Asus A8V

[Pointhore]

Originally Posted by Vogon

Yep, sandra mem bandwith is obviously hit, but game performance looks to be the same so thats not a concern for me
I just like nice round figures

You are right! Memory bandwidth takes a big hit, but actual gaming performance doesnt. Unless you can tell the difference between 1-2fps. Il2 runs smooth as silk with 2T enabled with perfect on.

Even in 3Dmark 2003 I only loose 200-300 points. I think I can live with that.

[:PopCorn]

I'm going to leave it at this because I don't want to get into a negative OT discussion here or have this appear to be flame material. I honestly don't mean it to come off that way, but I think Prime95 is the problem.

There's no way that I'm taking the opinion of one program over proven 24/7 stability in everything, and I mean everything else.

Super Pi 32m test is reliable and much faster (apprx. 25-30 minutes). This, along with a full 3D shake down can reveal limitations that Prime95 misses.

I have failed Prime @ stock and passed it when overclocked to 2650+. Now where's the logic in that? I have passed at speeds that aren't 3D stable and failed at speeds that are. I've tried all the different combos of stress testers and Super Pi + 3D + Memtest 86 is where it's at. Fast and effecient, the balance of these 3 reveal my true stable overclock speeds.

Using this method, I can run 24/7 @ 2600-2650 MHz, and I can bench up to 2700ish @ 1.7v. Using this method, Sierra Bound (who first said this was his way of testing back on page 6 of this thread I think) has acheived some of the best results I've seen.

Just my .02,

Good Luck

[Riptide]

Originally Posted by Blue078

Does everything else you, but prime, work good Rip ???

3dmark03 seems alright. 3dmark01 refuses to work on my machine any more, even after going back to stock clocks and doing a reinstall. I don't know why. Maybe it has something to do with Catalyst 4.9 since I just upgraded and it worked OK before that. Strange though..

Basically the way I've been doing it is using Prime as my main tool. If it fails I immediately move on to a different setting. If it passes I try and run 3dmark03 and see what happens.

I can't get past this requirement for so much vcore. My CPU seems to need 1.625v just to pass prime at 209x12. It runs hot at that setting (60c+) so it's unacceptable to me.

I might try switching from Prime over to SuperPi per PopCorn's advice. It does seem rather odd that Prime works so well at stock speeds but it's the first thing to freak out when I try and o/c. I guess part of the reason I've been so stubborn about using Prime is due to the fact that it works perfectly at stock speeds so it doesn't seem to be some sort of incompatibility between the AMD64 CPU and that program. I don't know what to make of this. It doesn't make sense.

One other thing worth noting. When I first got my Abit motherboard I didn't realize it set FSB to 204 by default. Folding@Home had been fine for weeks but one day I noticed it quit with EARLY_UNIT_END. It kept doing this about once/week. Eventually I did some reading and found out that this issue can come up due to overclocking. Guess what? I installed Prime95, it crashed within minutes. I went into the BIOS, found the fsb @ 204, and changed it to 200. System was perfect after that. What took over a week to crash the folding client took minutes to crash Prime. Prime is apparently *ultra* sensitive towards any instability.

[cromozone]

At 2.7GHz I can run everything apart from Prime95. And I mean, everything, 3DMark, Doom 3, SuperPi 32M test. Still, I don't need to run at 2.7GHz do I? So backed off to 2.56GHz for day to day use

[Sierra_abx]

One thing I've noticed about the 4.9 driver. When you open the Task Manager, 3DMark01 uses a lot more memory than with the 4.8 driver. Not sure if that is related to the problem you're having, Riptide. I'm also using the old ATI control panel. The new one strikes me as being a bloated POS.

Each computer user has different criteria for judging stability. For me, running programs with no random reboots, freeze-ups, etc. is all that I require. Even at stock speeds, an FX53 can crush just about anything out there. Enjoy your system, and have fun!!!

[Blue078]

Man Rip..I just hate to se you put your self through all this, just because one app dosnt work.

I was serious when I told you my Intel 2.8C would fail P95 at 3300Mhz and above...and if I had belived in P95 then I would have missed out on a years worth of 3860Mhz gameing goodness.....

Just try some of the other guys methods and see how it goes

[:PopCorn]

Riptide,

I'm sorry if you've already been asked this and answered, I may have missed it but have you exchanged your CPU? It's possible since you've had so many problems with various hardware setups that your chip is somehow defective, maybe the memory controller is freaking out.

My only point about Prime95 is that is has proven to be inconsistent and unreliable with this setup over my Intel rig. I just think Prime may have something funky going on with A64 setups of certain varieties.

