00000000.htm

一份很好的linux入门资料

&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;EDORAM:&nbsp;Extended&nbsp;Data&nbsp;Out&nbsp;RAM.&nbsp;Like&nbsp;DRAM,&nbsp;but&nbsp;allowed&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;reading&nbsp;of&nbsp;multiple&nbsp;bytes&nbsp;in&nbsp;parallel.&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;SRAM:&nbsp;Static&nbsp;DRAM.&nbsp;Usually&nbsp;faster&nbsp;than&nbsp;other&nbsp;methods,&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;but&nbsp;more&nbsp;power-hungry.&nbsp;Used&nbsp;for&nbsp;caches.&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;SDRAM:&nbsp;Synchronous&nbsp;DRAM.&nbsp;Operates&nbsp;in&nbsp;sync&nbsp;with&nbsp;the&nbsp;CPU&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;clock.&nbsp;Increases&nbsp;throughput&nbsp;by&nbsp;using&nbsp;pipelining&nbsp;to&nbsp;hide&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;setup&nbsp;times.&nbsp;This&nbsp;is&nbsp;the&nbsp;most&nbsp;common&nbsp;type&nbsp;of&nbsp;RAM&nbsp;in&nbsp;use&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;today.&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;WRAM:&nbsp;Windowed&nbsp;RAM.&nbsp;A&nbsp;dual&nbsp;ported&nbsp;memory&nbsp;that&nbsp;allowed&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;simultaneous&nbsp;reading&nbsp;and&nbsp;writing.&nbsp;Only&nbsp;ever&nbsp;appeared&nbsp;in&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;Video&nbsp;cards.&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;ECCRAM:&nbsp;Error&nbsp;Checking&nbsp;/&nbsp;Correcting&nbsp;RAM.&nbsp;Contains&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;several&nbsp;extra&nbsp;bits&nbsp;per&nbsp;byte;&nbsp;the&nbsp;memory&nbsp;controller&nbsp;uses&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;this&nbsp;to&nbsp;detect&nbsp;and&nbsp;correct&nbsp;all&nbsp;single&nbsp;bit&nbsp;errors&nbsp;and&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;detect&nbsp;most&nbsp;multiple&nbsp;bit&nbsp;errors.&nbsp;Typically&nbsp;only&nbsp;used&nbsp;in&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;servers,&nbsp;where&nbsp;the&nbsp;cost&nbsp;of&nbsp;an&nbsp;error&nbsp;is&nbsp;very&nbsp;high.&nbsp;&nbsp;<BR>
&nbsp;<BR>
See&nbsp;also&nbsp;Tom's&nbsp;Hardware&nbsp;Guide.&nbsp;&nbsp;<BR>
&nbsp;<BR>
4.3&nbsp;Parity&nbsp;Checking&nbsp;<BR>
&nbsp;<BR>
Standard&nbsp;main&nbsp;memory&nbsp;is&nbsp;Dynamic&nbsp;RAM&nbsp;of&nbsp;one&nbsp;sort&nbsp;or&nbsp;another,&nbsp;in&nbsp;<BR>
which&nbsp;each&nbsp;bit&nbsp;is&nbsp;represented&nbsp;by&nbsp;voltage&nbsp;on&nbsp;a&nbsp;capacitor.&nbsp;If&nbsp;a&nbsp;<BR>
