MP 1,1-5,1 My Mac Pro 5.1 (2009): plans, observations & questions

[ToniCH]

So, I bought a Mac Pro 2009 (4.1->5.1). Specs: X5660x2, 16GB RAM, 240SSD, HD4870, 10.13.x. Very clean machine.

I upgraded the RAM to 64GB with 8GB 1333MHz sticks I had in my box of old memory. Went smoothly. Also updated the 10.13 with all Apple had to offer. All good there too.

I have read Mac Pro -stuff for about 3 days now. Good thing I did. I did the ROM dump and health check on the NVRAM and the figures were very alarming. 😯 So alarming I don't want to boot it unnecessarily until its sorted out. And with tsialex's help it will be soon. 👍🏻

Other problems I have detected:
- the GPU fan seems noisy. Maybe bearing is bad or its scraping the plastics. Need to take a look.
- Booster 1/BoostA fan is not running at all according iStat M. and Macs fan control. Is it supposed to run all the time? And which one is it? techicians guide (2010) uses different names for the fans. Will peek inside to see if everything is connected and maybe spot a dead fan.
- DVD does not open from kbd, need to take a look

I am to not too bummed out because of the problems, pretty used to the fact that even though previous users say their machine is perfect, IRL they rarely are and need so love and care to be 100% operational. This was to be expected. Hence I prefer not to pay too much for vintage stuff.

Good things:
- to me temps look quite ok, at idle at least
- no weird behavior observed while testing
- processor performance is in line of what they should be, according to GeekBench
- everything else I've tested this far works incl. wifi and bt

What next:
1. ROM reconstruction (by tsialex)
2. maybe take the cooling towers off to see how the upgrade was made (lids still there?) and reapply heat paste as the upgrade was made many years ago. NB too. EDIT: ordered new plastic spring stands to the NB heatsink.
2b. sort out the booster 1 fan issue at the same time
2c. look at the GPU if anything can be done to the fan noise
3. ordered the Sapphire Pulse RX580 to replace the old GPU and a power cable for it
4. SSD or NVMe to PCIe -upgrade from existing parts
5. USB 3.2 -card (?)
6. possible cpu upgrade to X5690x2 when I see the machine is trouble free. The processor tray is the black variant with serial indicating manufactured in july 2009, so it should be ok. Not much of an improvement but I take all I can get.

Some pics below. Any comments on the iStat-pic - spot anything concerning?

 

[ToniCH]

Other problems I have detected:
- the GPU fan seems noisy. Maybe bearing is bad or its scraping the plastics. Need to take a look.
- Booster 1/BoostA fan is not running at all according iStat M. and Macs fan control. Is it supposed to run all the time? And which one is it? techicians guide (2010) uses different names for the fans. Will peek inside to see if everything is connected and maybe spot a dead fan.
- DVD does not open from kbd, need to take a look

maybe take the cooling towers off to see how the upgrade was made (lids still there?)

Ok, so - I took a look at these issues:

1. GPU fan is indeed noisy but goes quiet when I tap the fan. Will look at the card shroud more closely later.
2. Front, rear and PCI fans work. Didn't find any more fans but obviously there is one in the PSU. 2010 tech manual exploded view also has only 3 fans + PSU. What are these "booster" fans iStat reports?
3. DVD opens when I press the button on its face plate but does not open from keyboard even though the eject-pictogram visits the screen when I press the eject button on kbd?

4. Didn't have to take the coolers of the processors to see whats up. They have lids and no washers! 🙄 But, luckily they seem to be enough and not too much tightened as temps are low and it seems to be working. Will definitely order the new processors, delid them and replace with new thermal paste.

 

[tsialex]

What are these "booster" fans iStat reports?

There is a little fan inside the CPU heatsink assembly, so, a dual CPU tray, two CPU heatsinks, have two booster fans, one for each heatsink.

 

[ToniCH]

There is a little fan inside the CPU heatsink assembly, so, a dual CPU tray have two booster fans.

Ah, ok thank you. Are they both supposed to be running at all times? Ie. me seeing one run and another at 0rpm = normal or broken fan or loose connector? I am guessing that booster 1 is CPU A fan?

 

[tsialex]

Ah, ok thank you. Are they both supposed to be running at all times? Ie. me seeing one run and another at 0rpm = normal or broken fan or loose connector?

Yes, they run all the time.

It's very common to break the fan/sensor connector locking tab to install lidded Xeons to an early-2009 dual CPU tray, so, inspect the connector for the CPU with the booster fan not working before anything.

 

[ToniCH]

so, inspect the connector for the CPU with the booster fan not working before anything.

Thanks for the tip. 👍🏻 I took the tray out and pushed the connectors down with a flat tool between heatsink fins and replaced the tray. Now, the booster 1 works too! 😀

 

[ToniCH]

Alright, with tsialexes great expertise my NVRAM is now restored to fully working order and the bootROM is updated to the latest version!

Ps. for some reason my GPU fan is today much quieter than yesterday. Maybe because I was poking it? Well, not a permanent fix I am sure.

 

[ToniCH]

Took the OWC Accelsior S -Sata to PCIe -adapter card out of my G5 and installed it into the cMP with a 1T SSD. I pretty much tripled my write and doubled my read speeds vs. internal SATA connected SSD. 👍🏻

Also did a new BootROM flash to include the EnableGob.

EDIT: the DVD-drive started to work when I unplugged it and plugged it back. Now it opens normally. 🧐 Weird but...good.

 

[ToniCH]

This is interesting (GeekBench 5 comparison):

1. top is this machine with Linux Mint 21. The GB reports wrong processor speed, its 2.8GHz.
2. next is exactly same machine with High Sierra with the old MP51.0089.B00 BootROM.
3. then we have the same machine again with High Sierra but this time with the new 144.0.0.0.0 BootROM.

Do you think the speed difference between 2 & 3 is because the BootROM or what causes this?

 

[tsialex]

This is interesting (GeekBench 5 comparison):

1. top is this machine with Linux Mint 21. The GB reports wrong processor speed, its 2.8GHz.
2. next is exactly same machine with High Sierra with the old MP51.0089.B00 BootROM.
3. then we have the same machine again with High Sierra but this time with the new 144.0.0.0.0 BootROM.

Do you think the speed difference between 2 & 3 is because the BootROM or what causes this?

View attachment 2227100

Performance decrease is normal and expected with side-channel mitigations with the most recent Intel microcodes.

 

[ToniCH]

Performance decrease is normal and expected with side-channel mitigations with the most recent Intel microcodes.

Thanks for your reply. 👍🏻 What is "side channel mitigations" in laymans terms? And why does it effect only multiprocessor functions? Something between the cores? They obviously does not effect Linux.

 

[tsialex]

Thanks for your reply. 👍🏻 What is "side channel mitigations" in laymans terms? And why does it effect only multiprocessor functions? Something between the cores? They obviously does not effect Linux.

Please read the Wikipedia article about side channel attacks.

Side-channel attack - Wikipedia

en.m.wikipedia.org

Intel provide updated microcodes for minimizing the attack surface, the performance decreases around 3% to 5% for most real life workflows.

Apple recommended to also disable HT.

There are good threads here on the Mac Pro forum about Meltdown and Spectre.

Linux kernel works differently than Darwin and seems that some mitigations affect it less.

 

[ToniCH]

Ah, ok - I did read about the issue few days ago among hundreds of other things I read about cMPs. The term just didn't ring a bell. Thanks.

 

[ToniCH]

My GPU started again to sound like a buzz saw. So I lubed the fan bearing. At the same time I cleaned the board, there was quite a lot of greasy dust and the thermal compound had dried totally and reminded me of concrete. It didn't even all come off with isopropyl + qtip, I needed to scrape with a toothpick to get all off. I then reapplied new paste and put the thing together. No more noise of any kind! 👍🏻

Note that this model of Radeon fan didn't have the cover plug in the back for re-greasing/oiling, I had to pop the actual motor off the fan using 3 screwdrivers (even pressure to the triangle part). Wasn't difficult and it popped right back after greasing.

This Radeon HD 4870 will be my "recovery/emergency" GPU when its replacement (RX580) arrives.

I would recommend everybody to clean and reapply thermal compound if your GPU is old - lets say ~7 years old or more. And I am sure aftermarket compounds have better performance figures than what is used in most consumer grade cards nowadays. So you might see benefit in doing this even if card is not so old yet.

 

[ToniCH]

Today I installed the Inateck RedComets U21 USB 3.2 PCIe-card to the cMP. Its a ASM3142 -card with 3xUSB-A + 2xUSB-C.

It seems to work ok but not lightning fast. I don't have anything free now capable to decent speeds to test it with. I only had an old MBP Apple original SSD in a Transcend USB 3.0 case and with that I got 300mbps write and 400mbps read. If I remember it correctly the SSD wasn't any faster when it still was installed internally in the MPB. With a faster drive I might get max about 600-700mbps or so, according to manufacturer.

EDIT: found an Inateck USB 3.2 NVME -case with a cheap no name MBA (mid 2012) -compatible 512GB SSD. Connected to USB-C I am seeing w390/r510mbps.

Test transfer, copy 8.26GB Catalina installer from the Transcend/Apple SSD to the desktop:
1. Mac Pro USB 2: 3min 39s
2. Inateck USB 3.x: 0min 21s 👍🏻

Well, in any case its way faster in file transfer than the on board USB2 so I call this a success. 😉

PS. the RX580 arrived but the 2x6pin to 8pin is not here yet.

 

[ToniCH]

Next step: the 970 Evo plus 2TB NVMe. I installed it to the Aqua kryoM.2 PCIe -adapter and plugged it into my PC, started the Samsung Magician and verified that the drive has the newest firmware. The drive was made in march 2023. Then plugged it into the cMP and booted up. It works and seems pretty quick.

Benchmarks from left to right:
1. internal SATA Kingston 240GB SSD
2. OWC Accelsior S PCIe-card + OWC 1TB SSD -drive
3. external Inateck USB 3.1 NVMe case with no name 512GB NVMe-drive to Inateck USB 3.2 card through USB-C
4. Aqua kryoM.2 PCIe -card + 970 EVO Plus 2TB

 

[ToniCH]

I got frustrated to waiting the 2x6pin to 8pin cable which our glorious Postal service has been unable to deliver even as it has been in their hands since july 2nd. So, I bought another cMP which does have the cable - and some other parts I can use. So, my plan now is to take all the best parts I have and upgrade my first cMP to optimal combination and use the rest of the parts to make a good backup-machine from this new one.

Here is what I am at now:

1. the original 2009 4.1->5.1 cMP - my primary cMP
- 2x X5660 2.8GHz, have ordered 2x x5690
- 8x8GB 1333MHz RAM
- Sapphire Pulse RX 580 8GB
- Inateck RedComets U21 USB 3.2 PCIe-card
- OWC Accelsior S -Sata to PCIe -adapter + OWC 1TB drive
- Aqua kryoM.2 PCIe -card + 970 EVO Plus 2TB
- 144.xxx BootROM, today installed the RX580 and Mojave

2. the new mid 2010 5,1 cMP - the backup machine
- 1x x5680 3.33GHz (Geekbench 5 CPU: 656/3428)
- 4x16GB 1333MHz RAM
- Nvidia GTX 780 Ti 3GB
- 250GB SATA Samsung 850 EVO SSD
- Lycom DT-130 NVMe switch PCIe-card
- Sonnet Allegro 2x USB C 10Gps PCIe
- and it came with some other PCIe-cards, adapters, cables, heat sink, BT-dongle, orig. processor, more RAM, EVGA Powerlink, HD5770 etc. etc.
- 144.xxx BootROM and Mojave installed


So, here is the plan for the main cMP:
- install the RX 580 and upgrade to Mojave - done
- install the X5690-processors when they arrive
- Install the Lycom DT-130 with 2pcs of Samsung 970 Evo Plus 2TB NSVMe drives (did order another of these drives)
- RAM: install the 16GB sticks - done, put them into slots 1-2 and 5-6
- RAM: install additional 8GB sticks if they offer speed advantage when going to 3+3 memory configuration? To slots 3 and 7 or better to put them into 1 and 2? <- a question
- test the Inateck and Sonnet USB cards and choose the one that is better <- comments?

Plan for the backup machine:
- use rest of the parts to make a best possible combo

 

[ToniCH]

So, I installed the Lycom DT-130 and 2x Samsung 970 EVO Plus 2TB's. Works nicely. I made a RAID 0 out of them but I am not seeing any speed benefit over non RAID. The transfer speeds are pretty much the same as without RAIDing. At least in short Black Magic benchmark -runs. Maybe I am hitting a wall with the cards ability or maybe the PCIe-bus?

So, I might go without the RAID -config, unless there is a valid argument on behalf the RAID? Well, in any case a decent improvement by changing the card (+1000MB/s).

Very tight fit with the RX580. I put a cardboard spacer between the cards so the Lycom doesn't sag and touch the RX580 fan in any situation. I'll update the cardboard to something more appropriate later on.

BTW. I noticed that the EVO drive temps are 41°C for the one with Icy Box -sink and 50° for the one with smaller more simple sink. The difference between the 2 drives is quite steady 9-11° when stressed too.

The pics show each drive benchmarked individually and then in RAID 0 -configuration.

 

[ToniCH]

So, decided to tackle the wifi-card upgrade before night. I have the BCM94360CD, which was intended for my iMac 27" 2011 but now goes to the cMP with a chinese adapter kit. And I also had some antenna-cable extensions in my parts box which was useful for the BT-antenna connection.

Everything went smoothly, except the chinese kit didn't have a plug and play BT USB-cable. It came with 2 variants and both had the correct connector to the adapter side but neither was plug and play to cMP back plane side. One version had an incorrect connector and another came without the connector. So, I'll have to source the correct connector so I can plug it into the cMP.

But, I have a BT 4 -dongle which works just fine in the meantime. Brand/model: Kinivo Bluetooth dual mode dongle PANBT400.

Otherwise everything was totally plug and play and the cMP had 3 wifi-antennae cables even though the original card only had 2 connectors (weird).

So, after 15 minutes of fiddling I now have AC Wifi. 😎

The wifi card on top is the original one and the new one with the adapter is the other one. I connected my antennae according to the last photo I found by googling BCM94360CD connectors.

 

[ToniCH]

So, today I made the BT-cable by combining the original cable and the one that came with the cheap kit. And now I have BT 4.0 provided by the BCM94360CD wifi/BT -card. 😎

 

[ToniCH]

I have my new X5690 -processors and I have delidded them. I also received new pins for the NB. And I have tools, thermal paste etc. So, I am set for the upgrade. 👍🏻

Except that the previous owner used lidded processors in previous upgrade. So, I don't think I have the plastic processor spacer rings which originally were used with delidded processors in 4,1. Are they absolutely necessary or can I do the installation without them?

 

[ToniCH]

Well, I guess I continue my monologue. Too bad nobody is interested in commenting anything.

I did search the forum and read many threads and I now get the feeling from some of the posts that the spacers are not absolutely necessary. So, I will proceed without them.

 

[tsialex]

The spacers are to avoid any damage to the Xeon die when installing the heatsinks. If you take enough care is possible to install without it.

Let's never forget all the AMD Socket A Athlons and Durons with cracked dies. One trick from back then is to use a lot more thermal paste to act as a buffer/shock absorber while the heatsink is being installed, it will go to the sides of the die and will avoid damage.

 

[ToniCH]

The spacers are to avoid any damage to the Xeon die when installing the heatsinks. If you take enough care is possible to install without it

Thank you again for your help. 👍🏻

I am getting an impression that the cooler screw stands are dimensioned so that one cannot easily crush the processor die? So, when tightened until they feel snug the cooler and the Xeon die should be at ideal position to each other?

 

[tsialex]

Thank you again for your help. 👍🏻

I am getting an impression that the cooler screw stands are dimensioned so that one cannot easily crush the processor die? So, when tightened until they feel snug the cooler and the Xeon die should be at ideal position to each other?

Yes, but even with the spacers you need to screw one or two turns each screw in a X pattern to avoid crushing the die. It’s extremely easy to crack it when you are not taking care, I’ve saw several early-2009 Xeons with micro cracks at the edges and damaged socket pins over the years