DDR SDRAM – A Second Glance

Pages: 1 2

More Test Results

For Socket A based systems, what I’ve done is to use a KT133 system using PC133 CAS2 memory as a baseline and rather than give you the actual test result scores for each I’ve shown the difference in percentage. Quite often looking at actual scores can be confusing and sometimes even meaningless. I also picked the fastest Socket A CPU currently available – the Athlon 1.2GHz with a 200MHz FSB. The CPU was an OEM part and as luck would have it the L1 contacts were still connected! This allowed me to easily change the multiplier so I could run the CPU at both 200MHz FSB and 266MHz FSB, yet still keep the CPU at 1.2GHz for some good comparison testing. The Soyo K7VTA-B, K7VTA-Pro and K7ADA all allow overriding the multiplier if the CPU has the L1 bridges connected. The Gigabyte GA-7DX does not, and this one is an early ver. 2.4 which does not support (at least as listed on AMD’s site) 266MHz FSB CPU’s. I would have loved to use a true 266MHz FSB CPU, but there is just about no way to get a hold of one, and they aren’t expected to be available in the channel until at least late February to mid March. If I could get a 266MHz FSB CPU I could also test on the AD11 AMD 761 based mainboard that FIC supplied (Gigabyte supplied the GA-7DX mainboard)

What to compare to? I chose a new KT133A based Soyo K7VTA-Pro that would allow me to run with PC133 and a 266MHz FSB, an ALi MAGiK 1based Soyo K7ADA that would allow me to use PC2100 DDR with either a 200MHz or 266MHz FSB, and an AMD 761 (with a VIA 686B) Gigabyte 7DX that only allowed me to test with PC1600 and a 200MHz FSB. The only ‘issue’ I would like to point out is that the Soyo K7ADA Ali MAGiK 1 based Mainboard is a pre-production part (along with the BIOS) and it’s results may not be an accurate indication of what a final production part will perform like, especially when you look at it’s 200MHz FSB results (the production parts just can’t be that bad). Oh well, you sometimes can only test with what you have…

So here you are – Any result in Blue is that much faster in percentage than the base KT133 PC133 system, those in (Red) are that much slower than the base system. All tests used the following common components: Win98SE, 1024×768/64K @ 85Hz, 128MB memory, WD 200BB ATA/100 7200-PRM HD, ATI Radeon DDR 32MB AGP. I hope this chart makes sense for people:

Mainboard

K7VTA-B

K7VTA-Pro

K7ADA

K7ADA

GA-7DX

Chipset

KT133

KT133A

MAGIK 1

MAGiK 1

AMD 761

CPU

1.2GHz

1.2GHz

1.2GHz

1.2GHz

1.2GHz

FSB

200MHz

266MHz

200MHz

266MHz

200MHz

Memory

PC133 CAS2

PC133 CAS2

PC2100

PC2100

PC1600

Base

% different

% different

% different

% different

Winstone 2001

Business

45.90

2.61

(5.23)

2.83

0.65

Content Creation

44.50

3.82

(7.42)

3.15

2.70

3D Winbench 2000

Score

96.40

(0.31)

(0.62)

0.00

0.62

CPU

2.40

4.58

(13.33)

5.83

3.33

Winbench 99

CPUmark 99

106.00

2.83

(7.17)

3.77

2.83

FPU Winmark

6540.00

0.31

0.31

0.00

(0.92)

Business Disk

6550.00

(2.75)

3.05

(3.97)

2.44

High Disk

18300.00

0.55

1.09

3.83

3.83

Business Graphics

521.00

(1.34)

(9.21)

1.73

1.15

High Graphics

1470.00

2.04

(7.48)

3.40

0.68

3D Mark 2000

Score

4828.00

0.27

(2.90)

(1.37)

0.52

CPU

463.00

1.08

(6.48)

(3.67)

0.00

Quake III Demo 1

137.00

3.43

(13.28)

3.94

3.21

Wizmark 3.0 (lower better)

5.91

13.20

(14.38)

9.64

12.35

Tiartanium 1.90, D3D & 3DNow

60.90

0.33

0.33

0.33

0.49

Video2000

948.00

1.58

(5.27)

7.28

7.49

SYSmark 2000

Overall Rating

203.00

2.96

(7.88)

4.93

3.45

Internet Content Creation

216.00

2.31

(7.87)

5.56

2.31

Office Productivity

194.00

3.61

(7.73)

4.12

4.12

Kind of an interesting way of looking at the results, isn’t it? For now lets ignore the 200MHz Ali MAGiK 1 results (I’m going to assume it’s an issue with the MoBo not being a production part yet) and just look at the others. Also, we are beginning to max out the video card, which is why you don’t see any difference in the 3D Winbench 2000 scores but do in the CPU scores for that test. So where did DDR seem to show an advantage??

First thing that catches my eye is the Video2000 results, as both DDR MoBo’s seem so show their best in that test. Second would be the SYSmark 2000 scores, again with both DDR showing better scores, especially the MAGiK 1 with a 266MHz FSB.

Then there is the KT133A with a 266MHz FSB and its Wizmark score, although the AMD 761 even with its 200MHz FSB did almost as well, and the Ali MAGiK 1 wasn’t far behind. Quake is pretty close, with all comparison tests just about the same amount faster than the KT133 base.

I was surprised to see the 3Dmark 2000 results so close, and even more surprised that the Ali MAGiK 1 didn’t fare too well here. 3D Winbench 2000 CPU results are also a bit of a surprise as the MAGiK 1 did the best, the KT133A was next, and the AMD 761 was a close third.

So what do you think? Is DDR the wave of the future? Should you wait and buy it when it starts being generally available late next month? Should you buy PC1600 or PC2100? Use a 200MHz or 266MHz FSB? Buy the AMD 761 or Ali MAGiK 1… or even wait for the VIA KT266? Or just stick to a KT133? Buy the new KT133A?

Don’t ask me, I’m still trying to figure it all out as I don’t see any real trend in the results so far. What I will say is that there may be some applications where DDR indicates it will make a difference – such as those profiled by the Video 2000 test. Higher FSB speeds and faster CPU’s also appear to have some bearing. To me the jury is still out, but time should tell. The KT133A, PC133 and a 266MHz FSB CPU sure does look appealing though. I’ll put my money (at least for now) on that combo to rule the sales for mid Q1 2001 to mid Q2 2001, after that I don’t have a clue – yet.


Pages: « Prev  1 2  

Be the first to discuss this article!