Dry run for W3670 install.

[nanofrog]

It helps. Unfortunately, it was what i was afraid of. Too tight and "crack". Too loose and burnout.

I wonder if I use one of my mike's and measure the difference, I can put washers between the screws and the board. I have a device for making washers from 0.001" to 0.020" thick in varying sizes.

You don't really need to do this. It's more common sense than anything else. When the screw stops turnning, give it another 1/4, maybe 1/2 turn, and that's it. Quit turning. Don't keep applying torque to the screwdriver, or you will run the risk of breaking the semiconductor, and this is the same with the IHS versions as well.

An IHS unit has metal on the top, but underneath that, it's the same as a naked unit (think of it like a thin sheet of glass).

 

[melgross]

You don't really need to do this. It's more common sense than anything else. When the screw stops turnning, give it another 1/4, maybe 1/2 turn, and that's it. Quit turning. Don't keep applying torque to the screwdriver, or you will run the risk of breaking the semiconductor, and this is the same with the IHS versions as well.

An IHS unit has metal on the top, but underneath that, it's the same as a naked unit (think of it like a thin sheet of glass).

Yes, I understand. I do know how to tighten screws. When I designed electronics, and there were parts to be tightened like this, I always made certain to provide a "stop". That way, delicate parts wouldn't get damaged in assembly, or repair.

It just seems to make sense to have something set-up so that it's tightened exactly right, rather than going by feel. If I were replacing chips with the same thickness, there wouldn't be a problem, but it's different. As I said, even Anand damaged the system, and he's pretty good at this.

Have you done a 2009 model?

 

[nanofrog]

Yes, I understand. I do know how to tighten screws. When I designed electronics, and there were parts to be tightened like this, I always made certain to provide a "stop". That way, delicate parts wouldn't get damaged in assembly, or repair.

It just seems to make sense to have something set-up so that it's tightened exactly right, rather than going by feel. If I were replacing chips with the same thickness, there wouldn't be a problem, but it's different. As I said, even Anand damaged the system, and he's pretty good at this.

Have you done a 2009 model?

I didn't have any idea as to how much experience/knowledge you have, so I took the simple approach. As it happens, I'm a hardware designer myself, and also believe in stops (usually shoulders cut into the stand-offs for a precise depth).

I've not done a 2009/10, but if you've the tools (most people don't keep precision measurement instruments in their tool boxes ), you should be able to take some measurements if you want to make a stop based on the actual CPU used.

But as the common sense/cautious approach will work, it's up to you if you want to put in the time for the additional effort that will be involved.

 

[philipma1957]

I have thought about how this system cpu done by Anand got damaged. I first mentioned the lack of a top cover as he did and I suggested washers. while it is still a good idea. also believe that he may not have tightened each screw a half turn at a time.

maybe he did one screw fully this would certainly cause the cpu to slide off center since the dual cpu has no retention clip and no lid.

 

[melgross]

I have thought about how this system cpu done by Anand got damaged. I first mentioned the lack of a top cover as he did and I suggested washers. while it is still a good idea. also believe that he may not have tightened each screw a half turn at a time.

maybe he did one screw fully this would certainly cause the cpu to slide off center since the dual cpu has no retention clip and no lid.

I don't remember if he did the round robin approach, which is always the way I first tighten multiple screw sets (then it's opposite screws). But he did say that he tightened them by the same amount as the original screws. Of course, the newer chips need to be tightened to the same torque, not all the way down, which is what I suspect he did.

The problem here (and this is some pretty expensive equipment) is that a torque screwdriver doesn't measure the same torque going out as it does going in, so that's not an answer. It would be wonderful if someone had access to a service manual to find out if Apple does torque these screws, and if so, to what, or if they're tightened all the way down. the all the way down method is the way they would normally do this on an assembly line if using a torque driver wasn't being used.

But those hanging electric screwdrivers are almost always torque drivers set to a specific limit. So i imagine that's how they do it, and so they must have some way to spec this for repair.

I was going to give the guys at OWC a call about this as I have good relations there, but decided that it may be presumptuous.

 

[nanofrog]

I don't remember if he did the round robin approach, which is always the way I first tighten multiple screw sets (then it's opposite screws). But he did say that he tightened them by the same amount as the original screws. Of course, the newer chips need to be tightened to the same torque, not all the way down, which is what I suspect he did.

I'm not sure, as he had to disconnect the fan connector from the cooler. So he may have done something different in order to keep the now loose connector aligned with the fan socket on the daughterboard (i.e. initially tighten one side more than is normally performed as a means of holding the connector). But the uneven tightening process damaged the socket in the process (CPU drifted during the process - don't recall the CPU's actually being damaged in Anand's case, just the socket).

The problem here (and this is some pretty expensive equipment) is that a torque screwdriver doesn't measure the same torque going out as it does going in, so that's not an answer.

If you mean as a way to measure the torque of the screw, it's possible if you know the difference the tool uses (Wiha's are 50% more than the indicated setting in reverse for example). Still a PITA though, and can add additional error beyond the tool's rating of around 5 - 6% (i.e. actual value set at say 6.25 in-lbs, and it removes at 4.25 in reverse, so you calculate it to 6.375).

They're still handy tools though, and I've found them for between ~$115 - $220USD (new & adjustable types; Wiha, Proto examples).

I ended up finding a New Old Stock unit off of eBay a few years ago at a good price, so that may be an option if you're interested in getting one for personal use, and find the right deal.

It would be wonderful if someone had access to a service manual to find out if Apple does torque these screws, and if so, to what, or if they're tightened all the way down. the all the way down method is the way they would normally do this on an assembly line if using a torque driver wasn't being used.

Given there's apparently no stops, they'd have to use a torque driver to keep the waste/rework rates in check I should think. And given the additional speed and reduction of repetitive strain injuries, I'd guess electric models as the most likely type.

As per a torque value, I don't know if Apple or Foxconn deviated from Intel's spec, but Intel lists it as 8 in-lbs (found it in the very tail end of the 5500 series volume 1 datasheet).

 

[redarc]

about mac pro and cpu replace!!

ehy i see this guide so useful and i would try to do something with my mid 2010 mac pro westemere , what i ask is you know if everything u do work on x 2 xeon motherboard too? or there is something different? and what about this? i got x 2 2,4 ghz.. so low clock speed for most of software that use only 1-3 cores... so its possible to replace it with x 2 intel W3565 , u can find this on top single processor quad core 2010!

is possible?

thank you and sorry for disturb!

 

[nanofrog]

ehy i see this guide so useful and i would try to do something with my mid 2010 mac pro westemere , what i ask is you know if everything u do work on x 2 xeon motherboard too? or there is something different? and what about this? i got x 2 2,4 ghz.. so low clock speed for most of software that use only 1-3 cores... so its possible to replace it with x 2 intel W3565 , u can find this on top single processor quad core 2010!

is possible?

thank you and sorry for disturb!

Unfortunately, No.

The SP variants (35xx or 36xx) parts won't work in a DP system (has to do with the 2nd QPI channel on the 5520 chipset cannot be shut off). You'd have to swap out the daughterboard/CPU tray for an SP version to use the W3565, and then you'd only be able to run one of them anyway.

So you'll need to use either 55xx or 56xx series Xeons with your existing CPU tray (there is no consumer equivalent that could be substituted either).

 

[redarc]

thank you

Unfortunately, No.

The SP variants (35xx or 36xx) parts won't work in a DP system (has to do with the 2nd QPI channel on the 5520 chipset cannot be shut off). You'd have to swap out the daughterboard/CPU tray for an SP version to use the W3565, and then you'd only be able to run one of them anyway.

So you'll need to use either 55xx or 56xx series Xeons with your existing CPU tray (there is no consumer equivalent that could be substituted either).

really sure of this? it really a ****.. also no chance with hack , microcode nothing ? no chance?

also i can see in this site http://eshop.macsales.com/shop/turnkey/MacPro/2009_2010_Xeon_Processor/Apple_Mac_Pro_2010_2

that is possible put inside an xeon x3***? thats wrong please explain , i'm not sure to understand how can be this possible , its a fail.. my mac suck in most of software couse they use just 1-3 cores.. and if i want to replace cpu i have to spend over 2k $ what is this come on...

and still from single w*** series mac pro i can see here http://eshop.macsales.com/shop/turnkey/MacPro/2009_2010_Xeon_Processor/Apple_Mac_Pro_2010_1
they can replace it with xeon , so xeon can t use w*** series and w*** series can use xeon? Or as you say its just about link and multiprocessing? i can t believe.. so pissed :S

someone else can confirm?

