<http://arstechnica.com/gadgets/2016/01/intel-skylake-bug-causes-pcs-to-freeze-during-complex-workloads/> Skylake CPUs can hang during certain computations. Prime searching folks found this. Interesting that Intel says that new firmware can fix it. (The article reminds us that Haswell CPUs had a but in the TSX feature and that the firmware fix just removed TSX. I felt robbed because one reason I bought a system with that processor was that I wanted to play with TSX.))
On Tue, Jan 12, 2016 at 09:41:11AM -0500, D. Hugh Redelmeier wrote:
<http://arstechnica.com/gadgets/2016/01/intel-skylake-bug-causes-pcs-to-freeze-during-complex-workloads/> Skylake CPUs can hang during certain computations. Prime searching folks found this.
Interesting that Intel says that new firmware can fix it.
Firmware updates have been available for all intel CPUs since the Pentium Pro. I think they learned the hard way on the Pentium which did not have that feature and needed it.
(The article reminds us that Haswell CPUs had a but in the TSX feature and that the firmware fix just removed TSX. I felt robbed because one reason I bought a system with that processor was that I wanted to play with TSX.))
At least by turning it off on the broken ones, it means software can safely use it on the new ones where it is fixed. Better than having to declare the feature permanently broken. -- Len Sorensen
On 01/12/2016 12:11 PM, Lennart Sorensen wrote:
On Tue, Jan 12, 2016 at 09:41:11AM -0500, D. Hugh Redelmeier wrote:
<http://arstechnica.com/gadgets/2016/01/intel-skylake-bug-causes-pcs-to-freeze-during-complex-workloads/> Skylake CPUs can hang during certain computations. Prime searching folks found this.
Interesting that Intel says that new firmware can fix it. Firmware updates have been available for all intel CPUs since the Pentium Pro. I think they learned the hard way on the Pentium which did not have that feature and needed it.
Many years ago, I worked as a computer tech, maintaining various models of minicomputer. One thing some computers had then was accessable microcode. With the Data General Eclipse line, the microcode was stored in ROM, but if the Writable Control Store (WCS) option was installed, you could add new or even modify existing instructions. The DEC VAX 11/780 loaded it's microcode from a floppy at every boot, so it was a simple matter to provide an updated version. It was "fun" trouble shooting a problem by stepping through the microcode. ;-)
On 16-01-12 03:06 PM, James Knott wrote:
It was "fun" trouble shooting a problem by stepping through the microcode. ;-)
At least you could do that if you had to in order to track down a problem. -- Cheers! Kevin. http://www.ve3syb.ca/ |"Nerds make the shiny things that distract Owner of Elecraft K2 #2172 | the mouth-breathers, and that's why we're | powerful!" #include <disclaimer/favourite> | --Chris Hardwick
On Tue, Jan 12, 2016 at 03:06:27PM -0500, James Knott wrote:
Many years ago, I worked as a computer tech, maintaining various models of minicomputer. One thing some computers had then was accessable microcode. With the Data General Eclipse line, the microcode was stored in ROM, but if the Writable Control Store (WCS) option was installed, you could add new or even modify existing instructions. The DEC VAX 11/780 loaded it's microcode from a floppy at every boot, so it was a simple matter to provide an updated version.
It was "fun" trouble shooting a problem by stepping through the microcode. ;-)
The alpha couldn't do memory management without extra instructions loaded at boot (usually by the firmware, although milo could fake it). Those instructions were written in a special low level instruction language for the alpha which had access to internal registers for the MMU and TLB that normal instructions couldn't touch, and created instructions software could execute. So Digital kept that idea going. -- Len Sorensen
participants (4)
-
D. Hugh Redelmeier -
James Knott -
Kevin Cozens -
Lennart Sorensen