Xeon X5680s Now $45... October 2018

[HCMarkus]

Concern: Some say if you have the 2.26GHz "Nehalem" Mac Pro, don't upgrade to a processor with higher power consumption. The Nehalem is rated at 95W, and the 2010 Westmere at 130W. One post says the best upgrade for the dual 4,1 is the 5675. (A pair costs only around $100 now.) Pretty sure I've read accounts of people using 5690s in the 2.26, but I'd rather play it safe.

That post also recommends against de-lidding and provides what looks like a pretty straightforward solution: a 5mm thick thermal pad on the bottom of the heat sink; two 1mm thick washers on each post; and removing clips from the fan connector.

Bob

Moved to this thread to consolidate.

De-lidded CPUs allow the 4,1 dual to function as engineered and are the best way to go IMO. De-lidding eliminates the risk of bent CPU socket pins; the washer approach does not entirely avoid this potentially costly problem. De-lidding also allows a more direct path between CPU and the Mac Pro heatsink, as originally engineered by Apple. That said, many 4,1 > 5,1 upgrades have been performed using lidded CPUs, both with and without the washer stack. Thermal pad addition is definitely required if IHS is not removed.

I'm pretty sure the successful and trouble-free X5680/90 dual CPU Mac Pros number in the many thousands, including a significant number of MOTUNation members; concerns expressed by the referenced poster are, IMO, exaggerated.

As a DAW user, I'd go for the dual hex and de-lid. But that's just me...

[BobK]

Just thinking out loud here, people. Any feedback appreciated.

1. Power consumption

Somewhere - I forgot to note the link - Thomas Pindelski wrote that the 2009 Mac Pro (with 90W rated chips) and the 2010 Mac Pro (with 135W rated chips) use the same power supply, so there's no need to worry about putting the more power-hungry processors in the 2009 box. I don't recall if he included the source of this info.

This page at Apple Support says the 2009 8 x 2.26 CPU power consumption rating is higher than all subsequent models (though the only stats for 3 GHz or higher are the trash can Mac Pro - the towers are all rated with 2.x GHz chips).

https://support.apple.com/en-us/HT201796

Still, knowing nothing about this stuff, I wonder if there might be other components involved that could affect the ability of the 2009 MP to use higher-TDP chips. I realize lots of people are doing this - just trying to do some due diligence (or being overly cautious, depending on your point of view ).

And this parts supplier lists the same power supply for Mac Pro 2009-2010 models:

http://www.macpartsonline.com/661-5449- ... a1289.html

This all supports the notion that the 2009 can handle higher-power-rated chips, but I don't know if there are other hardware or software issues to consider.

2. Removing/installing heat sinks

The Apple Technician Guide for the 2009 Mac Pro gives detailed instructions on removal and replacement of the heat sinks, starting on page 156 (8-core) and page 162 (4-core). I don't think I've seen anyone quote these.

- Use the x-pattern not just for tightening the screws, but also for loosening them.

- Apply thermal grease in the pattern of an oval with a horizontal line through it.

- Use one hand to keep the heatsink level while tightening the screws.

- Use no more than two turns of the driver per screw at a time when tightening.

- Tighten to "finger tight", then a quarter-turn more.

3. Using non-delidded processors

I'm not sure that using washers with de-lidded CPUs is really necessary if you follow the above procedure for tightening the screws. Then again, they probably can't hurt, as long as each washer stack is the same height.

Even though the CPU lids add an extra layer between the CPU surface and the heat sink (vs de-lidded), that surface is much larger than the CPU. Makes me wonder if that might be an advantage.

Then I have to decide whether to stack another 3mm strip of thermal pad on top of the existing ones, or remove those and start with a fresh 5mm pad.

Anyone had problems with either approach?

[Michael Canavan]

Anyone had problems with either approach?

I de-lidded mine. I personally would not recommend using the CPUs with the lids. It's terrifying for sure to do the method in the video I posted earlier in this thread, but it works well, and really the tightening of the screws on the heat sinks is even with de-lidded CPUs the most sphincter clenching part of the operation. The only real pitfall I had was thinking that I had torqued the screws down enough to make contact and realizing when I turned on the Pro that they weren't! So, to me anyway, having to torque down on CPUs not designed for the machine with washers as your line of defense is a good way to possibly crack your chip. The stress of using an iron and cleaning CPUs of excess solder is nothing compared to that.

As far as power consumption, there's not a thing to worry about. The only concern might be heat, as the 12 core chips run hotter, there are cheap and free fan control apps you can use to boost the speed if you do not trust OSX to regulate the temperature. I do it because I also run a M.2 XP941 SSD as the boot drive, and a SM951 as a drive for samples. Both of those also run hot. Realistically OSX probably would adjust fan speed for heat, but caution isn't a bad thing considering that's three sources for extra heat.

[BobK]

Anyone had problems with either approach?

I personally would not recommend using the CPUs with the lids.

Why not?

