Re: Performance issues in 64-bit platforms
On Tuesday 20 March 2007 14:24, Johan@Elmerfjord wrote:
> Hi Dimitris,
>
> I have seen the same patterns - but I don't think it's reflecting normal
> workload - so don't be alarmed. :-)
>
> 1)
> You are comparing an Intel and a AMD processor accessing memory.
> Intel's Core-processors (and Intel in general) have always performed
> were well in this hdparm memory-tests.
> I don't know why.. but that's the case (Maybe because of there
> releatively large second level cache 4MB for Core2-duo) -
> compared to AMD that typically still runs with 512KB / 1MB.
> When it comes to real life usage the the difference is not that big. AMD
> is very fast on context-switching.
>
Well, thanx for the feedback, if this can be confirmed by someone else too,
then we can blame hdparm. But somehow I'm not convinced given that the Core2
reports 3 times faster times than the other. Its too much IMO.
> 2)
> I believe that this is syncronous reads.. The raptor disk is very fast
> in random reads as it has a low accesstimes.
> The bigger and slower disk may most likely have higher data- density -
> resulting in a high troughput anyway - as long as we are only reading
> one file at a time..
> If you compare random reads/writes between the disk you get what you
> payed for..
>
You're the second one (or actually the first) to give me this reply and I
totaly agree. This is not a problem after all.
Dimitris
PS: any hints/suggestions for benchmarking random r/w on the disks?
Reply to: