Enhance SATA disk performance

Toggle sidebar Toggle sidebar
S

SoFChef

Junior Member
Oct 28, 2006
24
0
0
copy that.
but does it make a discernable difference in performance?

BTW the hardware I'm using includes a 150GB Raptor
 
Nothinman

Nothinman

Elite Member
Sep 14, 2001
30,672
0
0
If it really is write caching then it should help a good bit, but as the article says if you lose power or the machine STOPs you lose all of the data in memory that's not been flushed to disk.
 
A

Aberforth

Golden Member
Oct 12, 2006
1,707
1
0
Originally posted by: Nothinman
If it really is write caching then it should help a good bit, but as the article says if you lose power or the machine STOPs you lose all of the data in memory that's not been flushed to disk.
Click to expand...

It might improve the performance only by 10% but don't do it on Raid0 setup- because raid0 is already fragile
 
Nothinman

Nothinman

Elite Member
Sep 14, 2001
30,672
0
0
It might improve the performance only by 10% but don't do it on Raid0 setup- because raid0 is already fragile
Click to expand...

RAID0 has nothing to do with it, if your drive dies and breaks your array disk caching will have no effect at all.
 
S

Shooks

Golden Member
Jun 19, 2001
1,428
0
76
Got two 36GB Raptors in RAID-0, thanks for the link.

edit: just realized it says "This device does not allow its write cache setting to be modified."

Oh well :(
 
Nothinman

Nothinman

Elite Member
Sep 14, 2001
30,672
0
0
Nothinman, I sure hope you were being sarcastic.
Click to expand...

Not at all, as long as both drives are working you have the exact same chance of problems with write caching as you do with 1 drive or a mirror or RAID5. If one drive fails in RAID0 the array is gone no matter what kind of write caching policy you're using. Write caching is a cause for alarm when you have flaky power, common box hangs, etc but that's it.
 
F

fierydemise

Platinum Member
Apr 16, 2005
2,056
2
81
So is it worth it? Is the performance boost significant enough to compensate for possible problems?
 
Nothinman

Nothinman

Elite Member
Sep 14, 2001
30,672
0
0
It's hard to say without knowing exactly what that does, but I couldn't imagine running without write caching enabled on Linux.
 
lxskllr

lxskllr

No Lifer
Nov 30, 2004
60,937
11,267
126
I'm not in front of my Vista box to confirm, but I believe that that tab isn't the same as write caching. You should have a separate write cache box, then the enhance performance box is in addition to that.

I don't know how much it speeds things up. I was having major memory problems when I had it enabled, and ended up with a bunch of corrupted files. Maybe in the next couple of days I'll bench my drives, and post the results. Then we'll see what kind of difference it makes.
 
lxskllr

lxskllr

No Lifer
Nov 30, 2004
60,937
11,267
126
Here's my benchmark results using Sandra. Computer was restarted after every change. I only ran 1 pass each time, and only used Sandra for testing, so you should take these with a grain of salt.

Write caching off

SiSoftware Sandra

Benchmark Results
Drive Index : 49 MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 2 ms
Results Interpretation : Lower index values are better.

Performance Test Status
Run ID : DAD-PC on Sunday, February 18, 2007 at 9:34:19 PM
System Timer : 14MHz
Use Overlapped I/O : Yes
IO Queue Depth : 4 request(s)
Block Size : 1MB