cosmic&nbsp;ray&nbsp;happens&nbsp;to&nbsp;dump&nbsp;too&nbsp;much&nbsp;charge&nbsp;onto&nbsp;one&nbsp;of&nbsp;those&nbsp;<BR>
capacitors,&nbsp;you&nbsp;can&nbsp;end&nbsp;up&nbsp;with&nbsp;a&nbsp;bit&nbsp;error.&nbsp;This&nbsp;happens&nbsp;very&nbsp;<BR>
infrequently,&nbsp;so&nbsp;desktop&nbsp;systems&nbsp;don't&nbsp;need&nbsp;to&nbsp;worry&nbsp;too&nbsp;much&nbsp;<BR>
about&nbsp;it.&nbsp;To&nbsp;detect&nbsp;bit&nbsp;errors,&nbsp;some&nbsp;memory&nbsp;chips&nbsp;come&nbsp;with&nbsp;a&nbsp;<BR>
parity&nbsp;bit&nbsp;on&nbsp;each&nbsp;byte;&nbsp;the&nbsp;parity&nbsp;bit&nbsp;is&nbsp;a&nbsp;simple&nbsp;checksum&nbsp;of&nbsp;<BR>
the&nbsp;byte.&nbsp;When&nbsp;the&nbsp;computer&nbsp;reads&nbsp;the&nbsp;byte,&nbsp;and&nbsp;detects&nbsp;that&nbsp;the&nbsp;<BR>
checksum&nbsp;on&nbsp;the&nbsp;byte&nbsp;is&nbsp;wrong,&nbsp;it&nbsp;declares&nbsp;a&nbsp;bus&nbsp;error,&nbsp;and&nbsp;halts&nbsp;<BR>
the&nbsp;current&nbsp;program.&nbsp;&nbsp;<BR>
&nbsp;<BR>
Not&nbsp;many&nbsp;computers&nbsp;use&nbsp;simple&nbsp;parity&nbsp;checking&nbsp;anymore;&nbsp;it&nbsp;doesn't&nbsp;<BR>
detect&nbsp;as&nbsp;many&nbsp;errors&nbsp;as&nbsp;full&nbsp;ECC&nbsp;memory&nbsp;(which&nbsp;has&nbsp;four&nbsp;or&nbsp;so&nbsp;<BR>
parity&nbsp;bits&nbsp;per&nbsp;byte,&nbsp;and&nbsp;can&nbsp;even&nbsp;correct&nbsp;single-bit&nbsp;errors).&nbsp;&nbsp;<BR>
&nbsp;<BR>
4.4&nbsp;Leveling&nbsp;the&nbsp;Cache&nbsp;<BR>
&nbsp;<BR>
In&nbsp;computing,&nbsp;the&nbsp;faster&nbsp;you&nbsp;want&nbsp;to&nbsp;access&nbsp;something,&nbsp;the&nbsp;more&nbsp;<BR>
expensive&nbsp;it&nbsp;is&nbsp;to&nbsp;manufacture.&nbsp;Have&nbsp;you&nbsp;ever&nbsp;wondered&nbsp;why&nbsp;we&nbsp;<BR>
don't&nbsp;see&nbsp;all&nbsp;hard&nbsp;drives&nbsp;replaced&nbsp;with&nbsp;RAM?&nbsp;Its&nbsp;cheaper&nbsp;to&nbsp;<BR>
produce&nbsp;the&nbsp;same&nbsp;amount&nbsp;of&nbsp;storage&nbsp;on&nbsp;a&nbsp;disk&nbsp;than&nbsp;it&nbsp;is&nbsp;in&nbsp;<BR>
memory.&nbsp;This&nbsp;price&nbsp;saving&nbsp;comes&nbsp;at&nbsp;a&nbsp;penalty&nbsp;-&nbsp;speed.&nbsp;A&nbsp;hard&nbsp;<BR>
drive&nbsp;is&nbsp;about&nbsp;100&nbsp;times&nbsp;slower&nbsp;to&nbsp;access&nbsp;than&nbsp;standard&nbsp;memory.&nbsp;&nbsp;<BR>
&nbsp;<BR>
Your&nbsp;CPU&nbsp;screams&nbsp;along&nbsp;at&nbsp;400&nbsp;million&nbsp;instructions&nbsp;a&nbsp;second&nbsp;<BR>
(obviously&nbsp;this&nbsp;depends&nbsp;on&nbsp;CPU&nbsp;clock&nbsp;speed).&nbsp;On&nbsp;a&nbsp;good&nbsp;day,&nbsp;that&nbsp;<BR>
means&nbsp;it&nbsp;needs&nbsp;to&nbsp;read&nbsp;one&nbsp;instruction&nbsp;every&nbsp;2.5&nbsp;nano&nbsp;seconds&nbsp;<BR>
