微星最近在AM5主板上加入了内存高性能模式,说是对SK海力士A-die颗粒的专属优化,实际上就是对内存小参动手了,通过优化这些小参让内存达到更高的传输速度以及更低的延迟,该模式支持的平台包括了X670、B650和A620主板。
想开启这模式首先得去微星官网下载对应型号的最新BIOS,然后开启EXPO或AXMP,接着在BIOS超频选项中开启High Efficiency
mode后,在其中Memory Timing Preset选项的四个等级里选一个,最高档的是Tightest,但未必可以成功。
我们拿了手头上两块微星的AM5主板来尝试过这一功能,一块是MEG X670E ACE,另一块是MAG B650M
MORTAR WIFI,用的是锐龙9 7950X处理器,两款主板均成功开启,但Tightest模式是挑内存的,未必所有内存都能开启这一档,而Tighter则基本没啥问题。
我们以MEG X670 ACE搭配七彩虹战斧·赤焰DDR5
6000MHz内存为例,记录了各个模式下的时序小参:
预览var tableid = settableid();
document.write('');
Table().init({
title: "各模式下时序对比",
header: ["模式","Tightest","Tighter","Balace","Relax","Auto","EXPO"],
width: '92%',
columnWidth: 'MAX', //表格单元宽度,"MAX"最大内容显示,""AVG"平均,"AUTO"自动,指定[15,16,20]
id: tableid,
data: [
["主时序",],
["tCL","30","30","30","30","30","30"],
["tRCD","36","36","36","36","36","36"],
["tRP","36","36","36","36","36","36"],
["tRAS","76","76","76","76","76","76"],
["tRCD","116","116","116","116","116","116"],
["tRW","60","60","60","60","90","90"],
["tRFC1","380","550","884","884","884","884"],
["tRFC2","480","480","480","480","480","480"],
["tRFCSB","390","390","390","390","390","390"],
["tREFI","65535","65535","65535","65535","11677","11677"],
["副时序",],
["tRTP","12","12","12","12","23","23"],
["tRRDL","8","8","8","8","15","15"],
["tRRDS","4","4","4","4","8","8"],
["tFAW","28","28","28","32","32","32"],
["tWTRL","28","28","28","30","30","30"],
["tWTRS","5","5","5","8","8","8"],
["Turn Around时序",],
["tRDRDSCL","8","8","8","8","8","8"],
["tRDRDSC","1","1","1","1","1","1"],
["tRDRDSD","8","8","8","8","8","8"],
["tRDRDDD","8","8","8","8","8","8"],
["tWRWRSCL","23","23","23","23","23","23"],
["tWRWRSC","1","1","1","1","1","1"],
["tWRWRSD","15","15","15","15","15","15"],
["tWRWRDD","15","15","15","15","15","15"],
["tWRRD","8","8","8","8","8","8"],
["tRDWR","20","20","20","20","19","20"]
]
});前面五个主时序是不会变的,因为这些都是写死在EXPO的预设里面,有变化的是主时序后面那些以及副时序,Turn Around时序变化不明显
,这里用的七彩虹这套内存Auto档和直接开EXPO基本没啥区别,但另一套测试过的芝奇焰光戟Auto档则和Balace一样,所以它真的是会跟随内存不同自动变化的。
接下来我们看看各个模式下内存带宽与延迟的变化:
预览var title_text = "AIDA 64内存与缓存测试"; // 标题文字
var subtitle_text = "www.expreview.com "; // 副标题文字
var unit_text = "MB/s "; //单位
var leftside = "8%"; //左边空白
var topside = 50; //每组间距
var mblur = 0.05; //模糊效果
var bgopacity = 0; //背景透明度,0为全透明
var mysort = 0; //是否排序
var myper = 1; //是否显示百分比
var mylengend = true; //是否显示图例
var toplegend = 50; //图例上端空白
var labelcolor = '#fff'; // 标签文字颜色
var legendwidth = '90%'; //图例宽度
var titlecolor = '#222'; //标题文字颜色
var barheight = 35; //柱高度,当为7倍数时,不显示刻度线
var bargap = 3; //柱间距,相当于百分值,即2%
var xaxismax = null; //指定X轴最大值,null时为自动
var autocolor = 1; //是否自动判断颜色,0为否,1根据品牌生成颜色 ,如果 >1 则为AMD NV Intel 标准色,
var mycolor = []; //自动生成随机颜色 也可ANIColor("NVIDIA",3)可以生成三种NV绿,可替换成AMD Intel
//var mycolor = ANIColor('AMD',4); //这样自动生成4种AMD红色,其它如NVIDIA、Intel同理
//var mycolor=['#953f61', '#72355f', '#d77169']; //自定义颜色 AMD红 '#9F1417', NV绿 '#82B41B', Intel蓝 '#0075C5',
//var mycolor = ['#0dcc47', '#969696', '#6f7b80','5A94DF'];
var data_arr =[
{"type":"内存读取","EXPO":"76843","AUTO":"76737","Relax":"84433","Balace":"84483","Tighter":"84712","Tightest":"84917"},
{"type":"内存写入","EXPO":"78339","AUTO":"77197","Relax":"87257","Balace":"88851","Tighter":"87271","Tightest":"89271"},
{"type":"内存复制","EXPO":"70458","AUTO":"70400","Relax":"77292","Balace":"77659","Tighter":"77650","Tightest":"78107"},
{"type":"内存延迟","EXPO":"65.1","AUTO":"65.7","Relax":"59.9","Balace":"59.3","Tighter":"59.5","Tightest":"56.4"}
];
TDInit(title_text, subtitle_text, unit_text, leftside, topside, mblur, bgopacity, mysort, myper, mylengend, toplegend, labelcolor, legendwidth, titlecolor, barheight, bargap, mycolor, data_arr, autocolor, xaxismax);测试的这套内存用Auto档和EXPO没啥区别,不过从
最低的Relax开始变化就很明显了,内存读取从7600MB/s提升到84000MB/s,内存写入从78000MB/s提升到87000~89000MB/s,内存复制从70000MB/s提升至77000~78000MB/s,基本上带宽提升了10%。
内存延迟也有明显变化,只开启EXPO时是65ns,而Relax、Balace、Tighter这三档则是59ns,最高级的Tightest则是56ns,延迟减少了接近14%,可见
这个内存高性能模式作用还是很明显的。
微星这个内存高性能模式其实对于那些不是A-die的内存也能用,只不过对A-die内存的效果会明显一点,内置的四挡性能最好的是Tightest,然后以此是Tighter、Balace、Relax,开启的时候建议直接从最高档开始尝试,如果失败的话就一档档降下去,兼容性的排序是与性能反过来的。
其实微星的这个内存高性能模式确实蛮使用的,内存性能能直接提升个10%,对于广大玩家来说是一个福利,不知道能不能进一步推广到Intel平台上。