Volume Information
Capacity : 298GB

Benchmark Breakdown
Speed at position 0% : 24MB/s (50%)
Speed at position 3% : 28MB/s (57%)
Speed at position 7% : 23MB/s (47%)
Speed at position 10% : 28MB/s (58%)
Speed at position 13% : 27MB/s (55%)
Speed at position 17% : 23MB/s (47%)
Speed at position 20% : 25MB/s (50%)
Speed at position 23% : 24MB/s (50%)
Speed at position 27% : 26MB/s (54%)
Speed at position 30% : 28MB/s (58%)
Speed at position 33% : 23MB/s (48%)
Speed at position 37% : 33MB/s (67%)
Speed at position 40% : 30MB/s (62%)
Speed at position 43% : 37MB/s (76%)
Speed at position 47% : 29MB/s (60%)
Speed at position 50% : 23MB/s (47%)
Speed at position 53% : 19MB/s (39%)
Speed at position 57% : 27MB/s (55%)
Speed at position 60% : 28MB/s (58%)
Speed at position 63% : 26MB/s (53%)
Speed at position 67% : 32MB/s (66%)
Speed at position 70% : 26MB/s (53%)
Speed at position 73% : 45MB/s (92%)
Speed at position 77% : 47MB/s (96%)
Speed at position 80% : 27MB/s (56%)
Speed at position 83% : 49MB/s (100%)
Speed at position 87% : 32MB/s (65%)
Speed at position 90% : 17MB/s (35%)
Speed at position 93% : 16MB/s (33%)
Speed at position 97% : 15MB/s (31%)
Speed at position 100% : 17MB/s (34%)
Random Access Time : 2 ms (estimated)
Full Stroke Access Time : 30 ms (estimated)

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.

Write caching on

SiSoftware Sandra

Benchmark Results
Drive Index : 73 MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 13 ms
Results Interpretation : Lower index values are better.

Performance Test Status
Run ID : DAD-PC on Sunday, February 18, 2007 at 9:29:16 PM
System Timer : 14MHz
Use Overlapped I/O : Yes
IO Queue Depth : 4 request(s)
Block Size : 1MB

Volume Information
Capacity : 298GB

Benchmark Breakdown
Speed at position 0% : 63MB/s (87%)
Speed at position 3% : 73MB/s (100%)
Speed at position 7% : 70MB/s (97%)
Speed at position 10% : 72MB/s (100%)
Speed at position 13% : 67MB/s (93%)
Speed at position 17% : 68MB/s (93%)
Speed at position 20% : 71MB/s (97%)
Speed at position 23% : 67MB/s (92%)
Speed at position 27% : 69MB/s (95%)
Speed at position 30% : 67MB/s (92%)
Speed at position 33% : 67MB/s (92%)
Speed at position 37% : 64MB/s (88%)
Speed at position 40% : 65MB/s (89%)
Speed at position 43% : 64MB/s (88%)
Speed at position 47% : 64MB/s (88%)
Speed at position 50% : 64MB/s (88%)
Speed at position 53% : 61MB/s (84%)
Speed at position 57% : 59MB/s (82%)
Speed at position 60% : 60MB/s (82%)
Speed at position 63% : 57MB/s (78%)
Speed at position 67% : 49MB/s (68%)
Speed at position 70% : 55MB/s (76%)
Speed at position 73% : 55MB/s (75%)
Speed at position 77% : 52MB/s (71%)
Speed at position 80% : 51MB/s (70%)
Speed at position 83% : 49MB/s (67%)
Speed at position 87% : 46MB/s (64%)
Speed at position 90% : 44MB/s (60%)
Speed at position 93% : 44MB/s (61%)
Speed at position 97% : 38MB/s (52%)
Speed at position 100% : 34MB/s (47%)
Random Access Time : 13 ms (estimated)
Full Stroke Access Time : 17 ms (estimated)

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.

Write caching and enhanced performance on


SiSoftware Sandra

Benchmark Results
Drive Index : 73 MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 13 ms
Results Interpretation : Lower index values are better.

Performance Test Status
Run ID : DAD-PC on Sunday, February 18, 2007 at 9:25:31 PM
System Timer : 14MHz
Use Overlapped I/O : Yes
IO Queue Depth : 4 request(s)
Block Size : 1MB