I have to credit Sierra for getting me off the Prime95 kick. Once I stopped screwing with that program, my life with this setup improved greatly. I've put this box through Hell and it is 110% stable. Prime95 might agree now but I really don't care enough anymore to test with it.

Now if super pi errors, then I'm not 100% stable and I've found that pretty soon some other App will agree, ending in a lock up usually. This has proven to be true time and time again. But I don't totally depend on Super Pi either, I still confirm with 3D testing and memtest.

[SupDawg]

Originally Posted by Riptide

3dmark03 seems alright. 3dmark01 refuses to work on my machine any more, even after going back to stock clocks and doing a reinstall. I don't know why. Maybe it has something to do with Catalyst 4.9 since I just upgraded and it worked OK before that. Strange though..

Basically the way I've been doing it is using Prime as my main tool. If it fails I immediately move on to a different setting. If it passes I try and run 3dmark03 and see what happens.

I can't get past this requirement for so much vcore. My CPU seems to need 1.625v just to pass prime at 209x12. It runs hot at that setting (60c+) so it's unacceptable to me.

I might try switching from Prime over to SuperPi per PopCorn's advice. It does seem rather odd that Prime works so well at stock speeds but it's the first thing to freak out when I try and o/c. I guess part of the reason I've been so stubborn about using Prime is due to the fact that it works perfectly at stock speeds so it doesn't seem to be some sort of incompatibility between the AMD64 CPU and that program. I don't know what to make of this. It doesn't make sense.

One other thing worth noting. When I first got my Abit motherboard I didn't realize it set FSB to 204 by default. Folding@Home had been fine for weeks but one day I noticed it quit with EARLY_UNIT_END. It kept doing this about once/week. Eventually I did some reading and found out that this issue can come up due to overclocking. Guess what? I installed Prime95, it crashed within minutes. I went into the BIOS, found the fsb @ 204, and changed it to 200. System was perfect after that. What took over a week to crash the folding client took minutes to crash Prime. Prime is apparently *ultra* sensitive towards any instability.

Riptide, if it fails Folding and Prime it's usually the overclock. But if it can pass Folding without crapping out, and fail Prime, I would consider it stable.

Also, some proteins are not stable. Please see if there is anything wrong with the protein before saying its not stable.

[Riptide]

Gentlemen, to answer a few of your questions...

I will remove Prime95 from my computer and start again, this time using SuperPi in place of it. I hate to say this but though Prime95 consistently works fine at stock speeds it does seem inconsistent when using it to judge the stability of an overclock. I've run it twice at the same settings and had different results. Not sure what's going on there but perhaps it's as some of you say: it doesn't like (for whatever reason) the athlon64.

I have not exchanged my CPU. There are a couple things kinda stopping me there. For one, newegg's exchange policy sucks and they won't cross ship so I'll be completely down. Also, my CPU runs perfectly at stock speeds so it's possible they may not even allow the exchange simply because it won't overclock. Plus, I'm just sick and tired of all these exchanges and rebuilds. It's beginning to get old...

When Folding@Home was going down, it was due to the slight fsb overclock. The first time I saw EARLY_UNIT_END come up I thought nothing of it. But it kept coming back every 4-7 days. That's when I took notice, fixed the fsb setting, and those units stopped ending like that. I haven't seen one since.

[:PopCorn]

I can understand where you're coming from Riptide. Hopefully you'll get things worked out. Don't take anything said as criticism but rather hopeful advice.

We are rooting for ya brother!

[Riptide]

How long does SuperPi typically take to finish the 32M calculation?

[Sierra_abx]

At 2.4GHz, probably 30-31 mins. Maybe closer to 31.

[Riptide]

Originally Posted by sierra_bound

At 2.4GHz, probably 30-31 mins. Maybe closer to 31.

Thanks. I'm trying these settings right now:

2.5-3-3-8-1T, 2.7vdimm
x4 HT
227x11 1.55vcore

If it finishes OK w/SuperPi what should I test with next? Memtest or 3dmark?

[Sierra_abx]

Maybe 3DMark. SuperPi really gives both your CPU and RAM a workout, so you may not need to worry about Memtest for the time being. There's another RAM testing program called Gold Memory. It's like Memtest, only I heard it works better with A64 processors. There's a shareware version.

http://www.goldmemory.cz/

[Riptide]

Man I wish I had a version of clockgen that worked with this motherboard. It might get around one little problem I'm having. When I get my FSB over about 220 it does strange things sometimes when going into Windows. Like re-detect my integrated NIC. Not sure why it likes to do that but it's annoying.

SuperPi errored out btw. "NOT EXACT IN ROUND"

I'll try 1.6vcore