new Kaveri – cheaper and more economical / processors and memory

Usually, when new APU generations were brought to the desktop market, AMD adhered to the tactics of mass announcements, when the complete lines of processors, beginning with the older ones and ending with the younger models, were produced almost simultaneously. The maximum that the company could afford was a small time gap between the actual appearance of various types of models in stores, but it usually took not more than a couple of months to wait for the sale of the entire set of processors.

With Kaveri, the situation has developed quite differently. The announcement of this processor design took place in January, and then only three quad-core models were officially introduced: A10-7850K, A10-7700K and A8-7600. And on sale this trio as a whole did not appear: the younger model A8-7600 to the counters then did not reach. As a result, until the most recent time, when eight long months passed since the announcement of desktop Kaveri, only the two oldest modifications could be bought.

A new technological process for AMD with a 28-nm standard became a stumbling block, which, as it turned out, did not allow easy and simple mass production of APU with reduced heat generation. The principal difference between the problematic A8-7600 and the other two Kaveri is its energy efficiency: it is designed for a 65-watt rather than a 95-watt thermal package, and AMD could not produce such APUs in sufficient quantities. Therefore, the supply of this modification was very limited, and only the partners of the company from among the large assemblers on special orders could receive these processors.

But progress does not stand still. Gradually, AMD, along with its production partner, Globalfoundries, was able to deal with technological problems, and it even gave AMD the opportunity to start releasing a mobile Kaveri with a calculated heat dissipation at 35W a month ago. And now, finally, the company decided to restore order and in its desktop line. As a result, a couple of weeks ago it was announced the release and fast availability of the long-awaited Kaveri varieties, complementing the existing short range with three new (almost new, if more accurate) processors: A10-7800, A8-7600 and A6-7400K. Novelties vary greatly in their characteristics and even have different numbers of computing cores and graphic computing devices, but they all have one thing in common: they are energy efficient and have a typical heat dissipation of 65 or 45 W.

Thus, the line of hybrid processors Kaveri, intended for use in desktop computers, acquires completeness and logical completeness. But not only: it is becoming much more relevant to end users. When we tested the senior representative of the Kaveri family, A10-7850K, we liked it with its fast graphics core, but at the same time it was frustrated by high power consumption, weak computational performance and obviously overpriced. The new processors can partially eliminate the flagship’s flaws. They have noticeably less heat release and lower cost, but the differences from the old models in terms of frequencies and internal structure of the graphics core are not very significant. Consequently, new items have a more favorable combination of performance and price, as well as performance and energy consumption: at least, so everything looks at first glance.

That’s why practical familiarity with the new Kaveri for desktop computers seemed to us quite interesting, and we took the senior Socket FM2 + processor from among the new products, A10-7800, for the test from AMD company.

⇡ # Read more about A10-7800

With the release of hybrid processors, Kaveri AMD has implemented the HSA (Heterogeneous System Architecture) architecture, which gives scalar general purpose kernels and shader clusters of the graphics core equal access to system memory. Since then, AMD’s marketing department has abandoned the separate calculation of x86-cores and operates with the concept of generalized cores, which now include graphic resources – shader clusters. In accordance with this paradigm, A10-7850K was represented as a 12-core APU, and A10-7700K was represented as a 10-core heterogeneous processor.

The new APs of the Kaveri family also support HSA, and therefore AMD refers to the A10-7800 as a 12-core processor, the A8-7600 as a 10-core, and the A6-7400K as a six-core processor. In fact, from the positions of the x86 architecture, the A10-7800 and A8-7600 are quad-core, and the A6-7400K is a dual-core processor. The full characteristics of the entire line of Kaveri for desktop computers look like this:

