The victory of the losers, or the story of the creation of FLASH memory (part2)
This type of memory was made in the form of a two-dimensional matrix of conductors, at the intersection of which a thin jumper (20-30 μm) was created from metal or…

Continue reading →

The victory of the losers, or the story of the creation of FLASH memory (part3)
We were very happy and proud; you know, young engineers with their own hands were able to make the first “flip” chip. After some discussion between Gordon and Andy, Gordon…

Continue reading →

"Bucket" Compression
An old army joke about a compression bucket that the "experienced" senior comrades are instructed to bring to a young soldier, checking his quick wits, is more relevant than ever…

Continue reading →

The oldest Athlon

No matter how unique a processor may be, there will always be tasks for which its performance will not be enough. Most often these are tasks of three-dimensional modeling (for example, 3D Max). As a rule, you can only improve your life a little in this situation by installing at least two processors in one machine. Therefore, a processor that cannot be installed in a multiprocessor configuration cannot be positioned as a processor for powerful workstations, not to mention servers. Recently, AMD has solved this problem. A new processor is released specifically designed to work in a multiprocessor configuration. This is Athlon MP and its latest modification is Athlon MPX. With the release of these processors, AMD begins to seriously lay claim to the market part of the most productive workstations and servers.

Today we will not give test results showing how many times the productivity increases. It is clear that in many ways the increase in productivity depends on the application running on a dual-processor machine. We only note that performance improvement is observed even for tasks that do not “understand” the second processor. Of course, the speed of the task itself in this case, perhaps, does not change. However, in the Windows operating system, several service tasks are always performed in parallel. In a dual-processor configuration, just the processor occupied by any application is less “distracted” by solving the service tasks of the operating system.

A more interesting question is how the Athlon MPX and Athlon XP processors differ. AMD’s position is clear: Athlon MPX is a processor for dual-processor systems, and Athlon XP is for single-processor systems. But Athlon XP also works in a dual-processor configuration. Is there any deception here?

Of course, you can’t do anything without deception. But in this case, more often we mislead ourselves. Why did we decide that the ability to work in a dual-processor configuration is determined by the processor? The specification of multiprocessor systems presents a lot of hardware requirements. But most of them actually determine not even the parameters of “iron”, but rather the conditions and method of its use. These are, for example, memory requirements (ordering records in memory, locking methods, allocation of service areas for specific tables necessary for the operation of two processors). The most serious hardware requirement is the requirements for the interrupt management method. Interrupts should only come to the processor to which they are addressed, or to the one that is currently capable of handling the interrupt. This leads to the fact that in a multiprocessor system interrupt control becomes distributed and requires special programmable interrupt controllers APIC (Advanced Programmable Interrupt Controller). The presence of such controllers is the main difference between a multiprocessor board and a uniprocessor one. As you can see, these controllers have nothing to do with the processor. As for processors, they only have to provide the chipset with the ability to synchronize the contents of their cache. Such an opportunity is inherent in Athlon processors initially. Moreover, a more powerful synchronization protocol is used than Intel processors. In Pentium processors, the cache can be in one of four states (data in the cache is the same as in memory; data in the cache line is modified; data in the cache of both processors and in memory is the same; data in the cache is incorrect). In Athlon processors, the cache may have another state indicating which processor is the “owner” of the data and, accordingly, which processor is responsible for updating the contents of the memory. This is due to the fact that the point-to-point architecture and the presence of additional information signals on the EV6 bus allow the chipset to synchronize the Athlon processor cache by directly transferring data from the cache of one processor to the cache of another, bypassing memory. But these great features are in the core of the processor. Therefore, any processor will work successfully in a dual-processor board, regardless of how AMD positions it. So there is no obstacle to the fact that any modern AMD processor, even Duron, works in the dual-processor configuration.

The victory of the losers, or the story of the creation of FLASH memory (part1)
It would seem that in common between a mobile phone, a GPS navigator, a digital camera and a modern voice recorder. The first thought that comes is that these are…

...

What is a model number?
Before comparing with competing processors, let's look at the performance of Athlon XP processors with different CPUIDs. The model number is simply the number of a specific model in the…

...

Athlon XP: Awaiting 64-bit Architecture
On April 22, AMD announced a server version of its 64-bit Opteron processor. Unfortunately, the server version was announced. In addition, most likely at the beginning the 64-bit platform will…

...

The oldest Athlon
No matter how unique a processor may be, there will always be tasks for which its performance will not be enough. Most often these are tasks of three-dimensional modeling (for…

...