Forum for gaming clan TKC, The Killing Committee
http://www.clan-tkc.com/phpBB/
Posted: Tue Oct 04, 2011 8:12 pmPosted: Tue Oct 04, 2011 10:01 pmPosted: Tue Oct 04, 2011 10:31 pmcountry_boy454 wrote:Cant wait to see the prices and performance in real world numbers not some good estimates! Could always use a second computer!!
Posted: Tue Oct 04, 2011 11:03 pm
Posted: Tue Oct 04, 2011 11:17 pmPosted: Wed Oct 05, 2011 1:31 pm
Posted: Wed Oct 05, 2011 1:33 pmPosted: Fri Oct 07, 2011 10:06 amPosted: Fri Oct 07, 2011 11:05 amPosted: Fri Oct 07, 2011 8:53 pmPosted: Mon Oct 10, 2011 6:33 amPosted: Mon Oct 10, 2011 11:23 amActually, we already have such an issue known for Bulldozer, and NO bench-marked system has the patch installed!
The shared L1 cache is causing cross invalidations across threads so that the prefetch data is incorrect in too many cases and data must be fetched again. The fix is a "simple" memory alignment and (possible)tagging system in the kernel of Windows/Linux.
I reviewed the code for the Linux patch and was astonished by just how little I know of the Linux kernel... lol! In any event, it could easily cost 10% in terms of single threaded performance, possibly more than double that in multi-threaded loads on the same module due to the increased contention and randomness of accesses.
Not sure if ordained reviewers have been given access to the MS patch, but I'd imagine (and hope) so! Last I saw, the Linux kernel patch was still being worked on by AMD (publicly) and Linus was showing some distaste for the method used to address the issue. One comment questioned the performance cost but had received no replies... but you don't go re-working kernel memory mapping for anything less than 5-10%... just not worth it!
Posted: Mon Oct 10, 2011 1:07 pmKnightwolf654 wrote:leak prices
FX-Series FX-6100 3.3 GHz 3.9 GHz 6 MB 8 MB 95 Watt $188.32
FX-Series FX-8120 3.1 GHz 4 GHz 8 MB 8 MB 125 Watt $221.73
FX-Series FX-8150 3.6 GHz 4.2 GHz 8 MB 8 MB 125 Watt $266.2
Posted: Mon Oct 10, 2011 11:09 pmKnightwolf654 wrote:yeah ive been watching a thread on another website on that poor excuse of a preview.Actually, we already have such an issue known for Bulldozer, and NO bench-marked system has the patch installed!
The shared L1 cache is causing cross invalidations across threads so that the prefetch data is incorrect in too many cases and data must be fetched again. The fix is a "simple" memory alignment and (possible)tagging system in the kernel of Windows/Linux.
I reviewed the code for the Linux patch and was astonished by just how little I know of the Linux kernel... lol! In any event, it could easily cost 10% in terms of single threaded performance, possibly more than double that in multi-threaded loads on the same module due to the increased contention and randomness of accesses.
Not sure if ordained reviewers have been given access to the MS patch, but I'd imagine (and hope) so! Last I saw, the Linux kernel patch was still being worked on by AMD (publicly) and Linus was showing some distaste for the method used to address the issue. One comment questioned the performance cost but had received no replies... but you don't go re-working kernel memory mapping for anything less than 5-10%... just not worth it!
the preview was thrown together in 30 min. half the graphs have the names switched and they did not install the patch so i'm treating that as a grain of poo.
Posted: Mon Oct 10, 2011 11:54 pm