In other words – 10% over-voltage leads to 20% over-consumption. Unfortunately, while this process is effective it’s not that power-friendly because the power leakage is proportional to the square of voltage increase. In order to avoid those “droops”, the system provides more power than needed to over-compensate the transient voltage. How does this work? It’s basically pretty simple – each SoC or CPU experiences drops in voltage or also known as droops. Starting with the Carrizo family, AMD has implemented several technologies that monitor the voltage of the CPU, making it more optimized in terms of power efficiency. In terms of energy efficiency, the engineers at AMD may have struggled a lot to optimize the old 28nm process, but as we all know, there is a lot more to power optimizing than just one manufacturing process. It also offers better performance per Watt. As mentioned above, the pUMA+ cores feature 28nm manufacturing process, 2 or 4 cores, GCN, and power consumption between 10 and 25W for the whole SoC. The new pUMA+ cores will be used in entry-level notebooks, ultra-thin laptops, tablets and other devices with smaller form factor. However, the hUMA feature will be available only on Carrizo series chips, while the Carrizo-L (in our case) uses the pUMA+ cores and what is believed to be the successor of the Jaguar cores. The hUMA lets both GPU and CPU cores use the same memory address, thus shrinking the amount of instructions required to complete the tasks which greatly decreases power consumption and increases overall performance. The new generation Carrizo lineup continues to use the so-called HSA architecture and the most notable feature is the hUMA (heterogeneous Unified Memory Access). That’s a risky move and only time will show if this strategy proves useful. It seems like AMD is optimizing the current technology process instead of moving on with 22nm or even 14nm like Intel did. Also, the new x86 Excavator CPU cores will provide 40% less power consumption which is really impressive if the numbers are actually met. However, AMD was able to fit 29% more transistors in the same die size compared to the Kaveri generation CPUs. The manufacturing process continues to be 28nm from which we can expect a bigger power consumption, some extra heat, and slightly bigger form factor compared to Intel’s chips. Still AMD is promising a huge performance leap over the last Kaveri generation. In some aspects, AMD is still falling behind Intel and NVIDIA’s manufacturing technology.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |