• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Tighten up my Ballistix

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

jason4207

Senior Member
Joined
Nov 26, 2005
Location
Concord, NC
I've got what's in my sig. Right now I'm sending 2.3v to my RAM. Enabling transaction booster in BIOS increased my RAM bandwidth from ~4800 to ~5200MB/s in memtest. I've just recently discovered I need to enable Static Read Control. I'll try that tonight.

My BIOS settings are here. Post #43

I'm basically trying to get my timings and bandwidth maximized. How much can I safely increase my RAM voltage to get the most out of them? I have the TT Armor w/ the 250mm side fan, so I have air blowing on them, but it's not like a few 40mm fans sitting right on the modules or anything. I've heard RAM can die fairly easily, and I don't want to risk that.

I read this article, and they are getting 8000-10000MB/s w/ similar FSB in Everest. Should Everest and memtest be about the same?

I'm pretty clear on the 4 main RAM timings. I've got them set to 4-4-4-12. I've tried 3 for the 1st two timings (at separate times), and it won't boot. But basically just try them out, see if it boots, and then check w/ memtest. How much can I increase the voltage and still be safe to keep trying better timings?

As for the sub-timings I don't know where to start. They're all on auto right now. Can anyone recomend a good starting point for them?
 
I copied this from my x32x0 OC thread. Hopefully someone here can help.

So far everything I do is causing memtest to fail, and I'm beginning to wonder if I have a bad stick. My 1st test may have passed, but I didn't take good enough notes. Those settings were 2.3v, 1:1, strap=333, 450FSB, 900MHz, 4-4-4-12, Transaction Booster disabled w/ relax level set to 3, and DRAM Static Read Control Disabled. Enabling DSRC, or TB is causing errors, and switching to 5:6, or 4:5 at 5-5-5-15 settings isn't helping either.

My results so far:
memtestjv8.jpg


The bandwidth reading is what memtest is reporting. That one that passed...I think it passed for over 8hours, but I didn't take notes very good on my 1st run. The one's w/ ? I only checked to see if they're weren't any immediate errors. I will have to go back and re-check them. Most of the one's that have failed have done so on the 1st or 2nd pass.

Any ideas on what I should try? More voltage, or perhaps less? Should I try 1 stick, and see if I possibly have a bad one?

Edit: I've updated the spreadsheet. I couldn't even get 900MHz 5-5-5-15 @ 2.3v w/ DSRC, and TB (relax set to 3) disabled to pass memtest! I'm guessing my RAM is bad. I know this is 800 RAM, but even its SPD table has 500 (5-5-5-15) 2.2v. I'm going to test 1 stick at a time, but I'll have to wait a bit. The wife is listening to Internet radio via my PC at the moment.

See sig for equipment in use.
 
well @ only 900mhz 2.1v should work fine for 4-4-4-12 you could try 4-4-4-4 and see if that helps the BW @ all or use a updivider and run them @ 500 5-5-5-x or even 600 depending on how the dividers work out
 
Well right now I can't even pass memtest. Do you think lowering the voltage might help this, or do you think it's a bad stick?
 
Try lowering voltage first, mine are running @ 2.1v and they actually are getting better scores at the lower volts. Check your timings with memset 3.3 this is what I run mine at with 2.1v

20070830memsetsb8.png
 
Thanks for the tips!

The wife is still cleaning the house for her guest this weekend and loves to pump the Internet radio (my rig is connected to the home theater). I'll try a lower voltage a bit later and report back. If I reboot now she'll be a little upset!
 
misterfix,
When I run memtest I'm booting direct from floppy so how can I use the memset settings? I will try to run at those, though, and see if I get better benches and don't lose stability in Windows. Is there a good program to test for memory errors in Windows?

I'm still having troubles. Haven't been able to pass memtest yet! I have tried every timing and sub-timing configuration imaginable...well close. I have to admit I haven't tried to go down to 2.1v yet (still running a 2.2v test, but I was at 2.3v several days ago).

I'll let you know how the 2.1v test goes if I fail the current one.

But so far i still get errors w/ 1 stick. I've recently reduced my FSB to 400 to make sure, but it seems these sticks can't even run at stock speeds at stock timings. 4-4-4-12-auto doesn't even work w/ DSRC, and TB (relax set to 3) disabled. I'm testing 5-5-5-12-(auto or something I found in a post) @ 800 2.2v right now. These sticks or the MoBo are messed up somehow.

