While AMD is rather cautious in its official statements
in regards to anything related to the upcoming CPU offerings based on the next-generation Bulldozer architecture
, unofficial sources leak more and more information. This is not a surprise, as the launch is drawing near. Now a Chinese website published excerpts from datasheets
explaining the technical specifications of various Bulldozer samples. These are not the usual engineering samples we are talking about here (those are ticking at 1.8 and 2.6GHz), but samples of actual products that are going to be launched on June 7, 2011 in Taipei, Taiwan
This code will be printed on the IHS (Integrated Heat Spreader) of each and every FX processor coming to market
The rundown of the product name above shows the various elements of the Original Part Number (OPN). The Z denotes engineering samples, so the final product should get another letter (F for FX, O for Opteron). The next letter relates to the target segment, which is D for Desktop. Next up is the model number, which basically specifies the clock speed. The next number is labeled roadmap and we are a bit unsure about its actual meaning. The detail description say there are two different numbers right now which have basically the same meaning. Next is the processor package, with W referring to AM3r2 - publicly AM3+ is usually used. This is followed by the core count as well as a letter encoding the cache size. The leaked table only lists cores with 2MB L2 cache and no L3 cache or alternatively the same L2 and 8MB L3 cache. Previous information alluded to additional cache size options, which seems highly likely considering the L3 cache is partitioned internally. Thus it is easy for AMD to use partially defective die for models with fewer cores and less cache. Last but not least is the Part Definition, which basically denotes different steppings. Currently, only available information is about A1 and B0 silicon.
OPN Explained: Cache sizes, CPU cores
AMD Zambezi clocks, i.e. Bulldozer-based FX CPUs
The most interesting thing about the leak is the information regarding clock speeds of the samples. According to this information, there are models starting at 2.4GHz up to 3.2GHz base frequency. Another column labeled "CPU Boost Frequency 0"
lists what can only be Turbo frequencies that are 500 – 1000MHz higher than the respective base frequencies. The 3.1GHz model gets Turbo clocks up to a remarkable 4.1GHz, which is a massive 32% increase in clock speed. For a given base frequency there are up to three different models featuring different turbo clocks. The table does not disclose as to why there are differences of the turbo clocks. The model numbers remain constant among different turbo options. It should be considered that in the final product these discrepancies could easily go away.
While this information is very interesting, the burning question regarding Bulldozers performance remains unanswered once again. However, it shows that the architecture has potential for high clock speeds. One can only hope that AMD has the power consumption in check as well. The base frequencies are not really Earth shattering compared to current offerings, but the turbo clocks are what make Bulldozer different. Currently the specific operating characteristics of Bulldozers turbo feature are unknown. All we know with utmost certainty that Bulldozer-based CPUs will not exceed their TDP of 95W and 125W respectively.
© 2009 - 2013 Bright Side Of News*, All rights reserved.