 800 MHz
Video Card GeForce FX 5200
Drive Type ST380013AS
CPU Test 118.60
GCD Recursion 72.87 2.85 Mops/sec
Floating Point Basic 212.40 755.35 Mflop/sec
AltiVec Basic 78.18 1.14 Gflop/sec
vecLib FFT 139.95 2.16 Gflop/sec
Floating Point Library 263.73 10.56 Mops/sec
Thread Test 81.59
Computation 60.26 485.43 Kops/sec, 4 threads
Lock Contention 126.31 1.59 Mlocks/sec, 4 threads
Memory Test 240.86
System 187.61
Allocate 327.47 110.41 Kalloc/sec
Fill 135.53 1078.84 MB/sec
Copy 179.91 899.55 MB/sec
Stream 336.30
Copy 332.93 1357.77 MB/sec [G5]
Scale 314.83 1307.83 MB/sec [G5]
Add 345.43 1457.68 MB/sec [G5]
Triad 354.71 1459.90 MB/sec [G5]
Quartz Graphics Test 155.00
Line 177.32 4.51 Klines/sec [50% alpha]
Rectangle 171.61 12.07 Krects/sec [50% alpha]
Circle 174.01 4.01 Kcircles/sec [50% alpha]
Bezier 144.57 1.57 Kbeziers/sec [50% alpha]
Text 123.02 2.01 Kchars/sec
OpenGL Graphics Test 183.62
Spinning Squares 183.62 128.49 frames/sec
User Interface Test 139.16
Elements 139.16 47.35 refresh/sec
Disk Test 105.70
Sequential 114.80
Uncached Write 147.33 58.65 MB/sec [4K blocks]
Uncached Write 137.63 53.72 MB/sec [256K blocks]
Uncached Read 76.28 12.08 MB/sec [4K blocks]
Uncached Read 130.20 52.61 MB/sec [256K blocks]
Random 97.93
Uncached Write 85.79 1.23 MB/sec [4K blocks]
Uncached Write 103.13 23.26 MB/sec [256K blocks]
Uncached Read 94.94 0.63 MB/sec [4K blocks]
Uncached Read 111.63 22.97 MB/sec [256K blocks]

[blueshift]

I'd ignore the overall score, it's meaningless and arbitrary...cause who knows how that's weighted...look at individual scores...and the 1.6 behaves just like i'd expect it to.

[Eug Wanker]

Originally posted by blueshift:
I'd ignore the overall score, it's meaningless and arbitrary...cause who knows how that's weighted...look at individual scores...and the 1.6 behaves just like i'd expect it to.

See my earlier post:

Stock G5 1.6: Processor PowerPC 970 @ 1.60 GHz - CPU Test 118.6
Eug's TiBook: Processor PowerPC G4 @ 1.00 GHz - CPU Test 121.73

[AB^2=BCxAC]

Originally posted by blueshift:
I'd ignore the overall score, it's meaningless and arbitrary...cause who knows how that's weighted...look at individual scores...and the 1.6 behaves just like i'd expect it to.

Well, testing a bottom rung G5 with only 256 MB of RAM (not to mention the slowest frontside bus and DDR RAM of the three new models) is not the best example of G5 brilliance.

That's why I'm not suprised a 1gig TiBook beat it. But I bet it outran the TiBook in quartz rendering.