 

[ActionableMango]

really sure of this? it really a ****.. also no chance with hack , microcode nothing ? no chance?

Nothing you are saying makes any sense to me, but I will confirm that you cannot use a single processor Xeon in a 2010 dual processor Mac Pro for reasons that have already been told to you.

 

[nanofrog]

really sure of this? it really a ****.. also no chance with hack , microcode nothing ? no chance?

It's hard to understand your post, but in the case of a Single Processor CPU (SP for short) in a Dual Processor board (DP for short), No. It's not possible at all.

The reason has to do with the chipsets, which are not the same between SP and DP systems (very similar, but not identical). Specifically, the X58 has a single QPI channel (chipset for the SP machines), while the 5520 (not to be confused with the E5520 CPU), has 2 QPI channels. Unfortunately, the 5520 chipset is unable to turn off an unused QPI channel, which is why the SP CPU's will not work in DP boards (hardware limitation in the chipset, not firmware).

The DP CPU's on the other hand, can turn off an unused QPI channel, so they can be used in a Single Processor board & X58 chipset (DP CPU's have 2x QPI channels; one to communicate with the other CPU, and the second to communicate with the 5520 chipset). In an SP board however, one is used to communicate with the X58, and the other is Disabled (turned OFF, as it has the additional circuitry needed to do so). But it's not exactly cost effective, as DP versions (same clock and core count) are more expensive.

that is possible put inside an xeon x3***? thats wrong please explain , i'm not sure to understand how can be this possible , its a fail.. my mac suck in most of software course they use just 1-3 cores.. and if i want to replace cpu i have to spend over 2k $ what is this come on...

and still from single w*** series mac pro i can see here http://eshop.macsales.com/shop/turnkey/MacPro/2009_2010_Xeon_Processor/Apple_Mac_Pro_2010_1
they can replace it with xeon , so xeon can t use w*** series and w*** series can use xeon? Or as you say its just about link and multiprocessing? i can t believe.. so pissed :S

someone else can confirm?

If you look closely, the first link are upgrades for DP systems ONLY. The second link are the available CPU upgrades they offer on the SP systems.

There's no "magic" here at all. They use 55xx and 56xx parts as upgrades in the DP models, while they offer 35xx and 36xx parts as CPU upgrades for the SP systems. Which are due to the 5520 chipset limitations I've tried to explain.

 

[redarc]

ok..

It's hard to understand your post, but in the case of a Single Processor CPU (SP for short) in a Dual Processor board (DP for short), No. It's not possible at all.

The reason has to do with the chipsets, which are not the same between SP and DP systems (very similar, but not identical). Specifically, the X58 has a single QPI channel (chipset for the SP machines), while the 5520 (not to be confused with the E5520 CPU), has 2 QPI channels. Unfortunately, the 5520 chipset is unable to turn off an unused QPI channel, which is why the SP CPU's will not work in DP boards (hardware limitation in the chipset, not firmware).

The DP CPU's on the other hand, can turn off an unused QPI channel, so they can be used in a Single Processor board & X58 chipset (DP CPU's have 2x QPI channels; one to communicate with the other CPU, and the second to communicate with the 5520 chipset). In an SP board however, one is used to communicate with the X58, and the other is Disabled (turned OFF, as it has the additional circuitry needed to do so). But it's not exactly cost effective, as DP versions (same clock and core count) are more expensive.


If you look closely, the first link are upgrades for DP systems ONLY. The second link are the available CPU upgrades they offer on the SP systems.

There's no "magic" here at all. They use 55xx and 56xx parts as upgrades in the DP models, while they offer 35xx and 36xx parts as CPU upgrades for the SP systems. Which are due to the 5520 chipset limitations I've tried to explain.

i have understood , thank you for every info , you have been so nice to explain me everything , so its hardware and about this damm QPI channel , thank you for the help and sorry for stupid thing i sayd.

