Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Petz who wrote (40848)	5/22/2001 4:25:42 AM
From: dougSF30	Respond to of 275872

Petz, Ali is wrong. Given that the effect does not occur with the P3, nor the Athlon, it pretty much seals it. And yes, the memory page size IS much larger than a cache line. You would also not expect a graph of that shape: constant high performance, sudden degradation to constant low performance. If it were memory fragmentation related, you'd expect a slow decay. And again, how do the P3 and Athlon manage to avoid this? Doug

To: Petz who wrote (40848)	5/22/2001 6:39:33 AM
From: Bill Jackson	Respond to of 275872

Petz, Seems we will have to wait for some tester at Tom's or Anands to slay this very significant dragon, if indeed they can slay it at all!!. Bill

To: Petz who wrote (40848)	5/22/2001 3:13:56 PM
From: Ali Chen	Respond to of 275872

Petz, "he thinks that an irreversible slowdown might be caused by memory fragmentation in Windows 2000, rather than thermal throttling." "Fragmentation" leads to "randomization" of TLB entries. It affects two things (at least)- cache allocation pattern, and memory access pattern. Depending on cache replacement algorithm, it can have sometimes a positive effect on performance as it was on early Athlons with crippled external L2 cache. If the memory subsystem does not like frequent _DRAM page_ switching (like Rambus), the randomization would negatively impact performance, until you re-boot the system. The effect of performance degradation due to memory fragmentation is well known. For example, the BAPCO Sysmark degrades if run several times in a row. As result, BAPCO recommends the special mode when having an "official" run - reboot after every iteration. Of course, the degradation is only 5-7% on P-III, not 40% as reported. Of course, it was only my theory. There are other mechanisms in Windows/Intel that may require a total reboot to restore performance. There are certain bits in ACPI hardware that can be reset only on system RESET#. So it is not entirely impossible that some paranoid Window/BIOS ACPI may engage some permanent throttling and was not able (or willing) to restore the initial state even if the danger is gone. For your safety I think :) Take care, - Ali