MP 1,1-5,1 Hex 3.2 update from 2010 quad 2.8 photos new! CPU to use!

[tpluth]

I believe the Mac Pro power supply is rated for around 1000W, maximum, but remember, its 2 to 5 years old at this point. Transformers and capacitors degrade with time.

Your GTX 680 draws up to 195W. Your processors draw up to 130W each. That's a total of 455W before you add in your hard drive, optical drive, chipset, fans, etc. I'm betting at full load you could hit close to 600 watts. You'll know if your PS fans start really speeding up.

By contrast, I'm hitting 480W including the displays (probably about 150 watts for both), so more like 330W for the box itself, if that tells you anything. It's not that it won't handle it, but extended use like that will reduce it's lifetime. When I'm building PC's I usually try to size the power supply to only run at 40% under load. That keeps things cool and quiet.

Get yourself a Kill-A-Watt and check what the idle and loaded power draw is.

http://www.amazon.com/P3-P4400-Elec...F8&qid=1417375792&sr=8-1&keywords=kill-a-watt

 

[Upgrader]

Hi Mike,

I’ve just done the same as you and successfully upgraded an 8 core 2.26ghz 2009 4,1 to a 12 core 5,1 with two x5690’s. It was a tense experience to say the least but it worked first time.

I’m only getting 30916 for the multi core score though. How are you getting 32000+ ? Are you somehow overlocking? From those Geekbench stats my MacPro is exactly the same as yours in terms of motherboard and components etc except I have 32gb RAM to your 24gb.

Cheers.

Thanks to this thread for the instructions on doing this. I recently did this upgrade on my 2010 MacPro, adding two x5690 processors and it flies!

Geekbench score attached.

 

[tpluth]

Hi Mike,

I’ve just done the same as you and successfully upgraded a 2.26ghz 12 core 2009 4,1 to a 5,1 with two x5690’s. It was a tense experience to say the least but it worked first time.

I’m only getting 30916 for the multi core score though. How are you getting 32000+ ? Are you somehow overlocking? From those Geekbench stats my MacPro is exactly the same as yours in terms of motherboard and components etc except I have 32gb RAM to your 24gb.

Cheers.

What kind of RAM and how many sticks, total?

With the x5690 you should be able to run 1333Mhz PC3-10600 ECC.

The chipset is set up for triple channel RAM. Filling all 4 slots for each CPU has a detrimental impact on that, though, or so I've heard. I'm running 6 sticks of PC3-10600R (Registered, ECC) and I get 29000 GB3 x64 on dual X5675's.

Cooling is also important. If the heatsinks don't have good contact with the CPU or the TIM isn't right (too much, too little), that could cause CPU throttling.

 

[Upgrader]

What kind of RAM and how many sticks, total?

With the x5690 you should be able to run 1333Mhz PC3-10600 ECC.

The chipset is set up for triple channel RAM. Filling all 4 slots for each CPU has a detrimental impact on that, though, or so I've heard. I'm running 6 sticks of PC3-10600R (Registered, ECC) and I get 29000 GB3 x64 on dual X5675's.

Cooling is also important. If the heatsinks don't have good contact with the CPU or the TIM isn't right (too much, too little), that could cause CPU throttling.

Hey,

Yeah, the 32gb RAM I have is actually 1600mhz DDR3 PC3-12800 ECC. I have 4x8gb sticks installed in the correct slots (1,2 and 5,6) for each proc. You’re right about filling all RAM slots having a detrimental impact - 3 is the sweet spot.

Re cooling, I’ve added extra thermal pads across the voltage regulators to fill the gap created by the new lidded CPU. Contact should be fine as I used 2mm pads and cut the strips to size. Actually, Mike, did you add extra thermal pads when you upgraded?

----------

...also, CPU_A is running roughly 10 degrees F hotter than CPU_B whilst just surfing a few web pages and generally idling. Is that normal?

 

[h9826790]

I’m only getting 30916 for the multi core score though. How are you getting 32000+ ? Are you somehow overlocking? From those Geekbench stats my MacPro is exactly the same as yours in terms of motherboard and components etc except I have 32gb RAM to your 24gb.