My scores for a Dual 867 MDD:
Results 101.06
System Info
Xbench Version 1.1
System Version 10.2.6
Physical RAM 1024 MB
ModelPowerMac 3,6
ProcessorPowerPC G4x2 @ 867 MHz
Version 7455 (Apollo) v3.2
L1 Cache 32K (instruction), 32K (data)
L2 Cache 256K @ 867 MHz
L3 Cache 1024K @ 217 MHz
Bus Frequency 134 MHz
Video Card ATY,RV250
Drive Type IBM-IC35L060AVVA07-0
CPU Test 109.12
GCD Recursion 108.25 4.23 Mops/sec
Floating Point Basic 110.37 392.50 Mflop/sec
AltiVec Basic 108.28 1.57 Gflop/sec
vecLib FFT 109.92 1.69 Gflop/sec
Floating Point Library 108.80 4.36 Mops/sec
Thread Test 108.18
Computation 108.82 876.67 Kops/sec, 4 threads
Lock Contention 107.56 1.35 Mlocks/sec, 4 threads
Memory Test 108.08
System 105.26
Allocate 108.37 36.54 Kalloc/sec
Fill 181.98 1448.53 MB/sec
Copy 72.57 362.86 MB/sec
Stream 111.06
Copy 111.45 454.53 MB/sec [altivec]
Scale 112.76 468.41 MB/sec [altivec]
Add 113.25 477.91 MB/sec [altivec]
Triad 107.01 440.45 MB/sec [altivec]
Quartz Graphics Test 106.22
Line 112.44 2.86 Klines/sec [50% alpha]
Rectangle 84.12 5.92 Krects/sec [50% alpha]
Circle 114.75 2.64 Kcircles/sec [50% alpha]
Bezier 111.31 1.21 Kbeziers/sec [50% alpha]
Text 116.40 1.90 Kchars/sec
OpenGL Graphics Test 76.83
Spinning Squares 76.83 53.77 frames/sec
User Interface Test 106.99
Elements 106.99 36.40 refresh/sec
Disk Test 101.74
Sequential 105.75
Uncached Write 97.82 38.94 MB/sec [4K blocks]
Uncached Write 97.08 37.89 MB/sec [256K blocks]
Uncached Read 152.74 24.18 MB/sec [4K blocks]
Uncached Read 92.99 37.57 MB/sec [256K blocks]
Random 98.01
Uncached Write 100.38 1.44 MB/sec [4K blocks]
Uncached Write 98.72 22.27 MB/sec [256K blocks]
Uncached Read 97.84 0.65 MB/sec [4K blocks]
Uncached Read 95.25 19.60 MB/sec [256K blocks]

[Eug Wanker]

Originally posted by AB^2=BCxAC:
Well, testing a bottom rung G5 with only 256 MB of RAM (not to mention the slowest frontside bus and DDR RAM of the three new models) is not the best example of G5 brilliance. That's why I'm not suprised a 1gig TiBook beat it.

1) RAM is irrelevant for the CPU bench.
2) G5 1.6 bus - 800 MHz. TiBook bus - 133 MHz.

But I bet it outran the TiBook in quartz rendering.

It did, but by only 20%.

Basically, even after the 1.1 update, Xbench is fubar'd on the G5.

[kupan787]

Originally posted by Eug Wanker:
Basically, even after the 1.1 update, Xbench is fubar'd on the G5.

I wonder if the author used Shark to profile it, or just recompiled it with gcc 3.3?

[krassy]

