3.05GHz on Air: Update (Now inside a case with CNPS9500)

[lopri]

After much researching I got this chip and I'd like to share the results.

Socket 939 Opteron 165
XXXXX 05XXXXXX
XXXXXXXLXXXXX


First attempt was 9x300 on stock vcore (CPU-Z 1.312V).
http://img150.imageshack.us/my.php?image=2700smallffta8uv.jpg
http://img69.imageshack.us/my.php?image=2700smallfftb1vv.jpg

Having toyed with so many Opterons, I now get the feeling of a chip's potential at first boot. First boot and the subsequent 3Hrs. dual-Prime95 gave me a very good feel about this chip.
So I went ahead and bumped the HTT to 311, making it 9x311=2799MHz. Still on stock vcore 1.312V.
http://img150.imageshack.us/my.php?image=2799smallffta0qx.jpg
http://img150.imageshack.us/my.php?image=2799smallfftb2nc.jpg

So far there was zero failure on anything. No trial and error, no vcore adjustment, no reboot due to failure on anything. Just the memory divider @DDR100 and bump of HTT. (First attempt 300, then second 311)
At that point I realized that I was dealing with something serious. First I made sure that I wasn't getting any bogus results, and checked all the BIOS settings. Everything's ordinary. Hmm.. What is this?

I was like, what the hell, let's try 9x333 (=2997MHz) on stock vcore! Well, it actually booted. But voluntarily rebooted itself in 5 secs.
So now I decided to up the vcore until I achieve the stability. Here comes my magic combinations.. 1.392V, 1.424V, 1.456V, 1.488V, 1.520V... And the final result is..
http://img69.imageshack.us/my.php?image=2997smallffta9al.jpg
http://img137.imageshack.us/my.php?image=2997smallfftb3fg.jpg

Did I just set a new record?
1.392V was a no-go, and 1.424V let dual-Prime95 run for 30 mins or so, which means with better cooling, the chip has a potential to run 3GHz @1.424V.
For the record, here are the components used to achieve the results.

Socket 939 Opteron 165 (DO NOT ASK ME THE STEPPING, I WON'T TELL YA With some researching and luck, you can easily figure it out. Tip: The 3rd line on the stepping is very important.)
DFI UT NF4 SLI-D
Generic Samsung PC2700 DDR 512MB x 2
Gigabyte GeForce 6200TC 64MB
ThermalTake Big Typhoon
Antec True-Control I (First generation) 550W 20-pin
Ambient temp below 20C (70F)
Open-air testbed

This chip runs extremely cool even with warm (well, winter-warm, not summer-warm) daytime temperatures. 1000MHz OC on stock vcore sort of proves it. I could never get an Opteron 144 to run stable @3GHz, so it's very, very refreshing that I'm running a dual-core @3GHz with relatively low vcore and low temps. Also I believe this chip still has more headroom, but unfortunately my board's max HTT is 335~340. Anything above 335 results in instability, and it won't even boot above 340. Maybe it's time for a motherboard upgrade.

I've been through so many Socket 939 Opterons, and have gotten somewhat bored of it. (and spoiled, of course) This chip was meant to be my last one, so this result will be kind of memorable for me. My previous best dual-core OC was 2900MHz @1.52V with a 170.
For mere $285, I've got myself an FX-64 (or will it be FX-70?) which might not even be out till next year!

Now I'm *definitely* done with Socket 939 Opterons. (No more impulse-buy, that is) They are the most exciting hardware I've experienced, and will probably remain such. Long live the Socket 939 Opterons!

P.S. #1 Please don't ask me to sell this particular chip. (Unless you're willing to pay $1,000 for it )
P.S. #2 My other Socket 939 Opterons will be available for sale in near future. I will first contact the fellow forummers who have been communicating with me, then will post @FS/FT forum. (or eBay, lol)


FOLLOWING IS AN EDIT ON 2/12/2006. I posted separately first, but I thought some people might not want to read through all the posts.

Whit the incredibly immature replies I've got from XtremeSystems members (Including a moderator, who threatend my account status just because I did not tell the stepping of a damn CPU?!), I've decided to reveal the stepping of this chip, along with my un-finished (and will not continue to finish, due to the negative feedbacks that I've been almost shocked by) observation on Socket 939 Opteron Dual-Core Processors' stepping/substrate/batch #s. I personally think the immaturity @XS is just as bad as that of HardOCP, and AT is where grown-ups can talk freely, and responsibly. I :heart: AT.

My phenomenal Socket 165 on this thread has the following stepping information:

OSA165DAA6CD
CCB1E 0550VPMW
XXXXXXAL505XX

Many people guessed it right, and my impression on this stepping is that it's mean to be FX-60s but somehow failed AMD's validations. The fact this stepping isn't as consistent as some others when it comes to OC'ing potentials, also proves it. So again, it's basically a luck of draw. (Still there are people who are jealous about it.. Please folks, get over it! It's just a piece of sillicon!)