AMD A10-7850K AMD A10-7800 AMD A10-7700K AMD A8-7600 AMD A6-7400K
] The code name Kaveri Kaveri Kaveri Kaveri Kaveri
Kernels 4 cores (2 modules) 4 cores (2 modules) 4 cores (2 modules) 4 cores (2 modules) 2 cores (1 module)
The microarchitecture Steamroller Steamroller Steamroller Steamroller Steamroller
The processor socket Socket FM2 + Socket FM2 + Socket FM2 + Socket FM2 + Socket FM2 +
The unlocked multiplier There are No There are No There are
The clock speed 3.7 GHz 3.5 GHz 3.4 GHz 3.1 GHz 3.5 GHz
The frequency in the turbo mode Up to 4.0 GHz Up to 3.9 GHz Up to 3.8 GHz Up to 3.8 GHz Up to 3.9 GHz
L2-cache 2×2 MByte 2×2 MByte 2×2 MByte 2×2 MByte 1 MB
The graphic core Radeon R7 Radeon R7 Radeon R7 Radeon R7 Radeon R5
Architecture of the GPU GCN 1.1 GCN 1.1 GCN 1.1 GCN 1.1 GCN 1.1
The shader processors 512 512 384 384 256
Frequency of the GPU 720 MHz 720 MHz 720 MHz 720 MHz 756 MHz
Support for DDR3 DDR3-2133 DDR3-2133 DDR3-2133 DDR3-2133 DDR3-1866
TDP 95 W 65 W 95 W 65 W 65 W
The official price $ 173 $ 153 $ 152 $ 101 $ 77

The new A10-7800 processor, which, in fact, will be discussed further, at first glance, is very similar to the flagship in the Kaveri line, A10-7850K. It has the same number of computational cores and the same graphical core Radeon R7, and there are small differences only in the clock frequency, which is 100-200 MHz lower. At the same time the novelty is deprived of the unlocked multiplier, but it belongs to the 65-watt thermal package. Thus, the A10-7800 is presented as an excellent Socket FM2 + processor for those users who do not need overclocking. It is slower than the flagship by only 5 percent, but it is 30 percent more economical and 12 percent cheaper.

The processors A8-7600 and A6-7400K presented along with A10-7700 also seem quite interesting suggestions. A8-7600 we already had the opportunity to test in detail in January, and the A6-7400K is an unlocked dual-core that plays in the same niche as the recently introduced Intel Pentium G3258. However, we can not make any conclusions about the overclocking potential of the A6-7600K, but nevertheless, as a base for cheap computers, it will be interesting regardless of anything.

As for the A10-7800, it is comparable in price to the older Intel Core i3 processors. And this, perhaps, is a more just positioning. As we could see from the example of A10-7850K, AMD’s quad-core processors based on the Steamroller microarchitecture noticeably lose in performance in common tasks with the four-core Haswell. But for Intel dual-core processors, they can be a very good alternative, so the price of A10-7800 does not look as unjustified as the older Kaveri.

However, there is one nuance: Kaveri processors compete not only with Intel processors, but also with their predecessors – hybrid processors Richland, which continue to remain on the market. Of course, Kaveri is a more relevant product, manufactured using the most advanced 28 nm process technology for AMD. The computing cores in these processors have a more perfect Steamroller microarchitecture, and the graphics are built on the same GCN 1.1 architecture as the modern flagship graphics cards. However, all this does not make Kaveri definitely better than Richland. The fact is that with the introduction of a new design into heterogeneous processors, AMD has slightly revised its approach to their design. First, the 28nm process introduced into use began to be optimized not for conquering high clock speeds, but for more dense placement of transistors and lower heat dissipation. Secondly, although this production technology allowed us to significantly increase the complexity of the semiconductor crystal, the main priority in the development of the potential opened up strongly towards the graphic component. She in Kaveri is given about 47 percent of the total transistor budget.

As a result, the A10-7800, which by all formal features should be a replacement for the old 65-watt quad-core A10-6700, its full-fledged alternative is still not. The A10-6700 has significantly higher clock speeds: the base clock is 3.7 GHz and the maximum is 4.3 GHz. Of course, one can hope that the lower frequencies of A10-7800 will be compensated by the advantages of the Steamroller microarchitecture, which assumes the appearance of an individual instruction decoder in each computational core. However, as practice shows, all changes made in the new microarchitecture in most cases increase the number of instructions processed per clock by only a few percent and do not cover the backlog in the clock frequency to the full.

