Considering replacing AMD 4200+ with Opty 165. Opinions??

[Bangwhiz]

Boy - the chip I bought looks HOT. Here's prior overclock validation: http://valid.canardpc.com/show_oc.php?id=518099 I had not considered a multi of 8.5 like previous owner used. I think my memory can handle higher FSB speed - but only tested board to 333Mhz before I pulled it apart to put in a new chipset cooler. Was solid at 333.

 

[crunchie]

From what I understand, that generation of chips do not run half multis all that well. I have steered away from them myself, so cannot verify.

 

[Bangwhiz]

From what I understand, that generation of chips do not run half multis all that well. I have steered away from them myself, so cannot verify.

Did you see his overclock of 2939 Mhz with this chip at 8.5? However, I am planning to overclock with a multi of 9. I worried about limiting my FSB to 333 to stay under a 1000 on my HT link, but I can turn HT link down to 2 with no real performance impact. Can't wait to get my hands on that chip. I think it is all going to come down to the chip and my cooling as to how it runs.

 

[crunchie]

Different hardware can and will give different results. Does your HT have a 2.5 multi? I've been able to run my HT at 3X with the bus speed @ 350. That put the HT link @ 1050.

 

[Bangwhiz]

Different hardware can and will give different results. Does your HT have a 2.5 multi? I've been able to run my HT at 3X with the bus speed @ 350. That put the HT link @ 1050.

I have got mobo torn down putting in chipset cooler, but I would bet a fiver it does. It had .5 on everything above it. "Different hardware can and will give different results." Totally agree.

See this article for interesting study about HT Link settings: http://www.neoseeker.com/Articles/Hardware/Guides/athlon64oc/2.html

See excerpts below:

"All in all, we can see that a reduced HT link speed has an extremely limited impact on overall system performance............"

"The synthetic SiSoft Sandra scores take a slight hit as we drop the HT link speed, with ALU performance taking the largest hit with its measly ~1.5% drop. All in all, the scores remain pretty constant, as the performance hits are quite insignificant (although reproducible)."

"It's apparent that XviD encoding isn't dependant on the HT link at all. There was absolutely no change in the frame rate going from 1000 MHz to 200 MHz."

"In total, Half-Life 2 performance drops a total of 1 FPS. Much like Half-Life 2, Doom 3 dropped a total of 1 FPS.

 

[Parad0x420]

all this talk about Optys makes me want to replace the FX60 in my A8N32-SLI Deluxe with the unopened Opty 170 I got and play.. Now I just need to get another case to put it all in lol.

 

[QuietIce]

all this talk about Optys makes me want to replace the FX60 in my A8N32-SLI Deluxe with the unopened Opty 170 I got and play.. Now I just need to get another case to put it all in lol.

You need a case for a machine to play around with??? O^vo
I only use cases for rigs people actually use all the time - everything else is caseless ...

 

[Bangwhiz]

Alright, got the chip today, installed the new chipset cooler (Thermaltake Extreme Spirit II as replacement for Zalman Passive chipset cooler), used OCZ Freeze thermal paste instead of Arctic Silver because I couldn't find my Arctic Silver, the local store wanted me to take out a 2nd mortgage to buy a new tube, and I liked the specs (and price) of the OCZ paste.

Both the chipset cooler and the CPU have just been seated. I have the CPU running at 9x200 idling along at 1800 for now. I don't know anything about the Opty 165. Come to think of it I think I've got the voltage way out of line 1.45 - excuse me.......!!

Edit: Boy, you think you've thought of everything and then you suddenly remember you haven't changed the cpu voltage from an overclocked AMD 4200+ to an Opty 165. Not the only screw up of the day. When I put the mobo back in a fan cable slid under the mobo and everything except the CPU was back in place until I suddenly wondered "where did that case fan cable go?" and saw a red, black and yellow wire disappearing under the motherboard. Could I just pull it out? Nope - so everything back out and then back in - except I needed a mobo fan connection splitter as well as a single fan controller and a 12" fan extension cable thank you!

SO it's running and I don't know beans about the operton 165. My first questions are: 1. How long should I let it idle along at 1800 before putting the pedal to the metal? I want to be sure the cpu and chipset cooler have seated properly. 2. What is max vcore voltage and when should I raise it as I overclock? 3. What maximum safe temperature is ok for the Opteron cpu?

When I start trying to oveclock it, I plan to set the CPU multiplier to 9, the HT link to 3 to start, Memory to 133 Mhz (max speed for mine is 250), and then start cranking up the FSB in some reasonable increment to 333. If the chip is still stable At 333x9, I intend to drop the HT link to 2 and then keep on pushing the FSB up until it fails Prime95. Since I still consider myself a rookie overclocker, what else do I need to know or consider?

Edit: According to OCZ they claim no curing time is needed. http://www.ocztechnology.com/produc..._freeze_extreme_thermal_conductivity_compound

 

[QuietIce]

Push the clock until it fails the stress test, raise the voltage, repeat until you hit 1.55 vCore or 60°C under load ...

