Conroe/MB Issue with Raid5

[imported_haris]

Interesting new problem just cropped up over at the Inq. http://theinquirer.net/default.aspx?article=32818

Any chance someone can do some testing to replicate the issue they seem to have found?

 

[Keysplayr]

Originally posted by: haris
Interesting new problem just cropped up over at the Inq. http://theinquirer.net/default.aspx?article=32818

Any chance someone can do some testing to replicate the issue they seem to have found?

AND please state the chipset, revision, and model on the mobo that you use... Seems like they said it may be southbridge related. That would be a good thing if that's all it is. If nvidia makes any Conroe boards, it might have a ULi Southbridge on it. Just thinking out loud here.

 

[CyborgNinja117]

Its the chipset... not conroe, and wtf has a cpu got to do w/ raid 5? (except for providing 1% cpu power)

 

[Keysplayr]

Originally posted by: CyborgNinja117
Its the chipset... not conroe, and wtf has a cpu got to do w/ raid 5? (except for providing 1% cpu power)

True. And I personally don't know anybody on these boards that has a RAID 5 setup in their desktop. RAID 0 or 1, yes. 5? No. It is typically used in server environments anyways and most likely the serverboard for Woodcrest will not be using the 975 or 965 chipsets anywho.

 

[Duvie]

Originally posted by: keysplayr2003

Originally posted by: CyborgNinja117
Its the chipset... not conroe, and wtf has a cpu got to do w/ raid 5? (except for providing 1% cpu power)

True. And I personally don't know anybody on these boards that has a RAID 5 setup in their desktop. RAID 0 or 1, yes. 5? No. It is typically used in server environments anyways and most likely the serverboard for Woodcrest will not be using the 975 or 965 chipsets anywho.

MarkFW900 has a Raid 5 in his dual opteron setup....ofcourse it is running off a 133mhz PCI-X slot so that will rule out any Conroe chipsets...

 

[LittleNemoNES]

The inquirer's getting worst. If it was crap before now its...fermented crap?



What a disgracefully misleading headline.

 

[kknd1967]

That is normal INQ. Not worse not better. Just the way it always is.
If I can post there, I also would like to post "Pentium-M just burned a Coaster on my expensive Verbatim media"

Originally posted by: gersson
The inquirer's getting worst. If it was crap before now its...fermented crap?



What a disgracefully misleading headline.

 

[Check]

If you want RAID5 buy a RAID card, that's what they are made for. How the hell are you supposed to run any sort of RAID5 array off of the two SATA ports that they give you for RAID anyway?

 

[Markfw]

Originally posted by: check
If you want RAID5 buy a RAID card, that's what they are made for. How the hell are you supposed to run any sort of RAID5 array off of the two SATA ports that they give you for RAID anyway?

1) Raid5 requires at least 3 drives
2) That motherboard probably has more than 2 sata ports.
3) The raid5 controller is probably on the motherboard. Either way, its the same chip....

 

[Madwand1]

Originally posted by: keysplayr2003
I personally don't know anybody on these boards that has a RAID 5 setup in their desktop. RAID 0 or 1, yes. 5? No. It is typically used in server environments anyways and most likely the serverboard for Woodcrest will not be using the 975 or 965 chipsets anywho.

There are tons of storage servers being used in home environments. Some of them use server boards with real HW raid, some of them use cheap consumer NAS boxes, and some of them use something in between. Some environments even use two such servers -- one for the data, and another for backups.

When consumer boards come with nice gigabit and a large number of SATA ports (which many current boards do, 4/6/8 ports), they become very attractive for this sort of usage, where price is a significant factor. Such boards also have PCIe slots that can be used for some higher-end controllers for those willing to potentially spend more. Server MB's and their paraphanelia are generally a big step up in pricing.

RAID 5 can be a tough nut to crack for write performance, but you pretty much always get good read performance, so such boards are still a good option, and the ports and RAID capability are on the boards just for this reason.

Whether or not I ended up using on-board RAID, or OS RAID, or an add-on controller, I'd look for a MB with hopefully a good on-board RAID, and then bench that myself against some other options and go from there. So this "Conroe" MB issue could be a real issue. Shame on INQ for not providing any real details or numbers though, or even identfying a site that tends to know what they're talking about to handle this issue.

