CPU Ultimate Guide

The ratings & benchmarks are at the end, but I have to explain myself first:

This is about an article posted in the “RAM” magazine (Sept 2001) – a Greek magazine about PC hardware, with more than 10 years of experience in hardware comparative testing. (I hope they will not sue me for the copyright, I know they are watching THG).

Those guys do hardware tests in their laboratory the same way as THG does, but they go on step further: they produce indices in the scale of 1-10 for the average user to understand not only which product is best, but how much better than the next, while printing their benchmarks for the more experienced users to modify the ratings more closely to their needs.

They did so for the complete series of Intel & AMD CPUs, and I found that very interesting.
Unfortunately I cannot post any link, they don’t have any. Nor do I have the time to make an Excel file with all the numbers and post it somewhere for you to download. So I’ll try to post most of their results:

SETUP

P4 + Intel D850GB (i850) + 256 MB RDRAM
P3 & Celeron + MSI Pro266 Master (Via Apollo Pro 266) + 256 DDR RAM
P3 & Celeron + ASUS CUSL2 (i815E) + 256 SDRAM

Athlon & Duron + EpoX EP-8K7A+ (AMD 761 & VIA 686B) + 256 DDR RAM
Athlon & Duron + MSI K7T &utbo-R (Via KT133A) + 256 SDRAM
Athlon Cooling : FQP38 Global WIN HSF

Common hardware: Hercules 3D Prophet III (GeForce3) , HDD IBM DTLA-307030 30GB Ultra ATA/100

No OC whatsoever! (neither CPU, nor GeForce3)

Win Me, Linux Mandrake 8.0 (dual boot)


Indices

*Office Index is tested with a program of their own who puts the CPU in a series of heavy duty tasks in Word, Excel, PowerPoint & Access 2000 + PhotoShop 6, CorelDraw10, WinMediaEncoder 7. This is very reliable, when rerun the difference is less than 1%.

*3D Gaming Indices is tested with Quake 3 Arena & Mercedes Benz Trucking in 640*480*32b (this + geForce3 ensures CPU is the bottleneck) & 3Dmark2001.

*MPEG-2 to DivX MPEG-4 Index with VirtualDub & DivX 4 (heavy duty for CPU).

*Linux Index: Core 2.4.8 - make config - make dep clean - time make bzimage (don’t ask me!)

*Total Score Index = 45% 3D Games, 35% Office Software, 15% DivX MPEG-4 compression, 5% Linux Core processing (or compiling- I have no idea what that is)

This is their estimate of how a CPU performance should be measured regarding the average user. Of course one might think differently, and change the percentages. A mad gamer may as well ignore all other indices and focus on games + 3Dmark2001. As far as I am concerned, it suites me just fine as it is.


These marks are generated like this: the first product is assigned a mark 10 (perfect), then the others get a mark proportionate to how their score was compared to the first, i.e. if A scored 500, B 490 and C 375 fps, then A gets 10, B:9.8 and C:7.5.

I believe this very useful, since the reader can easily find out that he can buy product C for let’s say half the price of A but still get 75% of the performance.



O.K. then, here is the numbers:

