Opteron 146/DFI Lanparty SLI overclocking Question

Taz777

Junior Member
Dec 20, 2005
11
0
0
Hello everyone,
I've been an anandtech reader and overclocker since Celeron 300A, but this my first post. I recently decided to sell my socket 478 3.2GHz P4EE (running at 3.8) and AGP 6800GT after hitting an upgradability deadend, and with the proceeds I was able to get an Opteron 146 CABYE 0540 (new), DFI Lanparty SLI DR (used), and a 6800 Ultra (used, plan to get a 2nd one for SLI later).
I don't yet have all of the components to finish the install but I've been playing around on the bench with CPU/MB using the stock cooler (with the thermal pad), an older PCI video card, and an old Enermax 430W Power supply and using built-in Memtest to check stability (no OS).
As of this morning I was able to complete a full cycle of memtest at the stock voltage (1.38 according to BIOS) running at 2.9GHz (FSB=305, LDT/FSB=3, CPU/FSB ratio=9.5) with the temperature never exceeding 40 deg C (takes a few seconds to escape out of Memtest and check temp. in BIOS).
The question I have is, is this a good indicator of what the cpu will do once I have my H20 cooling setup/OS installed? The fact that the cpu temperature is so low, makes me wonder if memtest is really stressing the cpu?
I plan on using my old Antec True550 power supply that does not have a 24 pin power connector) , can I use the 20pin connector on this MB? I do have a 20 to 24 adaptor but it is very flimsy and would prefer not use it if I don't have to.
Thanks





 

Tweakin

Platinum Member
Feb 7, 2000
2,532
0
71
Memtest isn't going to stress the cpu, just the memory. Without being able to run Prime or some other 100% cpu utility, the temps are going to stay low as there is little load on the cpu. I don't know of a cpu utility that will run from a prompt that can stress a cpu, but I'm sure if you google it there is one.
 

otogrim

Senior member
Sep 4, 2005
336
0
0
The board you have very picky. For stablity i would recommend a 24pin psu. Oh, and someone correct me if i'm wrong but as far as i know memtest doesnt stress the cpu much. Oh and i think it would be better off sticking with 10x multiplier on the cpu. also checking temps that was isnt really going to give you an accurate reading as temps will drop almost instantly ones cpu is no longer stressed. but with the right wc setup, that cpu will do wonders probably hitting 3.0 rock stable. Good luck on your rig and that stepping makes me jealous :frown:
 

nealh

Diamond Member
Nov 21, 1999
7,078
1
0
If you go over to DFI streets they will tell you must run a 24 pin psu for max stability..all 4 connecotr need to be plugged in

This mobo os excellent but you must you use a 24pin connector..I guess an adapter is ok

http://www.dfi-street.com/forum/index.php?

to stress the cpu you need to run dual prime 95 or someother intensive program
memtest will allow to make sure your ram is up to speed

http://www.dfi-street.com/forum/showthread.php?t=16446&highlight=prime95

go slow with your overclocking
 

xenolith

Golden Member
Aug 3, 2000
1,588
0
76
The very most memtest will "stress" in a CPU is its mem controller, and I don't think it stresses it very much at that.

I had an opty 144 CAB2E 0540 that could do 3GHz error free in memtest all day long @ default vcore, but it wouldn't even boot to the Windows desktop @1.5vcore.

In other words, memtest will provide no indication as what your CPU is capable of.
 

lopri

Elite Member
Jul 27, 2002
13,314
690
126
I use a 20-pin PSU with SLI-D and it hasn't given me any problem even with OC'ing.
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Damn that's is'nt much of an upgrade from a $1000 EE northwood @ a whopping 3.8.. that's like a 4.3 prescott..or about what you got now but with HT. I'm an AMD fan an all but thats a sidegrade if I ever seen one.