1) Use the 64bit version (paid version) will have better score.

2) 32G RAM may be the problem, no matter it's 4x8G or 8x4G, both are not the optimum config for the cMP. Use only 6 stick and make sure they run at 1333MHz should has better result.

 

[Upgrader]

1) Use the 64bit version (paid version) will have better score.

2) 32G RAM may be the problem, no matter it's 4x8G or 8x4G, both are not the optimum config for the cMP. Use only 6 stick and make sure they run at 1333MHz should has better result.

Thanks,

Yeah, using the paid 64bit version to get my score.
If I get another 8gb (2x4gb sticks) taking me to 40gb you think that will help the score difference?

 

[Upgrader]

...ok, I’m running a 30 minute Geekbench stress test. These are my temperatures after 19 minutes. Are they ok?

 

[h9826790]

Thanks,

Yeah, using the paid 64bit version to get my score.
If I get another 8gb (2x4gb sticks) taking me to 40gb you think that will help the score difference?

Technically it will give you more bandwidth. However, use different size, spec, manufacture's... RAM may cause unexpected outcome occasionally.

Anyway, IMO, unless your whole target is getting max score in this particular benchmark software. There is no need to get extra RAM, the real world performance difference usually very very small for a normal application.

----------

...ok, I’m running a 30 minute Geekbench stress test. These are my temperatures after 19 minutes. Are they ok?

The temperature and fan speed looks very normal to me.

 

[Upgrader]

The temperature and fan speed looks very normal to me.

Even the fact that two are marked in red?

 

[h9826790]

Even the fact that two are marked in red?

Yes, The IOH heatsink temperature is not that important, the real important one is the temperature of the IOH Tdiode. I think you prefer to have a hot heatsink, but a cool north bridge, which means the heatsink is working hard to cool down the system. Rather than a cool heatsink, but a hot north bridge. 79C is normal with that ambient temperature. It may go above 85C with the native fan control.

Also, Apple tends to make the cMP as silent as possible. It basically won't speed up your fan until the CPU temperature get close to it's max temperature. In your case, the fan speed may going up and down a bit, and keep the CPU temperature at around 85C under full load.

My fan speed and CPU temperature under stress almost identical to your's. I've already upgrade my CPU for more then 6 months, running it 24/7, sometimes let handbrake do it's job 7 days non stop. I did monitor my CPU's performance, never experience any thermal throttling, no hang, no auto restart, no thermal warning in terminal.

However, technically, my W3690's TCASE max is 67.9C, about 10 degrees lower than your X5690 (78.5C). So, if my CPU works fine with those temperature, your CPUs should have no problem at all, plus with some extra buffer.

P.S. The red number in iStat Pro not necessary means danger. If you idle you Mac, and spin up all the fans (e.g. Use the Medium fan setting in iStat Pro). Your CPU's temperature will drops a lot. And the "CPUx core 0 relative to ProcHot" will increase and then turn red at some stage. However, for that particular temperature, the higher the better, it should turn red when it's very low, but not very high.

 

[Upgrader]

ok, thanks, that’s good to know.

I’m concerned because the heatsink screws on CPU_A were much tougher to tighten to the required number of turns than the heatsink screws on CPU_B. I’m concerned that CPU_A heatsink is too tight and getting too hot because of it. I thought I might need to loosen each screw by a quarter turn.

I’m just running the fans on high as you suggest and it’s doing what you said it would. I can see that the red temps in iStat aren’t something to worry about.

 

[Upgrader]

As a side note, I’m running the Terminal command: pmset -g thermlog whilst running Geekbench. The CPU_Speed_Limit doesn’t change so CPU throttling isn’t occurring. I’m guessing from this that my processors must be seated and installed ok. It must be my 4x8gb Ram stick configuration causing the lower than expected Geekbench score.

 

[tpluth]

ok, thanks, that’s good to know.

