DDR SDRAM Comparison – Is it Worth Waiting For?

Pages: 1 2 3 4 5 6

Test results table

Below is the complete table of results. As usual, I ran each test three times and if not within 3% of the others, the run was rejected and run over. I’ve also added an FIC AD11 mainboard using the AMD 761 Northbridge and VIA 686B (also an early production Mainboard and BIOS) just to make sure none of the results were out of line with the Soyo K7ADA based on the ALi MAGiK1 (they don’t appear to be).

Win98SE, 1024×768/64K @ 85Hz, 128MB memory, WD 200BB ATA/100 7200-PRM HD, ATI Radeon DDR 32MB AGP

Mainboard

K7ADA

AD11

BIOS ‘Normal’, auto optimized settings – By SPD

BIOS ‘Max’ – CAS 2, Fast

BIOS ‘Normal’ – CAS 2.5, Fast

BIOS ‘Normal’ – CAS 2, Fast

BIOS ‘Normal’, auto optimized settings – By SPD

BIOS CAS2.5, Fast

BIOS ‘Max’ – CAS 2, Fast

8.5×142=1207 BIOS – CAS 2, Fast

BIOS ‘Normal’ – CAS 2, Fast

BIOS Man,, CAS2, rest default

CPU

A 1.2

A 1.2

A 1.2

A 1.2

A 1.2

A 1.2

A 1.2

A 1.2

A 1.2

A 1.2

FSB

200MHz

200MHz

200MHz

200MHz

266MHz

266MHz

266MHz

266MHz

200MHz

200MHz

Memory

PC1600

PC1600

PC1600

PC1600

PC2100

PC2100

PC2100

PC2100

PC2100

PC1600

Winstone 2001

Business

44.7

45.5

45.4

45.9

46.8

47.3

47.5

46.7

43.5

47.3

Content Creation

42.3

44.1

43.3

44.1

45.4

46.5

46.8

46.2

41.6

45.2

3D Winbench 2000

Score

95.5

95.2

95.1

95.3

96.5

96.4

96.4

96.6

95.8

96.9

CPU

2.25

2.35

2.31

2.36

2.46

2.52

2.55

2.59

2.08

2.45

Winbench 99

CPUmark 99

102

105

104

105

108

110

110

112

98.6

109

FPU Winmark

6560

6560

6560

6560

6560

6560

6560

6630

6560

6630

Business Disk

6710

6900

6800

6740

6640

6620

6750

6750

6750

6820

High Disk

19400

19100

18400

19500

19300

18800

19700

19400

17800

20000

Business Graphics

513

512

519

517

546

551

552

538

492

525

High Graphics

1430

1440

1440

1440

1510

1520

1530

1520

1370

1500

3D Mark 2000

Score

4678

4652

4646

4677

4783

4767

4775

4801

4690

4848

CPU

437

431

433

440

465

463

463

473

444

479

Game 1 Low

96.1

96.1

95.9

96.2

96.6

96.6

96.5

96.6

96.5

96.8

Game 1 Med

65.1

64.9

64.8

65

65.5

65.4

65.4

65.5

65.4

65.9

Game 1 High

28.4

27.7

27.8

28.1

29

28.7

28.7

28.9

28.8

29.7

Game 2 Low

91.4

90

89.9

90.6

92.8

92.1

92.2

92.9

91.6

94.9

Game 2 Med

66.9

66.8

66.5

67.2

70.2

70

70.3

70.9

66.6

71.4

Game 2 High

41.9

42.2

42.1

42.6

44.5

44.5

44.8

45.2

41.9

45.2

Quake III Demo 1

124.3

130

128.3

131

137.2

140.5

142.2

144

118.8

140

Final Reality

Score

7.47

7.97

7.7

7.93

8.32

8.64

8.86

8.89

7.24

8.12

2D

7.97

8.74

8.31

8.93

9.26

9.8

10.32

10.41

7.31

9.72

3D

6.36

6.43

6.41

6.43

6.47

6.52

6.54

6.54

6.32

6.47

Bus

10.53

12.06

11.19

11.43

13.2

14.13

14.46

14.43

10.47

10.98

AGP

321.28

327.87

327.55

328.65

335.76

338.72

342.12

342.29

313.06

336.51

Video2000

920

935

965

975

986

1009

1007

993

902

980

SYSmark 2000

Overall Rating

196

202

201

201

207

210

212

214

187

209

Internet Content Creation

218

215

213

215

220

225

226

228

200

222

Office Productivity

186

193

193

191

198

200

202

204

177

200

Summary

What I see is not a question of using PC1600 and PC2100, or CAS 2 and CAS 2.5. What does seem to make a difference is 200MHz vs. 266MHz FSB (100MHz vs. 133MHz HostClk) along with going from PC1600 to PC2100. Looking over all the scores I would say the overall difference between the best and worst is not quite 10%. Not a whole lot, but enough to make me think twice about using PC1600 DDR SDRAM with a 200MHz FSB CPU. Sometimes, its the sum of what you use that gives you the performance edge, in this case it appears to be a combination of PC2100 and a 266MHz FSB, while the CAS setting is not as important. Using the SPD to auto set the memory timing settings seems to be the best choice, no real gain over manually setting them.

For the time being, at least to me, it would only make sense to use PC2100 DDR SDRAM along with a 266MHz FSB CPU. I consider DDR SDRAM to be for high-end performance systems, and I see no reason to sacrifice any performance in order to cut system cost. If you don’t want to pay for the top of the line parts to get top of the line performance you might be better off sticking with PC133 SDRAM and a 200MHz FSB.


Pages: « Prev  1 2 3 4 5 6  

Be the first to discuss this article!