I understand. If I were you I would be feeling the same way...I'm not frustrated, maybe just a little A-D-D!I want to help as much as possible in getting to the bottom of these sleep issues people are having.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
3.33GHz Xeon (Server CPU) Transplant in a 2009 Quad
- Thread starter AppleWorking
- Start date
- Sort by reaction score
I understand. If I were you I would be feeling the same way...
... really and why is that?
All I did was to disconnect cables from the system. Waited at least 15 seconds. Plugged the power cord in. Then keyboard/mouse etc... Powered on... This is per Apples guidelines...
No real work necessary.. No problems yet... System behaves as it should.
Bad News...
Well, despite my confidence in having solved my wake from sleep issues I am now having login/logout issues. I have spent the better part of the night diagnosing the problem. It appears my login/logout issues are directly tied into whether I put the computer to sleep. So while the computer is sleeping and waking without issue, once I perform one or more sleep/wake cycles I then start to experience login/logout issues. On last attempt to login, I got my desktop but no icons, computer appeared to try to reboot itself with the spinning wheel, but it would be caught in an infinite loop and I would have to perform a hard shut down.
While I did not have these issues with Core i7 + Leopard, it appears Core i7 + Snow Leopard may be a bad combination. There is something in SL that is causing my system to hose itself. My Xeon W3580 is on order, I will report back next week whether that has solved my problem. It would appear from last night's experience that while sleep/wake is working, it's causing corruption elsewhere that is crashing login/logout.
Stay tuned.
Well, despite my confidence in having solved my wake from sleep issues I am now having login/logout issues. I have spent the better part of the night diagnosing the problem. It appears my login/logout issues are directly tied into whether I put the computer to sleep. So while the computer is sleeping and waking without issue, once I perform one or more sleep/wake cycles I then start to experience login/logout issues. On last attempt to login, I got my desktop but no icons, computer appeared to try to reboot itself with the spinning wheel, but it would be caught in an infinite loop and I would have to perform a hard shut down.
While I did not have these issues with Core i7 + Leopard, it appears Core i7 + Snow Leopard may be a bad combination. There is something in SL that is causing my system to hose itself. My Xeon W3580 is on order, I will report back next week whether that has solved my problem. It would appear from last night's experience that while sleep/wake is working, it's causing corruption elsewhere that is crashing login/logout.
Stay tuned.
I may be the last man standing.... Still working great... I guess I could try a power down startup to see if it changes anything....
*** UPDATE ***
Turned off auto login... Shut the machine down... Started up. Logged in. Put it to sleep.. Woke it up and wrote this... Put it to sleep again. And just woke up... No problems!
Looking into the Console messages everything is working fine.:
9/26/09 9:04:46 AM kernel System Sleep
9/26/09 9:04:54 AM kernel Wake reason = HDEF EHC1 EHC2 ((I moved my PS/2 Mouse - yes I use one of these))
9/26/09 9:04:54 AM kernel System Wake
9/26/09 9:04:54 AM kernel Previous Sleep Cause: 5
9/26/09 9:04:54 AM kernel USB (EHCI)ort 2 on bus 0xfa has remote wakeup from some device
9/26/09 9:04:54 AM kernel EIR is supported.
9/26/09 9:04:54 AM kernel SSP is supported.
9/26/09 9:04:57 AM kernel Ethernet [Intel82574L]: Link up on en0, 1-Gigabit, Full-duplex, Symmetric flow-control, Debug [796d,af08,0de1,0200,cde1,3c00]
9/26/09 9:04:59 AM kernel GbE en1: en1 Link down
9/26/09 9:05:01 AM configd[14] network configuration changed.
9/26/09 9:05:01 AM configd[14] network configuration changed.
9/26/09 9:05:04 AM kernel SerialATAPI device reconfiguration did not complete successfully. (failedCommandInfo = 0x1)
9/26/09 9:05:05 AM mDNSResponder[16] Note: DeregisterInterface: Frequent transitions for interface en0 (192.168.1.8); network traffic reduction measures in effect
9/26/09 9:05:09 AM ntpd[59] time reset +0.427245 s
*** UPDATE ***
Turned off auto login... Shut the machine down... Started up. Logged in. Put it to sleep.. Woke it up and wrote this... Put it to sleep again. And just woke up... No problems!
