AMD Opteron 6000 series issues

B

bot

n00b
Joined
Aug 12, 2014
Messages
6
I just tried updating a dual CPU system with two 6176 Magny-Cours CPUs to two Interlagos CPUs (one ordinary 6274 and one 6274 'Engineering Sample' edition).

Firstly, I get past BIOS but the system won't boot with both of these CPUs inserted, it just resets on boot. When installing only one of the Interlagos CPUs in the first socket the system boots up fine, but not with both of them together. It works with both CPUs separately but not with them together in each CPU slot. The system works fine with both of the old Magny Cours chips installed.

Secondly, the first two DIMM-slots of the first socket are now dead. That means that no memory modules inserted in those slots are found, either by the BIOS or the operating system. This applies to both Interlagos chips when they are used separately. Also, when both of the Interlagos CPUs are plugged in, the BIOS will detect all memory modules except the modules in the first two DIMM-slots of the first CPU. So, the CPU in the second socket doesn't have any problems detecting the modules. Swapping memory modules between slots doesn't change the situation. When changing back to the old Magny Cours CPUs all memory modules / DIMM-slots work normally again.

Does anyone know what's going on?
 
Distributed computing section has great troubleshooters for these processors.
My gut says possible power supply issues. Just a guess.
 
But why is the first pair of DIMM-slots dead even when the motherboard holds only one of the 6274 CPUs? It can't be the motherboard because they work fine with the older Magny Cours chips. It can't be a "different stepping" issue because the CPUs are separate and don't have anything to do with each other in such a setup, it can't be the memory modules because they are dead no matter what memory module I insert into them. These slots are dead with both of the 6274 so this issue is not isolated to one specific CPU.

The power supply is a brand new 1050W Seasonic, so I can't see how the PSU would cause any issues.

I've looked into the distributed computing sections and the DC guides but cannot find any useful troubleshooter. I did use the search function as well but with no relevant results.
 
Yeah, I've had brand new power supplies have issues. Usually takes me weeks to come to that conclusion. Because it's new :)

A lot of the heavy troubleshooters hang out in chat. Worth a try. They really are pretty good.
 
dwdawg said:
Yeah, I've had brand new power supplies have issues.
A lot of the heavy troubleshooters hang out in chat. Worth a try. They really are pretty good.
Click to expand...
I'm sorry, what do you mean by "chat"? There is no subsection in these forums named "chat" and I cannot see any chat functionality in these forums. And how would I identify said troubleshooters?
 
IRC. The info is in the DC forum.
The guys you might want to talk to are Tear, Musky, Grandpa.

Or just post the original request you have in the DC forum.
 
bot said:
But both of the newer CPUs are of the same model (6274) and generation (Interlagos), how could they possibly have different steppings? I cannot see more than one stepping in e.g. the following information page:

http://www.cpu-world.com/CPUs/Bulldozer/AMD-Opteron 6274 OS6274WKTGGGU.html
Click to expand...

A lot of the time not all ES steppings are known. Trust us, an ES CPU will not have the stepping as the retail one.

Steppings do matter as they will have different revisions of the processor die on them, as the die gets closer to production more of the errors are fixed. An early ES will have a lot more errors than a retail cpu, which is usually why they do not work together.

Intel recommend that CPU's are of the same stepping and state that steppings must be now more than one apart e.g C1 & C2 etc. Not sure about AMD but it will probably be the same.
 
Thanks for your replies. I have now bought another copy of the retail version of the 6274. I checked that those two retail 6274s have the same stepping and now they run just fine together.

They now also detect all memory. It makes me wonder whether it is the DIMM socket that is the problem. The sockets seem a bit loose and I can actually move the memory modules even though they are secured in the sockets.
 
not sure what youre asking about the dimms exactly, and maybe im incorrect, but arent the memory controllers on the cpu? As in, those "dead dimm sockets" just not active without the cpu in the slot associated with it? thats how it is on socket 940 opterons, though im dating myself saying that.
 
not sure what youre asking about the dimms exactly, and maybe im incorrect, but arent the memory controllers on the cpu? As in, those "dead dimm sockets" just not active without the cpu in the slot associated with it?
Click to expand...

You are correct. If you do not have a CPU in the second socket you can not use the DIMM slots next to that socket. There also may be devices connected to the HT bus on the second slot as well that will not work in single CPU configurations.
 
The motherboard has two sockets; CPU1 and CPU2. Next to CPU1 are the sockets P1-DIMM1A, P1-DIMM1B, P1-DIMM2A, P1-DIMM2B, ..., P1-DIMM4A, P1-DIMM4B. Next to CPU2 are the sockets P2-DIMM1A, ... ..., P2-DIMM4B. In short; each CPU socket has 8 DIMM slots denominated as above.

If I populate only CPU1 with one of them CPUs and populate two of the P1 DIMM slots with memory (usually P1-DIMM1A and P1-DIMM3A as recommended by the motherboard manual) the P1-DIMM1x sockets are dead. I tried both of them among an innumerable amount of combinations of DIMM slots that are connected to CPU1 while only populating the CPU1 socket.

If I populate both sockets with CPUs, the very same DIMM slots work. In fact all memory slots work.

Yes, I'm fully aware that the memory controller is in the CPU. But the slots are just as dead nonetheless. Just like the parrot in that John Cleese Monty Python sketch.
 
You must log in or register to reply here.
Back
Top