I’m concerned because the heatsink screws on CPU_A were much tougher to tighten to the required number of turns than the heatsink screws on CPU_B. I’m concerned that CPU_A heatsink is too tight and getting too hot because of it. I thought I might need to loosen each screw by a quarter turn.

I’m just running the fans on high as you suggest and it’s doing what you said it would. I can see that the red temps in iStat aren’t something to worry about.

CPU A is the back one and it always runs hotter. It's got the heat from the northbridge chip getting sucked into it.

Nothing to worry about.

 

[Upgrader]

CPU A is the back one and it always runs hotter. It's got the heat from the northbridge chip getting sucked into it.

Nothing to worry about.

Thanks, that’s a useful insight.

 

[EdDuPlessis]

As a side note, I’m running the Terminal command: pmset -g thermlog whilst running Geekbench. The CPU_Speed_Limit doesn’t change so CPU throttling isn’t occurring. I’m guessing from this that my processors must be seated and installed ok. It must be my 4x8gb Ram stick configuration causing the lower than expected Geekbench score.

Don't worry about the score it can be effected by such small things such as background processes, os version etc. The 12 core 6.1 gets around the same score and it's the fastest Mac in the current line up and costs a stupid amount of money.

 

[Upgrader]

The 12 core 6.1 gets around the same score and it's the fastest Mac in the current line up and costs a stupid amount of money.

Yeah, this was the thing that prompted me to source a 2009 4,1 tower and upgrade it. I’ve spent £2000 for about 95% of the performance of the 12 core nMP which costs £6000. Of course, the GPU performance isn’t close but I don’t really use the GPU.

 

[PowerMike G5]

Hi Mike,

I’ve just done the same as you and successfully upgraded an 8 core 2.26ghz 2009 4,1 to a 12 core 5,1 with two x5690’s. It was a tense experience to say the least but it worked first time.

I’m only getting 30916 for the multi core score though. How are you getting 32000+ ? Are you somehow overlocking? From those Geekbench stats my MacPro is exactly the same as yours in terms of motherboard and components etc except I have 32gb RAM to your 24gb.

Cheers.

Hi Upgrader-

I'm not sure. They are standard x5690 processors.

My entire system is fully upgraded though, so maybe that is contributing somehow to the speed (Accelsior_e2 PCIE boot SSD, Dual SSD PCIE Scratch drives for video editing, Nvidia GTX 680 GPU, etc.).

----------

Re cooling, I’ve added extra thermal pads across the voltage regulators to fill the gap created by the new lidded CPU. Contact should be fine as I used 2mm pads and cut the strips to size. Actually, Mike, did you add extra thermal pads when you upgraded?

----------

...also, CPU_A is running roughly 10 degrees F hotter than CPU_B whilst just surfing a few web pages and generally idling. Is that normal?

I only used Thermal paste. Arctic Silver 5 Thermal Compound.

 

[Upgrader]

Cheers Mike,

Of course, you wouldn’t need to add extra thermal pads as your machine is a 2010 version. Mine is a 2009. My bad. Re your RAM, do you have 6x4gb sticks installed, 3 per proc?

I’ve also added the Accelsior as my system disk. I’m impressed with it so far but slightly disappointed to discover the Barefeats shootout between the Tempo SSD Pro card and the Accelsior. The Tempo beat it across the board though to configure it to match the Accelsior would’ve cost me £50 more, plus the Tempo can’t hold a recovery partition.

Thanks.

 

[tpluth]

Having the system on a fast SSD helps GB3 scores.

My system disk is 2 Crucial MX500 mSATA cards on a Syba PEX-SD40079 card in a RAID0 configuration. It's a Marvell 88SE9230 controller and works OOB on the Mac Pro. You can only access the configuration through Bootcamp, though.

 

[tinycg]

Trying to justify doing this to my 2009 MP, I'm conflicted over how much more life this system has left, its been running for a long time, although not often taxed to limit as much anymore. I'm looking to do more CPU intensive stuff and wondering what advice you guys have for people that have done it, or are thinking of doing it. For the price, it seems like a good move compared to the cost of equivalent performance - granted that performance is new vs old haha.

Any advice would be awesome.

 

[tpluth]

Trying to justify doing this to my 2009 MP, I'm conflicted over how much more life this system has left, its been running for a long time, although not often taxed to limit as much anymore. I'm looking to do more CPU intensive stuff and wondering what advice you guys have for people that have done it, or are thinking of doing it. For the price, it seems like a good move compared to the cost of equivalent performance - granted that performance is new vs old haha.

Any advice would be awesome.

To get the most out of it, you'd need to replace your RAM, too..

If you're looking for better application response, think about a SSD for at least system and applications, and maybe look into the x5687 or x5677 if you only need 8 cores.

 

[flowrider]

^^^^The X5687 while being clocked higher than the X5677 is, I believe a No Go in the 4,1 or 5,1 Mac Pro. I have seen one case where someone with a SP 5,1 tried it and it didn't work. I have also scoured the internet trying to find a successful installation on a Mac Pro with no luck. And, if you look at the GeekBench database, there is no one that has shared a score with the X5687 in a Mac. For all those reasons, I chose to go with an X5677 in my 5,1.

Lou

 

[Upgrader]

Trying to justify doing this to my 2009 MP, I'm conflicted over how much more life this system has left, its been running for a long time, although not often taxed to limit as much anymore. I'm looking to do more CPU intensive stuff and wondering what advice you guys have for people that have done it, or are thinking of doing it. For the price, it seems like a good move compared to the cost of equivalent performance - granted that performance is new vs old haha.

Any advice would be awesome.

As an example, I needed a powerful machine to replace my laptop. Top of the range nMP = roughly £6000. I bought a 2009 8 core and upgraded the procs to X5690's. It's now roughly 2.5 times as quick. I've added an OWC Excelsior PCIe SSD for my boot drive. I added 48gb RAM. All for about £2000 which gives me about 90-95% of the £6000 nMP performance for about 30% of the cost. I foresee the machine lasting me for the next few years until the nMP's can be got cheaply and upgraded in a similar fashion. To my mind and for my requirements (After Effects/Cinema 4D) it's definitely worth it. Also, I spent a month working on a £5000 8 core nMP in After Effects and C4D and I cannot tell the difference to my newly upgraded unit. The only difference would've been the GPU performance which I don't rely on.

If you decide to upgrade then here’s the very best guide there is specific to the 2009 dual proc cMP: http://pindelski.org/Photography/2013/07/08/mac-pro-2009-part-tbd/ which you need to read and re-read. The author says he’s upgraded many processors using this method and hasn’t failed once. I believe him. There’s one point in that article that I think need amplification which was the varying friction of the heatsink screws and how it differs between loosening them and tightening them. Also, how the friction can differ between screws on the same heatsink. You must stick to the logic of the correct amount of turns regardless of friction. The other thing that’s easy to do is to lean your wrist on heatsink B as you loosen/tighten heatsink A, which I think must be avoided.

This forum holds some incredibly decent and helpful people who will help you too as they did me.

Cheers.

 

[tinycg]

Awesome advice guys. I know to get the most out of the upgrade it's recommended to upgrade ram as well. I already have 32GB of 1066, is that something that is required to upgrade (will I wreck havoc with the new chips and slower ram?) or is this something that I could do in the future?

What kind of performance gains are realistic with the 1333 is it noticeable in normal operation or just under high memory tasks?

 

[Upgrader]

Awesome advice guys. I know to get the most out of the upgrade it's recommended to upgrade ram as well. I already have 32GB of 1066, is that something that is required to upgrade (will I wreck havoc with the new chips and slower ram?) or is this something that I could do in the future?

What kind of performance gains are realistic with the 1333 is it noticeable in normal operation or just under high memory tasks?

RAM will just run at its 1066 speed, it wouldn’t week havoc at all. You could upgrade RAM at a later date. I don’t know the real world difference between the two speeds, i’d imagine it’s minimal so it’s up to you. For me it was a new system so I got the fastest I could.

You’ll also want to upgrade the Firmware to 5,1 over your current 4,1 for the RAM to run at 1333 if you decided to go for it.