And below is the preliminary results from my observation. The steppings/substrates/batch #'s with the potential of extremely high OC (2.8~3.0GHz)

0530TPMW
0546XPMW
0550UPMW (170, not 165)
0550VPMW

Honorable mentions: 0544XPMW, 0546MPMW, 0547TPMW

And regarding the 3rd line:

XXXXXXXL501XX
XXXXXXXL503XX
XXXXXXAL501XX
XXXXXXAL505XX
XXXX7XAL50XXX
XXXX786L50XXX
XXXX408L50XXX

The above tends to show very good results. Don't ask me why, because I do not know.

Thanks for AT folks who've supported me and been patient so that I could offer something more meaningful than just a brag. I sincerely apologize if this post was offensive in any sense to any of you.

I originally intended to delve more into this stepping/substrate/batch # theory, but with all the negative feedbacks (from XS) and with my loss of interest in Opteron in general, I don't think I will carry it on. Hope even this little info could help some of you out and wish you all good luck. Happy OC'ing!

lop

2nd Edit on 2/12/06: LargeFFT

With a pair of loaner sticks (My OCZ4200s seem defective) I was able to do a LargeFFT. As expected, for higher mem clocks, it required a higher vcore and therefore higher temps.

RAM: OCZ 3200 Platinum Rev.2
http://img119.imageshack.us/my.php?image=largefftvcoretemp13zo.jpg
http://img119.imageshack.us/my.php?image=largefftvcoretemp27qa.jpg

Update (2/27/2006)
The CPU is now in my main rig as well as the board. I finally ditched the A8N-SLI Premium. After a while of thought, I mated the CPU with Zalman CNPS9500. While not as good a performer as the TT Big Typhoon, I thought the BT is somewhat too heavy for the board layout (CPU in the middle), and I'm getting a window panel for my Lian-Li V1000, so esthetic side played a role, too. The max dual Prime95 stable clock inside the case looks to be 9 x 330 = 2970MHz. I was able to get 339 HTT from my board for dual SuperPI 32M. Generally my board starts giving me troubles past 335 HTT. Oh and the memory is 2 x 1GB OCZ PC3200 Titanium.

Here is the dual 32M shot.
http://img386.imageshack.us/my.php?image=3050dualsuperpi32m8cg.jpg

 

[Makaveli]

 

[TraumaRN]

*pangs of jealous emotion*

 

[jimbob200521]

OMGWTFBBQ!!!!1!!11!!1!!

 

[JustStarting]

Originally posted by: lopri
I will first contact the fellow forummers who have been communicating with me, then will post @FS/FT forum. (or eBay, lol)

I think you have a few PM's from me??

 

[Shimmishim]

eh, 3 ghz is nothing special anymore

show me 3.2+ then i'll be impressed

 

[lopri]

JustStarting, I remember our conversations. I will PM you as soon as I get some time.

 

[lopri]

Originally posted by: Shimmishim
eh, 3 ghz is nothing special anymore

show me 3.2+ then i'll be impressed

I don't think I started this thread to impress you. And 3Ghz is special to me, and to some others, almost certainly. What's the motive/intention of your reply? I could think of a couple but I'd rather not mention them. :laugh:

 

[Elfear]

Nice results lopri. Especially the fact that it's Prime stable.

On a side note, I figured out my problem is the board. Won't boot into Windows with anything over 320HTT even with a 5x multi and 2 hours of tweaking. Recommendations on a board that will hit 350-375 reliabley? After two Ultra-D's that maxed out at the same speed, I'm needing something more.

 

[CP5670]

I think the newer revision AD0 boards can go quite a bit higher than the older ones. Mine does 400 comfortably, which was as high as I tried setting it.

And that's a nice OC. But what's with hiding the week/batch code?

 

[lopri]

Elfear: Actually I was gonna PM you soon about your problem, after my own experience of high HTT on the motherboard. Yes, I think it's your board is holding your chip most likely. A8N32-SLI seems to be the best bet for high HTT right now. Have you tried some weird LDT multies? Like x2.5, x3.5, etc. I once had a problem with high HTT with a 144, and playing with the LDT multi eventually lead me to fix the problem. Just like certain CPU multies work better than others, and certain vcore/special combos work better, with DFI board it seems like you have to play with every single setting to find out the optimal configuration. Very annoying.

 

[lopri]

Originally posted by: CP5670
I think the newer revision AD0 boards can go quite a bit higher than the older ones. Mine does 400 comfortably, which was as high as I tried setting it.

And that's a nice OC. But what's with hiding the week/batch code?

Is it possible to check the revision (AD0?) of the board before making a purchase? Or is it a different model (i.e. Expert)?

Oh yah.. I'm hiding the Stepping information to mystify my chip and in case I decide to sell it on eBay.. Well, but it's basically draw of luck as you know. Even if you know what to look in the stepping, you won't get to see them until you pay for it. I was lucky to get a chance to look at the steppings before I make a purchase, so I picked the one (in this thread) based on my knowledge from here and there.

