Need Opteron165 / Neo2 Plat Help

[ferret99gt]

I need some help. I've just now upgraded from my trusty old Athlon-XP 2500, which for a long time solidly ran at 2.3ghz on aircooling, which was good for my purposes. I've recycling ALL of my parts except the motherboard and CPU, which is why I went with MSI Neo2 Platinum rather than an nForce4 board, cause I'm stuck on AGP for now.

Full specs:
Opteron 165
MSI K8N Neo2 Platinum (Latest official BIOS)
Windows XP Pro SP2
2x 512meg Mushkin PC3500 Dual Channel Pack w/ Heat Spreaders
BFGTech GeForce 6800 OC with pixel/vertex shaders unlocked
1x Maxtor 60gig
1x IBM 60gig
1x Maxtor 160gig
1x Pioneer DVD-RW

I see to be having a couple of issues, and maybe some of them are related to not reinstalling Windows XP fresh. I did repair the installation and re-update through SP1-SP2. When I'm just using the system by itself, and particularily during start-up, it seems like the HDD is spinning a lot and thrashing. At boot-up, my normal start-ups will appear in Systray fairly quick, but desktop icon images load super-slow. Windows don't always fully render on opening and sometimes pause for a second or two.

I grabbed the Opteron patch from AMD and applied to MS hotfix, and it seemed to help some, but the issue remains.

Now on to the second issue... I can't seem to overclock at all. Maybe I just haven't read enough. If I touch the HTT settings at all, the system doesn't boot and I have to clear the CMOS to get it back. I'm not talking putting it up to 250 and it not working, I mean, 201, 210, anything and everything, the system refuses to boot.

About the only thing I've managed to do in the BIOS is set my RAM from PC3200 to PC3500, but my understanding is dual channel on this board runs at DDR333 no matter what?

Any help appreciated.

 

[Ricemarine]

dual channel = 400.

You should lower the HTT to 4X, and then raise the fsb very slowly... What you could try is the modified bioses made by sideeffect (google it). Also, your power supply might not support it... Another factor could just be that your ram cannot go past pc3500.

 

[Captante]

I just ordered an Opteron 170 myself (upgrading from A64 3200+) and everything I've read says that if you are going from a single to dual-core CPU you really need to do a format & clean re-install of Windows or your performance will suffer quite a bit, so I strongly suggest that you back up your data if you havn't already and get to it.
Also I'd suggest you either kick your DDR back down to PC3200 speed or try lowering the memory divider as well as the Hyper-threading if you continue to have problems overclocking... your CPU should be able to do a minimum of 2.4ghz (& maybe up to 2.8ghz if you get lucky) so my bet is that either your RAM or your mb is the limiting factor, not the CPU... you might also want to read the sticky at the top of this forum for a lot of useful info about A64/Opteron overclocking.

 

[ferret99gt]

I've tried raised the FSB slowly from 200 with HTT at 4x and 3x with the same result, whether or not I've changed any of the DRAM settings. After the CMOS saves and the system should reboot, it just hangs and I have to clear it and try again.

The Neo2 Plat bios doesn't seem to allow you to adjust the memory divider, so I'll have to look at the modified bioses. The power supply is an Antec TruPower 430, which I would hope is still sufficent these days.

Can WinXP be reinstalled clean without a format? It'll be major grief to clean this HDD off but I can do it if I have to.

 

[ferret99gt]

Never mind about the memdivider, I think I understand that now.

I'm currently running at 280htt with no issues for general opteration. Have not done Prime95 yet. Backing up system files right now for new WinXP install.

 

[ferret99gt]

Ok, I've gotten Windows re-installed, with the AMD driver but not the hotfix, and the stutters are gone.

I've tried some of the things in the overclocking guide and ran some various tests. For example, when trying to find my memory speed, I can get successfully memtest at stock settings (9x200htt,200spd ram, DDR400), and I can get a successful memtest at 9x216,200spd (DDR433).

Everything else, memtest seems to fail at the last few percentage of test 5. Perhaps my memory just isn't good, but a year ago I was under the impression it was pretty decent stuff. Not the absolute top of the line, but not cheapo either.

I can boot into Windows at htt280 with ram spd@133, so about 2.5ghz. The ram timings under that end up at 2-2-2-5 t1, with everything on "auto" but the cas and command rate. However, prime95 immediately fails due to a rounding error.

Running stock, I had a question about Prime95 anyways. I've tried two things while running two seperate copies. I set affinity both from Prime95's menu, and within Taskmanager. However, the cpu activity seemed odd.

When I start the core0 prime, I would see core 0 spike to 100% and stay there. When I start core1, it would spike to 100% and almost immediately go back down to idle. After a few minutes, it would seem to start back up, and both would run at 100% for a while. Then a test would finish and as the next one started, that core would idle out for a few minutes. Is this some sort of loading delay? When both Primes are running, task managers shows my page file in use at 1.6gigs. The swap file and the Primes are on seperate partitions, but the same disk. Would it help to move Prime to a second disk drive?

 

[aka1nas]

If you use RAM dividers on the NEO 2, you should set ALL RAM timings to safe levels. If the stick has SPD settings for lower RAM speeds, it may set the RAM to tighter timings than is safe if you leave it on auto. Or to clarify, if you se the 133Mhz divider, it will assume it is running at 133Mhz for SPD purposes.

 

[ferret99gt]

Alright, we're getting somewhere now.

I looked at CPU-z.. Under the tab about the RAM, it showed the timings by SPD as 2-4-4-7 and 2.5-4-4-7 for DDR434. So I set those in bios instead of auto. I don't think they're the best timings in the world, but all memtest issues seem to have disappeared now.

