2008 upgrade path

[hyram]

I currently have a quad 2007 mp, but have a possible good deal on a quad 2008 mp, and I'm wondering about the upgrade possibilities for the 2008 mp. Specifically, the 2008 quad comes with a single quad processor, but the motherboard has 2 sockets. Does the 2008 mp quad come with both heatsinks or would I have to find one of those if I want to use both sockets??? Are there any other potential pitfalls I should be aware of upgrading a 08 quad to oct???

[J the Ninja]

The "You need to find the heatsink yourself" is probably the biggest one. The other one is that Apple had a lapse of judgement and forgot that the Mac Pro is supposed to subsidize the pro apps and not be sold on a thin margin. As a result, the machine wound up with CPUs that cost around $850 apiece when new. The CPU you need in an Intel Xeon E5462. There is a 2.83Ghz Penryn Xeon with a 1333Mhz FSB, that many people think is the 2008 MP CPU. It is NOT! The E5462 is a different chip, that uses a 1600Mhz FSB. You must get this one. If you get one the other one, the machine won't to boot. Oh yeah, the other reason people mix them up is that the E5462 is hard to find loose, almost all of them wound up in Mac Pros.

[Umbongo]

The heatsinks can be found for around $100. Apple Palace (http://www.applepalace.com/apple-mac/apple-parts/mac-pro-parts/mac-pro/heatsink-kit-processor-with-bumpers-and-top-gasket-076-1233.asp) and Mac Palace (http://macpalace.com/Scripts/prodView.asp?idproduct=2782) sell them.

There are two steppings (revisions) of the Xeon E5462. Most on eBay currently seem to be the C0 stepping (SLANT) where the revision is E0 (SLBBN). I don't know what Apple used (i.e if the E0s were available before they shipped) or if it matters if they are different. You are looking at $500-$600 for a used processor Unless you find a bargain. Should be able to get a new one for $800.

[hyram]

Thanks for the help guys. Know of anyone that has actually done this???

hyram

[nanofrog]

The heatsinks can be found for around $100. Apple Palace (http://www.applepalace.com/apple-mac/apple-parts/mac-pro-parts/mac-pro/heatsink-kit-processor-with-bumpers-and-top-gasket-076-1233.asp) and Mac Palace (http://macpalace.com/Scripts/prodView.asp?idproduct=2782) sell them.

There are two steppings (revisions) of the Xeon E5462. Most on eBay currently seem to be the C0 stepping (SLANT) where the revision is E0 (SLBBN). I don't know what Apple used (i.e if the E0s were available before they shipped) or if it matters if they are different. You are looking at $500-$600 for a used processor Unless you find a bargain. Should be able to get a new one for $800.
IIRC, Apple used the SLANT (C0 steppings), as the E0 parts weren't out when they started production. Not sure if they did a switch later though.

Last I checked, new went for ~$830USD new. They're still a little high, but it might be an option, as it's presumably less expensive to go this route over an '09. Not so sure compared to an '08 Octad, assuming a 2.8GHz model can be located. I'm only aware of 3.2GHz versions available on Apple's refurbished site for $3299USD.

[gugucom]

From what I have read about the dual CPU Intel chipsets different steppings should not be a fundamental problem, but it can increase undocumented errata. The best would be checking the other CPU which isn't such a big deal if you plan to work on the CPUs anyway. The long handled hex key is sometimes an issue to obtain locally.

I'm wondering if someone has ever upgraded a MP3,1 with two 3,4 GHz X5492 CPUs?

[ZennZero]

I'm wondering if someone has ever upgraded a MP3,1 with two 3,4 GHz X5492 CPUs?

Curious about the same thing, but they are ungodly expensive -- $1400 each seems to be the going rate.

[nanofrog]

From what I have read about the dual CPU Intel chipsets different steppings should not be a fundamental problem, but it can increase undocumented errata. The best would be checking the other CPU which isn't such a big deal if you plan to work on the CPUs anyway. The long handled hex key is sometimes an issue to obtain locally.

I'm wondering if someone has ever upgraded a MP3,1 with two 3,4 GHz X5492 CPUs?
You'd definitely want a matched set to avoid any issues with steppings. So that would mean pulling the existing part in a Quad to get the P/N off it.

AFAIK, no one's done a transplant on the '08's, given the cost of the CPU's. But it would be possible.