Code:
Results               126.13                         119.82 
System Info 
Xbench Version        1.1                            1.1 
System Version        10.2.7                         10.2.7 
Physical RAM          1280 MB                        2048 MB 
Model                 PowerMac7,2                    PowerMac7,2 
Processor             PowerPC 970 @ 1.60 GHz         PowerPC 970 @ 1.60 GHz 
L1 Cache              64K (instruction), 32K (data)  64K (instruction), 32K (data) 
L2 Cache              512K @ 1600 MHz                512K @ 1600 MHz 
Bus Frequency         800 MHz                        800 MHz 
Video Card            GeForce FX 5200                GeForce FX 5200 
Drive Type            ST380013AS                     ST380013AS 
CPU Test              91.85                          73.46 
GCD Recursion         72.97                          74.15
	                  2.85 Mops/sec                  2.90 Mops/sec
Floating Point Basic  213.26                         195.83 
                      758.44 Mflop/sec               696.43 Mflop/sec 
AltiVec Basic         78.43                          77.96 
                      1.14 Gflop/sec                 1.13 Gflop/sec 
vecLib FFT            51.25                          30.43 
                      789.76 Mflop/sec               468.95 Mflop/sec 
Floating Point Library 
                      264.65                         264.56 
                      10.59 Mops/sec                 10.59 Mops/sec 
Thread Test           80.68                          81.74 
Computation           59.27                          60.45 
                      477.48 Kops/sec, 4 threads     487.01 Kops/sec, 4 threads 
Lock Contention       126.33                         126.17 
                      1.59 Mlocks/sec, 4 threads     1.58 Mlocks/sec, 4 threads 
Memory Test           251.00                         259.39 
System                199.11                         205.66 
Allocate              327.59                         320.70 
                      110.45 Kalloc/sec              108.13 Kalloc/sec 
Fill                  155.75                         152.77 
                      1239.80 MB/sec                 1216.08 MB/sec 
Copy                  178.75                         203.12 
                      893.76 MB/sec                  1015.58 MB/sec 
Stream                339.49                         351.13 
Copy                  335.86                         350.36 
                      1369.69 MB/sec [G5]            1428.83 MB/sec [G5] 
Scale                 319.23                         332.01 
                      1326.10 MB/sec [G5]            1379.19 MB/sec [G5] 
Add                   347.79                         359.41 
                      1467.62 MB/sec [G5]            1516.68 MB/sec [G5] 
Triad                 357.52                         364.54 
                      1471.49 MB/sec [G5]            1500.38 MB/sec [G5] 
Quartz Graphics Test  156.39                         163.04 
Line                  174.55                         178.20 
                      4.44 Klines/sec [50% alpha]    4.54 Klines/sec [50% alpha] 
Rectangle             174.20                         171.60 
                      12.26 Krects/sec [50% alpha]   12.07 Krects/sec [50% alpha] 
Circle                175.11                         177.85 
                      4.04 Kcircles/sec [50% alpha]  4.10 Kcircles/sec [50% alpha] 
Bezier                148.88                         149.27 
                      1.62 Kbeziers/sec [50% alpha]  1.62 Kbeziers/sec [50% alpha] 
Text                  123.84                         144.78 
                      2.02 Kchars/sec                2.36 Kchars/sec 
OpenGL Graphics Test  188.62                         182.49 
Spinning Squares      188.62                         182.49 
                      132.00 frames/sec              127.71 frames/sec 
User Interface Test   141.20                         141.32 
Elements              141.20                         141.32 
                      48.04 refresh/sec              48.08 refresh/sec 
Disk Test             105.75                         99.70 
Sequential            114.68                         106.96 
Uncached Write        146.61                         131.21 
                      58.36 MB/sec [4K blocks]       52.23 MB/sec [4K blocks] 
Uncached Write        138.07                         123.69 
                      53.89 MB/sec [256K blocks]     48.28 MB/sec [256K blocks] 
Uncached Read         75.75                          75.20 
                      11.99 MB/sec [4K blocks]       11.90 MB/sec [4K blocks] 
Uncached Read         131.30                         119.16 
                      53.05 MB/sec [256K blocks]     48.15 MB/sec [256K blocks] 
Random                98.12                          93.36 
Uncached Write        86.38                          81.53 
                      1.24 MB/sec [4K blocks]        1.17 MB/sec [4K blocks] 
Uncached Write        102.33                         96.68 
                      23.08 MB/sec [256K blocks]     21.81 MB/sec [256K blocks] 
Uncached Read         95.15                          94.64 
                      0.63 MB/sec [4K blocks]        0.62 MB/sec [4K blocks] 
Uncached Read         112.26                         103.42 
                      23.10 MB/sec [256K blocks]     21.28 MB/sec [256K blocks]

[jrod7350]

CPU Test 97.48
GCD Recursion 92.90 3.63 Mops/sec
Floating Point Basic 101.02 359.25 Mflop/sec
AltiVec Basic 100.25 1.46 Gflop/sec
vecLib FFT 98.47 1.52 Gflop/sec
Floating Point Library 95.25 3.81 Mops/sec


This is for an 800Mhz Titanium with 1GB RAM and Safari running in the background. I'm really puzzled how my laptop is beating a 1.6Ghz G5 desktop in a CPU test

[jrod7350]

Hey, I was scrolling throught the previous posts, and I think I've noticed something interesting. While the G5 is getting wupped on most of the CPU tests, it's Floating Point Basic score of 213.26 is way faster than anything out right now, including the dual 1.42Ghz Mac. See:

1.42Ghz G4

Floating Point Basic 176.85 628.95 Mflop/sec


1.6Ghz G5
Floating Point Basic 213.26 758.44 Mflop/sec

I'm questioning the Xbench results myself, but it does seem like in FPU performance that the G5 is much faster than any G4 out right now.

[jrod7350]

Ahh, sorry for all these dumb replies, but here's more interesting stuff:

1.42Ghz G4
Floating Point Library 177.81 7.12 Mops/sec


1.6Ghz G5
Floating Point Library 264.65 10.59 Mops/sec

Again, floating point performance much better on the G5. Is the floating point benchmarks on Xbench aware of dual processors, or is this just for a single processor (My single 800Mhz Tibook has a floating point library score of 95.25)?

