|Jan 9, 2013, 01:30 PM||#51|
Hey KB, did you get the bad CPUs from IMS?
I ordered 2 from there and spent a bunch of time yesterday trying to get them to work.
Was a complete waste of time.
I finally remembered that my PC flashing rig was same CPU socket so I had another testing option.
Neither 5680 worked in there, and when I moved the i5 from PC into the CPU A socket on Mac Pro, it worked fine as a single CPU until the fans started ramping.
When I couldn't get Mac to boot with the 5680s in either slot and tried switching them, I eventually just tried running them in CPU A slot singly, neither worked there. A red LED lit up to the right of the RAM slots.
Oddly, when I ran the i5 in CPU A it thought it was a "2.8 Xeon".
I have senn plenty of GPUs die between when I test and when they arrive at customers place. The 8800GTs die all the time. But to have gotten 2 bad 5680s is just too much. They were $750/ea, which I guess falls in the "too good to be true" category.
Going to RMA them, guess I need to pony up and buy them from a different vendor.
Did anyone else get them from IMS in Michigan?
The fact that the lidded i5 worked on first try tells me that I did the fan connector mod correctly and that it shouldn't be so difficult.
I have the feeling that someone ended up with a bunch of "iffy" 5680s and decided to ship them out and see which ones would "stick" and which ones would come back. Still hard to imagine that both could be dead.
I even tried putting a 2.26 in the PC rig, booted up fine, while either of the 5680s gives a "FF" on the motherboard readout.
Talk about anticlimactic. Back to the screaming 2 x 2.26.
|Jan 9, 2013, 04:05 PM||#52|
I ordered from the ITCreations link I posted here before. One of the two CPU's I got was defective and one worked. I contacted the seller and they shipped me a replacement pair for overnight which worked without any real issues, was just a little afraid to tighten them as far as I needed to for the memory channels all to be recognized. The defective one I received had a Red Light next to it's ram light up so I assume that was for CPU Fail, when I switched it to CPU A i got the odd boot screen and after asking around here confirmed one of the 2 CPU's was faulty.
I paid 700 each for X5680's from ITCreations, for the ones they sell which they told me were "New Server Pulls" which they claim to cover for 3 years if anything goes wrong. Their customer service was responsive and promptly corrected the issue by sending me replacements so I feel pretty confident keeping the ones I received the 2nd time.
Just a note on the fan connector Modification, I didn't mod the cord that connects to the board and instead once I had the heat sink tightened applied pressure to the connector from the back slightly. I have 3 of the washers that Tutor recommended in place on each post.
Gluck with the upgrade if ur still going to try it, I spent about 6 hrs for mine but pretty sure I could do it under 45 mins now if I knew the CPUs both were functional. Thanks to all the great help I received on this forum
27" + 24" LED ACD; 2 x 3.33GHz X5680 Mac Pro, 16 GB RAM, 512 GB Samsung 830 SSD ,EVGA Geforce Titan Super-clocked; Early 2013 2.8Ghz 15in Retina Macbook Pro 16GB Ram 756GB
|Jan 11, 2013, 10:38 PM||#53|
X5680 Heaven !!!
OK, it is with some embarassment that I need ot admit that the 5680s I got were A-OK.
I have switched my Octo 2.26 to a 12 Core 3.33 !!!!
The biggest issue turns out to be with those $#%**!! heatsinks Apple uses. Why, oh WHY do they rely on precise screw torque for proper engagement? My PC flashing rig is also LGA1366 but uses an infinitely cheaper and INFINITELY easier to deal with clamp system.Flip the clamp over, chip is snugged with perfect tightness every time.
It turns out, that despite getting 2mm Nylon washers and thermal pad for chips, I was over tightening the screws. I still find it odd that I had no trouble getting my i5 working with a lid on it.
BUT...when I tested in flashing PC and got "FF" on the display, this seemed to correlate with the chips not booting in Mac. I then fired off an angry email to the IMS folks. They patiently wrote back with an RMA but pointed out that 2 dead Xeons seemed unlikely and that I should see if there were firmware updates for my rigs.
My PC flash rig is based on a EVGA e758 Tri-SLI CLassified board. I soon discovered that Xeon compliance was enabled in board Rev 1.2. I of course had board Rev 1.1. And, as noted in various forums, the Rev 1.0 and Rev 1.1 boards would show "FF" until a specific mod was done to board.
I spent hours looking for how to do this mod. Found pix for e759 and e760, but not for e758. Turns out that I needed to do one of the mods from those other boards, but components were in a different place. They were also numbered differently to add to the fun. I spent a cranky day finding the images, disassembling the PC and removing heatsinks to get to the two pads that needed to be soldered together. After putting it all back together with one of the 5680s in, the darn thing booted right up. Tried with other 5680 and same exact thing. Amazing how soldering together 2 little pads solved this issue.
In any case, I now was convinced that Xeons were fine and that "Operator Error" was the issue. Put them back in the CPU tray, still got red light. On a whim, I tried LOOSENING the screws. BOING !!! Got CPU-A working, loosened the screws on B and it started working too. I then had to TIGTHEN the screws on B to get all the RAM to show up.
But 12 cores now running at 3.33 Ghz. Machine feels MUCH faster now, like adding an SSD.
On a side note, at some point in the fiddling I pulled CPU-B and noticed a diagonal SMEAR of pins, almost exactly like the ARS-Technica article damage. And I mean EXACTLY like it, right down to blue pins from shorting. I ran over to DV Warehouse and grabbed a new CPU tray. To my surprise I was able to get the original one working again by sitting outside in California sun with a magnifying glass and a pin and bending the pins back up. It was working with the 2.26s before I needed the Heatsinls for the 5680s. SO, I'm going to buy more Heatsinks and see what happens. If I can get them torqued exactly right again, I 'll have a spare Octo 2.26 tray for quick back to back benches.
I already ran Cinebench OpenGl test and finally got a decent score, showing how CPU dependent that test is.
In short, upgrading a 3 year old Mac into something faster than you can get at the Apple Store for $6,199 (!!!!!!!) was very rewarding.
The day and a half spent jacking around with screw tightness and buying a new unneeded CPU tray, researching an e758, disassembling that, modding it, etc....not so much fun.
|Jan 11, 2013, 11:31 PM||#54|
16 self-built multi-OS CUDA rigs and 4 Mac Pros with a total GTX Titan RD Octane Rendering TE of >55. AKA-TheRealTutor. Benches: CB11.5-48.5; CB15-3,791; GB2-58,027; GB3-71,691; LuxMark/Sala-12,330.
|Jan 12, 2013, 12:08 PM||#55|
MVC, thanks for sharing your interesting experience with us. One thing I have learned from my own upgrade of the 2009 DP model is that without the protection of a clamped CPU cover, the CPU is stuck to the heatsink surface as soon as it is laid down over the CPU socket. Therefore any lateral movement of the heatsink should be avoided before and during tightening the screws. The same also applies when loosening the screws. Those pins in the CPU sockets are rather fragile and could be bent easily due to heatsink lateral movement or overtightening.
|Thread Tools||Search this Thread|
All times are GMT -5. The time now is 07:06 PM.