3.33GHz Xeon (Server CPU) Transplant in a 2009 Quad

[AppleWorking]

By the way, just adding a drop in the middle and hoping to compress it out evenly via the heat-sink is wrong. You do need to use a piece or cardboard or something and smooth it out as evenly as possible. This is fairly important! So this:

almost all techs will agree, is wrong!



.

I guess you would know better... Make sure you use your method on your next CPU upgrade!





For the rest, I might be wrong, but I still feel...


The pea size method is easiest and produces the same result as the line method. Both, however, are good and either will produce good results. Just know, spreading it all over is a bad idea, not only because spreading it is time consuming and messy, but simply because it doesn't work as well. I have attached a chart comparing the two ways - spread versus dollop/line. And also a quote directly from Arctic Silver Incorporated.

Here is a chart comparing the effectiveness of spread versus dollop/line:

.

 

[nanofrog]

The pea size method is easiest and produces the same result as the line method. Both, however, are good and either will produce good results.

Yours is rather centered.

I'm a fan of "works well with little effort" school of thinking.

I've even gone thought the effort of spreading it with a plastic bag over the end of a finger and credit card methods. Not any discernable difference when it was done correctly. If incorrect, it's worse.

So the "centered pea dollop" seems to be the best of the bunch IMO.

 

[AppleWorking]

If anyone is interested, just saw this over @ apple...

"Our company acquired three Quads and five Octos. I tried to install a Core i7 975 in all three Quads and I also receive kernel panics, system hangs, etc. on wake. Two of the MPs are 2.66GHz Quads, and one is the 2.93GHz Quad. All three machines have clean install, leopard/snow leopard, SMC/PRAM reset, nothing loaded, no external devices. We are going to try the Xeon CPU next, someone has reported success on MR."​

 

[AppleWorking]

Yours is rather centered.

I'm a fan of "works well with little effort" school of thinking.

I've even gone thought the effort of spreading it with a plastic bag over the end of a finger and credit card methods. Not any discernable difference when it was done correctly. If incorrect, it's worse.

So the "centered pea dollop" seems to be the best of the bunch IMO.

IMO ditto

 

[AZREOSpecialist]

I had no wake from sleep issues with Leopard, only after upgrading to SL did I start having wake from sleep problems. I've seen a lot of posts on the Apple support forums about Mac Pro wake from sleep issues with SL, so it's not confined to those with Core i7 975 CPUs. Who the hell knows.

 

[nanofrog]

IMO ditto

At least if you go multiple methods, it can confirm if one of the cores is hotter than the others.

I actually did this recently on my system. It confirmed Core 1 is a few degrees warmer than the others (i7-920 btw).

 

[AppleWorking]

Who the hell knows.

Certainly not me, I don't have an i7... I believe you, though... Did you try sleep/wake in L and not SL? Did you try sleep/wake with the 2.66 before the swap? I wonder why swapping the CPU back solved it for Msbeezy. I also wonder if he changed anything software or hardware wise before swapping the stock CPU back in there?

At least if you go multiple methods, it can confirm if one of the cores is hotter than the others.

I actually did this recently on my system. It confirmed Core 1 is a few degrees warmer than the others (i7-920 btw).

I never thought of that.

 

[Msbeezy]

I had no wake from sleep issues with Leopard, only after upgrading to SL did I start having wake from sleep problems. I've seen a lot of posts on the Apple support forums about Mac Pro wake from sleep issues with SL, so it's not confined to those with Core i7 975 CPUs. Who the hell knows.

I pulled my 3.33 yesterday after wake issues and imo lack of performance versus financial consideration and it once again wakes fine..returned the processor this afternoon for full refund.

 

[AppleWorking]

I pulled my 3.33 yesterday after wake issues and imo lack of performance versus financial consideration and it once again wakes fine..returned the processor this afternoon for full refund.

Congrats on getting your money back! That's always nice to hear. But I have a quick question, do you remember changing anything at all (software or hardware) before swapping the stock CPU back in there?

Also, was your original CPU the 2.66 or the 2.93?

 

[AZREOSpecialist]

I pulled my 3.33 yesterday after wake issues and imo lack of performance versus financial consideration and it once again wakes fine..returned the processor this afternoon for full refund.

I'm glad you were able to get a refund. I am wondering about the lack of performance you've mentioned. Did you run the latest GeekBench on your 975? What numbers did you get? How did you quantify this lack of performance?

The numbers I'm getting are 15-20% higher than 2.93 GHz numbers posted online, and my CPU clock runs 14% higher at 3.33 GHz. I've read that on Nehalem, when you increase clock speed you actually get back more than what you'd expect from just a linear scaling of clock speed. What I see with my numbers seems to support this, but obviously there are many variables involved. My Cinebench scores are pretty much comparable to those of AppleWorking, who is using a Nehalem Xeon @ 3.33 GHz. Your 975 should have been able to deliver comparable results.

