Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

nanofrog

macrumors G4
May 6, 2008
11,719
3
It always happens on first cold boot (after night in the morning) ;)
Which is why I suspected it was power related...

Though what Areca mentioned sounds like a timing issue (and I suspect that's the case), it could still be power related. Specifically, the capacitors (PSU, mainboard, ARC-1231ML) need time to charge (data is sent later than the system is supposed to receive it due to insufficient power to generate that data). It's a very brief period of time by human standards (ms), but if something isn't ready when the clock cycle hits, erroneous data results, and all hell breaks loose.

In such a case, a restart would work, as the caps are already charged on the second attempt to boot the system. What happens, is capacitors drain out over time, and won't contain sufficient energy on the next clock cycle unless they're charged again before the next clock cycle is reached. So when you turn off the system, the capacitors are discharged below the threshold value (minimum needed for the circuits to operate correctly), and you get this particular error.

Since the restart works, this is the most logical explanation based on experience.

As per a solution, you can either boot it twice in order to get the card running at x8 lanes, or leave it on over night.

Either way works, and is simple enough that either the additional aggravation or power consumption should be an acceptable compromise. ;)
 

Spacedust

macrumors 6502a
Original poster
May 24, 2009
999
160
Which is why I suspected it was power related...

Though what Areca mentioned sounds like a timing issue (and I suspect that's the case), it could still be power related. Specifically, the capacitors (PSU, mainboard, ARC-1231ML) need time to charge (data is sent later than the system is supposed to receive it due to insufficient power to generate that data). It's a very brief period of time by human standards (ms), but if something isn't ready when the clock cycle hits, erroneous data results, and all hell breaks loose.

In such a case, a restart would work, as the caps are already charged on the second attempt to boot the system. What happens, is capacitors drain out over time, and won't contain sufficient energy on the next clock cycle unless they're charged again before the next clock cycle is reached. So when you turn off the system, the capacitors are discharged below the threshold value (minimum needed for the circuits to operate correctly), and you get this particular error.

Since the restart works, this is the most logical explanation based on experience.

As per a solution, you can either boot it twice in order to get the card running at x8 lanes, or leave it on over night.

Either way works, and is simple enough that either the additional aggravation or power consumption should be an acceptable compromise. ;)

I'll still be investigating this issue. I got a BBU, so my card should be always powered on ;)
 

Spacedust

macrumors 6502a
Original poster
May 24, 2009
999
160
Actually, all the BBU keeps powered is the cache, not the entire card.

Wouldn't help with your issue anyway if it did.

Unplugged my Mac Pro from electric wall for whole night, booting in the morning and I got 8x ;)
 

nanofrog

macrumors G4
May 6, 2008
11,719
3
Unplugged my Mac Pro from electric wall for whole night, booting in the morning and I got 8x ;)
You might want to check your system memory (malfunctions from bad RAM can cause strange behavior during boot - seen this before).
 

Spacedust

macrumors 6502a
Original poster
May 24, 2009
999
160
You might want to check your system memory (malfunctions from bad RAM can cause strange behavior during boot - seen this before).

You were right ! One of my modules died in this week. I have a replacement, because I thought the other stick was dead, but accually it was the other one. Good I didn't sold that old one as a dead one ;)

I will stick keep looking at this and maybe it will be fixed permanently now.
 

nanofrog

macrumors G4
May 6, 2008
11,719
3
You were right ! One of my modules died in this week. I have a replacement, because I thought the other stick was dead, but actually it was the other one. Good I didn't sold that old one as a dead one ;)

I will stick keep looking at this and maybe it will be fixed permanently now.
Hopefully there aren't any others (bad RAM sticks), thus eliminating your issues. :)
 

Spacedust

macrumors 6502a
Original poster
May 24, 2009
999
160
Hopefully there aren't any others (bad RAM sticks), thus eliminating your issues. :)

Still working ok ;)

I need to RMA another Kingston 16 GB (2 Bad Clusters), so probably I will get money back for 3 SSD drives. These A-data 30 GB (32GB) are really powerful (280/250 read/write speeds). Four such drives in RAID 0 should max out this controller. Currently I got horrible write speeds on RAID 5 (about 13-20 MB/s).
 

nanofrog

macrumors G4
May 6, 2008
11,719
3
Still working ok ;)

I need to RMA another Kingston 16 GB (2 Bad Clusters), so probably I will get money back for 3 SSD drives. These A-data 30 GB (32GB) are really powerful (280/250 read/write speeds). Four such drives in RAID 0 should max out this controller. Currently I got horrible write speeds on RAID 5 (about 13-20 MB/s).
What are you using for the RAID 5 disks, and how is the card configured (could be a number of things)?

Logs and detailed settings data would go a long way, as would the model of disks being used.
 

Spacedust

macrumors 6502a
Original poster
May 24, 2009
999
160
What are you using for the RAID 5 disks, and how is the card configured (could be a number of things)?

Logs and detailed settings data would go a long way, as would the model of disks being used.

Ch01 Windows 7 32.0GB ADATA SSD S396 30GB
Ch02 Windows 7 32.0GB ADATA SSD S396 30GB
Ch03 Windows 7 16.0GB KINGSTON SS100S216G
Ch04 Windows 7 16.0GB KINGSTON SS100S216G

One 48 GB RAID5 volume.

JBOD/RAID Configuration RAID
Max SATA Mode Supported SATA300+NCQ
HDD Read Ahead Cache Enabled
Volume Data Read Ahead Aggressive
HDD SMART Status Polling Enabled
Disk Write Cache Mode Enabled
Disk Capacity Truncation Mode No Truncation

Volume Raid Level RAID5
Volume Stripe Size 128 kB
Volume Cache Mode Writeback
Tagged Command Queuing Enabled

Using latest drivers on Windows 7 64-bit Ultimate. I can't switch to RAID0 now, because I will be replacing both Kingston drives to A-data 32 GB then I will resize the volume and convert to RAID0 :)

Current speeds:

yjtzwxdo.jpg
 
Last edited:

nanofrog

macrumors G4
May 6, 2008
11,719
3
  1. Do those disks support TRIM or other form, such as Garbage Collection?
  2. What about their reliability?
  3. How long have they been in use, how full is it (relates to wear leveling), and in particular, in a RAID5 configuration?
I ask, as consumer grade SSD's are not suited to RAID5 (parity writes burn through the memory cells' write cycle limit).

The other factors for questions 1 & 2 are important, but ultimately, it will come down to #3 (write cycle limit).

There are enterprise grade SSD's, but they're very expensive, so only have a limited market (SLC based rather than MLC).
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.