(2.5&nbsp;x&nbsp;10^-9).&nbsp;An&nbsp;average&nbsp;hard&nbsp;drive&nbsp;has&nbsp;an&nbsp;access&nbsp;time&nbsp;of&nbsp;around&nbsp;<BR>
8&nbsp;milliseconds&nbsp;(8&nbsp;x&nbsp;10&nbsp;-3).&nbsp;If&nbsp;the&nbsp;CPU&nbsp;was&nbsp;going&nbsp;to&nbsp;read&nbsp;every&nbsp;<BR>
thing&nbsp;from&nbsp;the&nbsp;hard&nbsp;drive,&nbsp;for&nbsp;each&nbsp;instruction&nbsp;it&nbsp;executes,&nbsp;<BR>
there&nbsp;are&nbsp;about&nbsp;3.2&nbsp;million&nbsp;clock&nbsp;cycles&nbsp;that&nbsp;it&nbsp;doesn't&nbsp;do&nbsp;<BR>
anything&nbsp;for.&nbsp;Pretty&nbsp;obviously&nbsp;that's&nbsp;a&nbsp;big&nbsp;waste&nbsp;of&nbsp;resources.&nbsp;&nbsp;<BR>
&nbsp;<BR>
In&nbsp;order&nbsp;to&nbsp;keep&nbsp;the&nbsp;CPU&nbsp;feed&nbsp;with&nbsp;instructions,&nbsp;we&nbsp;need&nbsp;memory&nbsp;<BR>
that&nbsp;is&nbsp;going&nbsp;to&nbsp;keep&nbsp;up.&nbsp;The&nbsp;problem&nbsp;now&nbsp;becomes&nbsp;one&nbsp;of&nbsp;how&nbsp;to&nbsp;<BR>
get&nbsp;data&nbsp;from&nbsp;slow&nbsp;moving&nbsp;devices&nbsp;into&nbsp;the&nbsp;fast&nbsp;moving&nbsp;ones&nbsp;-&nbsp;<BR>
that&nbsp;is&nbsp;the&nbsp;job&nbsp;of&nbsp;the&nbsp;various&nbsp;levels&nbsp;of&nbsp;memory.&nbsp;On&nbsp;your&nbsp;typical&nbsp;<BR>
computer&nbsp;you&nbsp;have&nbsp;the&nbsp;following:&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;~&nbsp;Level&nbsp;1&nbsp;-&nbsp;The&nbsp;CPU&nbsp;cache&nbsp;built&nbsp;into&nbsp;the&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;microprocessor.&nbsp;Currently&nbsp;that&nbsp;is&nbsp;between&nbsp;8K&nbsp;and&nbsp;32K&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;bytes.&nbsp;Quickest&nbsp;as&nbsp;it&nbsp;operates&nbsp;at&nbsp;exactly&nbsp;the&nbsp;same&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;speed&nbsp;as&nbsp;the&nbsp;CPU&nbsp;core&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;~&nbsp;Level&nbsp;2&nbsp;-&nbsp;External&nbsp;Cache.&nbsp;The&nbsp;next&nbsp;level&nbsp;is&nbsp;much&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;bigger&nbsp;at&nbsp;between&nbsp;256K&nbsp;and&nbsp;2MB.&nbsp;This&nbsp;normally&nbsp;operates&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;at&nbsp;about&nbsp;half&nbsp;the&nbsp;CPU&nbsp;core&nbsp;clock&nbsp;speed.&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;~&nbsp;Level&nbsp;3&nbsp;-&nbsp;Main&nbsp;Memory.&nbsp;Your&nbsp;standard&nbsp;RAM.&nbsp;Anywhere&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;between&nbsp;16MB&nbsp;and&nbsp;a&nbsp;few&nbsp;gigabytes.&nbsp;Access&nbsp;to&nbsp;this&nbsp;is&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;around&nbsp;66&nbsp;or&nbsp;100MHz,&nbsp;or&nbsp;probably&nbsp;a&nbsp;quarter&nbsp;of&nbsp;the&nbsp;CPU&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;core&nbsp;clock&nbsp;speed.&nbsp;&nbsp;<BR>