Did a couple calculations, and basically for a 12% increase in processor speed you're getting a 48% increase in Mops/sec (that's if the floating point library is only benchmarking for a single processor). Hmm, maybe someone can make some sense out of what this could mean for the potential of the G5. I'm still curious why the Altivec and Veclib FFT scores are so bad.

[Eug Wanker]

Originally posted by kupan787:
I wonder if the author used Shark to profile it, or just recompiled it with gcc 3.3?

Xbench 1.1 Version History [2003-08-23]

* Optimized for the PowerMac G5, compiled with GCC 3.3
* Avoid use of vec_dst on PowerPC 970 for better performance
* Changes to the AltiVec test to reduce dependence on memory bandwidth
* Removed use of vec_dstt
* Changes to floating point test
* Added vecLib FFT test
* Added automatic machine type detection
* Fixed an issue where the results submission sheet caused the results window to become catatonic
* Prettier submit button
* Recalibration of CPU tests, Disk tests
* Fixed spinning rainbow cursor during OpenGL test

[kupan787]

Originally posted by Eug Wanker:
Xbench 1.1 Version History [2003-08-23]

* Optimized for the PowerMac G5, compiled with GCC 3.3
* Avoid use of vec_dst on PowerPC 970 for better performance
* Changes to the AltiVec test to reduce dependence on memory bandwidth
* Removed use of vec_dstt
* Changes to floating point test
* Added vecLib FFT test
* Added automatic machine type detection
* Fixed an issue where the results submission sheet caused the results window to become catatonic
* Prettier submit button
* Recalibration of CPU tests, Disk tests
* Fixed spinning rainbow cursor during OpenGL test

Doesn't mention any use of shark for profiling. All that is mentioned is "Optimized for the PowerMac G5, compiled with GCC 3.3". That could just mean he turned on the "g5 settings" in gcc 3.3, and compiled. He could have done no g5 performance tuning at all.

I have read in many places that 7450/7455 optimazations can just KILL the 970 performance wise. I wonder if that is what we are seeing in the CPU test. I mean a 800MHz G4 should not be beating a 1.6GHz G5 in a CPU test.

Also if you look at krassys post, how can more ram give you worse numbers?

Code:
Physical RAM          1280 MB                        2048 MB 
Model                 PowerMac7,2                    PowerMac7,2 
Processor             PowerPC 970 @ 1.60 GHz         PowerPC 970 @ 1.60 GHz  
CPU Test              91.85                          73.46

That just makes no sense. I have a feeling that xBench is still not properly running on the G5, and potentially running G4 optimized code, which is just killing the G5 in some of these tests.

[Bill Harrison]

Not to be the voice of reason here, but maybe... JUST maybe, the 1.6 is not in fact alot faster than the G4's its replacing.

I mean, it will be at SOME things, but I believe the collective score is not going to be much higher than the current top end G4 systems, and as a matter of fact may be lower.

This is what we are seeing across the board. Its not just this bench, but many. At first the excuse was unoptomized for G5 bench's. Well, now this one is. Now the excuse is just that "it doesnt like the g5"~???

The g5 is a fabulous processor, but NOT what apple has made it out to be. Thats apples undoing. Instead of being proud of the fact that they now have a competitive processor architecture under the hood, they went and shouted from the tallest buildings that they had the fastest thing ever. Guess what?

It is competetive with current offerings. It is not a quantum leap forward, just BIG step in the right direction. Its apples fault for raising everyones expectations too high. Much better to be pleasantly surprised when something outperforms expectations than to be dissapointed when something does not meet them.

I would say the 1.6 is going to be slightly below current top end G4 offerings, the 1.8 should about match them or be SLIGHTLY faster, and the 2.0 dual will top them by a decent amount.

Compared the pc's, the story will be a little worse with the 1.6 lagging a good bit, the 1.8 by slightly less, and the dual 2.0 being behind the current fastest x86 processors.

This is simply a reality check. Now that said, the G5 has ALOT of potential, and I love macs for so much more than the processor, namely one gorgeous Operating System, OSX. Now this is something that PC's cant touch REGARDLESS of the speed!

