Difference in MB/s on RW vs Average Read Access on HDD's?

[Logikos]

Question for you guys..

In terms of broad performance, explain something to me.

I have benchmarked two drives, both 7200rpm , 1TB.

Drive A) WD Caviar Black 1TB:

Linear read (Begin) - 105.4 MB/s
Linear read (Middle) - 90.2 MB/s
Linear read (End) - 54.1 MB/s
Random Read - 78.5 MB/s
Buffered Read - 215.0 MB/s
Average Read Access - 12.19 ms

Drive B) Seagate Barracuda 7200.10, 1TB:

Linear read (Begin) - 133.5 MB/s
Linear read (Middle) - 111.7 MB/s
Linear read (End) - 70.8 MB/s
Random Read - 109.6 MB/s
Buffered Read - 246.3 MB/s
Average Read Access - 13.59 ms


See, one drive well out performs the other in MB/s on read across the board (didn't test write).. but falls an average of 1.5ms slower on the 'read access'.

So which drive is better? I would have thought those numbers would be tied into each other.. the drives are same capacity.. roughly same use.. so I would think such a bench the lowest latency would read the highest MB/s read times. But that, clearly, isn't the case.

In this case the Seagate Barracuda 7200.12 1TB outperforms the WD Caviar Black 1TB in sheer MB/s read speed, .. yet the WD is an easier 1.5ms faster on average (sometimes greater) on average, random, access time.

Help me make sense of it.

Both drives are under a year old.. both drives have under 200gigs used.

Thanks,

 

[Mr Alpha]

How did you do the random read test?

 

[Logikos]

How did you do the random read test?

With Everest Pro edition,.. started them both at the same time with each hand (not that , that matters).

 

[tuskenraider]

Platter size. 180MB platters for the Seagate, 334MB for the Black...........

 

[Logikos]

Platter size. 180MB platters for the Seagate, 334MB for the Black...........

Huh?

The Seagate has two 500GB platters..

The Black has three 334GB platters, you were correct there. But your logic is backwards given this factual change on the Seagate.

But, nonetheless, you are probably right here.. the platters 2 vs 3 are likely responsible for the transfer rates vs seek time.

So I suppose the equation would be,.. smaller platter size = faster access/seek time but slower throughput. Conversely.. large platter size = slower access/seek time (avg/random) but considerable faster MB/s throughput.

Interesting.

Which outweighs the other.

I am inclined to think the 20-50MB/s throughput increase outweighs the 1-1.5ms latency difference.

Making the Barracuda the winner over the Caviar. I know the Caviar gets alot of praise, but comparing the two, I am happier with my Seagate.

 

[Mr Alpha]

With Everest Pro edition,.. started them both at the same time with each hand (not that , that matters).

I don't know exactly how Everest does their random read benchmark, but I think they use quite a big block size, meaning their random read isn't very random and so depends mostly on the linear read performance. There are, of course, other things that matter like queue depth. Try running some other benchmark like Crystal Disk Mark to get more data.

Platter size. 180MB platters for the Seagate, 334MB for the Black...........

180GB platters in the Seagate doesn't make any sense. That means they would have to have six platters to reach 1TB. I always thought the 1TB 7200.10 was four platter, which means at least 250GB platter size.

 

[Firestrider]

Is that drive a 7200.12 or 7200.10 (you said both in your first post)? That might be where the confusion is on platter sizes.

 

[tuskenraider]

180GB platters in the Seagate doesn't make any sense. That means they would have to have six platters to reach 1TB. I always thought the 1TB 7200.10 was four platter, which means at least 250GB platter size.

My typo, 188MB is what I read. I just googled it and that's what they stated sharkyextreme and I posted. Either way, different platter sizes.

 

[Mr Alpha]

Huh?
But, nonetheless, you are probably right here.. the platters 2 vs 3 are likely responsible for the transfer rates vs seek time.

So I suppose the equation would be,.. smaller platter size = faster access/seek time but slower throughput. Conversely.. large platter size = slower access/seek time (avg/random) but considerable faster MB/s throughput.

Interesting.

Which outweighs the other.

I am inclined to think the 20-50MB/s throughput increase outweighs the 1-1.5ms latency difference.

Making the Barracuda the winner over the Caviar. I know the Caviar gets alot of praise, but comparing the two, I am happier with my Seagate.

Generally works out like: More platters = slightly lower access time, and higher platter density = higher throughput. Which is why the biggest drives are the fastest since they combine many platters with high density.

I personally prefer access time over throughput since it helps more in most desktop usage scenarios. But things like firmware tuning also matters in complex workloads which isn't revealed by these kinds of synthetic benchmarks.

My typo, 188MB is what I read. I just googled it and that's what they stated sharkyextreme and I posted. Either way, different platter sizes.

I suppose they could be talking about density per square inch. Then 188 would make more sense.