[ZennZero]

AFAIK, no one's done a transplant on the '08's, given the cost of the CPU's. But it would be possible.

I've recently done a bit of research into this, and it looks like the only MP3,1 owners who *might* be able to handle the 3.4GHz Xeon would be existing 3.2Ghz owners. :(

The 2.8/3.0GHz machines had different motherboards and heatsinks, so unless you wanted to replace those parts as well (at which point you might as well buy a new machine), only the 3.2Ghz machine is equipped to handle a 150w TDP part like the X5492.

I doubt anyone could justify the cost of the X5492s for a mere 200Mhz increase in clock speed. And of course there is also the possibility that the logicboard wouldn't recognize its CPUID and refuse to boot.

[gugucom]

What is the difference in heat sink design?

[nanofrog]

I've recently done a bit of research into this, and it looks like the only MP3,1 owners who *might* be able to handle the 3.4GHz Xeon would be existing 3.2Ghz owners. :(

The 2.8/3.0GHz machines had different motherboards and heatsinks, so unless you wanted to replace those parts as well (at which point you might as well buy a new machine), only the 3.2Ghz machine is equipped to handle a 150w TDP part like the X5492.

I doubt anyone could justify the cost of the X5492s for a mere 200Mhz increase in clock speed. And of course there is also the possibility that the logicboard wouldn't recognize its CPUID and refuse to boot.
I never noticed any differing P/N's for the logic boards or coolers. It wouldn't make sense either in financial terms (from Apple's POV). Maybe different firmware was created to make changes to the VR controller settings (allows the VID to be changed), but not the hardware.

If you can give some links or more information to the details, it would help.

Thanks. :)

[J the Ninja]

I never noticed any differing P/N's for the logic boards or coolers. It wouldn't make sense either in financial terms (from Apple's POV). Maybe different firmware was created to make changes to the VR controller settings (allows the VID to be changed), but not the hardware.

If you can give some links or more information to the details, it would help.

Thanks. :)

The heatsink is definitely different, it has more heatpipes on it, I think:

http://www.welovemacs.com/0761303.html

http://www.welovemacs.com/6614458.html

[gugucom]

Jesus, 230$ for a heat sink is steep.

[Tom Sawyer]

Curious about the same thing, but they are ungodly expensive -- $1400 each seems to be the going rate.

x3 here... They SHOULD work as they are the same series, cache etc etc... mighty steep price to take one for the team.

I would LOVE to drop 2 of these babies in my Octomac, that's for sure!!!

[Tom Sawyer]

This may be a good way to acquire a pair of X5492's:

Dell Workstation (http://cgi.ebay.com/DELL-WORKSTATION-T7400-2-X-INTEL-QUAD-CORE-X5492-3-4GHZ_W0QQitemZ370287310490QQcmdZViewItemQQptZDesktop_PCs?hash=item5636d7f69a#ht_2921wt_1167)

Buy the workstation, swap your cpu's out for the 3.4's and then sack off the workstation with dual 2.8's for whatever it gets...

Tempting... But that is a surprise that the 2.8's have a different motherboard/PSU than the 3.2's? I wonder what the difference is...

[TheStrudel]

...Apple had a lapse of judgement and forgot that the Mac Pro is supposed to subsidize the pro apps and not be sold on a thin margin. As a result, the machine wound up with CPUs that cost around $850 apiece when new...Oh yeah, the other reason people mix them up is that the E5462 is hard to find loose, almost all of them wound up in Mac Pros.

They may have had a lapse of judgement there, but those of us who bought the machines were pretty satisfied with that lapse of judgement. Unfortunately, it resulted in a lot of teeth gnashing when the 09 machines dropped and we're all hoping that the '10 machines return to a similar "lapse of judgement."

[nanofrog]

The heatsink is definitely different, it has more heatpipes on it, I think:

http://www.welovemacs.com/0761303.html

http://www.welovemacs.com/6614458.html
A different cooler makes more sense than different boards though. But as gugucom indicated, $200 or $230USD (depending on which needed), OUCH!

I'll continue to look, but I'm not spotting different boards between the two models.

Thanks for the heads up, btw. :)

[ZennZero]

A different cooler makes more sense than different boards though. But as gugucom indicated, $200 or $230USD (depending on which needed), OUCH!

