it may be smart of hold off on sandy bridge...

[deadrats]

...and not because of the chipset bug:

http://www.xbitlabs.com/news/cpu/display/20110209222557_Intel_Plans_to_Update_Enthusia...e_in_2011.html

For example, the chips will carry "extra large cache" to maximize speed of single-threaded or dual-threaded applications, quad-channel memory controller and a number of other enhancements.

http://www.xbitlabs.com/news/cpu/display/20110203150914_Intel_s_Next_Gen_Ivy_Bridge_to...ge_Report.html

The Intel HD Graphics core of Ivy Bridge will not only feature higher performance (according to some sources, it will have 16 stream processing elements (up from 12 of SNB), some claim that it will have 24 stream processors), but will also support DirectX 11, HDMI 1.4 output and probably some other improvements.

honestly, if you consider when properly working chipsets are likely to enter the market and how long it will be before we see full and proper support for quick sync, the best move may be to hold off until the 4th quarter of this year before diving into a new platform.

[mh2360]

24 stream processors?

Even the lower spec cards from Nvidia and ATI have stream processors numbering in the 80s and 90s. If anyone is serious about GPU assisted encoding, then a simple graphics card upgrade would be much cheaper than a new CPU/Motherboard (and possibly RAM and PSU).

[deadrats]

24 stream processors?

Even the lower spec cards from Nvidia and ATI have stream processors numbering in the 80s and 90s. If anyone is serious about GPU assisted encoding, then a simple graphics card upgrade would be much cheaper than a new CPU/Motherboard (and possibly RAM and PSU).

nobody is going to buy a sandy bridge cpu, or any of it's descendants, for gpu assisted encoding, quick sync only uses the integrated gpu for motion estimation (so yes, there will be some benefit from the faster gpu), i was referring to the new gpu's dx11 capabilities, the "extra large cache *) and the quad channel memory.

*i remember reading an article on anandtech some time ago where an intel engineer was interviewed and he said that intel was working on making the L1 and L2 caches out of dram instead of sram, and that they believed that if they could get to the point where they had 512mb to 1gig of L2 that such a dual core processor would be about 10 faster than a core 2 just from the cache alone.

i wonder how big a cache they are talking about with these cpu's and if this could be the first step in making that a reality.