Dual Core Opteron Thread #2

[QuietIce]

Agreed if that's the LOAD temp.

Toledo, Denmark - no difference as long as it's 2 Mb of L2 ...

[cgs26]

im feeling a bit funny about this one, got it today, bought it off ebay, dont think its been oerclocked as the guy i bought it from dident even know what stepping it was/ what the stepping code is, and he has owned it from new, however, 260 x4 x10 @ 1.35v and 1:1 mem settings (200mhz,un divided, 1t, 3448, gskill pc4000), fine, about 46C in orthos, however as soon as you try for 270 x3 x10, you need 1.4v to get it stable (im still unsure of that as im testing right now, but its an hour in and looking ok) and im running around 52C!!!!, is this acceptable? im on air, Arctic Cooling Freezer 64 Pro V2. i thought this was a 3.0ghz chip, fair enough thats probably on water, but i was expecting atleast 2.8 @ 1.4V, at 280 (x3 x10), it does boot into windows ok, and is fine for browsing, but hates orthos, crashing after a few seconds, HELP!!!!
Cheers, Craig

[cgs26]

i know its early to judge, but it would appear another chunk of this mystery has appeared, its now running 2.6 @ 1.325, so god knows why it takes 1.4 to get 2.7 stable..., temps jumped to 54, and i wasent prepared to leave it unonitored, so i turned it down, i might se if it boots into windows at 2.6, stock volts, that would be interesting to see. Cheers, Craig

[QuietIce]

Not all 0617's are 3.0 GHz chips - I've seen several examples of 2.8 @ 1.5 vCore on water. The only older chip that seems consistent is the 0610, 2.9-3.1 GHz depending on cooling.


You might consider stripping it. Naked cores tend to run ~100 MHz higher than their IHS brothers ...

[Cheffy]

my cpu scales like that, only much higher up - 2.85 @ 1.35, 2.87@1.375 2.92 @ 1.425 3@1.55 3.1 @ 1.65+ (becomes very temp dependant here)

it just seems to be the way the X2's scale. suddenly, and badly

[Mr Frustrated]

Hello!

I signed up because of this excellent thread and it seems I'm just in time to follow on from someone who's had similar experience...

My chip is a CCBBE 0615EPMW (one of the "good" opteron 1xxs).

I can get to around 2.58 gHz stable on stock voltage (will play games at a little faster, but his is the lowest I've not been able to find any errors), and up to around 2.75 gHz at 1.5v. I daren't push it any higher on my cooler.

I noticed was that, when tying a copy of SP2004 to each core (setting affinity in task manager), one of my cores crapped out nearly 90 mHz sooner than the other. The less capable core also registers significantly higher temperatures than the better one (have tried reseating the HS and using different thermal pastes - no change).

Amazingly (and confusingly), looking at individual core temperatures I can see that a core's temperature goes up by more when you load the OTHER core than when you load the core itself! Anyway ...

My memory is running well below its top speed, and so is my fsb. I'm going to try and "burn in" the lesser of the cores (at over a year old, I've read this won't work) and I'm experimenting with my board's VCPURef voltage (alters the "high-low voltage"[??]). It'll forever bug me if I can't get this thing rock solid stable at 2.6 gHz at stock!

P.S. My lesser core does better in torture tests and goes a few mHz higher when the other core is loaded too - I've noticed that CPU voltage readings go up when this is the case. Anyone else ever noticed anything like this?

[QuietIce]

Hello!

I signed up because of this excellent thread and it seems I'm just in time to follow on from someone who's had similar experience...

My chip is a CCBBE 0615EPMW (one of the "good" opteron 1xxs).

I can get to around 2.58 gHz stable on stock voltage (will play games at a little faster, but his is the lowest I've not been able to find any errors), and up to around 2.75 gHz at 1.5v. I daren't push it any higher on my cooler.

