Pentium 4 Price/Performance Comparison

Pages: 1 2 3 4 5 6 7

Basic Price/Performance Comparison

Next let’s look at how the components break down as a percentage of the overall system cost. I’ll use a iP4 1.8A here as a baseline, since it appears to be the ‘sweet spot’ in price using the above graph:

For the above components I’ve used the following:

  • Soyo P4S Dragon Ultra (SiS 645 chipset, Integrated LAN and C-Media Sound, HPT RAID)
  • Boxed Intel Pentium 4 1.8A GHz CPU (Northwood – 512K L2, includes CPU cooler)
  • One 256MB PC2100 SEC DIMM
  • EVGA GeForce 3 Ti200 AGP card
  • 3.5" Teac Floppy Disk
  • Maxtor 40GB ATA/133 7200 RPM Hard Disk
  • Afreey 16x/40x DVD Drive
  • In Win S500 Case with a 300w Power Supply and 80mm Case Fan

Pretty average complete upper end system (minus an OS, KB and Mouse, monitor). As you can see, the CPU is a bit over 24% of the complete system cost and is the most expensive part of the system. But let’s take a look at some different component options and their effect on the overall system cost. Here we will use the iP4 2.0A as a base. The 2.0GHz 256K L2 cache (Willamette) iP4 I have for testing is unlocked, so I can set it for any speed to get a good idea what the effect different CPU speeds will have on performance. But when it came to testing the 512K L2 cache (Northwood) CPU’s I only had locked 2.0A and 2.2GHz parts to test with. I would have liked to test at least a 1.8A also, but it could just not be done without obtaining more CPU’s.

The actual numbers didn’t fit well on the graph so I’ll give them to you in the table below. As noted, the baseline is the iP4 2.0A with 256MB of memory and a single hard disk. For the Winstone ‘Index’ I combined the two scores and divided by two, for the game ‘Index’ I used the 3DMark 2001 score divided by 100, the Quake score multiplied by .3 and the Unreal Tournament score, added them up and divided by three. This bought all the scores within the same basic range so no one score would overly influence the ‘index’.

1.5, 256MB

1.8, 256MB

2.0, 256MB

2.0A, 256MB

2.0A, 512MB

2.0A, 256MB, RAID 0

2.2, 256MB

2.2, 512MB

2.2, 1.0GB

2.2, 1.0GB, RAID 0

MoBo

 

140

140

140

140.00

1.40

140.00

140.00

140.00

140.00

140

CPU

 

136

199

357

357

357

357

541

541

541

541

Memory

 

70.5

70.5

70.5

70.5

137

70.5

70.5

137

274

274

AGP

 

192

192

192

192

192

192

192

192

192

192

3.5″ FD

 

10

10

10

10

10

10

10

10

10

10

40GB ATA/133

89

89

89

89

178

178

89

89

89

178

DVD

 

58

58

58

58

58

58

58

58

58

58

Case

 

57

57

57

57

57

57

57

57

57

57

Total

 

752.5

815.5

973.5

973.5

990.4

1062.5

1157.5

1224

1361

1450

Cost %

77.30%

83.77%

100.00%

100.00%

101.74%

109.14%

118.90%

125.73%

139.80%

148.95%

WS Performance %

78.81%

87.10%

93.04%

100.00%

106.38%

108.72%

105.21%

112.46%

119.94%

126.98%

Game Performance %

83.42%

88.60%

91.20%

100.00%

100.12%

100.12%

102.79%

102.79%

102.79%

102.79%

WS Performance Index

53.75

59.4

63.45

68.2

72.55

74.15

71.75

76.7

81.8

86.6

Game Performance Index

32.13

34.12

35.12

38.51

38.56

38.56

39.59

39.59

39.59

39.59

So what do we see? Well I think a couple of things stand out. First is that the line for cost has a nice steady gain, the Winstone ‘Index’ also shows a nice steady gain but not quite as much and the game ‘index’ is fairly flat. The second thing is that while more memory and faster disk access show benefits in Winstone, they have just about zero affect on gaming. Third is that there is a pretty healthy performance difference between the 2.0 and 2.0A CPU’s. Doubling the size of the L2 cache does make a difference in performance with no real impact on cost.

Let’s take a look at the 2.0A test setups first. There are three different variations shown – the base with 256MB of memory, the same setup but with512MB of memory, and RAID 0 with 256MB of memory. As noted, the different memory size and RAID have no affect on gaming so we’ll just look at the Winstone test here. Doubling the memory gave a 6.38% gain in performance but at an 18.62% premium in cost, while doubling the disk and setting up as a RAID 0 array gave an 8.72% gain in performance while costing a 27.26% more. But before you make the conclusion that more memory would be a better choice then RAID, let’s think about those results and how they would affect different actual real world usage.


Pages: « Prev   1 2 3 4 5 6 7   Next »

Be the first to discuss this article!