6282 ES CPUs won't oc

What I've got here (an example)...
As you can see; both EPS are served by single rail.

img_2144.jpg
 
Last edited:
I finally got hold of a Corsair 1200AX and some ECC/reg RAM which -alias- kindly let me borrow.

I tested with the new PSU and the same old Kingston RAM, no change, except that the PSU didn't turn off when I raised vcore above 1.2. With the Chieftec PSU it turned off, with the Corsair it just rebooted.

No improvements with the ECC RAM either.

I also tried tear's new TPC rc, no difference (thanks tear for putting in all that hard work, btw).

So, my 4 8262 ES on a Supermicro board won't overclock at all. They run for maybe 20 minutes at 2.7 and 2.8, but then the rig reboots. All frequencies above that, the machine reboots the moment all CPUs are at 100% load.

I can't help to feel that I must have overlooked something trivial. From what I have read, most people have no problems overclocking their Interlagos cpus.

Could someone please read through my setup in the first post to see if I am doing something wrong (especially the BIOS settings). I am totally new at this, so I don't really know what I'm doing. :confused:
 
How are you trying to OC the chips? The easiest way to OC those ES chips is by changing the multiplier, this will be a setting in the bios.
 
tkam, there is no setting in a server board bios to change the multiplier.

This is the reason for Tear's BIOS for non-ES chips (to change FSB) as well as TurionPowerControl for ES chips (to change the multiplier)
 
Are there alternative ways of using TurionPowerControl other than the one I described in the first post?
 
Pretty the ones listed are pretty much proven. Have you tried running a single chip at the time and OC it?
 
No, but that's a good idea. Maybe there is a rotten apple in there. Will do, Thanks!
 
I'd also recommend checking ht-retries now that symptoms have changed (you may wish to add -i 1
parameter for quick refresh).

Script's at http://darkswarm.org/ht-retries.sh
(remember to chmod +x and call it via sudo)

I presume now it takes longer for the machine to collapse?
 
Yes, now it doesn't go down immediately, but more like a minute or two, after thekraken has gotten all the cores.

I am currently in the process of testing a single cpu as per 402blownstroker's suggestion. With just one CPU, I can oc to 3GHz (didn't go further), and it runs for about 30 minutes before it rebootes. Don't know what to make of that.

Will try ht-retries.
 
Last edited:
ht-retries -i 1 displays this over and over again:
Cumulative, Thu Dec 20 01:07:43 CET 2012 (1355962063)
L0S0 L1S0 L2S0 L3S0 L0S1 L1S1 L2S1 L3S1
Node 0
Node 1
Node 2
Node 3
Node 4
Node 5
Node 6
Node 7

Delta, Thu Dec 20 01:07:43 CET 2012 (1355962063)
L0S0 L1S0 L2S0 L3S0 L0S1 L1S1 L2S1 L3S1
Node 0
Node 1
Node 2
Node 3
Node 4
Node 5
Node 6
Node 7
Click to expand...
 
Tested all four cpus in socket 1. Same behaviour for all 4: Runs fine @3GHz for about half an hour, then the server reboots.
 
Tried to use win 7/server 2k8? Just to see if the same thing happens there. If you don't experience any reboots it's probably related to ubuntu.
 
Nilsch said:
Tested all four cpus in socket 1. Same behaviour for all 4: Runs fine @3GHz for about half an hour, then the server reboots.
Click to expand...
Was only one socket (socket 1) populated during these tests?

I'd also recommend checking dmesg for Machine Check Exceptions/Hardware Errors
(after FAH starts but before crash).
 
Yes, only one socket was populated during tests. Will check dmesg, thanks.

knopflerbruce: I don't have Win server, and I don't think Windows 7 will utillize the 16 cores.
 
Last edited:
win 7 Pro or above will, win 7 home won't.

edit:

Actually win 7 home would see one processor and I think it would see all all 16 cores. Win 7 pro will do dual processors.
 
Last edited:
With one CPU populated, memest through one complete pass. With 4GB or 8GB it should not take too long. If that is clean I am guessing bad motherboard or 4 bad CPUs. Any chance you have a non-ES CPU to try? A non-ES CPU help determine if the MB is bad.
 
That's under load, yes. I got two new motherboards at the same time, one was nothing but trouble (wouldn't recognize more than 16GB RAM and fan sensors didn't work). I will get a replacement within the next two weeks. When I get that the replacement, I will try oc on that mobo instead.

I haven't run memtest, but have tried 3 different sets/types of RAM and ECC/reg RAM.

I do however have a set of 4 6128, which I believe are non-ES. Can I run TPC the same way with those?
 
Nilsch said:
I do however have a set of 4 6128, which I believe are non-ES. Can I run TPC the same way with those?
Click to expand...

No. Try one as-is and see if it runs stable. If it does, try all 4 and see if it runs stable. If they do, then the MB is mostly good. If not, looks like that MB will need to be RMAed also.
 
The CPUs run stable as long as I don't try to oc. They ran for two weeks uninterrupted at stock speed (2.6) before I decided to overclock. They are still stable at stock speed. Come to think of it, this is actually the mobo I ran the 6128s on. The board I put the 6282 in was defect and I am waiting for a replacement. Switched the 6128 with the 6282 as they give more PPD while I wait.

Is there any way I can make them run full turbo without overclocking? From what I read, the 6282 CPUs should do 3GHz over all cores in turbo mode. How do I start turbo mode?
 
Last edited:
In order to accomplish this, enable PowerNow and Turbo/CPB in the BIOS.
Then boot the machine and call TurionPowerControl -psmax 1.

Start folding (on all cores, ofc), then verify that all cores are continuously
in P-State 1. Use TurionPowerControl -CM for that purpose (always use 0.44-rc1).
 
THANKS!!! :) Power draw from the wall went from 600W to 800W and TPF from 11:55 to 10:32 for 8101, so it definetely did something right. :D

Now it got even more puzzling why the overclocking didn't work, but who cares, 3GHz is what I aimed for 24/7 anyway.

Again, many thanks!!!
 
Nilsch said:
THANKS!!! :) Power draw from the wall went from 600W to 800W and TPF from 11:55 to 10:32 for 8101, so it definetely did something right. :D

Now it got even more puzzling why the overclocking didn't work, but who cares, 3GHz is what I aimed for 24/7 anyway.

Again, many thanks!!!
Click to expand...

You should try lowering the voltage. Default for 6282@3000 is ~1.23 . Mine are running 3GHz @ 1.15. Power draw dropped almost 100W from default (less heat and fans run slower=less noise also).

EDIT: It's also not linear (just like IvyBridge). My chips would not run 3000 @ 1.20v or 1.875 (but they would @ 1.2125 and up). They would run below that (they will run 3000 @ 1.1 for Prime95, but FAH crashes, 1.15 is the lowest I can stably run FAH). IvyBridge does something similar. I have a 3770k hackintosh that runs at 4.3Ghz on 1.10v (stock max turbo is [email protected]). It would not run 4.3 if voltage was between 1.12 and 1.18. There must be a point where the temps get too high internally, requiring more voltage to make stable, whereas if voltage is low enough, heat is less and it will be stable. Just a guess, but bottom line is it's worth trying seemingly improbably under clocks, it might just work.
 
Last edited:
I started out even lower than that, but with the Chieftec PSU. Worth trying. Right now I don't mind the extra heat, as it is -6C (21F) outside...
 
You must log in or register to reply here.
Back
Top