 

[crunchie]

Vcore of 1.6 to 1.65 or higher if you are game and temps are ok. 24/7 though, I wouldn't go above 1.5v personally.
Temps of ~65C are acceptable, but obviously the lower the better.
Keep vcore at stock and increase the ht frequency until you become unstable. Basically the same principle as other processors.

 

[Bangwhiz]

Push the clock until it fails the stress test, raise the voltage, repeat until you hit 1.55 vCore or 60°C under load ...

Thanks. I'll do exactly that. I cranked it up to 2.7 (9x300) started running prime and northbridge temp was 55C. Shut it down for the night so I could google max nforce4 temperature and I read that the nforce4 max operating temp was 90C, safe max temp was 80C, read a lot of threads with nforce4 chips in overclocked boxes running in the 70C's, but tomorrow I will put the ghetto fan I rigged up back into the case. It blows air right on my passive cooled GPU and the chipset cooler. Probably overly worrried, but all I need to buy is a one molex to fan connector and then fasten the fan back into place.

CPU Temps were fine - maybe mid-50's. Now I know why you like water cooling QuietIce.

 

[QuietIce]

WC is expensive to set up but once you've got your basic gear it's as cheap as anything else to update. Sometimes you need a new CPU block, sometimes not - and they run about the same as high-end air coolers. The pump and rad can be used forever and migrated from system to system ...

 

[Bangwhiz]

Alright, got in a lot of trouble but got out of it with a bios reset. Board is back to the day it was born. Everything was hunky dory at 310x9. Decided to push it up to 333x9 and post reappeared telling me to reduce the cpu clock and that it was running in some sort of MB safe mode. However, when I tried to enter bios to reset, box went nuts, monitor said "unsupported mode", motherboard displayed error code 26 on led. After fiddling around for a while I could get post to display for a 30-40 seconds and get into bios - then video would go nuts and that would be the end of it.

There were only two jumpers on the board that looked like bios reset, and the second set of jumpers I tried reset the bios. How I'll update the bios I don't know, but I'll worry about that later. EPOX still has a website, but the bios download wants a FTP user name and password and doesn't like the user name and password EPOX's site gave me so I emailed them and who knows?

Board ran prime95 at 333 fine as a test, memory was no where near the top end, and I don't know why the bios didn't like 9x333. I had the voltage on the CPU pumped up to 1.47 or so for 333 attempt, but that was a .25 (or is it .025? anyway) volt increase over the setting I had used for successful 310x9 prime95 test. Wasn't near max.

Aside from losing the bios updates, what do you think QuiteIce? Any thoughts on the video. I have a Gigabyte 7600GS budget video GPU normally running 450 GPU, 400 memory, like 256 MB video ram. HT was at 3 (maybe should have reduced to 2?). Memory voltage 2.7 or 2.8 - don't remember - but it ran memtest at 250. Memtest is not a system test however.

Right now box is a virgin running 1.8 - which isn't bad considering it wasn't running at all .

EDIT: Couple of other things happened. At 300 or 310 box ran chksk for some reason when I booted. I think at 310 1st attempt it said a windows file was missing. When I had trouble booting 310 I upped voltage slightly (maybe 1.41 or 1.45 or so) and it booted fine and ran prime ok. That's when I decided to try 333 and BOOM!

 

[QuietIce]

IDK anything about that board but from what you're describing I'd check your chipset temps. If they're OK then raise voltage on the NB - that might help. It's where PCIe and HT Link meet.

Or it could be the PCIe itself. Sometimes it helps to bump the PCIe bus just a little at those high clock speeds but, again, that depends on the board ...

 

[Bangwhiz]

IDK anything about that board but from what you're describing I'd check your chipset temps. If they're OK then raise voltage on the NB - that might help. It's where PCIe and HT Link meet.

Or it could be the PCIe itself. Sometimes it helps to bump the PCIe bus just a little at those high clock speeds but, again, that depends on the board ...

Chipset temps were 57C or so. I've been trying to learn what max safe temps for an nforce4 chipset is and don't know what to believe. Guess I better go get another molex to fan connector and put the ghetto fan back in the case also. Everything in bios went to auto for the most part. I'll start by running chkdsk and a registry check, then see what I can do to get bios set up as best I can with respect to all the auto settings. I suppose the first hurdle is to get it set up as closely as I remember and the start pushing up the FSB again and see wtf happens. I also better go get the mobo manual out of a storage box also - hadn't needed it for a while but I think now would be a good time to get it.

