Games against CPUs. Part II.

We decided to write an update to our CPU gaming performance analysis. In this article we will focus on high graphics quality settings and will try to find out what CPU speed is sufficient for comfortable gameplay. We will also pay due attention to the budget processor models and overclocking efficiency.

by Ilya Gavrichenkov
11/11/2005 | 01:12 PM

A few days ago we posted an article covering a detailed test session of 28 processors in the latest gaming applications (for details see our review called Contemporary CPUs and New Games: No Way to Delusions!). This article allowed us once again to make our conclusions about the processor performance in contemporary games and the best CPU choice for gamers (thanks to our article a lot of AMD fans out there got one more reason to be proud). However, the obtained results caused some concerns about their actual value. The thing is that having analyzed the results obtained in the latest gaming titles, a lot of questions arose, which we couldn’t answer immediately. Firstly, we have pointed out that the tests in gaming applications following our well-established time-proven methodology are somewhat synthetic. The gaming settings we used during our CPU test session would never actually be used by the real gamers. That is why the number we obtained are very far away from what the gamers would observe during real gaming experience. Why don’t we test the CPUs with the more realistic gaming settings then, when the image quality is adjusted so that the gamer could get a nice picture and at the same time enjoy acceptable fps rate ensuring comfortable experience. We decided to give it a try.


Secondly, the main conclusion drawn from the last article was the fact that the CPU speed doesn’t actually matter that much during actual gameplay. We could clearly see that contemporary 3D shooters require a powerful graphics subsystem in the first place. As for the CPUs, contemporary games do not take real advantage of the potential of high-speed processors. However, despite all those great conclusions one question remained unanswered: what CPU would be sufficient for comfortable gaming experience. In the previous article we tested high-end and mainstream processors. All of them performed well enough. Now, hoping that we could finally get to find some real losers, we decided to include a few value CPU models into our comparative test session. So, our second goal would be to see if the Value and Budget CPUs can actually do well in the latest gaming applications.

In other words, we felt like one article about the CPU performance in contemporary games was far not enough to cover all the interesting topics, so we would like to offer you the second part of our discussion on the matter. This investigation will help us to answer the question: what CPU of all the models available in the market today should be powerful enough to ensure comfortable and enjoyable gaming experience in the latest 3D shooters.

Testbed and Methods

First of all we have to select the CPUs, which will take part in the second round of our tests today. The major theory proven in our previous article states that all the high-end and mainstream processors provide pretty much the same level of gaming performance. Therefore, this time we will not all the 28 CPUs retested. This time we will only go for the eldest and youngest CPU models in the Athlon 64 and Pentium 4 processor families, and will also add here two dual-core processors, one from each of the rivalry camps.

As for the Budget CPUs, which will participate in this test session alongside with they high-end fellows definitely boast a much more numerous representation. The thing is that these CPUs tend to have much greater influence on the overall gaming performance. We took three models from the AMD Sempron and three from Intel Celeron D processor families: the top mode, the slowest model and one model in-between. Moreover, to make sure we get very illustrative results we also added Athlon 64 for Socket 754 to our group of testing participants.

Of course, we couldn’t disregard overclocking this time. Among the obtained results you can find some numbers taken for the systems with the overclocked Sempron, Celeron D and Pentium 4 CPUs. We tried to overclock these processors to some typical values, such as 3.8GHz for Celeron D, 2.6GHz for Sempron and 4.2GHz for Pentium 4. Here we didn’t test the overclocked Athlon 64, because the typical value for Athlon 64 overclocking is 2.8GHz, which is the working frequency of Athlon 64 FX-57. And this CPU participates in our tests anyway.

I should also say a few words about the graphics settings in the gaming applications we used this time. Since our primary goal was the measure the CPU performance with the real settings used by gamers out there, we started with 1280x1024 screen resolution with enabled 4x full-screen anti-aliasing and anisotropic filtering with the maximum quality settings. Most games worked quite fine with these settings on the NVIDIA GeForce 7800 GT graphics card we had in our testbed. However, if we failed to reach the acceptable fps rate with these settings in any of the games (we considered 40-60 fps to be the minimum acceptable gaming performance), we reduced the screen resolution to 1024x768 and repeated the tests with the same image quality settings.

