KxngDemon
2021-08-26T03:52:38+00:00
hwinfo64可显示zen3 imc报告的实时内存读写带宽信息
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-l88fK18T1kSfc-b9.png[/img]
接下来验证这个数据的可靠性。
5900x设置内存频率3800,理论带宽上限为60.8G/s。
运行Intel Memory Latency Checker的读取带宽测试,实时读取带宽53G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-gxllK4ToS8a-1g.png[/img]
Memory Latency Checker的测试结果也是53G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-6xxnKeT3cSii-2v.png[/img]
运行MemTestPro内存烧机,实时带宽约为50G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-37opK4ToS85-1b.png[/img]
运行prime95 large ffts烧机,实时带宽约为49G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-bsruK4ToS7y-18.png[/img]
以上测试表明该实时带宽的数值是可靠和准确的。
而运行aida64内存跑分的读取带宽测试中,实时带宽仅有29G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-gyvcK5ToS80-1h.png[/img]
写入带宽测试中,实时带宽仅有27G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-8m82K4ToS8d-1f.png[/img]
复制带宽测试中,实时带宽只有28G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-4dyoK4ToS7x-1a.png[/img]
刚好分别大约是aida64测试结果的二分之一
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-fqtyZeT3cSif-hq.png[/img]
可以认为aida64的内存跑分只实际使用了内存带宽的一半左右,跑分结果是其直接乘2倍。
难道aida64的内存跑分实际是在跑单通道?
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-l88fK18T1kSfc-b9.png[/img]
接下来验证这个数据的可靠性。
5900x设置内存频率3800,理论带宽上限为60.8G/s。
运行Intel Memory Latency Checker的读取带宽测试,实时读取带宽53G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-gxllK4ToS8a-1g.png[/img]
Memory Latency Checker的测试结果也是53G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-6xxnKeT3cSii-2v.png[/img]
运行MemTestPro内存烧机,实时带宽约为50G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-37opK4ToS85-1b.png[/img]
运行prime95 large ffts烧机,实时带宽约为49G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-bsruK4ToS7y-18.png[/img]
以上测试表明该实时带宽的数值是可靠和准确的。
而运行aida64内存跑分的读取带宽测试中,实时带宽仅有29G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-gyvcK5ToS80-1h.png[/img]
写入带宽测试中,实时带宽仅有27G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-8m82K4ToS8d-1f.png[/img]
复制带宽测试中,实时带宽只有28G/s:
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-4dyoK4ToS7x-1a.png[/img]
刚好分别大约是aida64测试结果的二分之一
[img]https://img.nga.178.com/attachments/mon_202108/27/9aQ8j2d-fqtyZeT3cSif-hq.png[/img]
可以认为aida64的内存跑分只实际使用了内存带宽的一半左右,跑分结果是其直接乘2倍。
难道aida64的内存跑分实际是在跑单通道?