I noticed was that, when tying a copy of SP2004 to each core (setting affinity in task manager), one of my cores crapped out nearly 90 mHz sooner than the other. The less capable core also registers significantly higher temperatures than the better one (have tried reseating the HS and using different thermal pastes - no change).

Amazingly (and confusingly), looking at individual core temperatures I can see that a core's temperature goes up by more when you load the OTHER core than when you load the core itself! Anyway ...

My memory is running well below its top speed, and so is my fsb. I'm going to try and "burn in" the lesser of the cores (at over a year old, I've read this won't work) and I'm experimenting with my board's VCPURef voltage (alters the "high-low voltage"[??]). It'll forever bug me if I can't get this thing rock solid stable at 2.6 gHz at stock!

P.S. My lesser core does better in torture tests and goes a few mHz higher when the other core is loaded too - I've noticed that CPU voltage readings go up when this is the case. Anyone else ever noticed anything like this?

Which version of CoreTemp are you using? The latest "stable" (still beta) is v0.95.3 IIRC. Older versions sometimes reversed the cores.


What are you using for cooling? Also, is the IHS still on? (If so, why? )


Odd that your vCore will go up under load - usually it goes down slightly. Are you using CPU-Z, Everest, or something else ...?

[Mr Frustrated]

Well, this is certainly interesting!

Thanks for pointing me towards CoreTemp - I'd been using RMClock's core temperature readings to check temperatures. There were the same numbers that CoreTemp is giving me ... but for the wrong cores. Now, the right core is shown to heat up the most when it's loaded.

With voltage management in place (1 gHz, 1.1v) I'm getting readings of 15C and 25C for the cores. 15C seems absurdly low. That's probably lower than room temperature. This is for the worse of the cores too.

This reminds me actually that occasionally in RMClock's temperature graphs the cooler core would occasionally spike up to 68C (instanly) then drop down again (instantly) to some very low level. Had put it down to being a bug in the software or the A-bit uGuru bios stuff. Oh dear, what's actually going on ...

My cooler is an XP-90 (originally bought for my 3000 Venice) and to measure voltages I use the A-bit uGuru tools and CPU-Z. They both show the same thing. Will check out Everest ...

Oh yeah, forgot to say, I'm too chickenly to remove the IHS. I'm scared my big, clumsy fingers would bend pins while I'm trying to pry the IHS off and then I'd crush the chip trying to lever the heatsink back on (the XP-90 kind of rolls on from the back before being clipped down at the front). If I could work out how to do it safely I might have a crack at it though.

[Mr Frustrated]

Everest confirms - as the load goes up, so does the voltage. No management software is running that could make this happen. 1.37 (should be 1.35) becomes 1.38 with one core loaded, and 1.38 - 1.40 with two cores loaded.

With both Core loaded by Orthos, Core 0 = 40C and Core 1 = 50C. Puzzling.

[QuietIce]

Most uneven core temps come from one of two sources:

1. The temp sensors in the K8 chips are not officially supported and, as such, the sensors are not tested/calibrated at the factory. While this is usually just an index error that applies to both core's sensors there are a few rare instances when the sensor readings on the same chip are different even if the actual temps are the same.

2. The IHS application can also cause uneven core temps. While the IHS application is always good enough for normal operation not all IHS's are equal. Some just don't get as good a contact with the chip from side-to-side resulting in uneven cooling of the chip between cores. This is another reason many of us remove the IHS, in addition to the increased cooling efficiency we get from direct chip contact.



I know removing the IHS is scary - I've been there. The real secret for first timers is taking your time and removing a small bit of the adhesive at a time so you don't end up cutting into the electricals. I went around my first chip four times before even trying to "punch through" the adhesive and then I did it on the corners first and took it real slow. Seemed to take forever but it worked like a charm. By my third chip it was easy, but I don't get over-confident either - that's a sure way to ruin!

