Another Opteron 165 Thread 0547 tpmw

[nealh]

Originally posted by: Lark888
I'm declaring victory on the 2700 Mhz speed 9x300 for the Opteron 165. Played BF2 for hours last night and then ran 12 hrs 21 minutes of dual Prime95 without a glich. Even had the low speed rpm on CPU cooler XP-90.

By the way, my low temperatures on the DFI Chipset are from a modified Alpha Pal CPU cooler that I chopped up to fit around an ArcticCooling Silencer video fan. Overall sound is very low on this setup and can only be heard in a quiet room.

Thanks to the OP for posting - I was holding off on getting an Opteron and am very pleased with performance/value of this one.

Info on Stability fun

damn sweet chip low vcore as well...I am impressed

you got IMHO once of the golden chips

 

[Chumster]

So I finally received my chip (0547 TPMW) and have been able to get 2500 (278x9) for over an hour with dual Prime95 running (still running the test) at stock voltages. Thing is, SmartGuardian is reporting the CPU temp @ 35C. This seems too low under load, even at stock voltage. At idle, I've found that it's about 2-4 degrees off from the BIOS. Is this normal? I installed the nTune Monitor which reports the same temps.

In case you're curious, I'm using an SI-120 with a Cooler Master 3 blade (~32 CFM). The setup is currently out-of-box and setup in open-air on my desk. Ambient temp is about 20C.

 

[The Sauce]

I'm running 2.6 GHz @ 1.5V and getting max load 38C temps with an XP-90. Your temps are right.

 

[lopri]

Chumster, what's the CPU-Z reading when you set stock (I'm assuming Auto x Auto) volrage?
Your chip looks good to be 2.7GHz @~1.4V. 2.8GHz @~1.5V.

 

[Chumster]

lopri: I'll check once I finish with my current Prime95 run. I tried 2610 (290x9) at stock and core 0 decided it wanted to take a nap after 30 minutes. So I bumped Vcore to 1.4 and am through about 6 hours of priming (large). Temps are still reporting low at 38C under load but I'm not complaining.


[edit] Forgot to ask - anyone have anything to say about running small vs. large in Prime95? I understand that "Blend" isn't useful, but which of the other two is best?

[edit 2] lopri: My Vcore on Auto/Auto displays as 1.344 in CPU-z. I just finished about 9 hours at the above clock. Going for 9x300 next....

 

[lopri]

SmallFFT for pure CPU stability
LargeFFT for CPU+RAM stability

So basically for max CPU OC I use SmallFFT. Memtest for memory max. SuperPI for memory configuration. (timings et. al) LargeFFT for CPU+Memory stability.

 

[Chumster]

I figured it was something like that given the description. Funny thing is that my PWMIC temps are about 3-4 degrees warmer when I run LargeFTT.

My chip wouldn't do 2700 (300x9) @ 1.425Vcore, so I went ahead and bumpbed it to 1.425 * 104%. It's about 3 hours SmallFTT stable so far - looks like my 2700 might be your 2800. Once I get 8 hours under my belt, I'll see if I can push it more.

Chum

 

[Job]

I am on the verge of buying the Opteron 165, but according to the specifications on the website, this chip requires registered ram. Is this entirely necessary? I've read some posts earlier in this thread where people have used un-buffered ram with their opterons, and I really don't wanna shell out the extra cash for needless ram!

Help!

 

[Chumster]

Originally posted by: Job
I am on the verge of buying the Opteron 165, but according to the specifications on the website, this chip requires registered ram. Is this entirely necessary? I've read some posts earlier in this thread where people have used un-buffered ram with their opterons, and I really don't wanna shell out the extra cash for needless ram!

Help!

The 940 pin Opterons require registered memory (as well as the 940 pin FX chips). The 939 pin Opty's, however, do not. This includes any of the single core series (14x) as well as the dual cores (165, 170, 175 & 180). Hope that helps!

Chum

 

[Job]

That's a massive relief! Thanks!

 

[Job]

Just another thing (that I can't seem to find on any sites):

I'm moving from an Abit AV8 AGP motherboard to the DFI Sli variety, will I require any power adapters for the new board and PCI-E card or will the regular ATX and Molex power connectors do? I keep reading that Sli boards and pci-e cards use different power connectors, but I never find any specific relation to the connectors I already have

Many thanks in advance,

 

[The Sauce]

Just so you guys know I went at my board with a digital multimeter cause I was concerned over the differences between the software reported voltage (CPU-Z) and the BIOS voltage. Just so you all know - DONT TRUST THE REPORTED VOLTAGE in cpu-z or any other software. When I had BIOS set to 1.500, at idle my multi was reading 1.500 and CPU-Z was reading 1.456. With full load measured Vid went up to 1.530 and CPU-Z stayed at 1.456. The reported Vcore is garbage. Dont trust it.