# / _ CPU _ / Total / MSOf / Pic Proces / Of tot / Quake3 / Merced / DivX / Linux / 3D

1. P4 2000/RDRAM: 10__8.79 __ 9.13__ 9.11 __ 10 __ 10 ___ 10 ____ 8.9 ___ 10
2. Ath1400/133/DDR 9.87 __ 10 __ 10 __ 10 __ 8.6 __ 9.6 __ 9.3 __ 10 __ 9.3
3. Ath1333/133/DDR 9.64 __ 9.62 __ 9.74 __ 9.68 __ 8.5 __ 9.4 __ 9 __ 9.7 __ 9.2
4. P4 1800/RDRAM 9.42 __ 8.23 __ 8.72 __ 8.6 __ 9.3 __ 9.5 __ 9.2 __ 8.1 __ 9.7
5. Ath1200/133/DDR 9.13__ 9.02 __9.25 __ 9.12 __ 8 __ 9 __ 8.5 __ 8.9 __ 8.9
6. P4 1700/RDRAM 9.11 __ 7.93 __ 8.42 __ 8.29 __ 9 __ 9.3 __ 8.7 __ 7.9 __ 9.5
7. Ath1400/133/SDRM 8.77 __9.05 __9.71 __ 9.35__ 7.1 __ 8.1 __ 8.2 __ 8.5 __ 8.3
8. Ath1300/100/DDR 8.63 __ 8.97 __ 9.39 __ 9.15 __ 7.3 __ 7.8 __ 7.7 __ 8.6 __ 8.4
9. Ath1333/133/SDRM 8.62__ 8.95 __ 9.43 __ 9.17 __ 7 __ 8 __ 8 __ 8.3 __ 8.3
10.P4 1500/RDRAM 8.43 __ 7.24 __ 7.75 __ 7.59 __ 8.2 __ 8.8 __ 7.9 __ 7.3 __ 9.1
11.Ath1000/133/DDR 8.33 __ 8.14__ 8.48 __ 8.27 __7.2 __ 8.4 __ 7.6 __ 7.8 __ 8.5
12.Ath1300/100/SDRM 8.24
13.Ath1200/133/SDRM 8.23
14.P4 1400/RDRAM 8.12 __ 6.88 __ 7.68 __ 7.35 __ 8 __ 8.3 __ 7.6 __ 7 __ 8.9
15.Ath1100/100/DDR 7.96
16.P4 1300/RDRAM 7.75
17.Ath1100/100/SDRM 7.65
18.Ath1000/100/DDR 7.6
19.Ath1000/133/SDRM 7.51 __ 7.41__ 8.17__ 7.73 __ 6.2 __ 7.2 __ 6.9 __ 6.9 __ 7.6
20.Ath1000/100/SDRM 7.23
21.Ath 900/100/DDR 7.18
22.P3 1000/133/SDRM 6.97 __ 7.11__ 7.60 __ 7.39 __5.9 __ 6 __ 5.8 __ 6.8 __ 7.6
23.Ath 900/100/SDRM 6.97
24.P3 1000/133/DDR 6.95
25.Duron 950/100/DDR 6.88 __ 6.37 __ 7.73 __ 6.95 _ 5.6 __ 6.3 __ 6.8 __ 5.9 __ 7.4
26.Duron 900/100/DDR 6.72
27.Duron 850/100/DDR 6.53
28.Duron 950/100/SDRM 6.52
29. P3 866/133/DDR 6.46
30. P3 866/133/SDRM 6.42
31. Duron 900/100/SDRM 6.38
32. Duron 800/100/DDR 6.32
33. Duron 850/100/SDRM 6.19
34. Duron 750/100/DDR 6.07
35. Duron 800/100/SDRM 6.01
36. Duron 750/100/DDR 5.79
37.Celeron 850/100/DDR 5.45
36.Celeron 850/100/SDRM 5.33
37.Celeron 800/100DDR 5.26

Sorry for the difficulty in reading, I tried to line up the columns as much as I could.

Here are some benchmarks that generate the previous marks
(fps & 3D Mark 2001:higher is better, seconds: less is better)

# / CPU /Office total / Quake3 / Merced / DivX / Linux / 3D Mark 2001
1. P4 2000/RDRAM 627.5sec __ 164.9fps __ 161.2fps __ 111sec__ 249sec __ 7047
2. Ath1400/133/DDR 571.5__ 141.4__ __154 __ 120 __ 221 __ 6574
3. Ath1333/133/DDR 590.5 __ 140.4 __ 152 __ 124 __ 228 __ 6485
4. P4 1800/RDRAM 664.6 __ 152.9 __ 153.8 __ 121 __ 273 __ 6867
5. Ath1200/133/DDR 626.9 __ 132 __ 144 __ 131 __ 247 __ 6284
6. P4 1700/RDRAM 689.1 __ 148 __ 149 __ 127 __ 281 __ 6709
7. Ath1400/133/SDRM 610.98 __ 117 __ 130 __ 136 __ 260 __ 5868
8. Ath1300/100/DDR 624.45
9. Ath1333/133/SDRM 623.4
10.P4 1500/RDRAM 752.87 __ 135.5 __ 141.9 __ 140 __ 303 __ 6386
11.Ath1000/133/DDR 691.47
12.Ath1300/100/SDRM 644.48
13.Ath1200/133/SDRM 664.1 __ 111 __ 125 __ 146 __ 284 __ 5661__
14.P4 1400/RDRAM 777.1 __ 131 __ 133 __ 147 __ 317 __ 6239
.
.
25.Duron 950/100/DDR 822.5 __ 92 __ 102 __ 164 __ 377 __ 5185

© Sept 2001 “RAM” magazine – Greece
(hey, RAM guys don’t get mad, look at it as publicity !)

Sorry, I couldn’t fill all marks and benchmarks but I think you get the point.
(Maybe later I will edit it, adding some more numbers).



Summary

1. This is as much close to perfect testing as one can expect. Yeah, sure I would like more chipsets and motherboards, but these are already a huge load. Remember that Tom only tests as much as 5-6 setups and without bothering to add up the results in a quantitative comparison. I can already hear people screaming that with the X or Y motherboard P4 or Athlon might have scored better, but I don’t think there would be great changes in the overall ranking. Anyway, those where their choice.
2. AMD must use this test as a commercial. Not only Athlons are faster than P4 clock by clock, but also they can beat those with 400-500MHz higher clocks. Even in applications like games and Divx where SSE2 & higher MHz and RDRAM transfer rate should make a difference, Athlons are able to beat even higher clocked P4s. Even SDRAM Athlons can match or beat RDRAM P4s for crying out loud!
3. Finally, P3 and Celerons are dead buried and long forgotten. Their architecture is so old that they cannot even benefit from DDR over SDRAM!


Phew… that’s all.




just keep on feeding me those numbers!