To answer your question you need to run prime95 to test that 2.9 is stable.. I doubt it will be w/o adding more volts even on water. Gluck:)
 

Taz777

Junior Member
Dec 20, 2005
11
0
0
Yes the P4EE was an awsome processor and I'll miss the HT feature:( but the primary purpose of the rig was gaming (I have a dually with a pair of 2000+ XP-Ms modified to run as MPs @ 2.4GHz rock solid that I use for everything else) , and with the socket 478 setup I was pretty much limited to an AGP video card (fastest AGP video card out there that I know of is the 6800 Ultra). I know that there are a couple of sckt 478 MBs out there that support PCIe but there is no guarantee that they would allow me run the cpu @ 3.8 and no SLI.
I am an AMD fan too and the reason I swapped to Intel a couple years ago was because I got pissed off after cracking the die on a couple of XPs.
I agree that looking at the processing powre alone it may have been a sidegrade, but going from a single 6800GT to dual 6800Ultra setup for gaming would be considered as an upgrade.

I did a quick WinXP install yesterday and the opteron has been priming all night @ 2850 (9.5X300) without any errors @ stock voltage, although the temperature is a little high with the stock cooler @ 46-48C. I am hoping that with the WC setup (kept the P4EE @ 27 C idle and <40C under load @ 1.8v Vcc) and a voltage bump I should be able to get the opteron above 3 GHz:)

 

Taz777

Junior Member
Dec 20, 2005
11
0
0
Thanks everyone for you repliess and I hope you all had a nice Christmas. I got the final component for my setup, a Swiftech Storm, for Christmas and finished up my setup. According to nv Monitor system idles @ 30 deg. C and under load maxes out at 39 deg. C, pretty much regardless of the voltage ( stock to 1.6v).

A couple of you guys made me re-think my decision to go from a P4EE @ 3.8 to the Opteron 146, so as soon as I had it stable @ 3.0GHz I run some benchmarks and got
anywhere from 0 to 20% improvement (most improvement in 3D Mark 3, video card was not a factor since I was comparing a 6800GT run at Ultra speeds to a 6800 Ultra running at stock speeds).
Was it worth it? Money wise it did not cost me anything, but it was rather time consuming. Looking back, I should have probably waited for the new M2 socket, oh well..............

When I started with this project, my goal (maybe a little too optimistic :)) was to hit 3.2GHz, and as of now it seems like I am upto 3.1 @ 1.5 v+10%=~1.6-1.618v (primed all night last night @ 3.05 and now it's priming at 3.1 and has been stable for the last 2 hours). It boots into windows at 3.2 but errors out on super PI almost immediately.
Do you guys have any suggestions on how I can possibly hit 3.2 (I've already taken off the IHS). The MB seems to be stable beyond 325MHz FSB, and my memory, 1gig of Corsair PC3200 Twinx rated at 2.2.2.5 at 200MHz, is running at ~ 250 MHz with 166 divider (tested stable up to 285MHz @ 3.4.4.8 and 3.1v).
One thing that I am still uneasy about is my Antec True550 20 pin power supply. It is running with the cheesy 20 to 24 pin power adaptor and I am a little hesitant about running without the adaptor based on what I've read online (damaging memory??).
Do you guys think going to a true 24 pin power supply help me get to 3.2GHz?
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
My CABYE 0540 topped out at around 3.15GHZ even with 1.65V (1.67 in the bios) and load temps at 38C. At that speed it was barely bench stable.

That's impressive that your chip will Prime at 3.1GHz though. I'd imagine with some more volts you should be able to get OCCT/SuperPi 32M stable at 3.15-3.2GHz.
 

Taz777

Junior Member
Dec 20, 2005
11
0
0
Nice rig!!!

Sound like we have simlar WC setups, have you thought of tossing the water cooling and the headaches that go along with it for one of these new monstrous (sp?) heatsinks? I am sure we both could run out systems @ 3.0GHz relatively quietly?


 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
Originally posted by: Taz777
Nice rig!!!

Sound like we have simlar WC setups, have you thought of tossing the water cooling and the headaches that go along with it for one of these new monstrous (sp?) heatsinks? I am sure we both could run out systems @ 3.0GHz relatively quietly?