 

[nanofrog]

i have understood , thank you for every info , you have been so nice to explain me everything , so its hardware and about this damm QPI channel , thank you for the help and sorry for stupid thing i said.

NP.

It just seemed like there was still a little confusion (understood what CPU series went with what board, but not why).

 

[redarc]

!

NP.

It just seemed like there was still a little confusion (understood what CPU series went with what board, but not why).

yes , thank you for all info.

 

[redarc]

mate!

just 1 more question , sorry for disturb nanofrog , so if i buy x 2 Intel® Xeon® Processor X5570 i can use them ? they are 1 year old but at least they're clock speed its 500 mhz up..its not an problem the different tdp? the 2,4 is on 80 watt and x5570 is 95 watt :S? sorry for nebbish question, thanks for patience.


also this? that i can t find in you're 5500 series list , the xeon 5590 3,33 ghz nehalem?
http://eshop.macsales.com/shop/turnkey/MacPro/2009_2010_Xeon_Processor/Apple_Mac_Pro_2009_2

also the W5580 130 watt?

 

[nanofrog]

just 1 more question , sorry for disturb nanofrog , so if i buy x 2 Intel® Xeon® Processor X5570 i can use them ? they are 1 year old but at least they're clock speed its 500 mhz up..its not an problem the different tdp? the 2,4 is on 80 watt and x5570 is 95 watt :S? sorry for nebbish question, thanks for patience.


also this? that i can t find in you're 5500 series list , the xeon 5590 3,33 ghz nehalem?
http://eshop.macsales.com/shop/turnkey/MacPro/2009_2010_Xeon_Processor/Apple_Mac_Pro_2009_2

also the W5580 130 watt?

Apple selected CPU's up to 95W TDP, but there has been another member that managed to get a pair of W5590's working in a 2009. So it seems the heatsinks were able to handle 130W TDP parts. The OWC link confirms they trust the heatsinks with 130W TDP parts as well.

BTW, performing the swap is easier in a 2010, as there are latching mechanisms on the slots in that model (meant for CPU's with the metal cover, aka Integrated Heat Sink, or IHS for short).

 

[redarc]

lastttt

very last question nanofrog.. or to all community , i was reading around here about to buy just 1 processor and put 2nd one inside when i have the money ... can t find any good news about the problem of fan , or about " what happen if you leave other socket with no cpu inside it" , put heatsink on it or not? And more

http://cgi.ebay.it/W5580-AT80602000...307?pt=CPUs&hash=item19c6dd9b03#ht_1013wt_907

what about this engineer sample? is that risky? i can read around that many peeps would prefer it for multiplier unlocked.. they cost just 200 euro , so its so interesting? you can give me any good tip? i would also love an you're personal tip .

thanks again! sorry if i can seems repetitive !

 

[nanofrog]

very last question nanofrog.. or to all community , i was reading around here about to buy just 1 processor and put 2nd one inside when i have the money ... can t find any good news about the problem of fan , or about " what happen if you leave other socket with no cpu inside it" , put heatsink on it or not? And more

Yes, it's possible to purchase a single DP CPU and use it in a DP board (adding the second one later). In such a case, you leave the second CPU socket empty (no CPU, no heatsink).

As per what the fans will do, I'm not sure. Generally speaking, they don't rev up, but Apple might be different as they sell the systems with a pair of CPU's (don't care what it does with a single CPU long term, as that wasn't what they planned for). So I'll leave this to others who may have tried (seem to recall they do spin up with only a single CPU, but not 100% certain).

what about this engineer sample? is that risky? i can read around that many peeps would prefer it for multiplier unlocked.. they cost just 200 euro , so its so interesting? you can give me any good tip? i would also love an you're personal tip .

Though Engineering Samples have gotten better than they were (fewer bugs), they're not necessarily the final product. Which means they can be a bit risky (hardware bugs).

So I tend to be a bit old school here, and would recommend staying away from them if you don't know the exact issues and how they'd affect you (research out the P/N, and see if there's any information).

In their favor (link), they are B0 stepped, so I would expect they're close to the final release (also B0 stepped).

 

[MrAverigeUser]

Thank you very much for your excellent help by your expertise!!