&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;~&nbsp;Level&nbsp;4&nbsp;-&nbsp;Storage&nbsp;devices.&nbsp;Hard&nbsp;drives,&nbsp;CDs,&nbsp;Floppies&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;etc.&nbsp;Access&nbsp;speed&nbsp;is&nbsp;around&nbsp;10-50MHz&nbsp;depending&nbsp;on&nbsp;the&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;device&nbsp;and&nbsp;connection&nbsp;(eg&nbsp;IDE&nbsp;is&nbsp;16MHz,&nbsp;UW&nbsp;SCSI&nbsp;is&nbsp;<BR>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;80MHz,&nbsp;FireWire/FibreChannel&nbsp;is&nbsp;around&nbsp;300MHz).&nbsp;&nbsp;<BR>
&nbsp;<BR>
There&nbsp;are&nbsp;some&nbsp;real&nbsp;over&nbsp;simplifications&nbsp;here,&nbsp;but&nbsp;it&nbsp;should&nbsp;give&nbsp;<BR>
you&nbsp;a&nbsp;rough&nbsp;idea&nbsp;of&nbsp;how&nbsp;each&nbsp;level&nbsp;slows&nbsp;down&nbsp;compared&nbsp;to&nbsp;the&nbsp;one&nbsp;<BR>
below&nbsp;it&nbsp;as&nbsp;we&nbsp;move&nbsp;away&nbsp;from&nbsp;the&nbsp;CPU.&nbsp;As&nbsp;another&nbsp;gross&nbsp;<BR>
simplification,&nbsp;for&nbsp;each&nbsp;unit&nbsp;at&nbsp;each&nbsp;level,&nbsp;the&nbsp;cost&nbsp;stays&nbsp;<BR>
roughly&nbsp;the&nbsp;same&nbsp;of&nbsp;around&nbsp;US$250&nbsp;so&nbsp;you&nbsp;can&nbsp;see&nbsp;how&nbsp;cost&nbsp;<BR>
influences&nbsp;size&nbsp;and&nbsp;speed.&nbsp;&nbsp;<BR>
&nbsp;<BR>
Now,&nbsp;that's&nbsp;a&nbsp;long&nbsp;introduction&nbsp;to&nbsp;this&nbsp;point.&nbsp;As&nbsp;you&nbsp;can&nbsp;see,&nbsp;we&nbsp;<BR>
get&nbsp;to&nbsp;smaller&nbsp;and&nbsp;smaller&nbsp;memory&nbsp;sizes&nbsp;the&nbsp;closer&nbsp;we&nbsp;get&nbsp;to&nbsp;the&nbsp;<BR>
CPU.&nbsp;That&nbsp;means&nbsp;we&nbsp;fit&nbsp;less&nbsp;in,&nbsp;and&nbsp;the&nbsp;chances&nbsp;of&nbsp;having&nbsp;to&nbsp;<BR>
fetch&nbsp;something&nbsp;from&nbsp;the&nbsp;next&nbsp;level&nbsp;up&nbsp;increases.&nbsp;Since&nbsp;this&nbsp;next&nbsp;<BR>
level&nbsp;is&nbsp;slower,&nbsp;we&nbsp;pay&nbsp;a&nbsp;penalty&nbsp;each&nbsp;time.&nbsp;&nbsp;<BR>
&nbsp;<BR>
The&nbsp;simplistic&nbsp;approach&nbsp;to&nbsp;tuning&nbsp;your&nbsp;memory&nbsp;and&nbsp;cache&nbsp;is&nbsp;to&nbsp;put&nbsp;<BR>
as&nbsp;much&nbsp;of&nbsp;the&nbsp;fastest&nbsp;memory&nbsp;that&nbsp;you&nbsp;can&nbsp;buy&nbsp;into&nbsp;the&nbsp;machine.&nbsp;<BR>
This&nbsp;is&nbsp;the&nbsp;reason&nbsp;the&nbsp;Xeon,&nbsp;Sparc&nbsp;and&nbsp;Alpha&nbsp;chips&nbsp;can&nbsp;have&nbsp;as&nbsp;<BR>