Volume Information
Capacity : 298GB

Benchmark Breakdown
Speed at position 0% : 69MB/s (95%)
Speed at position 3% : 73MB/s (100%)
Speed at position 7% : 70MB/s (97%)
Speed at position 10% : 72MB/s (100%)
Speed at position 13% : 59MB/s (82%)
Speed at position 17% : 68MB/s (93%)
Speed at position 20% : 44MB/s (61%)
Speed at position 23% : 29MB/s (40%)
Speed at position 27% : 27MB/s (37%)
Speed at position 30% : 33MB/s (46%)
Speed at position 33% : 30MB/s (42%)
Speed at position 37% : 49MB/s (68%)
Speed at position 40% : 65MB/s (89%)
Speed at position 43% : 64MB/s (88%)
Speed at position 47% : 64MB/s (88%)
Speed at position 50% : 64MB/s (88%)
Speed at position 53% : 61MB/s (84%)
Speed at position 57% : 59MB/s (82%)
Speed at position 60% : 60MB/s (82%)
Speed at position 63% : 57MB/s (78%)
Speed at position 67% : 55MB/s (75%)
Speed at position 70% : 49MB/s (68%)
Speed at position 73% : 55MB/s (75%)
Speed at position 77% : 47MB/s (64%)
Speed at position 80% : 42MB/s (58%)
Speed at position 83% : 49MB/s (67%)
Speed at position 87% : 43MB/s (59%)
Speed at position 90% : 39MB/s (54%)
Speed at position 93% : 40MB/s (56%)
Speed at position 97% : 38MB/s (52%)
Speed at position 100% : 34MB/s (47%)
Random Access Time : 13 ms (estimated)
Full Stroke Access Time : 17 ms (estimated)

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.


It appears to me that the biggest gains are seen by enabling write caching. The enhanced performance tab reduced results in some areas, and increased results in others. Further testing would be needed to really check enhanced performance out. I'm guessing that it doesn't make a huge difference though.
 
lxskllr

lxskllr

No Lifer
Nov 30, 2004
60,937
11,267
126
Sorry for the double post guys. The write caching off results were bothering me because they were so low. I retested without restarting the computer, and retested after a restart. The results were much better when testing before restart, SuperFetch doing it's thing? After the restart the results were a bit better, but still not up to the write caching on speeds.

Retest write caching off before restart

SiSoftware Sandra

Benchmark Results
Drive Index : 73 MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 15 ms
Results Interpretation : Lower index values are better.

Performance Test Status
Run ID : DAD-PC on Sunday, February 18, 2007 at 9:46:40 PM
System Timer : 14MHz
Use Overlapped I/O : Yes
IO Queue Depth : 4 request(s)
Block Size : 1MB

Volume Information
Capacity : 298GB

Benchmark Breakdown
Speed at position 0% : 69MB/s (94%)
Speed at position 3% : 73MB/s (100%)
Speed at position 7% : 70MB/s (97%)
Speed at position 10% : 72MB/s (100%)
Speed at position 13% : 67MB/s (93%)
Speed at position 17% : 68MB/s (93%)
Speed at position 20% : 71MB/s (97%)
Speed at position 23% : 67MB/s (92%)
Speed at position 27% : 69MB/s (95%)
Speed at position 30% : 67MB/s (92%)
Speed at position 33% : 67MB/s (92%)
Speed at position 37% : 64MB/s (88%)
Speed at position 40% : 65MB/s (89%)
Speed at position 43% : 64MB/s (88%)
Speed at position 47% : 64MB/s (88%)
Speed at position 50% : 64MB/s (88%)
Speed at position 53% : 61MB/s (84%)
Speed at position 57% : 46MB/s (64%)
Speed at position 60% : 60MB/s (82%)
Speed at position 63% : 57MB/s (78%)
Speed at position 67% : 55MB/s (75%)
Speed at position 70% : 55MB/s (76%)
Speed at position 73% : 55MB/s (75%)
Speed at position 77% : 52MB/s (71%)
Speed at position 80% : 51MB/s (70%)
Speed at position 83% : 49MB/s (67%)
Speed at position 87% : 46MB/s (64%)
Speed at position 90% : 44MB/s (60%)
Speed at position 93% : 44MB/s (61%)
Speed at position 97% : 38MB/s (52%)
Speed at position 100% : 36MB/s (50%)
Random Access Time : 15 ms (estimated)
Full Stroke Access Time : 16 ms (estimated)

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.


