Jump to content

Help with OCZ Gold rev 2 PC 4000


Indyxc

Recommended Posts

Well, some some success.

 

Apparently, the CPU wasn't stable, and was causing the problem. I guess it was stable enough to pass teh FFT test, but not the Blend test. I added more voltage, and now it passes all Prime 95 tests, with the memory at slow speed, at 3000mhz. The memory controller is running at HTT speeds (?), so NOW can I assume the chip is stable, and work with the memory?

 

At 3000 here are my settings:

HTT: 273

LDT: 3x

Multi: 11x

 

So I moved onto the memory. I tried every possible combination with your settings, and it was a no good at 273 1:1, with any voltage, and ram timmings. Threw errors very quickly, faster than everything on AUTO.

 

So I tried 4-5-5-10, and the computer failed to boot, so I reset the CMOS.

 

Loaded everything on memory auto, but 3000mhz changed the ram voltage to 2.9 and THOUGHT I set the ram divider to 180.

 

So I run prime 95 blend, and after about 20 mins, I check CPU-Z, and noticed I'm running the memory 1:1, so 273, with auto, which is 2.5-4-3-8. I got excited it was running, and let it run. It ran for 2 hours, and I was getting more excited. 2hours 32 mins in it threw an error. DAM YOU MEMORY, I was getting really excited. :P

 

So I upped the voltage to 3.0, same timings, error- 30 mins

Upped to 3.1, error even faster-20 mins

Upped to 3.2, error- 20 mins

 

Decided to change timing to 3-4-4-8.

All voltages, not prime 95 blend stable.

 

So I attempted 4-5-5-10 again, and it won't boot as those timings for some reason.

 

It's stable with the mem divider at 9/10, mem at 231. But that's too slow.

 

Off those settings you gave me, which individual ones make a big difference on the conservative side?

 

What else can I try?

 

Thanks again

Indyxc

 

Getting closer..

Share this post


Link to post
Share on other sites

  • Replies 27
  • Created
  • Last Reply

Top Posters In This Topic

4-5-5-10 is nuts!

 

Mine errors when I change it to 3-4-4-10.

 

The first number (cas) needs to be 2.5. If I slow it down to DDR400 speeds, it will run at 2.0. But regardless of frequency, I've never had 3.0 be stable. You wouldn't want 3 or 4 anyway. 2.5 for this ram works for a wide range of speeds. If you use 2.5-4-4-10, and can't get it stable then, there's either something wrong with the ram, or one of the many other settings is off. Well it could something else like voltage, or the motherboard too, I guess. Try testing each stick individual (I suggest the orange slots) using memtest and maybe some other bench.

Share this post


Link to post
Share on other sites

Try:

DRAM Configuration Settings:

Command Per Clock (CPC) - Enable

CAS Latency Control (Tcl) - 2.5

RAS# to CAS# delay (Trcd) - 04 Bus Clocks

Min RAS# active time (Tras) - 10 Bus Clocks

Row precharge time (Trp) - 04 Bus Clocks

Row Cycle time (Trc) - 10 Bus Clocks

Row refresh cyc time (Trfc) - 16 Bus Clocks

Row to Row delay (Trrd) - auto

Write recovery time (Twr) - auto

Write to Read delay (Twtr) - auto

Read to Write delay (Trwt) - auto

Refresh Period (Tref) - 3120 Cycles

Write CAS Latency (Twcl) - Auto

DRAM Bank Interleave - Enabled

 

DQS Skew Control - Increase

DQS Skew Value - 50

DRAM Drive Strength - Level 8 ------------ these 2 settings were important for me

DRAM Data Drive Strength - Level 1

Max Async Latency - 8ns

Dram Response Time - Normal

Read Preamble Time - Auto

IdleCycle Limit - 256

Dynamic Counter - Disable

R/W Queue Bypass - 16 x

Bypass Max - 07 x

32 Byte Granularity - Disable(8 Bursts)

 

Remove one stick and try these settings for memtest (and any more benching). Set the divider at 1:1, and HTT at 250 (this will make sure the memory is can run atleast stock). Lower the cpu multi a couple, so if it errors, it won't be because of this. If it passes, then start upping the HTT to you find THAT STICKS max. Then test the other stick the same. If you think it can pass at stock speeds, then test both sticks together, this may save some time, but if it fails, you won't know if it's because of one stick or the other, or some other problem(s).

Share this post


Link to post
Share on other sites

I've got some of an update:

 

Thanks for the slower settings fightgame! Thank you. :)

 

This setup is proving to be quite difficult and counter-intuitive.

 

With the memory settings you gave me, and these settings:

HTT 273

MEM: 1:1

CPU: 3003mhz

LDT= 3x

 

I run the Prime 95 blend test, it passes the 1024k test, then the 8k, 12k, and on the fourth test (896K), 3rd set, it ALWAYS fails around 51 mins in. Tried different voltages for the memory 2.8-3.2, and it's the same story.

 

So, I figured lets try HTT 273, HTT 1:1, CPU (4x) so ~1000mhz. I set the voltage to 2.9v on the mem, and it passed 2 hours of the blend test, including the 896k it normally fails. I quit after 2 hours to see how it would do with the cpu at full speed again, and it failed the 896k as usual.

 

So, I'm going back priming the CPU at HTT 273, cpu 3003, and the memory at 133mhz, so to see if the cpu is unstable. I'll go back and test the memory at 273 again over a longer span to see if that's stable on it's own too.

 

 

So it appears the CPU can do HTT 273, and so can the memory, but NOT together.