Therefore, in computing performance in common applications, the A10-7800 is likely to lose to its predecessor of the genus Richland. But the graphics core of the novelty, no doubt, is much more powerful. The number of shaders in the built-in GPU of the A10-7800 processor is 1.5 times higher and brought up to 512 pieces, plus the more efficient GCN architecture that replaced the VLIW4.

The A10-7800, like the other 65-watt Kaveri processors, has one more interesting feature – a configurable TDP. This means that any of the 65-watt APUs have two operating modes. One – standard, and the second – with increased energy efficiency. In the second mode, which can be enabled via the motherboard BIOS for any processor with a 65 watt thermal package, its TDP is limited to an even more serious 45W boundary. In this mode, the operating frequencies of the processor, of course, are further reduced, but AMD promises that the real average-weighted performance level does not suffer much. For example, for the A10-7800, such an estimate is given: in the 45-watt mode, the performance drops by only 6-7 percent relative to the 65-watt, but the power consumption drops by almost a third.

In other words, while Intel offers a whole lot of energy-efficient CPUs with different heat dissipation values, AMD decided to go the other way and offers processors whose level of profitability can be set by the user himself. This is a very good option for enthusiasts who are assembling compact and quiet systems like HTPC on the basis of AMD’s APU, but not very convenient for computer suppliers, which in the description of their systems will be forced to indicate not only the name of the processor, but also their frequencies, which can now Differences depending on the target TDP.

Here, for example, how the A10-7800 works. In a 65-watt mode, under a high multi-threaded computing load, its typical frequency is set at 3.6 GHz. This is 100 MHz above the nominal frequency, which is stated in the specifications.

With a small single-threaded load, the processor, as promised by the manufacturer, is capable of overclocking to 3.9 GHz.

If for the same processor to lower the target TDP to 45 W, then the typical operating frequency for a multithreaded load will be only 3.0 GHz.

At the same time, auto-acceleration with a small load on computing resources will be limited to a frequency of about 3.5-3.6 GHz.

The maximum frequency of the graphics core is not subject to such a pronounced change. Both in the 65-watt and in the 45-watt mode it remains equal to the claimed in the specifications 720 MHz.

However, if you select a 45-watt heat pack for the A10-7800, this frequency will also be unstable and will decrease depending on what tasks the graphics processor is busy with. So, the usual 3D game load can lead to its reduction to 654 MHz or in some cases even up to 554 MHz, depending on the complexity of the GPU work. Such “subsidence” is not incidental, but systematic, but if the A10-7800 chooses a standard TDP of 65 W, the graphics speed, unlike the cores, will remain constant: the GPU will always work at 720 MHz.

The most unpleasant in all the non-obvious algorithms for changing the frequencies of Kaveri is how the processing cores of this processor react to the activation of the built-in graphics core. The simple inclusion of 3D modes immediately leads to the fact that the real frequency A10-7800 independently of anything falls to 2.5 GHz. This is not throttling, not a reaction to temperature increase, but a rigidly fixed rule: any load on the built-in GPU causes the processor to significantly reduce the frequency of its CPU part and deactivate all auto-dispersal technologies. Such a drop in frequency will certainly become a very unpleasant surprise for gamers who will receive much lower computing power than the one described in official specifications.

Thus, AMD’s specifications in frequency specifications for Kaveri are not entirely true and do not reveal all the features of their functioning. In reality, the speed of work can vary over a much broader range. It’s very offensive, but it only touches the lower frequency limit. For example, in the A10-7800 considered in this review, the minimum operating frequency of the processing cores is not at all the promised 3.5 GHz, but only 2.5 GHz. At the same time, the maximum frequency in the turbo mode in the characteristics is indicated correctly – 3.9 GHz.

If you notice an error – select it with the mouse and press CTRL + ENTER.

Tags: , , , ,
Previous Post

we beat records of energy efficiency / Processors and memory

Next Post

breakthrough of the year / processors and memory

Leave a Reply

Your email address will not be published. Required fields are marked *