Looking into the Console messages everything is working fine.:
9/26/09 9:04:46 AM kernel System Sleep
9/26/09 9:04:54 AM kernel Wake reason = HDEF EHC1 EHC2 ((I moved my PS/2 Mouse - yes I use one of these))
9/26/09 9:04:54 AM kernel System Wake
9/26/09 9:04:54 AM kernel Previous Sleep Cause: 5
9/26/09 9:04:54 AM kernel USB (EHCI)
ort 2 on bus 0xfa has remote wakeup from some device9/26/09 9:04:54 AM kernel EIR is supported.
9/26/09 9:04:54 AM kernel SSP is supported.
9/26/09 9:04:57 AM kernel Ethernet [Intel82574L]: Link up on en0, 1-Gigabit, Full-duplex, Symmetric flow-control, Debug [796d,af08,0de1,0200,cde1,3c00]
9/26/09 9:04:59 AM kernel GbE en1: en1 Link down
9/26/09 9:05:01 AM configd[14] network configuration changed.
9/26/09 9:05:01 AM configd[14] network configuration changed.
9/26/09 9:05:04 AM kernel SerialATAPI device reconfiguration did not complete successfully. (failedCommandInfo = 0x1)
9/26/09 9:05:05 AM mDNSResponder[16] Note: DeregisterInterface: Frequent transitions for interface en0 (192.168.1.8); network traffic reduction measures in effect
9/26/09 9:05:09 AM ntpd[59] time reset +0.427245 s
My Xeon should arrive on Wednesday or Thursday of next week and I'll report back as soon as I've popped it in and had a chance to test it out. Good idea to check out the console... I just did on my MacBook Pro and found a launch daemon constantly starting and stopping for a program I deleted a long time ago. I think I've already purged my Mac Pro of anything that could be causing issues with Snow Leopard. I should have a more complete picture by week's end. I'm pretty anal and thorough with my troubleshooting, I'll file an update as soon as I have something to report.
I was having an issue at the office with my EyeTV software that was conflicting with a 4870 and not the Atari 2600xt... LOL.. I mean ATI ;-) After removing the software from startup... My sleep issues there were resolved... basically I put it to sleep and it would wake back up.. I would guess most sleep issues are related to a conflict and NOT the CPU!!!! Doh!
I honestly think people are just panicking...
I'm probably panicking as well, but since I rely on my rig for business on a daily basis I would rather be safe than sorry. There's a great market for Core i7 975 CPUs on eBay if I need to re-sell that baby. Ultimately I'll either add myself to the voices claiming that the Xeon fixed the issues or I'll continue having these weird login/logout issues regardless. One way or the other I'll figure out whether it's the CPU to blame.
As others have said, the Core i7 may just be different enough from its Xeon counterpart so as to cause issues. If it is due to lack of ECC support or slight voltage differences, I'm not sure how that relates to sleep, wake from sleep, or login/logout. It's most likely not solely the ECC support -- if it were, you would have the same issues as the others. If it is voltage related, it's possible that some Core i7 samples are drawing more or less than the 0.8v-1.225v specification of the Xeon part. Since the Core i7 can draw 0.8v-1.375v, the CPU could be drawing 12% more than the Xeon specification if the particular sample is requiring a full 1.375v to run at 3.3 GHz. Macs are notoriously sensitive with regards to RAM and whatnot, it wouldn't be unreasonable to think that an additional 12% voltage draw from the CPU may be "out of spec" and could cause something in memory to get hosed. Since the Core i7 doesn't have ECC support, there would be no chance of recovery from a memory error. It's easy to see how this could lead to a kernel panic or other type of crash under the "right" set of circumstances.
Snow Leopard also introduces new technologies and optimizations that may be sensitive to the CPU swap. This would be consistent with the fact that I had no issues prior to the Snow Leopard upgrade. This could also be nothing more than a rogue, incompatible task causing SecurityAgent to crash on login/logout. I'll keep tweaking and report my findings.
As others have said, the Core i7 may just be different enough from its Xeon counterpart so as to cause issues. If it is due to lack of ECC support or slight voltage differences, I'm not sure how that relates to sleep, wake from sleep, or login/logout. It's most likely not solely the ECC support -- if it were, you would have the same issues as the others. If it is voltage related, it's possible that some Core i7 samples are drawing more or less than the 0.8v-1.225v specification of the Xeon part. Since the Core i7 can draw 0.8v-1.375v, the CPU could be drawing 12% more than the Xeon specification if the particular sample is requiring a full 1.375v to run at 3.3 GHz. Macs are notoriously sensitive with regards to RAM and whatnot, it wouldn't be unreasonable to think that an additional 12% voltage draw from the CPU may be "out of spec" and could cause something in memory to get hosed. Since the Core i7 doesn't have ECC support, there would be no chance of recovery from a memory error. It's easy to see how this could lead to a kernel panic or other type of crash under the "right" set of circumstances.
