As reports associate, at the conference SC09 Justin rattner, chief Intel technologist , Â shared information about the known development – graphic chip from x86 core larrabee.
In SGEMM the multiplication test in the chip, which has 80 cores, showed the peak performance of 805 billion operations floating point per second. Which is interesting, rattner asked technicians to overclock Larabee, and productivity increased to 1 trillion operations floating point per second after this. With the use of only half accessible core it was possible to reach the peak productivity 417 billion operations floating point per second. One of the fundamental characteristics Of larrabee will become precisely the quantity of core in the chip.
In the tests of sparse matrices, which are used with the simulation of cloths and water,  Larrabee showed the result of 7.9 and 8.1 billion operations floating point per second, depending on test and matrix size  .
It became also known that in the future the processors core and Xeon can create the sections of common virtual memory, access to which will be both in processor and Larrabee. Common virtual memory will allow to work at one and same task consecutively, that must radically increase productivity and optimize simulations.