I'll continue to look, but I'm not spotting different boards between the two models.

Thanks for the heads up, btw. :)

The 3.2GHz mobo is p/n: 661-4676
The 2.8/3.0GHz mobo is p/n: 661-4449

As was suggested above, it may well just be a firmware or EFI change, but I read (on another forum) that there were additional heatsinks and the like on some of the components.

[nanofrog]

The 3.2GHz mobo is p/n: 661-4676
The 2.8/3.0GHz mobo is p/n: 661-4449

As was suggested above, it may well just be a firmware or EFI change, but I read (on another forum) that there were additional heatsinks and the like on some of the components.
Interesting.

The only thing that would truly make sense to me (from a HW POV), is the board used in the 2.8 and 3.0GHz models, is the Voltage Regs couldn't handle that CPU. Odd they didn't choose to use the more robust VR parts on all units, but not impossible I guess. The accounting may have made sense at the time, depending on the unit quantities manufactured.

Firmware = EFI btw, as is the SMC.

[gugucom]

Edit:
There is a jump in FSB from 1333 to 1600 in the Harpertown range. But Apple only used the E5462, E5472 and the X5482 CPUs.

Model Speed (GHz) L2 Cache (MB) FSB (MHz) TDP (W)

E5440 2.83 2x6 1333 80
X5450 3.00 2x6 1333 120
E5450 3.00 2x6 1333 80
X5460 3.16 2x6 1333 120
X5470 3.33 2x6 1333 120
E5462 2.80 2x6 1600 80
X5472 3.00 2x6 1600 120
E5472 3.00 2x6 1600 80
X5482 3.20 2x6 1600 150
X5492 3.40 2x6 1600 150

So the TDP jumped from 80 to 150 W which is significant. All MP3,1 had the 5400 Seaburg chipset with 1600 MHz FSB and 800 MHz RAM. The different p/n are definitely not based on different chipsets. Apple used no 5100 San Clemente chipsets which were limited to 1333 FSB. The different p/ns are probably related to different mechanical fixtures of the heat sinks or EFI differences in order to let the fan regulation cope with almost twice the heat.

[Tom Sawyer]

Thanks for the info guys... this is a bit disappointing as I've always had in the back of my mind the idea of bumping up to faster CPU's when they came down in price. Going to 3.0's from 2.8's is not worth it, only the 3.4's would really make a worthwhile upgrade for the cost/trouble. Guess it'll be a new machine when the time comes.

[gugucom]

I'm wondering if the MP3,1 would run with a mismatched E5462 and E5472 pairing.

[nanofrog]

Edit:
There is a jump in FSB from 1333 to 1600 in the Harpertown range. But Apple only used the E5462, E5472 and the X5482 CPUs.

Model Speed (GHz) L2 Cache (MB) FSB (MHz) TDP (W)

E5440 2.83 2x6 1333 80
X5450 3.00 2x6 1333 120
E5450 3.00 2x6 1333 80
X5460 3.16 2x6 1333 120
X5470 3.33 2x6 1333 120
E5462 2.80 2x6 1600 80
X5472 3.00 2x6 1600 120
E5472 3.00 2x6 1600 80
X5482 3.20 2x6 1600 150
X5492 3.40 2x6 1600 150

So the TDP jumped from 80 to 150 W which is significant. All MP3,1 had the 5400 Seaburg chipset with 1600 MHz FSB and 800 MHz RAM. The different p/n are definitely not based on different chipsets. Apple used no 5100 San Clemente chipsets which were limited to 1333 FSB. The different p/ns are probably related to different mechanical fixtures of the heat sinks or EFI differences in order to let the fan regulation cope with almost twice the heat.
I'd forgotten that the E5462/E6472's were 80W. I was thinking 120W for some strange reason.

Definitely explains the different cooler, and need for more robust Voltage Regs. Substantially so on the latter, as it's almost 2x the power output required.

[gugucom]

http://www.hexus.net/content/item.php?item=11938

It looks like Apple may not have made substantial technical changes to the heat sinks and the logic board in the final design. The guys at Hexus could run a pair of 150 W CPUs in a 80 W machine.

The funny thing is that beefier heat sinks with more heat pipes were reported in connection with MP3,1 systems for higher power. I wonder if the heat sinks with fewer heat pipes were prototypes and Apple used the stronger versions in the series.

