The memory at 2134 MT/s is the BO bin 3 limitation (I thought it must be 2133, strange 4). Also, the product will be at 1.4 GHz base frequency. I have a good "preliminary" news. I ran a single tile STREAM triad for two data points, right in MCDRAM footprint and well beyond that. In both cases I am getting exactly what was projected: a little over 13 GB/s. I will need to make more intensive benchmarking but I am happy so far. Memory before_update after_update 1373.3 MB 8598.3969 GB/s 13594.6362GB/s 22888.2 MB1849.2670 GB/s 13169.3377 GB/s I will try something else to see if this converts into performance. Vitali On 03/29/2016 04:46 PM, Allen, Benjamin S. wrote:

I did change two BIOS settings: the acoustic profile in the BIOS to favor performance over noise. As well as changed a the Performance Bias of the CPU to be "Performance" instead of Balanced. This is supposed to be a 1.3Ghz CPU. In addition, the new BIOS includes a number of microcode updates for the B0. I suppose it could be any of the above.

Note the memory still came up at 2134MT/s.

Ben

...

On Mar 29, 2016, at 4:39 PM, Vitali A. Morozov <[email protected]> wrote:

Hi Ben,

thanks a lot - will start a few tests right away. /proc/cpuinfo gives higher frequency now:

cpu MHz : 1299.644

it was lower before:

cpu MHz : 1200.519

I am curious if this is reporting fix or sensor fix or indeed changes in frequency.

Vitali

On 03/29/2016 04:17 PM, Allen, Benjamin S. wrote:

...

Vitali,

Upgraded the BIOS with A0 and swapped the B0 back in. Appears to be a lot more happy now. I can't reproduce the panic as I had before. Various release notes and updated manuals attached. Feel free to start testing.

Thanks,

Ben