Hi all, I posted this on someone's thread regarding a much slower boot time with his new WD740 Raptor (ADFD) on a 3rd party JBMicron controller. I thought it best I start a new thread on it. I'm hoping any of you with the WD740ADFD can check this.
This might explain something about the ADFD and 3rd party controllers. I never noticed this with my old WD740 (8mb buffer). But on this new one, I don't know if it's due to some kind of command or tagged or native queuing, or the 16mb buffer, a combo of both or something else. The ADFD is supposed to have NCQ (Native command queuing).
Most people may not be aware, but on my mobo (P4C800-E Deluxe) ANY HD on the 378 Promise 3rd-party controller gives PHENOMENAL buffered HD benchmarks, with PREVIOUS HD's. This may be the same on any mobo's non-native controller. On AIDA32 for example, the buffered read results are over 1800MB/sec. That's not 180, but eighteen hundred. This is NOT a burst or peak, but CONSTANT from start to finish. On the AIDA32 random write tests, there's massive transient peak spike lasting only a ms or two I would guess that's anywhere from 150-270MB/sec depending on how many times the test is run. It varies a lot.
On HDTach 2.61-2.62, for the "Maximum write" result, there's also this massive transient peak spike lasting only a ms or two a the very instant the test begins, that is over 200-250MB/sec. On it's "Maximum read" test there is also this transient spike, but not nearly as high, about 116mb/sec for the OLD WD740, and much lower for other HD's.
So when I first started testing the new WD740, the ADFD, these peak spikes were no where to be found. Much more troubling was on the AIDA32 buffered read tests, where all other drives were 1800mb+, the ADFD was only about 85mb/sec across the chart. That's even lower than an IDE Maxtor 7200rpm/8mb on the NATIVE controller (which is about 100mb/sec)! Also on HDTach, these massive spikes were also gone. So I'm thinking there's either something wrong with this ADFD I have, or it's just plain S-L-O-W on some tests, even MUCH MUCH slower than any other drive....on the Promise controller, and also like I said above, even lower than other drives on the NATIVE controller in that one AIDA32 buffered read test.
I hooked the old WD740 back up, and it still had these same fantastic scores. So, there's no issue with the controller, nor drivers. Out of desperation, since Asus still has not come out with any new Promise drivers since 2003, I tried the 2006 Promise RAID drivers. I new they would not work since I'm not running RAID, but I thought they may add "something" to the drivers that may help. Well, they did. After XP no longer recognized the HD's attached to the Promise controller (as expected), I then did a "Rollback" of the driver and restarted. I then noticed in MSCONFIG a new entry, another "Setwriteback" type of DLL file. Previously it was ptipbm.dll, but now it was that one plus ptipbmf.dll and it says "Setwritecachemode". These are some of the Promise controller drivers. The "f" is for "Fastrack", which is RAID.
I ran these aforementioned tests again, and the ADFD then had these great results like all the other HD's I've tested had. But, it stopped! WTF??
:Q A couple of days later when I realized I forgot a few tests for the ADFD on the Promise controller, I noticed it was back to these comparatively horrible results on these specific tests, yet both of the ptipbm* files were still loading with XP. So, I'm dumbfounded. :shocked: For the life of me I can't figure out what's going on. Why would they be great, like the other HD's, even better, then go back to bad like it was previously?
Interestingly enough, when I UNCHECKED one of those ptipbm* files in MSCONFIG, the OTHER HD's ALSO dropped from these 1800mb/sec+ scores in AIDA32 Buffered Read. If I HAD to GUESS, forced to guess, I'd have to say we're dealing with these Promise drivers not correctly "working with" (for lack of a better term) with these new WD740's. But what's so new about it? It's just a HD, and just like the old WD740 except for NCQ and the 16mb buffer. I don't see how that would throw-off the controller and confuse it, and even if it did, why intermittently??? Test results should either be good all the time, or bad all the time; not bad, then great, then back to bad again! And through the bad/good/bad results, they are all totally consistent!
Maybe I do have a flaky ADFD after all, but on every other benchmark, (PCmark04 & 05, Winbench, HDtune, SANDRA, etc., etc.), on the other portions of AIDA32, HDTach, and on the Native controller, it pretty much blows away the other 12 HD's I've tested, even the old WD740. So it doesn't seem too flaky there.