 

[imported_haris]

They already have another followup at the Inq: http://theinquirer.net/default.aspx?article=32842

This looks like it could be a serious problem for Intel's Woodcrest processor. I just wish there were more details on what configurations were causing problems.

 

[Markfw]

Originally posted by: haris
They already have another followup at the Inq: http://theinquirer.net/default.aspx?article=32842

This looks like it could be a serious problem for Intel's Woodcrest processor. I just wish there were more details on what configurations were causing problems.

WoodCrest systems were delivered for an evaluation for one of the US government departments, and it all ended up in tears, with Woodcrest being rejected in the first round of trials. The reason was that there were severe problems when Woodcrest was paired with a 1E RAID field when using IBM ServeRAID controllers. The problems didn't occur just in benchmarking, it was the every-day usage model that produced unexpected errors.

Worse of all, several problematic situations occurred during the trial period and Intel was heavily criticised in internal memos, all ending up in a really unexpected manner. Since the company has allegedly been aware of the problems and didn't responded in a timely manner, the supply contract was then lost and awarded to a different product coming from a competing company, which passed the field test with flying colours.

Woa ! Sounds like AMD got a big contract !

 

[zsdersw]

Woa ! Sounds like AMD got a big contract !

Maybe.. or maybe they simply went with someone other than IBM and some system other than a Woodcrest-based one (perhaps Dempsey). The INQ article is absent of those details.

 

[Markfw]

Originally posted by: zsdersw

Woa ! Sounds like AMD got a big contract !

Maybe.. or maybe they simply went with someone other than IBM and some system other than a Woodcrest-based one (perhaps Dempsey). The INQ article is absent of those details.

It did say "different product coming from a competing company,", so not an Intel. Who else makes compatable chips ? AMD

 

[dmens]

The CPU cannot see the external world, *if* there is a problem (I say if because Inq is really full of crap lately), it's with the RAID controller paired with the motherboard.

 

[HurleyBird]

Same problem, two different RAID controllers. It may not be a coincidence.

 

[zsdersw]

It did say "different product coming from a competing company,", so not an Intel. Who else makes compatable chips ? AMD

There are server vendors other than IBM, you know. The article doesn't state whether or not they went with another company's non-Woodcrest Xeons or another company's AMD-based offerings.

 

[Keysplayr]

Originally posted by: Madwand1

Originally posted by: keysplayr2003
I personally don't know anybody on these boards that has a RAID 5 setup in their desktop. RAID 0 or 1, yes. 5? No. It is typically used in server environments anyways and most likely the serverboard for Woodcrest will not be using the 975 or 965 chipsets anywho.

There are tons of storage servers being used in home environments. Some of them use server boards with real HW raid, some of them use cheap consumer NAS boxes, and some of them use something in between. Some environments even use two such servers -- one for the data, and another for backups.

When consumer boards come with nice gigabit and a large number of SATA ports (which many current boards do, 4/6/8 ports), they become very attractive for this sort of usage, where price is a significant factor. Such boards also have PCIe slots that can be used for some higher-end controllers for those willing to potentially spend more. Server MB's and their paraphanelia are generally a big step up in pricing.

RAID 5 can be a tough nut to crack for write performance, but you pretty much always get good read performance, so such boards are still a good option, and the ports and RAID capability are on the boards just for this reason.

Whether or not I ended up using on-board RAID, or OS RAID, or an add-on controller, I'd look for a MB with hopefully a good on-board RAID, and then bench that myself against some other options and go from there. So this "Conroe" MB issue could be a real issue. Shame on INQ for not providing any real details or numbers though, or even identfying a site that tends to know what they're talking about to handle this issue.

Oh, well my mistake then. I guess everybody uses RAID5 then.
No seriously, If this is an Intel chipset problem, they better address it very quickly. That being said, if I had bought a mobo that supposedly had RAID0,1,5 capability, and I never opted to use RAID5, I still would want the option to work. But truthfully, I would never really see myself using a RAID5 setup. Hook up 3 250GB drives and you still only get 500GB of storage. Not many desktop users "with the exception of Mark" are willing to "donate" 1/3 of their total storage to parity bits. Servers? Hell yes. I wouldn't do it any other way. With differential/incremental backups every night and full backups every week. My favorite way to setup a server is RAID 5 with hot spare. Desktops? meh.