A good way to avoid damaging the pins is to put the chip back in it's shipping case. That hard black foam combined with the plastic case protects them very well while you're holding the chip - but try to keep your fingers off them anyway, just to be safe. Use the sides to hold it when possible ...

[Cheffy]

some boards will overvolt on load, I-wills were loved by all for this FEATURE! it played havoc with the +5v rail however (KK266 anyone remember that bad boy?)
its not gonna cause harm. temps are not to bad, certainly not worth worrying. as said, the probes are not to be considered "accurate"

[Mr Frustrated]

Unfortunately I passed on the Opteron packaging to the guy who bought my Venice, so that option is out for "de-lidding"! I'll have to find something else ...

Thanks for all the advice folks. Seems I may never get a good overclock out of this thing but it's all good fun.

[tojo15]

Hi everyone. I just received my Opteron 170. The stepping is LCB9E 0704TPMW. I was wondering if someone has this stepping and if it is a good overclocker. So far I got it to 2.6 on default voltage stable, but before i push it any higher I gotta get a better heatsink.

[QuietIce]

I've the exact same chip in an Opty 165 flavor running 2880 (320x9) @ 1.40 vCore. The stock 4-heatpipe Opty cooler w/stock TIM is giving me +30°C load over ambient.

I'm sure it'll go higher w/WC ...

[biohead]

Hi everyone. I just received my Opteron 170. The stepping is LCB9E 0704TPMW. I was wondering if someone has this stepping and if it is a good overclocker. So far I got it to 2.6 on default voltage stable, but before i push it any higher I gotta get a better heatsink.

had a 170 lcb9e 0704xpmw, did 3200mhz 1.35v on air.

edit: oops, meant 180.

[tojo15]

had a 170 lcb9e 0704xpmw, did 3200mhz 1.35v on air.

No way... I wouldn't believe that. Basically on stock voltage 3200mhz. Thats gotta be a record on stock voltage. Show some proof...

[QuietIce]

I can believe it - I've seen what mine will do with the stock cooler ...

[knopflerbruce]

I can believe it - I've seen what mine will do with the stock cooler ...

He had an Opty 180 with the same stepping that was very good. Maybe he was talking about that one.

Need some screens for that 170, yes.

[biohead]

He had an Opty 180 with the same stepping that was very good. Maybe he was talking about that one.

Need some screens for that 170, yes.

oops lol. i meant 180.

i said 'had' because i'm not running the rig in sig for a while now. mobo dead and @ rma, taking ages. socket ate a few pins off the 180 when i took it out, that was a great experience :/ would probably run again if I could find someone to fix the pins for cheap. any ideas?

been on a lappy for a while now. i'm fine as long as i can visit XS when mobo gets back i still have a 170 ccbwe 0550upmw 2.8ghz to pop in.

[squito]

Just playing around with a recently purchased NF4 Ultra-D (disappearing 939 boards made me do this)
3120 (240x13) 3-3-3-5

http://valid.x86-secret.com/show_oc.php?id=287078

http://img179.imageshack.us/my.php?i...n3120xsdo5.jpg

[Pisu87]

Just playing around with a recently purchased NF4 Ultra-D (disappearing 939 boards made me do this)
3120 (240x13) 3-3-3-5

3120 MHz at 1.325 V? That chip's gotta be from wafer downtown

[squito]

3120 MHz at 1.325 V? That chip's gotta be from wafer downtown

We had to bump it to 1.375 for absolute stability (prime and gaming) - will update sig when the dust settles

[boblemagnifique]

Opteron isn't Dead !!!




LCB9E 0742 is very very recently , the last Opty ???



I don't test , not for me

[Lastviking]

Opteron isn't Dead !!!




LCB9E 0742 is very very recently , the last Opty ???



I don't test , not for me

When will you send them to me then?, its good weather outside...cold....

[boblemagnifique]

When will you send them to me then?, its good weather outside...cold....

Exchange the €€€ , why not