As mentioned in the chapter on VIA C3, when VIA released the Nehemiah core, they could very well have changed the name of their processor, bacause this core was really different from the previous ones.
But they didn't do, probably because it kept the same socket as the previous cores, and was more or less backward compatible. But when VIA developed a new iteration of the C5, and finally abandoned socket 370 and all compatibility with Intel's P6 bus, then this time they changed the name of their processor. Despite the fact that the evolutions were less important...
And since its competitors were already selling their 7th generation of x86 (and even 8th for AMD), VIA decided to name it C7.
It was not exactly on their own initiative that VIA abandoned the P6 bus and socket 370. Indeed, this bus and this socket were owned by Intel, and VIA had been able to use it through intellectual property exchange agreements with them. But these agreements ended in 2006, and VIA now having little to offer Intel in return, there were no plans to renew them.
Worse, the first iteration of the C7, initially named the C5I core, should have been compatible with the Pentium M Banias bus, in addition to a VIA proprietary bus.
But the trade negotiations certainly did not come to an agreement, since in the end, the C7, released in May 2005, used a C5J core, or Esther, without the Pentium M bus.
Regarding the new features of this new processor, there was finally the addition of the SSE2 and SSE3 instructions, a doubled L2 cache, at 128KB, and a 0.09µm process from IBM (for first models).
If we except the new security functions, which we will describe below, everything else, the pipeline, the branch prediction, the amount of execution units, the FPU... etc., was directly carried over from C5P Nehemiah.
These processors also inaugurated a brand new package, much smaller and more consistent with the die size, the nanoBGA2 format, which succeeds the nanoBGA inaugurated on the VIA Eden-N (based on the Nehemiah core).
On the other hand, even if they are very small, they are a little bigger than the latter, with 2,1cm of side, against 1,5cm for the nanoBGA !
In any case, it was much smaller than the package of the CPUs they replaced, the EBGA of the C3 and C3-M :
Regarding the security features, very important to VIA, the C7s went a step beyond compared to the C3, with HW support for SHA-1 and SHA-256 algorithms encoding and a Montgomery multiplier. Everything was still marketed under the VIA Padlock name.
The C7s also got the NX bit functionality, which VIA's competitors had introduced since 2003-2004, under the name Execute Disable bit (or XD bit) for Intel and Enhanced Virus Protection (EVP) for AMD.
Regarding VIA's main advantage, namely power consumption, the reduction in the process helped to keep it very low, despite the increase in frequencies. However, in order to further optimize this, VIA integrated into the C7s an interesting feature which they called TwinTurbo. It consisted of the integration of a double PLL, one for the high frequency and one for low frequency, in order to solve the problem of the VIA PowerSaver function of the C3 (similar to the Intel SpeedStep) concerning the (long) switching time between these two frequencies modes (and the corresponding voltages). Thus, switching from one frequency to the other could be done in just one cycle, as long as the OS was fast enough for that.
With all this, and with the Quad-pumped bus at 400 (4x100), 533 (4x133) or 800MHz (4x200), performances increased significantly compared to the C3, without being exceptional either, at a time where AMD had already released their overkill Athlon 64s. But for the market targeted by VIA, which was no longer such a small niche in 2005, namely that of light client and other set-top box, this represented a credible choice, especially since Intel and AMD had not yet invested much in this market.
And we must not forget either that VIA, with its C3 EBGA, its Eden and other CoreFusions, were the forerunners in the market for mini motherboards and had invented the mini-ITX format.
VIA therefore did not hesitate to offer a whole line of motherboards based on the new C7 nanoBGA2s. In 2006, if you wanted mini-ITX, there was only VIA .
Even if then, this market was still far from being as important as it is today.
On the other hand, concerning the second market targeted by VIA, of long battery-life and inexpensive ultraportables, it was more complicated. Because, even if AMD was not really present in this market, there was on the other hand Intel, which had entered in force with their Pentium M, and especially its Low-cost version, the Celeron M. Not only were the latter well established and unavoidable in the market, after lean years with the Pentium 4-M, but in addition, they were quite powerrful. Much more than the C7.
The only area on which VIA could compete in the Mobile market was still consumption. And while the C7s were intended primarily for desktop PCs (but could also be used for laptops), VIA also annonced, just a few days after the C7, the C7-M , with still low consumption, reduced compared to C7.
This decrease in consumption was mainly due, in addition to the lower frequencies, to the use of a higher number of ACPI P-States (8 instead of the 2 basic ones, described more top), thus specifying more frequency / voltage combinations steps, in order to adapt them as closely as possible to the actual load.
And unlike the C7s, some C7-Ms were also available, in addition to the nanoBGA2 package, in mPGA-479, the same package as the Pentium M and Celeron M.
But as they did not use exactly the same bus, only special motherboards, with VIA chipsets, like the VN800s, were able to handle both the Pentium/Celeron M and the C7-M, giving the artificial illusion of intercompatibility .
A few months later, in March 2006 , VIA released Ultra Low Voltage (ULV) versions of their C7-M, with voltages and consumptions further reduced, below 8W .
Finally, six months later, in September 2006, VIA released a last version of C7, still based on the Esther core, the C7-D. Unlike the very simple C7s, whose market was not really clear (Desktop, Mobile, Set-Top Box ... etc), these were clearly intended and sold for the Office Desktop PC market. Although in practice, nothing really distinguished them from the C7s except that they had a very slightly higher TDP (+2W). Perhaps due to less aggressive energy management (switch to LFM - Low Frequency Mode - less frequent?).
VIA tried with these models a rather interesting marketing approach, perhaps a little too ahead of its time, by comparing the C7-D to its competitors on the axis of CO2 emissions .
VIA reported that the C7-D was the world's first carbon-free computing component, helping individuals and organizations reduce their carbon footprint. To do this, for every VIA C7-D processor sold, VIA worked with environmental experts to calculate the electricity used over the lifespan of the processors (they took a favorable case of 3 years - duration of the "first life" of PCs in business in general). Then, from the amount of electricity used, they calculated the amount of CO2 emissions that would be released to the environment primarily from fossil-fueled power plants, and then work with regional offsetting organizations to 'offset' this amount of CO2 through projects such as the reforestation, development of alternative energies, energy saving.
And in order to give an indicator of comparison, they defined what they called the TreeMark. This indicator, developed with the independent consulting firm Best Foot Forward Ltd, took into account the average use of a PC and derived the amount of carbon dioxide created to produce the electricity required to power the processor during its operational life - and therefore, the number of trees required to compensate for that amount in the environment.
With its energy efficiency, the VIA C7-D had a TreeMark score of four trees, against an average of 18 trees - so more than 80% - for competing desktop processors (a Pentium 4, necessarily ...).