 

[evilharp]

Before this thread turns into a Flame thread (or Inquirer bash session) can we please get back to haris' request:

Any chance someone can do some testing to replicate the issue they seem to have found?

Now, there are quite a few Conroe and Woodcrest in circulation (most in xtremesystem.org hands) so it should be possible to either support or dispel this "issue".

Ideally:

1) Tests with onboard Raid-5
2) Tests with good "hardware" Raid-5 controller(s)
3) Tests with onboard software goodies (quoting the inq "how will a two-meg cache Conroe work on a fully integrated motherboard, with Gigabit Ethernet, software-based sound-card ? ever popular AC'97 codec and so on." Inq Link )

I am hoping that it is a myth, but you never know. The Inq is right on the money every now and again.

 

[Markfw]

Well, somebody loan me a Woodcrest or Conroe motherboard with a 133 mhz PCI-X slot, and I will test ti out. I have hardware raid card from LSI, with on-board controller (Intel no less) and 5 drives that I can stripe in raid 5 (right now in 0, but I can blow it away, just a extra boot partition) and I will test it out.

 

[Brahmzy]

3 drives(minimum) for RAID 5.

 

[mhahnheuser]

Interesting thread guys. Here's something that's come up in my shed that might be of some interest. My son and I have been slowly assembling a server based on 2 X P6 866MHz i865 platform where I tried to add a 120Gb SATA 1 Seagate HDD. Gave up after being frustrated by system unable to complete successful format of the drive. I've been blaming the SATA PCI card that I purchased for the job, but reading this Conroe post has got me taking another look since Conroe is based on P6 architecture. Also just added a 1000mb LAN PCI card to above server which is running currently with 10Gb Seagate IDE HDD, and guess what? System has suffered a massive slow down that until know haven't been able to explain. Hmmmmm??????? I wonder?

 

[Keysplayr]

Originally posted by: Markfw900
Well, somebody loan me a Woodcrest or Conroe motherboard with a 133 mhz PCI-X slot, and I will test ti out. I have hardware raid card from LSI, with on-board controller (Intel no less) and 5 drives that I can stripe in raid 5 (right now in 0, but I can blow it away, just a extra boot partition) and I will test it out.

Hehe, you could probably get a setup through your company easier than trying to pry one from the hands of someone her who finally gets one.

I plan on getting the E6600 and a 965 board when available and I believe I have 3 WD 120GB SATA drives to test with. It'll be a while before I get this done, so we should see many more web reviews before anyone sees my results.

 

[Keysplayr]

Originally posted by: mhahnheuser
Interesting thread guys. Here's something that's come up in my shed that might be of some interest. My son and I have been slowly assembling a server based on 2 X P6 866MHz i865 platform where I tried to add a 120Gb SATA 1 Seagate HDD. Gave up after being frustrated by system unable to complete successful format of the drive. I've been blaming the SATA PCI card that I purchased for the job, but reading this Conroe post has got me taking another look since Conroe is based on P6 architecture. Also just added a 1000mb LAN PCI card to above server which is running currently with 10Gb Seagate IDE HDD, and guess what? System has suffered a massive slow down that until know haven't been able to explain. Hmmmmm??????? I wonder?

I don't know that you can compare 2x866 PIII's on an 865? chipset did you say? How on earth did you get/find an 865 chipset on a PIII board?

 

[mhahnheuser]

Sorry no 865 should have been as below, slack in that i didn't chk chipset before writing.

Serverworks? Serverset III HE-SL chipset
? NB-HE-SL + OSB4 + CIOB2

Just interesting to read symptoms as described and having never before had SATA setup probs on other similar aged pc's and have done a few. HDD is definitely ok so have been blaming PCI SATA card although I also tried to install SATA drive alongside current IDE drive (same card) and it installed without conflict with other devices including SCSI interface. Drive was recognised, but continued to fail format. Also interesting that Giga LAN PCI card caused system to slow to almost unusable level, but haven't had time to troubleshoot yet. System also has wireless LAN and onboard LAN activated so I still have some exploring to do on this level. But interesting that wireless PCI card did not seem to impact on system performance at all. Conroe has its roots in P6 architecture, so maybe some of these issues may have been dormant within the architecture itself? Now the answers could be nothing at all to do with this problem, but it's given me a bit of heart that mabe the PCI card may be ok. I was intending to use this computer as a small network server, (20 workstations) but this issue has somewhat stonewalled the project.