What type of memory do you have installed in your system? Do you have a RAID card? Were any USB devices attached at the time you had wake from sleep issues? Have you checked your system for corrupted fonts and caches?

 

[Msbeezy]

Congrats on getting your money back! That's always nice to hear. But I have a quick question, do you remember changing anything at all (software or hardware) before swapping the stock CPU back in there?

Also, was your original CPU the 2.66 or the 2.93?

Well before I removed the cpu as it was not waking I removed one of the only two programs I had installed following the 3.33 install..they were a fan utility and geekbench for 64 test.The removal of the fan utility did nothing to remedy the issue. As I stated the wake issue resolved with the re-install of the 2.93
My last stab with the 3.3..at geekbench 64..netted an 11415. My score right now with the stock 2.93 is 10204......cost for the cpu retail ver with fan was about $1,400 cdn...not worth it in my books

 

[handheldgames]

Just upgraded.. Took a ton of photos... I've uploaded the most important one 1st. After the install, as a precaution, I reset the SMC.

Happy to report sleep mode works..

 

[Msbeezy]

I'm glad you were able to get a refund. I am wondering about the lack of performance you've mentioned. Did you run the latest GeekBench on your 975? What numbers did you get? How did you quantify this lack of performance?

See my post below.......

The numbers I'm getting are 15-20% higher than 2.93 GHz numbers posted online, and my CPU clock runs 14% higher at 3.33 GHz. I've read that on Nehalem, when you increase clock speed you actually get back more than what you'd expect from just a linear scaling of clock speed. What I see with my numbers seems to support this, but obviously there are many variables involved. My Cinebench scores are pretty much comparable to those of AppleWorking, who is using a Nehalem Xeon @ 3.33 GHz. Your 975 should have been able to deliver comparable results.

I purchased the 64 ver and ran it and posted them but don't recall offhand..not that impressive if I recall

What type of memory do you have installed in your system? Do you have a RAID card? Were any USB devices attached at the time you had wake from sleep issues? Have you checked your system for corrupted fonts and caches?

I filled the ram with 16g of OWC ram.2 intel x25m in soft raid0. Yes there were usb connected but never changed usb config with old stock cpu or new and now with the stock back in a I mentioned the system recovers fine again..
The way I see it at $1,500 for the cpu upgrade I can add another 1K or so and get an 8 core 2.93 refurb from Apple Canada now listed..that will smoke this one..sorta

 

[handheldgames]

I filled the ram with 16g of OWC ram.2 intel x25m in soft raid0. Yes there were usb connected but never changed usb config with old stock cpu or new and now with the stock back in a I mentioned the system recovers fine again..
The way I see it at $1,500 for the cpu upgrade I can add another 1K or so and get an 8 core 2.93 refurb from Apple Canada now listed..that will smoke this one..sorta

$1500?? OUCH! My CPU i7 975 was purchased for $770 on Ebay and is guaranteed to the specs of the item through PayPal. Also had a 60 day no questions asked return policy. Cant beat that.

 

[Msbeezy]

couple pics

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

[/URL][/IMG]

[/IMG]

[/IMG]

 

[Msbeezy]

$1500?? OUCH! My CPU i7 975 was purchased for $770 on Ebay and is guaranteed to the specs of the item through PayPal. Also had a 60 day no questions asked return policy. Cant beat that.

That's a good buy but $770 is near $1k to Canada and remember this is an retail kit..I couldn't get the unit just by itself..

 

[Tesselator]

I guess you would know better... Make sure you use your method on your next CPU upgrade!

Yes, of course I did. And 80% or 90% of all home-made PCs have done so as well Shop builds too. Macs from Apple also uses a machine applied spread and not a pea!

I dunno about that graph. I guess it was tested with either too much or not enough paste.

Anyway, it's no big deal. If they put the pea-ball in the middle the worst that will happen is they'll have to do a lot of cleaning when they change the CPUs next or maybe the system will run a few degrees hotter. Same with spreading I guess. If it's not on properly the same clean-up or temperature difference applies. So, whatever...



.

 

[handheldgames]

I'll consider this mission Accomplished.....

 

[AZREOSpecialist]

I pulled my 3.33 yesterday after wake issues and imo lack of performance versus financial consideration and it once again wakes fine..returned the processor this afternoon for full refund.

So you're not claiming that you weren't getting better performance, just that the performance gains were not worth the investment. Were you expecting more than the 14% speed increase through clock speed alone? Also, an 8-core will only "smoke" a 4-core if you are utilizing software that makes use of the additional cores (a/v encoding, 3-D rendering, etc.). But for all other purposes, such as web browsing, CS4, Office, web authoring, print/publishing, etc., you won't see an improvement. Once applications are optimized for Snow Leopard this situation may change, but by then we'll have a whole new slew of Mac Pro offerings that will be even faster than the current generation.