I'm looking for specific reasons to not do this. I'm already aware of the argument "the 2009 Mac Pros were designed for de-lidded CPUs". I'm curious if there are other reasons. I don't think the risk of crushing the CPU is any worse with non-delidded ones; and I don't see how the risk of damaging the socket pins is any greater with the lid than without.

FWIW, according to these listings, the same heat sink is used in 2009, 2010, and 2012 Mac Pros:

https://www.macpartsdepot.com/076-1338- ... RBEALw_wcB

http://www.applecomponents.com/items/07 ... gxEALw_wcB

There could still be differences in the mounts, though. Maybe the CPU retaining clips help protect the CPU in the 2010 and later models, and without them, there's greater risk of damage (?).

[HCMarkus]

The real risk is not crushing the CPU, it is crushing the CPU socket pins.. they are delicate and, on the 4,1 Dual, are calibrated to the height of a lidless CPU. The web is replete with stories of folks who trashed a CPU tray. Here's the first one of which I am aware, many years back:

http://www.anandtech.com/show/2800/upgr ... mac-pro/11

I'd rather waste a $80 CPU than a $400 tray. https://www.dvwarehouse.com/661-4998-Ap ... 38517.html

Just sayin'.

By the way, this guy has a $130 for 2 CPUs de-lidding service (he has a foolproof method) if you are loath to try it yourself:

http://www.ebay.com/itm/2009-Mac-Pro-Du ... SweW5VGXm2

OTOH, many have successfully done the 4,1 mod with lidded CPUs, like Mr. Pidelski. Another approach is to sell your 4,1 and buy a 5,1, then upgrade that.

Taking a lead from the Moody Blues... "You decide which is right, and which is an illusion!"
However you proceed, the upgrade is well worth it if you need more CPU HP for DP.

[BobK]

The real risk is... it is crushing the CPU socket pins

Yes, I'm aware of that!

[the socket pins]on the 4,1 Dual, are calibrated to the height of a lidless CPU.

Do you mean the pins shorter than in the 2009?

I checked the 2010 Mac Pro Technician Guide and the instructions for securing the heat sink are the same as for the 2009 - finger tight plus 1/4 turn.

So I would think that this results in the same amount of pressure being put on the CPU and pins, unless there's something else stopping the heat sink from going down too far.

In the 2010 manual, it looks like the CPU retainers (absent in the 2009) are below the CPU lid, in which case they wouldn't stop the heat sink.

Just trying to get as much hard info as possible to assess the risk, given that I don't have 2009 and 2010 machines to compare in person.


EDIT: Now that I think about it, when the kit from DM computing not only gets you the de-lidding service, but a guarantee that the chips work. Probably worth it.

[Michael Canavan]

I'm looking for specific reasons to not do this. I'm already aware of the argument "the 2009 Mac Pros were designed for de-lidded CPUs". I'm curious if there are other reasons. I don't think the risk of crushing the CPU is any worse with non-delidded ones; and I don't see how the risk of damaging the socket pins is any greater with the lid than without.

FWIW, according to these listings, the same heat sink is used in 2009, 2010, and 2012 Mac Pros:

https://www.macpartsdepot.com/076-1338- ... RBEALw_wcB

http://www.applecomponents.com/items/07 ... gxEALw_wcB

There could still be differences in the mounts, though. Maybe the CPU retaining clips help protect the CPU in the 2010 and later models, and without them, there's greater risk of damage (?).

The heat sinks are the same, the screws themselves are different. I was cautious about over tightening as I said, but I recall it seeming as if they were designed to only go so far. You simply have more chances of things going wrong in my mind with lids that are higher and washers, or modified specs VS engineered. With the lid on you have to modify the setup with washers to help prevent messing the hardware up. Once the lid is gone you're simply swapping out a chip. Or to flatly state it, I haven't heard of someone crushing the pins in a de-lidded mod. Maybe it's possible? but I haven't heard of it.

That's what I recall from reading on it and doing it myself, it's up to you which method you use, personally I like being able to see my possible areas for mistakes, so de-lidding and installing to spec is much less frightening to me than playing with washers and hoping it's all tight enough. Like I said, I was shocked that I had to actually open up and apply more torque to the screws to get contact, it had already felt like contact was made. Tightening to "finger tight , then a quarter-turn more" was not enough, probably due to remnants of the Apple applied thread lock adhesive in the screw holes I would guess? but it messes up the perceived 'tightness'.

[BobK]

Just ordered two 5680s from what appears to be a decent vendor on eBay. Will probably de-lid using a combination of a razor blade around the edges to saw through some of the glue, and then a vise (without hammer).

EDIT: I ended up using razor blades plus a clothes iron, like Michael Canavan did. See my later posts.

I didn't realize at first how many videos there are on YouTube of people de-lidding various CPUs (including a 5690) with vises. Doesn't look too bad - you just have to take it slow and pay attention.

If I trash a CPU in the process, I'll have lost $80. If I left the lid on and trashed a socket, I'd be out $450 for a new logic board.