Snow Leopard also introduces new technologies and optimizations that may be sensitive to the CPU swap. This would be consistent with the fact that I had no issues prior to the Snow Leopard upgrade. This could also be nothing more than a rogue, incompatible task causing SecurityAgent to crash on login/logout. I'll keep tweaking and report my findings.
Quick question, did you do a fresh install of snow kitty first? I would assume you did because of your thoroughness...
I contacted OWC and they said they would test it but they never got back to me.
Our best bet would be if one of us contacted Robert at barefeats and asked him to test this for us. He would need to use mem with the specific JEDEC SPD programming for 1333 in it. And of course he would have to test this in one of the 2009 octos not quad, since only those xeons support the 1333 mem (I am assuming, I haven't looked). If it does work, we can now be relatively certain that it will work with the W3580 since it also supports the spec.If you like, you can be our guinea pig and test it for us.
Like nano said,Please start a new thread, and let the forum know what the results are.
I think there's a few that would appreciate it.
Good luck bud!
http://ark.intel.com/Compare.aspx?ids=37106,40200,39719,39723,
The 2.66 and 2.93 quads don't support 1333, nor does the 2.26 octo. The whole point of testing is to see if the speed depends on apples motherboard or not. Either someone who installed the 3.33 or who has the 2.66 or 2.93 octo needs to test it. And considering you only need 3 1GB sticks in the quad, it would make a lot more sense to test it with the quad rather than an octo.
Quick question, did you do a fresh install of snow kitty first? I would assume you did because of your thoroughness...
.
.
.
If you like, you can be our guinea pig and test it for us.
Good luck bud!
After replacing my Core i7 later this week with a Xeon part, I'll be done with any guinea pig behavior for a good, long while.
It would be fun to know whether the Mac Pro mobo supports 1333. I wouldn't put it past Apple to disable the functionality in firmware even though the hardware could theoretically support it.Regarding snow kitty, I just did an upgrade install. I wanted to save the potential hassle of re-installing all of my software. There's something about clean installs that tends to wipe out my serializations. I chose the path of least resistance. I think I've only ever done one clean install of Leopard due to a hard drive failure, but other than that all of my OS X installs have been upgrades. Never had any issues until now.
I'll report my findings on Friday with regards to the Xeon replacement.
The hardware itself, does support it. Intel has published this information.
The real question is if Apple intentionally yanked it in the firmware.
The hardware itself, does support it. Intel has published this information.
The real question is if Apple intentionally yanked it in the firmware.
This is why we need access to the firmware! Why is this such an impossible task? I can't believe in all of these years nobody has released any information on accessing and changing firmware settings on the Mac. I'll put up a $100 bounty for anyone who can crack the firmware and find out how to change settings that would allow us to enable 1333, higher CPU clock rates, voltages, etc. If a few more of us ante up, it may end up being a worthwhile project for some industrious man or woman...
It is truly a darn shame that the same W3580s and W5590s that we are putting into our Mac Pros can be over clocked in excess of 4 GHz on much less expensive PC motherboards. But on this brilliantly designed, well engineered Mac Pro we are stuck at the core clock frequency without any options for experimenting.
Reverse engineering firmware isn't easy. You need a disassembler/decompiler, and knowledge of the machine code (instruction set used) it usually doesn't exist for the current standards used. So you have to write them.This is why we need access to the firmware! Why is this such an impossible task? I can't believe in all of these years nobody has released any information on accessing and changing firmware settings on the Mac. I'll put up a $100 bounty for anyone who can crack the firmware and find out how to change settings that would allow us to enable 1333, higher CPU clock rates, voltages, etc. If a few more of us ante up, it may end up being a worthwhile project for some industrious man or woman...
It is truly a darn shame that the same W3580s and W5590s that we are putting into our Mac Pros can be over clocked in excess of 4 GHz on much less expensive PC motherboards. But on this brilliantly designed, well engineered Mac Pro we are stuck at the core clock frequency without any options for experimenting.
Recompiling is easier, as compilers exist for the EFI byte code, but they're not free.
It's a lot of effort, and is usually done for patent reasons. Either to nail some company that's infringed on your IP, or make sure they can't nail you. At least the legal reasons.