By the way, my Mac Pro w/ Core i7 975 was happily sleeping since 6pm last night and it woke up as expected this morning at 9am. No issues at all. I also removed SMC Fan Control since I had sleep issues, so that could be the culprit for both of us. Whatever is the cause for sleep issues, it's not the Core i7 975.

 

[Tesselator]

Yeah, CPUs are a lot like horses or cars or something. If you ride or drive at 5 Mph then it doesn't matter if you have a race-horse/car. At 5Mph a old nearly dead pinto goes just the same. Step on the gas or ask it run fast tho and there's a very big difference!

 

[Msbeezy]

So you're not claiming that you weren't getting better performance, just that the performance gains were not worth the investment. Were you expecting more than the 14% speed increase through clock speed alone? Also, an 8-core will only "smoke" a 4-core if you are utilizing software that makes use of the additional cores (a/v encoding, 3-D rendering, etc.). But for all other purposes, such as web browsing, CS4, Office, web authoring, print/publishing, etc., you won't see an improvement. Once applications are optimized for Snow Leopard this situation may change, but by then we'll have a whole new slew of Mac Pro offerings that will be even faster than the current generation.

By the way, my Mac Pro w/ Core i7 975 was happily sleeping since 6pm last night and it woke up as expected this morning at 9am. No issues at all. I also removed SMC Fan Control since I had sleep issues, so that could be the culprit for both of us. Whatever is the cause for sleep issues, it's not the Core i7 975.

Yes I pretty much anticipated somewhat better performance gain..had the investment totalled like $800 I would have appreciated the gain for $.
I realize the performance improvement in the *core to be relative to the task at hand and how it's structured so I'm still on the fence and thinking I'll land staying with my current machine and adding more ssd,maybe a 30 cinema until the next gen cpus arrive and appropriate software..
cheers

 

[AppleWorking]

New post over @ apple.

"I just wanted to report that we have not had any sleep issues in all five machines since returning them to the stock CPUS. I contacted another company that tried this, he started seeing sleep issues after a couple weeks. Says tried everything humanly possible but only fix was going back to the stock CPU. It looks like the Xeon is the only stable alternative. Some on MR report success with the 975, personally I don't believe um, too quick to defend their investment. Will try Xeons next... If you don't hear back from me just assume success, will try and post, but very busy."​

Does anyone think the VID Voltage Range has something to do with it? The Xeon has a different range from the i7.

Xeon = 0.800V-1.225V

i7 = 0.800V-1.375V

Could this explain why some with the i7 have success and others don't?

BTW Xeon is running perfectly, just wish we could help some of those with the i7s.

Edit: I just thought of this... What if no ECC support on the Core i7 has something to do with it? If OSX is checking for ECC functionality, wouldn't it explain why some have the problem and some don't? Seeing that we are using different memory from different sources?

Who knows, at this point I would say get an Xeon, it's more of a sure fire, no hassle, upgrade. Not worth the headache just to save a few bucks. It just works.

 

[nanofrog]

New post over @ apple.

"I just wanted to report that we have not had any sleep issues in all five machines since returning them to the stock CPUS. I contacted another company that tried this, he started seeing sleep issues after a couple weeks. Says tried everything humanly possible but only fix was going back to the stock CPU. It looks like the Xeon is the only stable alternative. Some on MR report success with the 975, personally I don't believe um, too quick to defend their investment. Will try Xeons next... If you don't hear back from me just assume success, will try and post, but very busy."​

Does anyone think the VID Voltage Range has something to do with it? The Xeon has a different range from the i7.

Xeon = 0.800V-1.225V

i7 = 0.800V-1.375V

Could this explain why some with the i7 have success and others don't?

BTW Xeon is running perfectly, just wish we could help some of those with the i7s.

Edit: I just thought of this... What if no ECC support on the Core i7 has something to do with it? If OSX is checking for ECC functionality, wouldn't it explain why some have the problem and some don't? Seeing that we are using different memory from different sources?

Who knows, at this point I would say get an Xeon, it's more of a sure fire, no hassle, upgrade. Not worth the headache just to save a few bucks. It just works.

Voltage is by far the more likely culprit compared to the lack of ECC.

 

[AppleWorking]

Voltage is by far the more likely culprit compared to the lack of ECC.

I knew you would be able to give us your insight on this. Thanks.

 

[nanofrog]

I knew you would be able to give us your insight on this. Thanks.

NP.

Slightly off topic, but is still applicable as a CPU upgrade for '09 MP's:
Wiki claims that the Gulftowns will have a VID of 1.055V, but the Process Eval chips Intel's run aren't quite there (1.10V). It's close, and they should work on the boards, as I doubt Intel veered off their Design Guidelines to make the board in the '09's (0.8 - 1.350V). What we don't know, is if the Gulftown parts will work as a Drop-In replacement without a firmware change (updated microcode for the CPU ID's).