So in my experience here under full load the board does overvolt.

 

[Lark888]

I was also going to use meter to check the voltages but couldn't find the test points. I recently saw an excellent photo layout showing where to measure voltages and, of course, now I can't find it.

The CPU is running very well. I may try to go up a notch or two again after some additional burn-in. I was able to play BF2 for a few hours at 9x305 at same voltage last night.

What really surprised me was that my Corsair memory clocked much higher than I was able to get before. Once I had established 9x300 on the CPU, I started to select different memory dividers and was successful at 300Mhz with 2-2-2-6 timing. With my last setup I only ran at 250Mhz. May be the memory controller on the die is also better than my 3200+ Winnie.

 

[mrkun]

Originally posted by: Lark888
What really surprised me was that my Corsair memory clocked much higher than I was able to get before. Once I had established 9x300 on the CPU, I started to select different memory dividers and was successful at 300Mhz with 2-2-2-6 timing. With my last setup I only ran at 250Mhz. May be the memory controller on the die is also better than my 3200+ Winnie.

DDR 600 at 2-2-2 timings? I think you made a typo.

 

[Lark888]

I agree that there is something amiss here. I think more than a typo - but trying to figure this one out. I can't take the PC down tonight to do some checking. I did not intend to set up and run at FSB speed. CPUz reports out memory at 300 Mhz. It doesn't make sense to me either. Probably a bug in CPUz. However, the couple memory speed benchmarks have been very nice regardless of the true speed. I've probably got this goofed up but don't have time to sort it out tonight.

 

[mrkun]

Originally posted by: Lark888
I agree that there is something amiss here. I think more than a typo - but trying to figure this one out. I can't take the PC down tonight to do some checking. I did not intend to set up and run at FSB speed. CPUz reports out memory at 300 Mhz. It doesn't make sense to me either. Probably a bug in CPUz. However, the couple memory speed benchmarks have been very nice regardless of the true speed. I've probably got this goofed up but don't have time to sort it out tonight.

You should just go by what your BIOS settings are. If you're on a 2/3 divider, that would make sense.

 

[lopri]

Originally posted by: Snatchface
Just so you guys know I went at my board with a digital multimeter cause I was concerned over the differences between the software reported voltage (CPU-Z) and the BIOS voltage. Just so you all know - DONT TRUST THE REPORTED VOLTAGE in cpu-z or any other software. When I had BIOS set to 1.500, at idle my multi was reading 1.500 and CPU-Z was reading 1.456. With full load measured Vid went up to 1.530 and CPU-Z stayed at 1.456. The reported Vcore is garbage. Dont trust it.

So in my experience here under full load the board does overvolt.

I do believe your results and knowing the competition between the mobo manufacturers, it's also quite expected. But however, I still have this one mystery that I can't seem to understand: CPU-Z

OK, take a look at this screenshot closely.
http://img59.imageshack.us/my.php?image=cpuz1vx.jpg

My two boards, A8N-SLI Premium and DFI SLI-D, use same 'sensor' chip for volt, temp, etc - ITE 8712. Thing is, on the ASUS board, the VCore reading in CPU-Z literally dances. Like from 1.36 to 1.52 to 1.44 to 1.23 to 1.62, etc. You get the idea. This voltage dancing also shows up in Asus Probe, also.

But on DFI board, I hardly see this voltage fluctuations. They're usually rock solid. Oh, btw, we're talking about DUAL-CORE CPUS. I've heard single-core CPU's VCore's are OK on Asus board.

If both DFI and ASUS use same ITE 7912 chip for sensoring temperatures/vcore/voltages, etc. then I don't quite understand why on ASUS platform the ITE chip seems refelct the voltage flucuation so well, but not on DFI board?

 

[nealh]

Originally posted by: lopri

Originally posted by: Snatchface
Just so you guys know I went at my board with a digital multimeter cause I was concerned over the differences between the software reported voltage (CPU-Z) and the BIOS voltage. Just so you all know - DONT TRUST THE REPORTED VOLTAGE in cpu-z or any other software. When I had BIOS set to 1.500, at idle my multi was reading 1.500 and CPU-Z was reading 1.456. With full load measured Vid went up to 1.530 and CPU-Z stayed at 1.456. The reported Vcore is garbage. Dont trust it.

So in my experience here under full load the board does overvolt.

I do believe your results and knowing the competition between the mobo manufacturers, it's also quite expected. But however, I still have this one mystery that I can't seem to understand: CPU-Z

OK, take a look at this screenshot closely.
http://img59.imageshack.us/my.php?image=cpuz1vx.jpg

My two boards, A8N-SLI Premium and DFI SLI-D, use same 'sensor' chip for volt, temp, etc - ITE 8712. Thing is, on the ASUS board, the VCore reading in CPU-Z literally dances. Like from 1.36 to 1.52 to 1.44 to 1.23 to 1.62, etc. You get the idea. This voltage dancing also shows up in Asus Probe, also.