 

[Elfear]

Originally posted by: lopri
Elfear: Actually I was gonna PM you soon about your problem, after my own experience of high HTT on the motherboard. Yes, I think it's your board is holding your chip most likely. A8N32-SLI seems to be the best bet for high HTT right now. Have you tried some weird LDT multies? Like x2.5, x3.5, etc. I once had a problem with high HTT with a 144, and playing with the LDT multi eventually lead me to fix the problem. Just like certain CPU multies work better than others, and certain vcore/special combos work better, with DFI board it seems like you have to play with every single setting to find out the optimal configuration. Very annoying.

Thanks for the advice lopri. I tried 1x, 2x, and 3x LDT multi. Nothing seemed to help. I think I tried about everything I could. Maybe I'll give the weird LDT multipliers a try. Can't hurt I guess.

Almost all of the new Ultra-D boards from the last 3-4 months have been revision ADO. That's what my latest board is (first one was an ABO) but it seems to hit a ceiling at the same spot.

Are the A8N32-SLI boards all that good? Seems like all the records over at XS are being broken with DFI's.

 

[lopri]

I would even try Half-CPU multies. You know, those x7.5, x8.5, x9.5, etc. I really have no idea regarding those 2 boards. Between A8N-SLI Premium and NF4 SLI-D, it was easy. If you *actually* SLI, go with the ASUS. If you're more into CPU OC, go with the DFI. But even with the DFI's excellent NF4 series, these potent Opterons can't get enough HTTs. Who would've thought that 340HTT is not enough to fully realize a CPU's potential, before these Socket 939 Opterons came out? Between A8N32-SLI and DFI NF4 Expert, I'm exactly 50/50. Only problems with the ASUS board seem the lack of usable PCI slots and not-yet-perfect BIOSes. Oh and another thing is, A8N-SLI Premium (and A8N-SLI Deluxe) has a terrible vcore fluctuation problem with dual-core CPUs. If they have fixed that problem, I'd get the ASUS. But on the other hand, I'd like to know which board can raise HTT higher. Does anyone know?

 

[dguy6789]

Is there any particular reason you won't tell us the cpu stepping other than the fact that you feel a bit more secure with the thought that you know it and we don't?

 

[PremiumG]

Originally posted by: dguy6789
Is there any particular reason you won't tell us the cpu stepping other than the fact that you feel a bit more secure with the thought that you know it and we don't?

yeah really

 

[GOREGRINDER]

Originally posted by: lopri
But on the other hand, I'd like to know which board can raise HTT higher. Does anyone know?

a few sites say the dfi lanparty sli-dr expert "purrs" along at 250 with a 5x htt multi still intact,..lol,.. hows that for bandwidth headroom?,.. closeing out at around 370.but that usually becomes more flexable as bios updates hit

heres one PCStats overclocking results

 

[RichUK]

That is absolutely amazing. I don?t care what anyone says, to get 3Ghz stable from a 165 at 1.456 is just immense. lopri, that is one hell of a chip you have there, I mean some FX60?s like Anands sample couldn?t go past ~2.8Ghz.

I think this chip should be kept it is not worth selling, it is definitely a keeper.

I must say lopri you are very lucky that you get to select your chip?s before you buy, all of these good chips you have had, and I?ve been lucky to purchase my Opty 148 from you

Just a side not, my board the other day was able to post at 360HTT, and boot into to windows at 350HTT, and I don?t even have the ability to increase the HTT voltage :shocked:. The MSI has done me proud, but it is time to go, a new board is on the horizon

 

[Greenman]

Whats the point of not telling us the stepping? To be honest, it seems a bit childish to me.

 

[n7]

Wow...

 

[sparkey]

i would love to know the stepping but his chip his rules i gotta respect him for that
great chip, really wish it were mine but i'm happy with the two that i have
congrates (you lucky devil)
tom

 

[TheRyuu]

The stepping is either a 550, 551 or 552.

I have yet to see 552, but ya never know. And what is the point in not telling us the stepping? Afraid to loose your bragging rights?

 

[potato28]

:Q :thumbsup:!!!

 

[nealh]

This an amazing result..do not forget my PMs too

BTW the vcore is cpuz which under reports the vcore from dfi NF4 mobos....but with that said I bet his vcore setting is around 1.48-1.5v...amazing

 

[Giscardo]

Originally posted by: lopri

Originally posted by: CP5670
I think the newer revision AD0 boards can go quite a bit higher than the older ones. Mine does 400 comfortably, which was as high as I tried setting it.

And that's a nice OC. But what's with hiding the week/batch code?

Is it possible to check the revision (AD0?) of the board before making a purchase? Or is it a different model (i.e. Expert)?

Oh yah.. I'm hiding the Stepping information to mystify my chip and in case I decide to sell it on eBay.. Well, but it's basically draw of luck as you know. Even if you know what to look in the stepping, you won't get to see them until you pay for it. I was lucky to get a chance to look at the steppings before I make a purchase, so I picked the one (in this thread) based on my knowledge from here and there.

Yeah whatever, so what's the stepping?