Thought about it. Watercooling can be a hassle sometimes but I think the Opteron I'm currently running (CAB2E 0540) would give high-end aircooling a run for its money. I get load temps almost 10 degrees higher than my 3700+ SD and I think with aircooling I'd be in the 50's pretty easy even with low volts.

I've got my video cards in the loop too but I think they'd be fine with some aftermarket aircoolers.

I might switch sometime soon. The nice thing about watercooling is that it seems to give more consistent results. Aircooling can hit about the same speeds as watercooling but generally isn't as stable/reliable. That's just my opinion though.
 

Taz777

Junior Member
Dec 20, 2005
11
0
0

I guess I am having a hard time justifying WC for only a ~.1-2 GHz increase in overclock over air :(.
Looks like I'll have to finally settle for somewhere between 3.1-3.15GHz @ 1.65v. Prime errors out after 4.5 hrs at 3.15 and seems to go indefenately at 3.1.
Thanks for everyones help.
 

Elfear

Diamond Member
May 30, 2004
7,165
824
126
The fact that it Primes at 3.15GHz for 4.5 hours is pretty amazing. You have a nice chip there.
 

themusgrat

Golden Member
Nov 2, 2005
1,408
0
0
I used an adapter with the DFI SLI-DR board, and it gave me many troubles. It all depends on the PSU though, because using an adapter is like borrowing power from other components, and the PSU has to be able to handle the fact that you are pulling too much power from that rail, and those components need to not draw much power. I forgot what it steals power from, but hopefully this helps. Also, a 6800 Ultra isn't cost-effective at all. I would get a 6800GS or 7800GT. They'll do much better. Welcome to the forums.
 

tjpark1111

Senior member
Oct 5, 2005
287
0
0
I heard that increasing voltage shortens cpu life. Is it solely because of the extra heat from the extra voltage? or some sort of internal damage to the cpu circuitry? I'm basically asking if you got excellent cooling, can you up the volts as much as you need without shortening life? does the same apply for memory and chipset as well?
 

Taz777

Junior Member
Dec 20, 2005
11
0
0
thanks themusgrat,

You are absolutely right about the power supply being a factor, and I am finding this out slowely as I play more and more with the system. Other than my Vmem fluctuating all over the place, system seems to very unpredictable, sometime it'll pime all night at 1.50v other times with the same settings won't even pass super PI 1.65v. I've decided to look for another power supply, any suggestions?? I was looking at the turbo cool 510 SLI, but that fact that it only has a single 12v rail concerns me?
As far as the video cards go, I've already got one 6800Ultra that seems to run very happily at 450 and 1200, and I figured if I get another used one for ~$200 I'll have a decent system. Frankly I am not familiar with the 6800GS, how does that compare to the 6800GT/Ultra?

 

Taz777

Junior Member
Dec 20, 2005
11
0
0
I am sure there are several different types failures that can occur when the voltage is increased, but I am only familiar with electromigration. Failures caused by electromigration happen when the current actually moves the metal molecules in the metal lines that interconnect the devices on the chips, eventually forming a void and opening the line (physically). The main factors that effect this are the amount of currently going through the metal lines or current density (more voltage=more current), temperature (which speeds up the reaction), and the type of metal used for the interconnect (Cu). I can not tell exactly how low you'd have to drop the temperature to make up for the increase in current current density as a result of the increase in voltage, but can tell you that you'll definitely shorten the life of your chip by increasing the voltage and keeping the temperature the same.
 

vanvock

Senior member
Jan 1, 2005
959
0
0
Originally posted by: Taz777
thanks themusgrat,

You are absolutely right about the power supply being a factor, and I am finding this out slowely as I play more and more with the system. Other than my Vmem fluctuating all over the place, system seems to very unpredictable, sometime it'll pime all night at 1.50v other times with the same settings won't even pass super PI 1.65v. I've decided to look for another power supply, any suggestions?? I was looking at the turbo cool 510 SLI, but that fact that it only has a single 12v rail concerns me?

I wouldn't worry about that. The Turbo cool is up to the task. I would rather have it that way than 2 rails with half the power on them.