 

[Bangwhiz]

Ok, its running 320x9 =2880 but prime95 stops executing. Why? Orthos fails and stops executing also. Says hardware failure "rounding error"
Type: Blend - stress CPU and RAM Min: 8 Max: 4096 InPlace: No Mem: 1791 Time: 15
CPU: 2880MHz FSB: 320MHz [320MHz x 9.0 est.]
8/11/2006 2:18 AM
Launching 2 threads...
Using CPU #0
Beginning a continuous self-test to check your computer.
Press Stop to end this test.
Test 1, 4000 Lucas-Lehmer iterations of M19922945 using 1024K FFT length.
FATAL ERROR: Rounding was 0.4993307822, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 0 minutes 3 seconds - 1 errors, 0 warnings.
Execution halted.

 

[Bangwhiz]

Alright, with the case closed and just the existing cooling, working with the bios settings from memory, I have it running 312x9=2808 and running prime95 for quite a while. It would not run 315 with all the obvious tweaking I could do. If it fails Prime at 312 I'll start backing the FSB down until it passes. Vcore is at 1.47 or so, Vdimm 2.8 - Everything system wise within specs. Core temps mostly in the mid-50's with occasional spikes to the upper 50's max, Northbridge in upper 50's. I could bring down the northbridge temps with the case fan on a PCI cover if it fails at 312 as a final shot at 2800+, but probably not worth the noise.

I've got to sell the AMD 4200+ on ebay now and I might as well unload some other stuff I have at the same time. So assuming it runs ok one way or another between 2750 and 2800 (safe to assume I think), I think it will be worth it versus the 4200+. 200 more Mhz and a cache twice as big works for me. The AMD 4200+ was a disappointment.

Of course, with some help I might get the speed up higher, but probably going to be hard to pull off. I didn't want to put money into the RAM chips and since it seems to want to run below 315 FSB it is well I didn't blow in any money on the RAM. Now if I can just find a bios update for my mobo (having blown it away this morning) I'll be done with it for now.

What do you think QuietIce?

Edit: it is stable enough at 9x312 - ran Prime for 3.5 hours fine. Northbridge 60C, cpu 55-59C. I have to go on a trip. Will try to lower northbridge temps when I get back and see if I might push it up a little further. Ambient temps high 70's. I got a copy of the latest bios. I will flash it when I get back from trip. I did a database import and maintenance job and it ran really fast. Think the bigger cache helps.

 

[QuietIce]

Considering your RAM you should probably turn it down to 100 MHz. At 133 MHz with a CPU of 2800 MHz you're back to running 200 MHz on the RAM. Once you increase the CPU speed above 2800 your RAM is over-clocked again.

But you're also running into a heat wall. Your CPU is a little warm and so is your chipset. Try pulling the side of the case off and using a room fan to cool the case, see if it helps ...

 

[Bangwhiz]

Considering your RAM you should probably turn it down to 100 MHz. At 133 MHz with a CPU of 2800 MHz you're back to running 200 MHz on the RAM. Once you increase the CPU speed above 2800 your RAM is over-clocked again.

But you're also running into a heat wall. Your CPU is a little warm and so is your chipset. Try pulling the side of the case off and using a room fan to cool the case, see if it helps ...

I believe I had a the case cover off and a room fan blowing into the case when the box went completely nuts at 333. Temps were not as high as they are now - mid 50's I would guess. Not sure why it cracked up. I'll work on it when I get back. Raising pciE slightly as you suggested might help. Lost video real quick with monitor message "unsupported mode". Note up above described everything that happened.

RAm ran before overclocked at 235, and passed memtest at 250 running Prime95. At 316 ram is barely overclocked. Do you really think a mild 3-5 Mhz ram overclock will blow it up?

Edit: I tried running 316 with ram at 100 Mhz and HT link at both 2 and 3 with box fan blowing into the case. It would boot and run Prime95 for maybe 10 minutes max and freeze up. Temps for both core and northbridge looked fine - mid to low 50's. I might have pumped up the vcore voltage a little more, but running 312 again now it is reporting voltage at 1.49 now. At 316 it was reporting voltage at like 1.41. Also, all of a sudden northbridge temps dropped a ton during testing, booting and rebooting also. Like the sensor was stuck reporting 60C. Then max temp was like 52C after the big drop.

 

[Bangwhiz]

I would like to add a footnote to this thread. For me, the switch was well worth it. My primary application is a large Postgres database that I use in a real-time environment. Essentially, the database continually updates statistics from nine data inputs when in use. The larger cache makes a huge difference in the speed of the database updates - and particularly when I do database maintenance. Periodic database clustering, vacuuming and analyzing only takes about half as much time as before. I am sure I could try to push it up a few more Mhz but the box is very stable, no voltage or temperature overloads and 2.808 is could enough. Glad I made the switch.

Ironically, the person I bought it from got it for the same reason I did - he was unhappy with the overclock he got from an AMD 4400+. He had a better board than I did and could run it at 316x9- mine started panting above 312 - so that was good enough for me. Now I have to see what I can get for the 4200+ on ebay, but it will be enough to cut the cost of the $70 165 Opteron to minimal, or maybe break-even.