much&nbsp;as&nbsp;2MB&nbsp;of&nbsp;on&nbsp;board&nbsp;cache.&nbsp;For&nbsp;server&nbsp;applications&nbsp;where&nbsp;<BR>
there&nbsp;is&nbsp;huge&nbsp;amounts&nbsp;of&nbsp;number&nbsp;crunching&nbsp;to&nbsp;do,&nbsp;the&nbsp;more&nbsp;you&nbsp;can&nbsp;<BR>
cache&nbsp;the&nbsp;quicker&nbsp;everything&nbsp;runs.&nbsp;&nbsp;<BR>
&nbsp;<BR>
Buying&nbsp;big&nbsp;caches&nbsp;doesn't&nbsp;always&nbsp;&nbsp;gain&nbsp;you&nbsp;extra.&nbsp;For&nbsp;example,&nbsp;if&nbsp;<BR>
you&nbsp;are&nbsp;mail&nbsp;serving&nbsp;or&nbsp;doing&nbsp;only&nbsp;static&nbsp;web&nbsp;pages,&nbsp;you&nbsp;will&nbsp;<BR>
gain&nbsp;almost&nbsp;nothing&nbsp;compared&nbsp;to&nbsp;more&nbsp;reasonable&nbsp;&quot;standard&quot;&nbsp;<BR>
amounts.&nbsp;For&nbsp;many&nbsp;operations,&nbsp;information&nbsp;gets&nbsp;dumped&nbsp;straight&nbsp;<BR>
from&nbsp;the&nbsp;hard&nbsp;drive&nbsp;or&nbsp;main&nbsp;memory,&nbsp;straight&nbsp;to&nbsp;the&nbsp;output&nbsp;<BR>
device.&nbsp;For&nbsp;example,&nbsp;sound&nbsp;clips&nbsp;or&nbsp;image&nbsp;textures&nbsp;may&nbsp;go&nbsp;<BR>
straight&nbsp;to&nbsp;output&nbsp;devices&nbsp;for&nbsp;processing.&nbsp;If&nbsp;you&nbsp;are&nbsp;dishing&nbsp;up&nbsp;<BR>
files&nbsp;to&nbsp;a&nbsp;server,&nbsp;you&nbsp;are&nbsp;much&nbsp;better&nbsp;off&nbsp;trying&nbsp;to&nbsp;store&nbsp;the&nbsp;<BR>
files&nbsp;in&nbsp;main&nbsp;memory&nbsp;rather&nbsp;than&nbsp;on&nbsp;the&nbsp;hard&nbsp;drive.&nbsp;Tuning&nbsp;a&nbsp;file&nbsp;<BR>
server&nbsp;usually&nbsp;involves&nbsp;buying&nbsp;lots&nbsp;of&nbsp;standard&nbsp;RAM&nbsp;rather&nbsp;than&nbsp;<BR>
big&nbsp;caches.&nbsp;&nbsp;<BR>
&nbsp;<BR>
4.5&nbsp;Bus&nbsp;speed&nbsp;<BR>
&nbsp;<BR>
As&nbsp;you&nbsp;saw&nbsp;in&nbsp;the&nbsp;previous&nbsp;section,&nbsp;accessing&nbsp;a&nbsp;byte&nbsp;of&nbsp;<BR>
information&nbsp;is&nbsp;dependent&nbsp;on&nbsp;the&nbsp;rate&nbsp;at&nbsp;which&nbsp;you&nbsp;can&nbsp;access&nbsp;the&nbsp;<BR>
information&nbsp;in&nbsp;the&nbsp;device.&nbsp;However,&nbsp;that&nbsp;device&nbsp;needs&nbsp;to&nbsp;be&nbsp;<BR>
connected&nbsp;to&nbsp;the&nbsp;other&nbsp;devices&nbsp;so&nbsp;that&nbsp;means&nbsp;it&nbsp;must&nbsp;travel&nbsp;over&nbsp;<BR>
some&nbsp;intermediary&nbsp;connection.&nbsp;Like&nbsp;a&nbsp;dam&nbsp;with&nbsp;a&nbsp;straw&nbsp;allowing&nbsp;<BR>
water&nbsp;to&nbsp;escape,&nbsp;it&nbsp;doesn't&nbsp;matter&nbsp;how&nbsp;much&nbsp;or&nbsp;how&nbsp;quick&nbsp;the&nbsp;<BR>
device&nbsp;might&nbsp;be&nbsp;capable&nbsp;of&nbsp;delivering&nbsp;information&nbsp;if&nbsp;the&nbsp;bus&nbsp;<BR>