[osiris]

This is simply a reality check. Now that said, the G5 has ALOT of potential, and I love macs for so much more than the processor, namely one gorgeous Operating System, OSX. Now this is something that PC's cant touch REGARDLESS of the speed! [/B]

IMO the sad thing is that similar tasks will be slower on a G5 - at least until all these miraculous optimizations are made. Sounds awefully like the Altivec number I heard a few years ago.

The saving grace will be a 64-bit Panther, which (I hope) will double all test results/benchmarks and finally prove Apple true.

[Horsepoo!!!]

Xbench is a piece of ****. You can get wildly different results on the same computer everytime you bench it.

**** this ****!

[Hydra]

I think people need to take a step back and gain some perspective here. The biggest problem with the G5 (PPC 970) is that outside of Apple hardly anyone has had a chance to even touch one. Apple has made sure that certain companies (like Adobe) have had resources made available to them in order to develop optimized software. I'm not being an Apple apologist here but how can someone who has no access to the hardware work out the kinks that inevitably arise in software development.

I'm not a software developer but it seems like common sense. Please stop me if I'm wrong. As advanced as the aerospace industry is and the increasing amount of design that goes on in the digital realm they still must fire up the engines and go for a ride to see how thing shake out in the real world.

I think the G5 will prove to be a real performer but with a big change in architecture comes the inevitable teething problems. I think before people criticize and critique the G5 they should realize there are probably more forum threads about benches on the G5 around the web than there are G5 chips in existence That is the one thing that will change as production ramps up.

I think the biggest criticism leveled at Apple should be the overly secret nature of such a big introduction. I think allot of performance issues would have been moot had developers had access long before the general public. I'm still optimistic and think it won't be long before the performance issues are put to rest.

-Jerry C.

[Eug Wanker]

Doesn't mention any use of shark for profiling. All that is mentioned is "Optimized for the PowerMac G5, compiled with GCC 3.3". That could just mean he turned on the "g5 settings" in gcc 3.3, and compiled. He could have done no g5 performance tuning at all.

I'm not a programmer but over at Ars people are saying the avoidance of vec_dst alone may go a long way to speed up performance on the G5. Note the info from his version history:

* Optimized for the PowerMac G5, compiled with GCC 3.3
* Avoid use of vec_dst on PowerPC 970 for better performance
* Changes to the AltiVec test to reduce dependence on memory bandwidth
* Removed use of vec_dstt

Here is a blurb from an Ars poster about vec_dst:

I think the highly optimized Altivec apps like Photoshop, Final Cut and Logic will get updates very soon after release. Adobe already mentioned that one would come out this month, but I haven't seen to many other similar claims yet. The 'optimization' for the G5 is really more like turning off the G4 optimizations. Here's the exact thing involved:

G4:
vec_dst();

G5:
#ifdef PPC970
vec_dst();
#endif

That's pretty much all that's involved with avoiding the nasty vec_dst issue. Now there are other cache control ops like dcbz that apparently do work on the G5. The 970 has a hardware prefetch, but it's effectiveness has yet to be established.

OTOH, the same Ars people say that Xbench, even if optimized, is not a very good performance benchmark, because of the design. ie. It would be OK in some situations to compare chips of the same class, but definitely not to compare chips of different classes.

[kakashi]

Synthetic benchmarks are a joke. Especially when comparing between different processors. As has been mentioned before, each processor has a unique way of working, which means the programmer should code to the design of the CPU. And yes, although the G4 and the G5 share the same instruction sets, their architecture differs. Secondly, XBench has always, and always will in the forseeable future be one of the flakiest and most inconsistent benchmark suites out there. Results differ greatly on the same machine when run several times; clearly something is wrong. Real world applications are the only way to properly tell whether or not a computer is better than another. Even then, some real world apps aren't a great way to 'benchmark' a computer. Take Quake3 and Unreal Tournament for an example, manufacturers have been known to tweak their hardware for these apps alone. Basically, take these Xbench results with a grain of salt..okay maybe a cup of salt.

[jrod7350]

Originally posted by Bill Harrison:
Not to be the voice of reason here, but maybe... JUST maybe, the 1.6 is not in fact alot faster than the G4's its replacing.