I put in sideeffect's 1.C3 bios as well, so I could put the memory SPD at 150mhz so that 288HTT would be 216mhz for the RAM. That seemed to go well in memtest. I tried a little higher and had no memtest issues with the ram at at about 223, but higher it started to fail at the end of test 5.

I was able to boot into WinXP with this, and began Prime95. No more rounding errors now that the RAM timings have been fixed. I didn't now ram could cause bad number crunching, unless the numbers were being corrupted by the bad timings as things were fetched from memory or something.

I really want to hit 2.6 stable and HTT 288. I can eat memtest all day at this, but stable booting doesn't always occur. Prime95 can run for about a minute before the system fails and reboots.

I don't think the vcore on the CPU is changing from stock. I've tried 1.4, 1.425, and 1.45... But the vcore monitor in the BIOS still states between 1.3 to 1.35. This BIOS has two vcore options...

The first is CPU VID. The options are "Startup" which I assume just means "Stock", then it goes up from around 1.2 or so to 1.6 or so. I currently have this set to 1.400

The next is CPU Voltage, and its either "By CPU VID" or a percentage. I have not tested this setting.

Has anyone had issue with getting their vcore to change on this board?

 

[ferret99gt]

I know i know, posting without any responses... I changed CPU VID back to Startup, and changed CPU Voltage to 5.0%. 1.325 x 1.05 = 1.4

After doing this, the system became a ton more stable. Prime95 is not failing, but neither is it really stressing the system.

The issue I have now is that Prime95 isn't running most of the time. I'll start it on one core, and it'll spike that core to 100%. I'll start the second core and it will spike to 100% then drop to idle. After the first core finishes a few tests, it drops to idle and the second core runs for a few tests.

They won't both run at full load for more than a few seconds, and at times both idle for 10+ minutes with Prime doing no work. The system still behaves as if CPU is under high load though.

Also, whenever an instance of Prime95 begins torture test, Page File usage jumps by about 800megs. When both are running, usage is at 1.6gigs...

The system seems stable and completes tests whenever one of the cores decides to actually run them, but I don't see how I can claim its stable if it only does 10-15 minutes of work each hour..


EDIT: And once again, right after posting, I finally figure out why I'm an idiot, and figure out Prime95. I'm going to stop posting now

 

[Jojo1971]

i got an opteron 165 @ 2.6ghz / k8n neo2 platinum /pc2700 512mb x4:

here are my settings: it's prime95 stable...

3x
disabled
x9
289
66
1.425
8.3%
1.65

 

[ferret99gt]

That seems like a really high vcore though... What do monitoring programs say is the vcore? When I left CPU VID at "Start up" with the CPU Voltage at 8.3%, I was seeing as high as 1.49, which from my reading is a little bit too dangerous for my tastes.

Right now, I've tested from 290HTT down to 280HTT. Testing 280 right now.

DDR set to 150 divider (4:3), 2-4-4-7 T1
3x HTT multi
9x CPU multi
280 HTT
66 AGP
1.325 CPU VID
8.3% CPU Voltage (With the above, results in about 1.44-1.46 in monitoring programs)
2.70 DDR voltage
AGP voltage at default

After about an hour, core 2 fails Prime95 with a rounding error, although Core 1 keeps on going. This has been without fail so far, core 2 only is failing me.

Before when I was letting the motherboard adjust RAM timings, I was getting rounding errors almost immediately. I'm starting to think my RAM is the major issue here. It passes all 8 tests at up to about 224mhz in memtest, but I haven't had time to let it run through more than 2 passes of all the tests.

At 280htt, the ram should be running at 210mhz, which is what I ran the FSB at on my old Athlon-XP system. I guess I may have to settle for a lower clockspeed for now, until I can afford better RAM.

 

[Tweakin]

Just as a note on the board, I also use Sideeffects modded bios, and installed the TCCD mod for my OCZ Memory. I run and Opty 146 and my settings are as follows:

133 divider : 2-2-2-10 1T @ 2.65v
3x HTT multi
10x CPU multi
301 HTT
66 AGP
1.475 CPU VID
5.0% CPU Voltage Everest and CPU-Z shows 1.52~1.57
2.65 DDR voltage
AGP voltage at default

 

[BoboKatt]

Very interesting. I have an Neo2 running my X2 3800+. I have 2 X 1024 OCZ ram. Might I ask if these sideeffects modded BIOS for this board just help in giving more options such as that extra memory divider or are they more stable?

Also does our board still suffer from the SATA 3 and 4 not being NOT locked OR our AGP speed not being locked unelss we put it to 67 Mhz instead of 66? I noted you all had 66 which I was told a LONG time ago there was a bug that if you did not set it to 67 it would actually scale high on OC'ing. However with the latest official bios maybe they fixed that.

Has anyone here tried to see what the highest HTT they got with a 3800+? Man I could never imagine 280 let alone 300. I just OC it to 2200 (220 X 10). Can anyone explain to me if I can get the SAME total speed so say 2500 Mhz, whether it's 250 X10 or 277 X 9... would one be faster or since memory controller is on chip and such it's just the FINAL total that matters? Or could someone confirm that since I am running at 277 is that more of a stress on the CPU than say 250 (albeit at X10)... of again it matters not but rather it's all in the voltage and temperaturs? In other words trying to figure out if it''s "WORSE" to have 250 X 10 or 277 X 9 on this board component wise?

thanks!~
!
Bob