We decided not to test the CPUs with a slower graphics card than NVIDIA GeForce 7800 GT. We believe these tests do not make any sense, because in case of a slower graphics card we will have to agree to less aggressive graphics quality settings in order to achieve the same acceptable performance rate of 40-60fps. So, the quality picture of the CPU performance in case of a slower graphics subsystem would hardly be any different.

So, let’s sum up and list all the hardware components that were involved in our today’s test session:

The tests were carried out with the BIOS Setup of the mainboards used adjusted for maximum performance.

Performance in Games

Battlefield 2

The high graphics quality settings in battlefield 2 didn’t affect the gaming performance that much. Systems with fast processors reach pretty high fps rates without any problems.

However, the obtained results still gave us some food for thought. Firstly, we can clearly see that the performance of systems with fast processors is limited by the graphics subsystem. Any Pentium 4 and Athlon 64 CPUs as well as the top Sempron processor model with the 3400+ performance rating produces the same results.

The actual performance differences can be observed only between the Celeron D and the remaining Sempron processor models. If we compare their results we will clearly see that in the budget segment AMD CPUs would be a better choice for gaming needs, especially in Battlefield 2. Even overclocking doesn’t help Celeron D to win. Of course, the overclocked Celeron D performs better than the CPU working at the nominal frequencies, however it still fails to get into the leading group. Sempron overclocking appears much more efficient, because once it reaches 2.6GHz its performance catches up with that of Athlon 64.


We have already mentioned that the new F.E.A.R. shooter is very sensitive to the graphics subsystem performance and requires a lot from it. In particular, it is impossible to get any acceptable fps rate with the highest graphics quality settings and highest screen resolution. Thus, we could only reach 31 frames per second in 1280x1024 with maximum graphics quality settings when running this game on our test platforms equipped with the GeForce 7800 GT graphics card. Therefore, we reduced the screen resolution to 1024x768.

The second indication of specifically high requirements to the graphics subsystem imposed by the F.E.A.R. game can be clearly seen on the diagram with the benchmarking results. All CPUs, including even the slowest Celeron D and Sempron models provide pretty much the same performance rate.

So, in this case the game doesn’t really care what CPU is installed in your system, so F.E.A.R. should definitely be excluded from the list of CPU dependent games of today.

Serious Sam 2

The performance in Serious Sam 2 is quite scalable depending on the CPU power. Moreover, the connection between the fps rate with high graphics quality settings and the CPU speed can be observed in all cases even with the most powerful processors. Although in the latter case this connection is not that strong any more. For example, the frequency difference between Athlon 64 FX-57 and Athlon 64 3200+ is 40%, while the performance difference is only 6.5%. The only comment I have to make right now is about the price difference between these two: it makes 665% and easily beats both previous differences.

The performance difference between the fastest and the slowest Pentium 4 CPU model is considerably greater. Serious Sam 2 shooter, like many other games, feels much better when working in a K8 based system than in a NetBurst based one. Somewhere in the middle of the NetBurst based CPUs line-up the processors can no longer provide sufficient computational power for comfortable gaming. Therefore, if we look at the results obtained in systems with slower Pentium 4 CPUs, we will see that the performance is limited not by the graphics card any more but mostly by the CPU.

Inexpensive Sempron and Celeron D processors, just like the youngest Pentium 4 solutions, limit the fps rate in this game when we run it with the maximum graphics quality settings. Of course, Sempron appears far ahead of Celeron D, and when overclocked - even gets into the leading group (these are the CPUs that load the graphics subsystem to the full extent). Overclocked Celeron D cannot boast anything like that and even falls behind Pentium 4 630.

I would like to specifically stress the performance of our dual-core testing participants in Serious Sam 2 game. The second test session carried out in this 3D shooter with the settings close to real gaming conditions indicates that dual-core CPUs perform very well here. They are evidently faster than their single-core fellows working at the same clock speeds. In other words, Serious Sam 2 knows to deal with Athlon 64 X2 and Pentium D in the most efficient way.