Maybe so, but I have to admit that I think something else is wrong here, especially when my 800Mhz Tibook is faster in a CPU test than the G5 which is twice the clockspeed.

[kcmac]

Could someone run some application software tests and stop with all the benchmark silliness?

I don't know that I will ever need to use X-bench during my day but I do run photoshop, etc.

Until then, I think the dual 2gig G5 will do at least as good as Apple said. I want to see some apps comparisons and will stop reading these idiotic threads until they start happening in a way that the majority of us can understand.

[jrod7350]

Originally posted by kcmac:
I don't know that I will ever need to use X-bench during my day but I do run photoshop, etc.

I agree with you here too, but apparently people are waiting for machines with greater than 256MB for benchmarking, esp. Photoshop. So that would mean that at the Apple Stores it would be at least a 1.8Ghz G5 for benchmarking purposes, since the 1.6 only come standard with 256MB (and I don't want to see a pathetic photoshop test with that little RAM. It would again be basically meaningless numbers and more bad publicity for the G5)

[MindFad]

Well, how's OS X on the machines, at least? Sorry to butt in on all the Xbenching. As EugMeaister said, I think Xbench is FUBARinated.

[kupan787]

Originally posted by Eug Wanker:
I'm not a programmer but over at Ars people are saying the avoidance of vec_dst alone may go a long way to speed up performance on the G5.

Yes, in altivec stuff. But that has no bearing on the CPU tests at all.

xBench is not a good tool to test architectural differences. There is alot of work to write a good optimized app, if you already have optimized 7450/7455 code. You can't simply just compile with gcc 3.3 and call it a day.

The xBench results coming out are so wild (a 800MHz TIbook beating a 1.6GHz desktop...come on). Plus why does it seem that everyone with a G5 only does this bench, and nothing else? Could the numbers be faked? I am starting to think maybe yes...

[Eug]

Originally posted by kupan787:
Yes, in altivec stuff. But that has no bearing on the CPU tests at all.

xBench is not a good tool to test architectural differences. There is alot of work to write a good optimized app, if you already have optimized 7450/7455 code. You can't simply just compile with gcc 3.3 and call it a day.

The xBench results coming out are so wild (a 800MHz TIbook beating a 1.6GHz desktop...come on). Plus why does it seem that everyone with a G5 only does this bench, and nothing else? Could the numbers be faked? I am starting to think maybe yes...

1) I agree it's not a good bench, esp. in this situation.
2) You have to deactive vec_dst manually, not just with a recompile.
3) The Altivec bench is a component of the CPU score.
4) I presume it's run because it's very short and easy to install on an in-store demo unit. It's not as if anyone can install Photoshop and all the necessary plug-ins. However, we'll see them soon enough.

[kupan787]

Originally posted by Eug:
1) I agree it's not a good bench, esp. in this situation.

2) You have to deactive vec_dst manually, not just with a recompile.

True. However, I wonder if he ran it through CHUD (and Shark) at all. If he #ifdef a few lines, that isn't profiling.

From the G5 developer tuning page (http://developer.apple.com/technotes/tn/tn2086.html):

A standard rule of optimization is to profile your application before tuning it.

I still wonder if the xBench author did this...

3) The Altivec bench is a component of the CPU score.

Ok, I didn't realize that. However, there are still other portions that I am sure coudl use fixing. I hate to keep harping it, but the 800 vs 1600 doesn't make sense.

4) I presume it's run because it's very short and easy to install on an in-store demo unit. It's not as if anyone can install Photoshop and all the necessary plug-ins. However, we'll see them soon enough.

Ok, but there are at least 3 people now with machines at home. They took time to post pictures, I hope they post a few realworld numbers (startup time or something).

I am betting that either there was no profiling or made up numbers being posted. I have seen 3 different tests of 1.6GHz machines and all are very widely different.

[lookmark]

Why don't those lucky (very) few with G5s post some real-world comparisons, e.g. Photoshop operations with and without Adobe's optimized plug-ins? Instead of the somewhat random XBench tests?

[kupan787]

Originally posted by lookmark:
Why don't those lucky (very) few with G5s post some real-world comparisons, e.g. Photoshop operations with and without Adobe's optimized plug-ins? Instead of the somewhat random XBench tests?

Once my 1.8 ships, I will post here. It is still listed as on or before 8/29. Hears hoping....