Hey Appleworking.... Any anyone else who jumped to 3.3 Ghz...
How are your fan speeds working with the W3580?
While I'm still not having any sleep speed issues, I've noticed that the BOOSTA fan takes off from 1041RPM and jumps up to 2000 - which is noisy - when my cpu hits around 39(c). This is with minimal CPU activity. I first noticed when working with the system for a long time this evening... I tried to control the maximum fan speed with SMCFANCONTROL, but was unable to change the F4Mx register on the command line.
This seems to be present in SL 10.6.1 and not Leopard...
I agree.. But someone has part of it figured it out.. This is why i emailed the author HERE @ coolbook.se last night
How are your fan speeds working with the W3580?
While I'm still not having any sleep speed issues, I've noticed that the BOOSTA fan takes off from 1041RPM and jumps up to 2000 - which is noisy - when my cpu hits around 39(c). This is with minimal CPU activity. I first noticed when working with the system for a long time this evening... I tried to control the maximum fan speed with SMCFANCONTROL, but was unable to change the F4Mx register on the command line.
This seems to be present in SL 10.6.1 and not Leopard...
Reverse engineering firmware isn't easy. You need a disassembler/decompiler, and knowledge of the machine code (instruction set used) it usually doesn't exist for the current standards used. So you have to write them.
Recompiling is easier, as compilers exist for the EFI byte code, but they're not free.
It's a lot of effort, and is usually done for patent reasons. Either to nail some company that's infringed on your IP, or make sure they can't nail you. At least the legal reasons.
I agree.. But someone has part of it figured it out.. This is why i emailed the author HERE @ coolbook.se last night
I haven't experienced any weird fan issues with my Core i7 that I know of, although my office is a somewhat noisy environment where I probably wouldn't notice anyway. Temps hover right around 38-40C under light use. I can drop that to 28-30C if I double the minimum RPMs with SMC Fan Control. You may want to boost your minimum RPMs to keep your CPU running cooler. I understand the Xeon part runs slightly cooler at the stock 3.3 GHz speed.
I've been doing some research into this as the fan noise is killing me on this new box compared to the MacPro 1,1. This looks like its either a bug with MacPro 4,1, and/or SL.... Was reading over at macupdate where a bunch of 4,1 owners were talking about high heat on when playing basic audio.. and the fans kicking in. Just trying to make sure this isnn't the CPU....
I installed iStat Menu (THX MacRumors) and is it really shows how BOOSTA goes into overtime... And temps really rise up.. With minimal activity... Guess this could be from the CPU jumping up in Mhz?
I installed iStat Menu (THX MacRumors) and is it really shows how BOOSTA goes into overtime... And temps really rise up.. With minimal activity... Guess this could be from the CPU jumping up in Mhz?
I just opened iTunes and activated the visualizer while playing a track, then opened simultaneously played two 1080p Quicktime video streams over the internet. This caused my Core i7 to quickly jump to 57C, but my BOOSTA fan speed remained at 857 RPM (it was at 855 RPM when temps were at 38C). I am attaching a screen shot of my iStat Nano showing fan speeds while CPU temp was at 57C.
No unusual fan activity here. I don't think this is necessarily CPU related, however, as all of the problems so far identified in this thread are also being experienced by other Snow Leopard users who haven't swapped their CPUs. However, it will be nice to finally eliminate the CPU as the culprit.
Attachments
Are you running SL or Leopard?
I'm not seeing any activity like this in Leopard... I guess I could to a clean install tonight and see if it's still happening.....
The W3570 is 3.2GHz and is $400 less. Why did you chose the W3580?
Please forgive me if I missed out something in the thread...
Please forgive me if I missed out something in the thread...
Please forgive me if I missed out something in the thread...
I'm not sure I agree with that price. At what reputable place can you find the W3570 for $600-$700? The lowest price I could find from an online retailer was $818, otherwise the price is pretty much identical to the W3580. All the prices are coming in from around $1,000-$1,100 per CPU from places such as newegg.com, buy.com, etc.
Perhaps they are cheaper on eBay, but personally I'd have an issue buying a CPU on eBay. I'm just not comfortable with that. I'm comfortable selling things on eBay though! Isn't that a hoot.
DOH! I bought my i7 9775 on ebay for $750... Couldn't resist.
I'm running on my 8x2.8 2008 MP at the office. This box is rock solid and just churns. I don't think I't ever noticed the fan spinning up in the last year. For that matter, neither did my 2006 that I replaced with the 2009...