But on DFI board, I hardly see this voltage fluctuations. They're usually rock solid. Oh, btw, we're talking about DUAL-CORE CPUS. I've heard single-core CPU's VCore's are OK on Asus board.

If both DFI and ASUS use same ITE 7912 chip for sensoring temperatures/vcore/voltages, etc. then I don't quite understand why on ASUS platform the ITE chip seems refelct the voltage flucuation so well, but not on DFI board?

this probably do to the design and performance of the DFI..it was designed more for overclocking....I would assume they tested to provide most stable voltage delivery possible without fluctuation(do not forget the DFI Ultra D/SLI DR do overvolt underload which helps with stablity)

Asus does not seem to be as good mobo anymore for overclockers...at least from what I see

 

[Chumster]

Just a follow-up...

After successfully hitting 2600 without only a small increase in voltage, it was time to aim for 2700. My first attempt I set the voltage to 1.425 * 103% (calculated voltage ~1.47V). After about 5 hours I was ready to declare it a success but let it continue running to hit my personal 8 hour criteria. It's a good thing, because at hour 6 core 0 gave up the ghost.

I tried a few other Vcore+VID combos and eventually settled on 1.35V * 110% (calulated voltage 1.485V). I started it yesterday afternoon and successfully primed (Small) for 14 hours!

Screenshot - 9x300 - 1.35v *110% - 14 hours Prime stable (small)

Given how close I am to 1.5V, this will be the fastest running speed I'll use.

Chum

 

[lopri]

Chumster,
That is a great OC. Anyone w/ dual-core is dreaming of 2.7GHz!
What is your cooling and ambient temperature? I think you can still crank out a bit more clock.

 

[RideFree]

Originally posted by: Snatchface
Just so you guys know I went at my board with a digital multimeter cause I was concerned over the differences between the software reported voltage (CPU-Z) and the BIOS voltage. Just so you all know - DONT TRUST THE REPORTED VOLTAGE in cpu-z or any other software. When I had BIOS set to 1.500, at idle my multi was reading 1.500 and CPU-Z was reading 1.456. With full load measured Vid went up to 1.530 and CPU-Z stayed at 1.456. The reported Vcore is garbage. Dont trust it.

So in my experience here under full load the board does overvolt.

I'll bet they all do in order to attain the highest possible "cutting edge" in the statistical wars without crapping out.

 

[Chumster]

Originally posted by: lopri
Chumster,
That is a great OC. Anyone w/ dual-core is dreaming of 2.7GHz!
What is your cooling and ambient temperature? I think you can still crank out a bit more clock.

Specs are as follows:

Opteron 165 (CCBWE 0547 TPMW)
Si-120 w/Cooler Master Tri-blade (20db/32CFM)
2x1024MB OCZ PC4000 OCZ5002048ELGE-K
1x 74GB WD Raptor 10K
1x 250GB Seagate 7200.8
NEC 3550A DVD-RW
eVGA 7800GT CO (470/1100)
Enermax Liberty 500W Modular
Windows XP Pro SP 2
Antec P180 (Black)

Ambient temperature sits right around 20C, though my wife has been known to bump the temp from time to time. I may consider going higher, but my next step is coaxing my RAM to run at 9/10 (~270Mhz). BTW, lopri, you'd apprecaite my current setup.

 

[Eirias]

any luck with the ccbwe 0550upmw?

 

[Lark888]

Ambient temperature sits right around 20C, though my wife has been known to bump the temp from time to time. I may consider going higher, but my next step is coaxing my RAM to run at 9/10 (~270Mhz). BTW, lopri, you'd apprecaite my current setup.

Thanks for the humor. Clicked your link and immediately started laughing - I've seen that rig before.... . At least there's no increased temperature from the case(s).

 

[Lark888]

I located the test points for CPU voltages and took a couple of readings with my digital voltmeter. Although I've never "calibrated" my multimeter, it has given reasonable data in the past. My CPU VID is set at 1.45V in BIOS. CPUZ and MBM5 both report about 1.41V. Measured voltages are 1.454v at idle and 1.486 with dual prime95s running.

My readings are consistent with data reported earlier. BIOS set voltages are likely to be the real nominal ones and voltage gets a ~ 0.03 volt boost under load.

Also appears that CPUz and MBM5 are reading about 0.04v lower than actual idle voltage and do not show any voltage changes for load.

One other question! Why do some individuals select a VID*overvoltage factor versus the same voltage available a direct VID selection?