Who needs these 64 bits?
On September 23, the new 64-bit processors from AMD, Athlon 64 3200+ and Athlon 64 FX-51 were officially announced. However, Internet users have long been familiar with these processors, since the first AMD64 (Opteron) architecture processors appeared at the beginning of this year. Yes, and some instances of Athlon 64 processors during this time, many were in the hands of.The EPOS also had the opportunity to comprehensively investigate the performance and capabilities of the new processors, as well as work out a new model of the graphics station. Accordingly, the results of their testing are available, and the architectural features of these processors are described in sufficient detail. We will also consider some technical solutions used in the new processors (as we see it) and give some test results.
Who needs these 64 bits?
The main difference between the new AMD processors from the previously released ones is that they are 64 bit. In addition, with the new processors from AMD, we are promised an easy transition to 64-bit computing. But do we want to go to them? Before discussing whether AMD64 architecture is good or bad in general and new processors in particular, we need to decide if we really need this?
First of all, we are always interested in speed, of course. However, the transition to 64-bit computing, it turns out, does not in itself promise us an increase in this very performance. Recall the story. For example, the first 32-bit processor is the i386 processor. At the same clock frequency, its performance turned out to be exactly equal to the performance of the i286 processor. It will be the same with the transition to 64 bit computing. Only very few tasks, such as cryptography, will immediately gain in the form of greater speed, since they operate with very large digits (megabit is now far from the limit in these problems).
The second advantage of 64 bit computing is the ability to directly address a large amount of memory. Here the gain from 64 digits is obvious. 32 bits allow you to directly address only up to 4 GB, and 64 bits allow you to address just the astronomical amount of memory.
The only question that arises is: why?
This question is quite common in reviews on 64-bit processors. Sometimes attention is drawn to the fact that now a gigabyte of RAM is not so common. And in most cases, the opinion is expressed that a large amount of directly addressed memory is not bad, but not a matter of prime necessity.
But in fact, this is not about the available amount of RAM, but about the total amount of addressable memory. And this is far from the same thing. And it turns out that 4 GB for some tasks is already not enough.
There was enough free space on the hard drive, there were no other open applications either, but, nevertheless, the Windows paging file size limit is 4095 MB, and this limitation is not due to the whims of developers, but with a limitation of 32-bit architecture. The screenshot shown in Fig. 1 clearly shows that already now I would like to remove this restriction on the amount of addressable memory of 4 GB.
Of course, ways to overcome the 4 GB barrier have been known for a very long time. Already in the aforementioned i386 processor, a memory manager was built-in, which allowed working with virtual memory up to 64 TB. However, all known methods for increasing the amount of addressable memory are based on dividing the available address space into pages. In this case, you must first select the desired page, and then specify the desired address within the page. This leads to a significant (sometimes more than 30%) decrease in processor performance. Therefore, such methods are applied only in case of emergency.
Thus, the expansion of directly addressed address space is now in demand and in demand for performance reasons.
In connection with the foregoing, another question implicitly arises: If maintaining high performance when addressing a large amount of memory is possible only in 64-bit applications, then where to get these applications, and what to do while there are no such applications yet? Subconsciously, doubts arise whether it is too early to introduce 64-bit processors everywhere. A strange situation: on the one hand, it makes no sense to purchase a 64-bit processor, while there are no 64-bit applications, and on the other hand, who wants to develop serious applications for processors that no one buys. Just some kind of vicious circle.
Let’s recall the story once again, how the transition from 16 to 32 digits went. But this transition did not occur very quickly, and almost imperceptibly for us. A 32-bit processor (i386) appeared in 1985. The first systems based on the i386 processor were released at the end of 1986. But most systems continued to operate under the 16-bit operating system (MS DOS) for a very long time.