connecting&nbsp;the&nbsp;two&nbsp;runs&nbsp;like&nbsp;a&nbsp;glacier.&nbsp;&nbsp;<BR>
&nbsp;<BR>
One&nbsp;immediate&nbsp;and&nbsp;fairly&nbsp;easy&nbsp;way&nbsp;to&nbsp;get&nbsp;information&nbsp;out&nbsp;quicker&nbsp;<BR>
is&nbsp;to&nbsp;start&nbsp;playing&nbsp;with&nbsp;the&nbsp;bus&nbsp;speeds&nbsp;for&nbsp;everything.&nbsp;A&nbsp;normal&nbsp;<BR>
computer&nbsp;comes&nbsp;with&nbsp;adjustments&nbsp;for&nbsp;PCI,&nbsp;ISA,&nbsp;Memory&nbsp;and&nbsp;a&nbsp;few&nbsp;<BR>
other&nbsp;items&nbsp;set&nbsp;to&nbsp;fairly&nbsp;conservative,&nbsp;safe&nbsp;settings.&nbsp;On&nbsp;the&nbsp;<BR>
other&nbsp;hand,&nbsp;the&nbsp;devices&nbsp;plugged&nbsp;into&nbsp;these&nbsp;busses&nbsp;usually&nbsp;have&nbsp;<BR>
some&nbsp;amount&nbsp;of&nbsp;tolerance&nbsp;for&nbsp;the&nbsp;bus&nbsp;speed&nbsp;moving&nbsp;around&nbsp;a&nbsp;bit.&nbsp;<BR>
This&nbsp;gives&nbsp;you&nbsp;room&nbsp;to&nbsp;tinker.&nbsp;&nbsp;<BR>
&nbsp;<BR>
Playing&nbsp;with&nbsp;bus&nbsp;speeds&nbsp;is&nbsp;a&nbsp;bit&nbsp;of&nbsp;a&nbsp;trial&nbsp;and&nbsp;error&nbsp;approach.&nbsp;<BR>
Good&nbsp;quality&nbsp;components&nbsp;usually&nbsp;have&nbsp;quite&nbsp;a&nbsp;margin&nbsp;to&nbsp;play&nbsp;with,&nbsp;<BR>
but&nbsp;el&nbsp;cheapo&nbsp;components&nbsp;(like&nbsp;your&nbsp;$10&nbsp;NE2000&nbsp;clone&nbsp;network&nbsp;<BR>
card)&nbsp;won't&nbsp;tolerate&nbsp;it&nbsp;too&nbsp;much.&nbsp;&nbsp;<BR>
&nbsp;<BR>
Bus&nbsp;speeds&nbsp;are&nbsp;usually&nbsp;only&nbsp;available&nbsp;in&nbsp;the&nbsp;BIOS&nbsp;setup.&nbsp;To&nbsp;tune&nbsp;<BR>
your&nbsp;bus,&nbsp;up&nbsp;the&nbsp;speed,&nbsp;reboot&nbsp;the&nbsp;machine&nbsp;and&nbsp;see&nbsp;if&nbsp;things&nbsp;<BR>
start&nbsp;locking&nbsp;up.&nbsp;Assuming&nbsp;the&nbsp;machine&nbsp;still&nbsp;boots&nbsp;OK,&nbsp;run&nbsp;<BR>
benchmarks&nbsp;over&nbsp;it.&nbsp;In&nbsp;some&nbsp;cases,&nbsp;increasing&nbsp;the&nbsp;bus&nbsp;speed&nbsp;will&nbsp;<BR>
_slow_down_&nbsp;your&nbsp;machine&nbsp;due&nbsp;to&nbsp;the&nbsp;devices&nbsp;not&nbsp;adequately&nbsp;<BR>
dealing&nbsp;with&nbsp;higher&nbsp;than&nbsp;specified&nbsp;settings.&nbsp;Don't&nbsp;just&nbsp;set&nbsp;it&nbsp;as&nbsp;<BR>
high&nbsp;as&nbsp;possible&nbsp;and&nbsp;assume&nbsp;everything&nbsp;will&nbsp;be&nbsp;better.&nbsp;Find&nbsp;<BR>
benchmarks&nbsp;for&nbsp;the&nbsp;particular&nbsp;subsystem&nbsp;that&nbsp;you've&nbsp;played&nbsp;with&nbsp;<BR>
and&nbsp;check!&nbsp;<BR>
&nbsp;<BR>
--&nbsp;<BR>
※&nbsp;来源:·BBS&nbsp;水木清华站&nbsp;bbs.net.tsinghua.edu.cn·[FROM:&nbsp;162.105.179.11]&nbsp;<BR>


<CENTER><H1>BBS水木清华站∶精华区</H1></CENTER>



</BODY>

</HTML>