Quake 4

In the previous article we said that the CPU speed affects the gaming performance in Quake 4 quite tangibly. This time we are about to take these words back. Once we increased the screen resolution and set the graphics quality to the maximum all high-end and mainstream CPUs started running equally fast. The performance difference between them never exceeded 6%. Therefore, I would warn you against using Quake 4 as a criterion for selecting the best gaming CPU, if you are not planning to go for a budget processor.

However, I have to admit that even in this game Celeron D and Sempron CPUs can easily compete with the more expensive solutions when it comes to the fps rate. It is solely due to efficient overclocking that allows these budget solutions to feel pretty powerful.

If we take a really close look at the diagram above, maybe even through a magnifying glass, we might be able to make some conclusions about the relative CPU performance. For example, we can once again confirm that Quake 4 uses the advantages of dual-core architecture.

Call of Duty 2

First of all, I would like to correct one mistake I made in the previous article. I said that Call of Duty 2 uses Quake 3 gaming engine. In fact I was wrong. The Quake 3 engine was used for all previous parts of this game, and for the new version of call of Duty shooter the developers introduced their own DirectX9 engine. As we have seen in the first part of our review, this engine needs powerful graphics subsystem. And we found one more piece of evidence proving this point when we working on the current article. While Quake 4 ran fine on our testbeds in 1280x1024 with enabled anti-aliasing and maximum graphics quality settings, Call of Duty 2 could hardly generate 30 fps in the same testing conditions and with the fastest CPU. Therefore, we decided to run the tests in lower resolution set to 1024x768 but still with maximum image quality settings.

However, lowering the resolution hardly brought us anything to discuss. All the testing participants perform more or less equal. The performance difference between the fastest and the slowest CPU in this test session made only 5 fps, which 15%. Of course, this is slightly higher than the performance difference we observed in F.E.A.R. game, but it is still not really impressive. I believe that the performance in call of Duty 2 can hardly be regarded as an argument in favor of any CPU.

Anyway, although we have just deprived Call of Duty 2 from the “CPU benchmark” title, I would like to draw your attention to the CPU that stands at the head of the diagram. It is Intel Pentium 4 overclocked to 4.2GHz. This is a true wonder: I cannot remember seeing anything like that for a long time.


Everything we said in our previous article called Contemporary CPUs and New Games: No Way to Delusions! was absolutely right. It is true: you don’t need a high-end processor for real gaming with realistic settings and high image quality. The gaming performance will still be limited by the graphics card. The recommended system requirements mentioned by all the game developers are absolutely correct. Do not be surprised that the game developers mention Pentium 4 3GHz+ and Athlon 64 2GHz+ processors as the minimum suitable CPUs for comfortable gameplay, even though today we can get 3.8GHz Intel CPUs and 2.8GHz AMD CPUs easily. It is true that faster CPUs than those mentioned in the minimum system requirements do not really stimulate and significant fps rate increase. So, the slower processors models from the Pentium 4 and Athlon 64 processor families can cope easily with the latest generation 3D shooters. So, if you already have one of those CPUs in your home system, then you shouldn’t worry about upgrading them for your gaming needs.

The low-end processors is a totally different story. Here we can actually see that the CPU speed does affect the fps rate in many latest gaming titles. Although, there are some opposite examples, such as F.E.A.R., for instance.

It is evident that Sempron processors perform better in games than Celeron D. moreover; overclocked Sempron CPUs can sometimes run almost as fast as the top processor models. Unfortunately, Celeron D cannot boast the same success from overclocking.

So, the best gaming system configuration from the price-to-performance point of view should definitely include a powerful graphics card, which will determine not only the image quality, but the gaming comfort in general. As for the CPU, you shouldn’t spend tons of money for a high-end model. You will be able to get almost the same performance from a CPU like Athlon 64 3200+ or Pentium 4 650.

And in conclusion I would like to say a few words about overclocking. The results suggest that only if you have a budget CPU overclocking might be a good idea for you. By increasing the CPU working frequency beyond the nominal value, you can really speed up your gaming system, sometimes even up to the high-end CPUs level. However, if you have a mainstream processor or better, overclocking will have no real effect on the frames per second you get in the latest 3D shooters.