If anyone's ADFD is your boot drive now, of course you can't do the write tests (AIDA32 and HDTach write tests destroy all data), but can someone please run the HDTach read (2.61 or 2.62, NOT the new version), and the AIDA32 "Buffered read" test and post the results? Also please check to see if you have any dll files or otherwise loading in MSCONFIG for your non-native 3rd party controller.
Any comments from anyone on what may be going on is appreciated. I'm hoping others will be as curious about this as I am.
Thanks.
This might explain something about the ADFD and 3rd party controllers. I never noticed this with my old WD740 (8mb buffer). But on this new one, I don't know if it's due to some kind of command or tagged or native queuing, or the 16mb buffer, a combo of both or something else. The ADFD is supposed to have NCQ (Native command queuing).
Most people may not be aware, but on my mobo (P4C800-E Deluxe) ANY HD on the 378 Promise 3rd-party controller gives PHENOMENAL buffered HD benchmarks, with PREVIOUS HD's. This may be the same on any mobo's non-native controller. On AIDA32 for example, the buffered read results are over 1800MB/sec. That's not 180, but eighteen hundred. This is NOT a burst or peak, but CONSTANT from start to finish. On the AIDA32 random write tests, there's massive transient peak spike lasting only a ms or two I would guess that's anywhere from 150-270MB/sec depending on how many times the test is run. It varies a lot.
On HDTach 2.61-2.62, for the "Maximum write" result, there's also this massive transient peak spike lasting only a ms or two a the very instant the test begins, that is over 200-250MB/sec. On it's "Maximum read" test there is also this transient spike, but not nearly as high, about 116mb/sec for the OLD WD740, and much lower for other HD's.
So when I first started testing the new WD740, the ADFD, these peak spikes were no where to be found. Much more troubling was on the AIDA32 buffered read tests, where all other drives were 1800mb+, the ADFD was only about 85mb/sec across the chart. That's even lower than an IDE Maxtor 7200rpm/8mb on the NATIVE controller (which is about 100mb/sec)! Also on HDTach, these massive spikes were also gone. So I'm thinking there's either something wrong with this ADFD I have, or it's just plain S-L-O-W on some tests, even MUCH MUCH slower than any other drive....on the Promise controller, and also like I said above, even lower than other drives on the NATIVE controller in that one AIDA32 buffered read test.
I hooked the old WD740 back up, and it still had these same fantastic scores. So, there's no issue with the controller, nor drivers. Out of desperation, since Asus still has not come out with any new Promise drivers since 2003, I tried the 2006 Promise RAID drivers. I new they would not work since I'm not running RAID, but I thought they may add "something" to the drivers that may help. Well, they did. After XP no longer recognized the HD's attached to the Promise controller (as expected), I then did a "Rollback" of the driver and restarted. I then noticed in MSCONFIG a new entry, another "Setwriteback" type of DLL file. Previously it was ptipbm.dll, but now it was that one plus ptipbmf.dll and it says "Setwritecachemode". These are some of the Promise controller drivers. The "f" is for "Fastrack", which is RAID.
I ran these aforementioned tests again, and the ADFD then had these great results like all the other HD's I've tested had. But, it stopped! WTF??
Interestingly enough, when I UNCHECKED one of those ptipbm* files in MSCONFIG, the OTHER HD's ALSO dropped from these 1800mb/sec+ scores in AIDA32 Buffered Read. If I HAD to GUESS, forced to guess, I'd have to say we're dealing with these Promise drivers not correctly "working with" (for lack of a better term) with these new WD740's. But what's so new about it? It's just a HD, and just like the old WD740 except for NCQ and the 16mb buffer. I don't see how that would throw-off the controller and confuse it, and even if it did, why intermittently??? Test results should either be good all the time, or bad all the time; not bad, then great, then back to bad again! And through the bad/good/bad results, they are all totally consistent!
Maybe I do have a flaky ADFD after all, but on every other benchmark, (PCmark04 & 05, Winbench, HDtune, SANDRA, etc., etc.), on the other portions of AIDA32, HDTach, and on the Native controller, it pretty much blows away the other 12 HD's I've tested, even the old WD740. So it doesn't seem too flaky there.
If anyone's ADFD is your boot drive now, of course you can't do the write tests (AIDA32 and HDTach write tests destroy all data), but can someone please run the HDTach read (2.61 or 2.62, NOT the new version), and the AIDA32 "Buffered read" test and post the results? Also please check to see if you have any dll files or otherwise loading in MSCONFIG for your non-native 3rd party controller.
Any comments from anyone on what may be going on is appreciated. I'm hoping others will be as curious about this as I am.
Thanks.