Retest Write caching off after restart

SiSoftware Sandra

Benchmark Results
Drive Index : 67 MB/s
Results Interpretation : Higher index values are better.
Random Access Time : 33 ms
Results Interpretation : Lower index values are better.

Performance Test Status
Run ID : DAD-PC on Sunday, February 18, 2007 at 9:52:15 PM
System Timer : 14MHz
Use Overlapped I/O : Yes
IO Queue Depth : 4 request(s)
Block Size : 1MB

Volume Information
Capacity : 298GB

Benchmark Breakdown
Speed at position 0% : 20MB/s (30%)
Speed at position 3% : 20MB/s (29%)
Speed at position 7% : 21MB/s (31%)
Speed at position 10% : 20MB/s (30%)
Speed at position 13% : 21MB/s (31%)
Speed at position 17% : 21MB/s (31%)
Speed at position 20% : 22MB/s (32%)
Speed at position 23% : 21MB/s (31%)
Speed at position 27% : 20MB/s (31%)
Speed at position 30% : 24MB/s (36%)
Speed at position 33% : 67MB/s (100%)
Speed at position 37% : 23MB/s (34%)
Speed at position 40% : 21MB/s (32%)
Speed at position 43% : 34MB/s (51%)
Speed at position 47% : 22MB/s (32%)
Speed at position 50% : 24MB/s (35%)
Speed at position 53% : 28MB/s (41%)
Speed at position 57% : 24MB/s (36%)
Speed at position 60% : 23MB/s (34%)
Speed at position 63% : 24MB/s (35%)
Speed at position 67% : 24MB/s (35%)
Speed at position 70% : 25MB/s (37%)
Speed at position 73% : 21MB/s (31%)
Speed at position 77% : 21MB/s (31%)
Speed at position 80% : 24MB/s (35%)
Speed at position 83% : 21MB/s (32%)
Speed at position 87% : 19MB/s (29%)
Speed at position 90% : 19MB/s (28%)
Speed at position 93% : 16MB/s (24%)
Speed at position 97% : 21MB/s (31%)
Speed at position 100% : 17MB/s (26%)
Random Access Time : 33 ms (estimated)
Full Stroke Access Time : 28 ms (estimated)

Performance Tips
Notice 5008 : To change benchmarks, click Options.
Notice 5004 : Synthetic benchmark. May not tally with 'real-life' performance.
Notice 5006 : Only compare the results with ones obtained using the same version!
Tip 2 : Double-click tip or press Enter while a tip is selected for more information about the tip.
 
Chaotic42

Chaotic42

Lifer
Jun 15, 2001
35,386
2,503
126
It's a good bit faster. My same-disk copy speeds were 11MB/s vs 14MB/s for a 190MB directory. Of course, since this is a 15K SCSI drive, I'm not really in love with either of those speeds.
 
X

xtknight

Elite Member
Oct 15, 2004
12,974
0
71
Originally posted by: Nothinman
Perhaps it's NCQ?
Click to expand...

Doubtful, there's no reason to disable NCQ as long as the drive supports it.
Click to expand...

NCQ on some nForce 4 controllers was not enabled by default (it caused data corruption due to a bug). I think in general it is not enabled by default.
 
Nothinman

Nothinman

Elite Member
Sep 14, 2001
30,672
0
0
NCQ on some nForce 4 controllers was not enabled by default (it caused data corruption due to a bug). I think in general it is not enabled by default.
Click to expand...

That should be the exception and not the rule. If there really was a data loss bug in those controllers then it should be disabled on them, but that doesn't mean everyone else should pay for nVidia's screwup.
 
A

aka1nas

Diamond Member
Aug 30, 2001
4,335
1
0
NCQ and read caching on NF4 chipsets tends to give poorer performance and significantly higher CPU utilization from the tests that I have done. My guess is that the chipset was buggy and they had to compensate for it in the drivers somewhat.
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom