MOTUnation.com

BobK wrote:

James Steele wrote:I could try to take the lidded hexcore CPU in my 4,1 quad I converted into a 5,1 hexcore and test it in the dual cpu machine, but I'd have to worry about tightening down the heatsink too much and messing up the socket.

I got a red LED more than once. Once it was because of a defective CPU; the other times, it was because the heat sink bolts weren't tight enough.

I just checked my notes, and in one case, tightening the bolts an additional 1/8 turn was enough to fix that. (BTW, if I recall correctly, the LEDs between the heat sinks are for RAM issues.)

Not to beat a dead horse, but the guy who runs DN computing, which sells de-lidded CPUs, said in a thread at MacRumors that it's impossible to tighten the bolts too much unless you have superhuman strength. Just make sure you do it in the order Apple recommends, and just 1/4 to 1/2 turn at a time.

I certainly tightened mine more than I expected to, but I realize that's subjective to the point of meaningless without a torque wrench reading.

(Yes, the guy at Anandtech somehow bent some pins in one of the sockets. I'm not sure what DN Computing would say about that.)

When I got the replacement processor (one of my first batch was defective), I tested it before de-lidding by installing in CPU slot A. With only one processor, my 2009 Mac Pro's fans went full blast, and a red LED went on, but it started up, and the new processor and 4 RAM sticks appeared in System Info. (There are 4 RAM slots per processor, so even if you have 8 sticks installed, only 4 will appear when there's only one CPU.)

I figured that the main risk of installing a lidded processor is overheating, because the thermal pad at the bottom of the heat sink won't fully contact the logic board; but I also figured that if you just start it up just long enough to check System Info, that shouldn't be a problem.

Also, in the 2009 MP, without a clip to secure the CPU, MAYBE it could slide around a bit if the bolts were tightened unevenly and/or too quickly, but there's very little (if any) wiggle room in the CPU socket.

I had one CPU where I turned the power on and the fan ran full blast but it did that for at least a full minute and I just gave up on that one and held the power button down until it went off. Maybe if I had let it continue to run it would have been okay... dunno.

I know I didn't under tighten it because I tightened it down as far as it would go, like I did with my single processor 4,1 when I upgraded it. I didn't tight it down 1/4 or 1/2 turn at a time though on each bolt, although I cant imagine with the notches in the socket, etc. that the CPU could slide around much at all, but maybe that's worth a try. I know I'll test before de-lidding next time.

Did you attach the heat sink before testing the lidded processor in CPU A slot? I know with the lid on the height is different and you COULD overtighten the heatsink.

Sigh... not sure what is going on here. I did do the firmware update, but I did it with the CPU tray in a different Mac running Mountain Lion. Doesn't seem like it should make a difference because it does show as MacPro5,1 in system info.

James Steele wrote: Did you attach the heat sink before testing the lidded processor in CPU A slot?

Yes.

I left CPU slot B empty, powered up, and as soon as I saw that the processor and RAM appeared in "About This Mac", I shut it down.

Well thanks to HC Markus's advice I found a guy online who will delid CPUs for $100/pair. Way better than the other eBay source. I wrote him and he offered to inspect them for me and tell me what may be wrong with them. I mailed them off to him today. In an odd way I'm really hoping he confirms that they're all bad (whether they came that way or I damaged them), because otherwise I might have other issues going on.

I think there's a slightly bent pin in the socket for CPU A, yet the machine boots no problem with the original quad core CPU. None of the hexcore CPUs would boot, however. I don't know if it's safe to assume that if a quad CPU will boot, there isn't some socket damage that only a hexcore might have an issue with??

If any of the CPUs test good maybe I'll see if I can send him my CPU tray. Otherwise maybe I'll just have an 8-core for the office and when 5,1s get really cheap when the modular Mac drops I'll go for one of those. The whole de-lidding thing introduces variables into this situation that are kind of a pain. I wonder why Apple chose to do this in just the 2009 dual CPU models? Maybe for cooling? Perhaps the heat sinks in the 2010 dual models are larger or more efficient?

Oh well. I'm where I was at before but that's not a bad thing. FYI, my single hexcore gets a Geekbench 4 score of over 15k which still beats the stock 4,1 8-core by over 2,000.

Sorry you're having to go through all of this, James.

James Steele wrote:The whole de-lidding thing introduces variables into this situation that are kind of a pain.

Yup! I was concerned that the heat of a clothes iron might damage the chips, but I guessed that if they could tolerate the heat of solder when the lid was first attached, maybe they could tolerate it one more time. And that if this were really risky, people would have mentioned it in comments on the YouTube videos or in forums.

In other news, I upgraded my 2009 hex-core from El Capitan to Sierra 10.12.6 yesterday. It seems to start up faster.

It also seems to be maybe running a little cooler. This is mainly based on casual glances at the iStat Menus temperature readings in the menu bar, but the "last 7 days" graphs seem to show a slight decrease too.

Well just a follow up. I heard back from the guy who does the de-lidding and he inspected the CPUs. I pretty much managed to trash 5 of the capacitors on one of the CPUs and 7 on another through my poorly executed attempt to do the wet-sanding of the die to buff all the solder off of it. Apparently it's an unnecessary step and scraping the solder off carefully with a single edge razorblade is adquate. He did say one CPU appeared okay, but didn't want to test in his machine since he didn't do the de-lidding. He could clean that one up and replace capacitors on the one with 5 bad ones for $60 and refund $50 if it didn't work ($10 for shipping/handling not refundable) when I got it. It would take 7-10 business days to turn around.