Will report back.

[James Steele]

If I trash a CPU in the process, I'll have lost $80. If I left the lid on and trashed a socket, I'd be out $450 for a new logic board.

I think your logic there is flawless.

[BobK]

If I trash a CPU in the process, I'll have lost $80. If I left the lid on and trashed a socket, I'd be out $450 for a new logic board.

I think your logic there is flawless.

If this music thing doesn't work out, maybe I have a future in risk assessment.

[BobK]

Almost there!

5680s delidded and recognized.

But RAM slot 1 shows as empty, and red LED on logic board is lit in between the CPUs. Pretty sure that LED is indicating the RAM problem.

Tried swapping RAM sticks with slot 2, and the RAM is fine.

EDIT: see my later post - this seems to be caused by a defective CPU.

I saw in one video where tightening the CPU screws resolved this, but I've tried this, and the bolts are tight enough that I'm getting scared to go any further.

Also, I can't reset NVRAM. EDIT: RESOLVED - needed to re-enable SIPin order for cmd-opt-P-R to work. See my post below for details.

Suggestions?

Thanks!

[Michael Canavan]

Almost there!

5680s delidded and recognized.

But RAM slot 1 shows as empty, and red LED on logic board is lit in between the CPUs. Pretty sure that LED is indicating the RAM problem.

Tried swapping RAM sticks with slot 2, and the RAM is fine.

I saw in one video where tightening the CPU screws resolved this, but I've tried this, and the bolts are tight enough that I'm getting scared to go any further.

Also, I can't seem to reset NVRAM. I think it might have restarted once when I held down cmd-opt-P-R, but then it fully booted. I'm used to it restarting until I've let go.

Suggestions?

Thanks!

Are you sure both CPUs are recognized? I seem to recall a problem like that... when one CPU wasn't being recognized. did you get all the solder off the chips? I'm not sure how RAM could be affected by the CPU's, maybe someone else can chime in about that..

[James Steele]

Almost there!

5680s delidded and recognized.

But RAM slot 1 shows as empty, and red LED on logic board is lit in between the CPUs. Pretty sure that LED is indicating the RAM problem.

Tried swapping RAM sticks with slot 2, and the RAM is fine.

I saw in one video where tightening the CPU screws resolved this, but I've tried this, and the bolts are tight enough that I'm getting scared to go any further.

Also, I can't seem to reset NVRAM. I think it might have restarted once when I held down cmd-opt-P-R, but then it fully booted. I'm used to it restarting until I've let go.

Suggestions?

Thanks!

Awesome that the delidding worked out. Please let us know what method you used. This is something I plan on attempting at some point. Sorry to hear you're having issues with your RAM. I hope you get it worked out. Hope you thought to run GeekBench before the swap. It's a lot of fun to see your before and after!

[BobK]

Please let us know what method you used.

Basically, the same method as Michael:

1. Used a single-edge razor blade to cut through the seal around the edges, taking care to not go too far in. One video recommended starting at the lower right corner, so I did. Probably any corner would be fine, but definitely start in a corner. As mentioned in a video, rock the blade back and forth, don't use a sawing motion.

2. Inserted one razor blade in each side, making a frame.

3. Put the frame on a roll of artist tape I had handy - the kind you put on mixing boards.

4. Set a clothes iron on top. In about 30 seconds, the CPU dropped down to the desktop. Then I realized I had a window fan right in front of me, so turned it on and grabbed the CPU with a double thickness of paper towel and held it in the breeze to cool it off quickly.

Here's the video Michael linked to showing the iron technique:

https://www.youtube.com/watch?v=eItYq6nxfJ4&t=24s

The iron was on the highest setting, but I only let it warm up for a minute.

I found it nerve-wracking, but it's not difficult if you have patience and a steady hand.

Hope this helps.

[BobK]

EDIT: You can skip this post unless you're having a problem resetting NVRAM after doing the firmware update. I resolved that issue (see below), but it didn't enable the computer to see the RAM stick in question. That seems to be due to a defective CPU - see my later post for details.

Troubleshooting steps taken to reset NVRAM:

- Plugged keyboard directly into computer instead of through a USB hub.

- Tried a terminal command I found online.

These didn't work. Then I thought maybe it has something to do with the fact that after disabling SIP to do the firmware upgrade, I didn't re-enable it.

I tried to boot into recovery mode to reset SIP from Terminal, but it kept booting to the regular system drive. Yikes.

Then I found online that if you hold down the option key, you can boot from a recovery disk on a Time Machine backup. I have a Time Machine backup in one of the internal drives. I restarted holding the option key, then chose a recovery volume that showed up in the list. It wasn't clear exactly which disk had that recovery volume, but I guess it doesn't matter - at least it worked!

I launched Terminal and re-enabled SIP.

After rebooting, the cmd-option-P-R command worked.

Phew.

But resetting NVRAM didn't help with the RAM recognition problem.