Jump to content

Decrease Hardisk performances in 10.4.8

El Tubaron

6 posts in this topic

Recommended Posts

Well, after the installation of the new 10.4.8 Jas DVD and correcting some problems (network, ATM, Sound and Nvida :) ), everything is much more stable than with the 10.4.7 (no doubt about that). But, there is always a but :P ---- I've got decreasing Hardisk performances in the new version ....


Disk Test 34.35

Sequential 59.14

Uncached Write 39.38 24.18 MB/sec [4K blocks]

Uncached Write 61.97 35.06 MB/sec [256K blocks]

Uncached Read 84.05 24.60 MB/sec [4K blocks]

Uncached Read 70.36 35.36 MB/sec [256K blocks]

Random 24.20

Uncached Write 7.95 0.84 MB/sec [4K blocks]

Uncached Write 56.20 17.99 MB/sec [256K blocks]

Uncached Read 83.56 0.59 MB/sec [4K blocks]

Uncached Read 103.03 19.12 MB/sec [256K blocks]


for a sata hardisk I think it could be better :D and with the 10.4.7 kernel the disk test was better... any ideas to boost up my hardisk?


Ps: Config in the signature - all hardisk ar sata drives

Link to comment
Share on other sites

On 10.4.8 Jas Intel SSE3 DVD using Semthex




Results 51.15

System Info

Xbench Version 1.3

System Version 10.4.8 (8L2127)

Physical RAM 1024 MB

Model ACPI

Drive Type WDC WD800JB-00FSA0

Disk Test 51.15

Sequential 71.54

Uncached Write 89.40 54.89 MB/sec [4K blocks]

Uncached Write 82.81 46.86 MB/sec [256K blocks]

Uncached Read 83.98 24.58 MB/sec [4K blocks]

Uncached Read 48.20 24.23 MB/sec [256K blocks]

Random 39.81

Uncached Write 14.19 1.50 MB/sec [4K blocks]

Uncached Write 96.07 30.76 MB/sec [256K blocks]

Uncached Read 90.55 0.64 MB/sec [4K blocks]

Uncached Read 116.89 21.69 MB/sec [256K blocks]

Link to comment
Share on other sites

It has been said a billion times, the HD benchmark of Xbench 1.3 just doesn't work. Even on real Mac it gives bas results.


The problem is with 4K blocks. Everybody has the problem and the standard Xbench procedure here is now NOT to enable the disk benchmark.


It seems that nobody cares/dares to contact the author to ask him why it does that, especially on our hackintoshes. Nothing in real life actually shows a slowdown, that's what actually matters. So even if the author has a technical answer, it's flawed since a benchmark should test the performance for people, not just for users.

Link to comment
Share on other sites

I agree with Bamboo. You shouldn't worry about it.


And I (personally) think that author knows about it (since there are thousans of results on his site) and the result is JUST FINE. Maybe reading these 4K blocks in these synthetic conditions is as it's supposed to be.


Nothing to worry about.

Link to comment
Share on other sites

  • 6 months later...

Think about it from a different point of view..... most users don't realize how much hard drive speed changes performance of many things....


AND he IS providing both 256k blocks... a more user oriented test... and 4k blocks.. a more server oriented test.


Think about reading a database... think about loading a bunch of small text config files.. think about serving up pages on websites, think about loading small font files... think about appending to log files...


And sure... you'll be loading most BIG things sequentially... if your hard drive isn't fragmented beyond belief....


So the VERY slow speed of some of the random reads and writes doesn't make much difference if you're only writing say 500k.... or even 5mb... since it passes it ALL to the hard drive at once... and the hard drive will write it over the next 1-10 seconds....


So yeah, for many USERS its no big deal.... but some people work day in, day out.... with THOUSANDS of tiny files at a time... and for them... it is a VERY BIG DEAL... and its an important comparison.


NCQ helps...... but SCSI still wins on reliability, speed, access times, and handling huge batches of disk requests.


Ask yourself... do you do that kindof work? If not... then don't worry about a low speed there.... but for everyone elses sake... please include those hard drive benchmarks, so that when people like me look at the benches... we can see what drives are doing what in what systems.... (and thus, on what controller...)



If one of your games got 150fps in all areas except a VERY IMPORTANT level... and it got 10fps on that level.. would you guys drop that level from the benchmark results just to get a higher score? or would you say: "Damn.. that level is slow as hell for everyone.... who's doing what to get through it faster..."

Link to comment
Share on other sites


  • Create New...