In the face of a variety of CPUs, how can we see the performance of a single CPU at a glance?
The following describes the factors affecting the performance of mainstream CPUs and the overall performance calculation formula.
1. There is no doubt that the performance of CPU of the same type is proportional to the clock speed, and it is basically linearly related. For example, the p4660 performance of 3.6g is about 20% stronger than that of p4630 of 3.0g.
2. Architecture is the most important factor affecting the performance of the same-frequency CPU. Of course, different architectures are suitable for different jobs. For example, p4 is suitable for video work, while amd X-Dragon is good at games. Here, we take the average value. Assume that your computer has 30% of the time to play games, 30% of the time for multimedia processing, and 40% of the time for office and computing. In this case, the architecture parameter of PD Based on P4, sai Yang and similar architectures of Prescott architecture is 1.0; athlon64, flash dragon, haolong, FX and athlon X2 Based on K8 architecture, the architecture parameter is 1.50, the Pentium M series architecture parameter is 1.6, and the latest Conroe architecture parameter is 1.9.
3. cache is the place where the CPU calls data directly. The cache speed is about 10 times faster than the current memory, and there is a lower latency. The greatest impact on CPU performance is the second-level cache. Based on the 2 m cache of P4 630, the cache parameter is 1.0. The performance is improved by about 5% every time the cache is doubled, and the performance is reduced by about 5% every half. For amd series Processors, because they integrate a memory controller, the standard parameter 1.0 is adopted as the K.
4. bus P4 uses the front-end bus 1.00 MHz as the standard, 533m is 0.95, 1066m is 1.03, amd series uses HT bus 1.00 as the standard, 800m is 0.97, Conroe is 1066 as the standard, 1333 is 1.02.
5. The first four items of multi-core basically determine the single-task performance of the CPU, while the dual-core U is almost doubled in multi-task processing. However, in most cases, we only use one core (or two cores are used in turn). Here, we assume that the computer has 1/3 of the time to run multiple tasks, in this way, each time the number of cores doubles, the practical performance will increase by 33%. As for the hyper-Threading Technology, tests prove that its multi-task performance will be improved by about 30%, taking advantage of the 1/3 usage time, which will increase by about 10%.
Let's first calculate several common CPU
Clock speed 1.4 architecture K8 cache 256 k bus 800 single core, its performance is 1400x1.5x0.95x0.97/30 = 65
Clock speed 1.6 architecture K8 cache 128 K bus 800 single core, its performance is 1600x1.5x0.90x0.97/30 = 70
Clock speed 2.53 architecture Prescott cache 256 k bus 533 single core, its performance is 2530x1.0x0.85x0.95/30 = 68
These three CPUs are 2500 +, 2600 +, and cd2.53 respectively. The prvalue is similar to the calculation result.
The remaining CPUs can be calculated by yourself. The following parameters are for your reference only.
CPU |
Clock speed |
Architecture Parameters |
Level 2 Cache |
Cache Parameters |
Bus |
Bus Parameters |
Multi-core |
Multi-core parameters |
Performance |
P4 630 |
3000 |
1 |
2048 |
1 |
800 |
1 |
HT |
1.1 |
110 |
P4 531 |
3000 |
1 |
1024 |
0.95 |
800 |
1 |
HT |
1.1 |
105 |
P4 670 |
3800 |
1 |
2048 |
1 |
800 |
1 |
HT |
1.1 |
139 |
P4 524 |
3066 |
1 |
1024 |
0.95 |
533 |
0.95 |
HT |
1.1 |
101 |
PD 1, 805 |
2666 |
1 |
1024 |
0.95 |
533 |
0.95 |
Dual-core |
1.33 |
107 |
PD 1, 820 |
2800 |
1 |
1024 |
0.95 |
800 |
1 |
Dual-core |
1.33 |
118 |
PD 930 |
3000 |
1 |
2048 |
1 |
800 |
1 |
Dual-core |
1.33 |
133 |
PE 965 |
3730 |
1 |
2048 |
1 |
1066 |
1.03 |
Dual-core + HT |
1.46 |
187 |
PM 740. |
1733 |
1.6 |
2048 |
1 |
533 |
1 |
Single-core |
1 |
92 |
PM 780. |
2260 |
1.6 |
2048 |
1 |
533 |
1 |
Single-core |
1 |
120 |
CD 1, 2.53 |
2533 |
1 |
256 |
0.85 |
533 |
0.95 |
Single-core |
1 |
68 |
A3200 + |
2000 |
1.5 |
512 |
1 |
1000 |
1 |
Single-core |
1 |
100 |
Op146 |
2000 |
1.5 |
1024 |
1.05 |
1000 |
1 |
Single-core |
1 |
105 |
A3400 + |
2200 |
1.5 |
512 |
1 |
800 |
0.97 |
Single-core |
1 |
107 |
A3500 + |
2200 |
1.5 |
512 |
1 |
1000 |
1 |
Single-core |
1 |
110 |
A3700 + |
2200 |
1.5 |
1024 |
1.05 |
1000 |
1 |
Single-core |
1 |
116 |
A3800 + |
2400 |
1.5 |
512 |
1 |
1000 |
1 |
Single-core |
1 |
120 |
Sp2500 + |
1400 |
1.5 |
256 |
0.95 |
800 |
0.97 |
Single-core |
1 |
65 |
Sp2600 + |
1600 |
1.5 |
128 |
0.9 |
800 |
0.97 |
Single-core |
1 |
70 |
2x3600 + |
2000 |
1.5 |
256 |
0.95 |
1000 |
1 |
Dual-core |
1.33 |
126 |
2x3800 + |
2000 |
1.5 |
512 |
1 |
1000 |
1 |
Dual-core |
1.33 |
133 |
2x4600 + |
2400 |
1.5 |
512 |
1 |
1000 |
1 |
Dual-core |
1.33 |
160 |
2x4800 + |
2400 |
1.5 |
1024 |
1.05 |
1000 |
1 |
Dual-core |
1.33 |
168 |
FX 57 |
2800 |
1.5 |
1024 |
1.05 |
1000 |
1 |
Single-core |
1 |
147 |
FX 62 |
2800 |
1.5 |
1024 |
1.05 |
1000 |
1 |
Dual-core |
1.33 |
196 |
FX 74 |
3000 |
1.5 |
1024 |
1.05 |
1000 |
1 |
Dual-core X2 |
1.76 |
277 |
E6300 |
1866 |
1.9 |
2048 |
1 |
1066 |
1 |
Dual-core |
1.33 |
157 |
E6600 |
2400 |
1.9 |
4096 |
1.05 |
1066 |
1 |
Dual-core |
1.33 |
212 |
X6800 |
2933 |
1.9 |
4096 |
1.05 |
1066 |
1 |
Dual-core |
1.33 |
259 |
Intel Xeon 5160 |
3000 |
1.9 |
4096 |
1.05 |
1333 |
1.02 |
Dual-core |
1.33 |
271 |
Q6700 |
2666 |
1.9 |
4096 |
1.05 |
1066 |
1 |
Quad-core |
1.76 |
312 |
If you want to view the performance of a single task, only the values before the multi-core parameters are used for calculation. If your multi-task runs for a long time, you can actually multiply a suitable multi-core parameter.