:sad:

 

Any ideas? Could increasing the LDT voltage or the Chipset voltage higher help? I think I've got both 1 notch about min.

 

Thanks for your help again fightgame.

 

Indyxc

Share this post


Link to post
Share on other sites

I did some more testing over night to verify the memory and cpu were stabile on their own.

 

Ran the CPU at 3003 mhz

HTT: 273

MEM: 166 mhz

LDT: 3x

 

Made it through 6 hours of Prime 95 blend test, and then I quit.

 

 

Ran the CPU at 1000 mhz

HTT: 273

MEM: 273

CPU: 1000

LDT: 3x

 

 

Made it through 2 hours of Prime 95 before i decided to quit.

 

BUT with the CPU at 3003 and the MEM at 1:1

 

It fails prime 95 blend on the third 896K test, 51 mins in every single time.

 

At this point I'm not sure what I can do.

 

It just passed me playing BF2 for 1 hour, so at this I'm thinking if it's game/everything else stable i'll take it.

 

Anyone else ever see this problem before? Searched around, but couldn't find anything.

 

Thanks

Indyxc

Share this post


Link to post
Share on other sites

try to go for a 250HTT first, you aim too high too early.

run Prime95 Torture test for 12hours without a glitch, THEN an only THEN you can assume that your setup is stable. increase memory timings until you can run this 1:1 for another 12 hours. then increase with 5Mhz.... etc... etc... A good OC takes time.. 10 mins to 4 hours of Prime mean nothing, i've learned the hard way. I had the REV2 kits as well. Mine maxed out about 265Mhz....

 

You can also work the other way around start with high HTT but very low multiplier.. in this way your processor is on the safe side and fails are quite likely to be ram.

Share this post


Link to post
Share on other sites

Everyone has their own definition of stable. As long as my system doesn't crash after being on 24/7 and weeks of no restarts, no errors, no freezing, I consider that 95% stable, and that's all that matters (to me). After I passed prime torture for 8+hrs (while folding), all the 3dmark's, a few others like occt, then I called this 100% stable. I mean what's more important, being able to do everything you want to do on your computer without problems, or being able to pass 10 tests, running at the same time, while blowing hot air into it...haha ok so I went a little overboard..

Share this post


Link to post
Share on other sites

Good news to report!

 

I upped the LDT voltage and Chipset voltage, and it passed the 896k test! :) That seems to be the problem. In reality, the cpu and memory were unstable at first, but it seems that was the final touch. I'm not 100% certain it's stable, but its made it farther than it ever has. I'm going to let it run for another couple of hours to be sure.

 

If it passes prime 95, I'm going to go back to work on the memory timings. With the settings you originally gave me Fightgame the memory bandwidth was around 8000mb/s, and with the slower timings it's only around 7500 mb/sec. The memory is only at 2.9v right now, so it should be able to do tighter timings.

 

My only concern now is the temps:

CPU: 41C

PWMIC: 51C

Chipset: 47C

 

I wish I could do something about the chipset, but it's jammed under the video card.

 

What do you guys think about the temps?

 

As far as stability:

 

I agree that if it is stable in what you usually do it's generally stable. But, my experience has been, if it won't pass prime 95 for hours on end, eventually the stability problem will rear its' head.

 

With my Pentium 4 3.2e. It was overclocked to 4166, and would fail prime 95 after a while, but was stable in games. Half Life 2 came out, and the stability began to rear its head, and the game would crash.

 

In the end the 3.2e was 100% stable at 4000.

 

Thanks for the help.

Share this post


Link to post
Share on other sites

My idea would be to use your processor's highest mulitplier and put your HTT on 250Mhz. with 1:1 timing, in this way you should be able to reach about 6500 in Sandra, i've benched 7300 as higest in Sandra, however wasn't able to get this much bandwidth stable. Maybe another guide for you is that very rarely you see memory benching 7000+ and being stable for day to day use.

Share this post


Link to post
Share on other sites

An update: Some more sucess

 

I still haven't given up. I flashed it to the 623 bios today, and worked around with that. Definantly more stable.

 

I noticed 2.9v is the sweet spot for memory. Any more and it fails prime 95 in 10 mins or less.

 

I also noticed the LDT voltage make a big difference in stability. On the lower settings prime fails 30 mins or so. On the 2nd to highest 2 hours.

 

On the highest LDT it made it 3 hours, and I was getting excited. It looped one whole test starting at 8k to 4096k in prime. I was getting certain it was stable. 3hr 45 mins in boom, failed a test it had passed already. DAM YOU, WHY MUST YOU TEASE ME.

 

So the issue has been the bios, and especially the LDT voltage. The chipset voltage is at the 2nd lowest, so I might work with that.

 

Put 3hrs 45 min Prime stable is pretty good in my book. Maybe I'll just take it.

 

Can I somehow cool anything else related to the LDT? Increased LDt voltage raised my PWMIC voltage. Might that help?

 

 

Anyone notice chipset voltage helped them?

 

So close, need 1 more small change and I'm there!

 

Thanks

Share this post


Link to post
Share on other sites

Well, I tweaked a few more things, but it's failing around the 3.5 to 4 hour mark. Same rounding error each time, different tests though.

 

I'm thinking maybe the motherboard is the problem? I searched around and rarely has anyone had issues with increasing LDT voltage (some but not many), where for me in makes the biggest single difference to run the Mem 1:1

 

Should I just be happy with 3.5 hours prime blend stable? It never crashes in anything else, but I want it to be 100% prime stable.

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now

×
×
  • Create New...