That seemed very fair, but time consuming and when I tested the CPUs individual in CPU A slot (with B empty)... NONE of them would boot, so it seemed to like the best thing to do would be to write this off as a very expensive learning experience. I got the last two CPUs from the same vendor and am having them mailed directly to him and he'll de-lid them with a special tool he has and test them, all for $100. If I felt like taking a risk at this point, I'd be brave and de-lid with the vise and just scrape the solder. The wet sanding thing was really my undoing here. So don't do this: https://youtu.be/aPhDfUkll-o?t=16m53s

Fortunately, I got such a good deal on the 4,1 8-core ($300) that I'm probably about what it would have cost if I paid normal going rate for the 8-core machine. Win some... you lose some. I think though, that if I do this for a friend, I'll probably be able to handle it. Again, I just wasn't careful enough with the sanding and from what I've read that's really some anal retentive stuff... not necessary and pretty risky if you're not SUPER careful.

Back on track... this is good.

Glad Box185 worked out!

HCMarkus wrote:Back on track... this is good.

Glad Box185 worked out!

Yes. He's a very nice and very trustworthy person. Thanks so much for telling me about him!

Well, hooray! Just a quick update. I'm now running a 12-core machine. I received my processors today from a guy who charges $100 to professionally delid and prep them (as i was a bit gun shy this time around), and popped them into my 8-core and everything went smoothly. All seems well, although I haven't checked all my plugs to make sure authorizations are okay. Native Instruments seemed to give me a hard time at first, but seem to work now.

FYI, I previously had a 4,1 single CPU 4-core machine in my studio that I'd upgraded to a 5,1 6-core. After I upgraded the 4,1 8-core to 5,1 12-core I just swapped CPU trays with my studio machine which was much easier than swapping out drives and PCIe cards, etc. That might be why NI plugs are still working? Dunno.

So, here are the results of the upgrade according to Geekbench 4. I noticed Geeekbench 4 seems to produce lower scores than Geekbench 3.

My previous 3.33 hexcore machine with 24GB RAM scored:
Single Core: 3008 Multi-Core: 13946

After the upgrade to dual 3.33 hexcore (12 cores) with 48GB RAM:
Single-Core: 2864 Multi-Core: 24343

So my Geekbench single-core score actually went DOWN 144 points from 3008 to 2864, whereas the multi-core score went UP a rather huge 10397 points from 13946 to 24343.

Here's the thing that's interesting to me. I loaded a project I'd been working on and it didn't SEEM like the processor meter really went down very much if at all. I'm thinking this may be because it wasn't a LOT of tracks and from what I've come to understand, whenever you have several plug-ins in series in a single audio track, they're all on ONE core. So in this case, given my single core score actually went DOWN, I may not see much difference until I'm working on a bigger project with lots of plug-ins spread across MANY tracks to take advantage of more cores.

Not exactly sure why my single core score would go down. Interestingly, I installed the hex-core CPUs in the 4,1 8-core machine one at a time. After installing CPU-A, I booted just to see that it worked, and it booted up as a single CPU hexcore machine. I actually ran Geekbench 4 on it at this point, and the single-core score was about 3000. After installing the second hexcore CPU in CPU B slot, as you can see, the single-core score dropped to 2864. I don't know what it is about dual CPUs that reduces the single-core score, but it seems to be a real phenomenon.

It does look like I lost my Native Instruments B4II authorization and since it's no longer supported, not sure how to get it back. Running Service Center (both old and newer version) shows that it's authorized, but every time I launch it it tells me it's not authorized and puts it in demo mode. Sigh.

James Steele wrote:It does look like I lost my Native Instruments B4II authorization and since it's no longer supported, not sure how to get it back. Running Service Center (both old and newer version) shows that it's authorized, but every time I launch it it tells me it's not authorized and puts it in demo mode. Sigh.

Send over your organ tracks and we'll record 'em thru my (real) Leslie.

I heard about this video and it has some good info on the 4,1 to 5,1 12-core upgrade including some drive options and wifi and Bluetooth.

Finally decided to upgrade the CPU on my 8 core 4.1 (5.1 firmware), I have 2 de-lidded X5690 3.46Ghz + 32GB 1333Mhz ram arriving next week. I have watched various videos and read many threads, so I hopefully know what to do.

But I see various opinions on the best method of applying the Arctic MX-4 thermal paste on the de-lidded (non square), some say a pea blob others a line. Which did you guys do?

stevenew wrote:But I see various opinions on the best method of applying the Arctic MX-4 thermal paste on the de-lidded (non square), some say a pea blob others a line. Which did you guys do?

Page 1770 of the Mac Pro 2009 service tech guide, downloadable from various sites, shows the paste applied in kind of an elongated/squished "theta" and says to use "approximately 0.1 cc".

I used the line method recommended by Arctic Silver for Xeon CPUs, but my CPUs were not de-lidded.

Thanks for the replies, sounds like the squashed Theta might be best for de-lidded but I imagine getting the right amount in to that shape will be tricky.

Arctic recommends the one point blob, is there a reason that would be unsuitable for de-lidded?