I guess we need a service manual to identify different parts there. One thing we can see in the linked story is the different heat sink on the chipset compared to the MP1,1. The Seaburg chipset runs much higher RAM frequencies and apparently needed a big heat sink.

[nanofrog]

http://www.hexus.net/content/item.php?item=11938

It looks like Apple may not have made substantial technical changes to the heat sinks and the logic board in the final design. The guys at Hexus could run a pair of 150 W CPUs in a 80 W machine.

The funny thing is that beefier heat sinks with more heat pipes were reported in connection with MP3,1 systems for higher power. I wonder if the heat sinks with fewer heat pipes were prototypes and Apple used the stronger versions in the series.

I guess we need a service manual to identify different parts there. One thing we can see in the linked story is the different heat sink on the chipset compared to the MP1,1. The Seaburg chipset runs much higher RAM frequencies and apparently needed a big heat sink.
I saw that artical yesterday, and it's a bit odd, given the different P/N's used. Obviously the board's VR's can handle the 150W chips, and presumably the coolers are fine with it too, though there's no thermal data to make a comparison with (Apple equiped 3.2's vs. the modded 2.8).

Your idea of a prototype is interesting, but both cooler P/N's are available as replacement parts, so that wouldn't make much sense to me. But a mid-model switch would. The earlier units may have used the smaller cooler (and even the different board). Apple then could have decided to change coolers, and even change the VR's to "beefier" parts to handle the additional load, also causing a P/N change. I never did locate two different logic board P/N's being available as replacement parts, but I may not have found it, or the places I checked were out of stock, and removed the item from the site.

I have seen this with one of my favorite suppliers for RAID cards (ewiz.com), as their stock shifts quickly. When out for some time, they pull the item from the site, and add it back if they get a shipment in.

[ZennZero]

I never did locate two different logic board P/N's being available as replacement parts, but I may not have found it, or the places I checked were out of stock, and removed the item from the site.


Both versions of logicboard are available here:

http://www.dvwarehouse.com

The 3.2GHz board p/n: 661-4676
The 2.8/3.0GHz board is p/n: 661-4449

There are a number of other vendors selling them as well.

I actually acquired a copy of the service manual, and while I haven't looked through it in depth, I did see something about the 3.2GHz logicboards having different "coatings" on their heat-exchanging components... Since this article confirmed that voltage and EFI aren't issues, I suspect that all changes are cooling-related.

[gugucom]

Both versions of logicboard are available here:

http://www.dvwarehouse.com

The 3.2GHz board p/n: 661-4676
The 2.8/3.0GHz board is p/n: 661-4449

There are a number of other vendors selling them as well.

I actually acquired a copy of the service manual, and while I haven't looked through it in depth, I did see something about the 3.2GHz logicboards having different "coatings" on their heat-exchanging components... Since this article confirmed that voltage and EFI aren't issues, I suspect that all changes are cooling-related.


The key thing here is the desciption. One board is for 3,2 GHz only, the other is for both 3,2 and 2,8 GHz. So there is no problem to upgrade, only to downgrade. It sure looks very strange but who knows.

[nanofrog]

Both versions of logicboard are available here:

http://www.dvwarehouse.com

The 3.2GHz board p/n: 661-4676
The 2.8/3.0GHz board is p/n: 661-4449

There are a number of other vendors selling them as well.
I tried there and macpalace.com using the P/N's posted previously. Nothing showed up. :confused:

I actually acquired a copy of the service manual, and while I haven't looked through it in depth, I did see something about the 3.2GHz logicboards having different "coatings" on their heat-exchanging components... Since this article confirmed that voltage and EFI aren't issues, I suspect that all changes are cooling-related.
I hadn't gone back and looked at the .pdf I have of it. :o Too much effort. :p

But for 80W to 150W, they could have changed the VR parts (cheaper VRs on the 80W) or added cooling for the 150W boards (heatsinks and TIM to cool them - maybe that's what their terminology means :confused:). Which would be sufficient, so long as the parts used could actually sustain the load. Otherwise I think of a spray on finish of some sort (aka conformal coating), which isn't capable of dealing with that much of a thermal difference. TIM, no matter what it is in makeup, isn't the most efficient thermal conductor (it's still a suspension), but it is better than bare metal to the part surface. It still needs to be used with a cooler of some kind.