It looks like RMA time unless anybody has any ideas.
 
Maybe you should contact Crucial for an RMA. If one stick works fine and the other one doesn't then I'd say your ram is fubar.
 
I didn't try every possible setting w/ 1 stick, but I'm pretty sure I tried 5-5-5-15 @ 900MHz 2.2v, and both sticks failed. I don't think I tested the single sticks @ 800MHz, but I will before I call Crucial. The pair just failed at 800MHz 5-5-5-15-6-42-11-4-6-2T DSRC disabled, and TB disabled w/ relax level set to 3.

I haven't had any problems in Windows yet. Also the fact I haven't even gotten 1 stick to work makes me think it might be the MoBo. I'll probably call Crucial 1st, though, b/c I know Asus will tell me to call them anyway.
 
Is there a good program to test for RAM stability in Windows? How do you guys determine if memset settings are stable?
 
Memtest is all I've ever used, but you can't check memset changes if you use that, right?

I can't get memtest to pass anything, so I'm just going to see what stays stable while I mess w/ memset w/n windows.

I'll check 1 stick at a time again, but I'm certain Crucial will be getting a call soon.
 
Memtest is all I've ever used, but you can't check memset changes if you use that, right?

I can't get memtest to pass anything, so I'm just going to see what stays stable while I mess w/ memset w/n windows.

I'll check 1 stick at a time again, but I'm certain Crucial will be getting a call soon.

I cannot seem to be able to get my 4x1GB to pass memtest86+ either above any speed above DDR2-800 speeds.
 
I use Orthos and set it to "stress some ram", if you don't test the rams one stick at a time then you can't really say if there bad or not. You may have a problem with the board where it won't run both sticks without giving errors in memtest. I don't like metest because it doesn't always give accurate results. I've had numerous occassions where memtest showed no errors but orthos would crash in seconds or my system would BSOD with a memory error.

Try running orthos from windows, if it doesn't crash or give an error stop message then your ram is probably fine. You did say you haven't had any problems while using windows, right? If you haven't had any BSOD's or crashing while playing games then it could be as simple as your computer dosen't like memtest.:eek:
 
Yeah, no BSOD, or problems. Overnight stable w/ P95 small fft. Should I try the P95 blend test overnight or maybe go w/ Orthos's "stress some RAM"...or both?

It would suck if my PC just doesn't like memtest...what a waste of time trying to trouble-shoot this!!! If I hadn't heard of memtest I wouldn't have known I had a "problem".
 
If your using prime then go with large fft's to stress the ram.

I'll try that tonight. These are my results from last night.

36ghztighttimingskv4.jpg


I went through a rigorous process w/ memtest, Everest, and P95-Blend last night.

Just using memset I was able to find the lowest possible timings that wouldn't lock up the PC, and allow P95-Blend to run at least 15min (I didn't test these settings longer than that). 4-4-2-4-1 4-20-7-9-9-1-7-2-4-2N

I was concerned, though, that even though they worked that may actually hinder performance, and Everest showed reduced bandwidth scores, so I started playing w/ the timings and then using Everest as a benchmark tool. I was getting odd results. Sometimes my write score would jump up pretty high when I reduced a timing. I'd reduce it another click, and the score would go down, so I'd put it back on the "sweet" spot, but when I retested again it was pretty much the same as the higher and lower timing settings. Inconsistent results to say the least. I tried using 3DMark06 as a benchmark, and it helped some, but it takes it a while to run, and it's results can vary a bit from run to run anyway. I did manage to get a new high (12,350), though!

Is there a good tool to test whether my timing changes are having a positive or negative effect...maybe SuperPi?

As you can see from the pic above I passed P95-Blend last night w/ these timings: 4-4-4-5 4-25-7-10-10-3-8-3-4-2-2N. They seem pretty tight to me, but I have to admit these sub-timings are all new to me. I am glad that I can set up my BIOS so that these timings pop up in memset...I don't have to use memset now unless I want to play w/ the timings some more.

Anyone care to comment on my methodology and can anyone offer any advice? Thanks! :beer:

Edit: I would also like to know if I'm OK since memtest never passed for me. I know GT wasn't able to get anything to pass except 800MHz, but I can't even get that. Has anyone else had problems getting memtest to pass, but otherwise their rig was solid as a rock?
 
Back