Ballistix PC8500

[Dawgdoc]

Blue rig in the sig.

Trying to OC this ram and tighten the timings.

OCing the ram itself is a snap. It will go SO much higher than I need, and am currently testing the difference on stock timings between 1:1 and 4:5.

From what I have read, it will likely not make any big difference, but Im giving it a go anyhow.

My question is really mostly related to timings. Im having trouble tightening these timings at all really.

Stock is 5-5-5-15, 2T. I can boot to windows (Vista 64 Ultimate) at 4-4-4-12, but havnt really gotten to test performance much yet on those settings. At stock 2.2volts and FSB 450 with 4:5 ratio it is memtest solid for 8+ hours.

At any timings I try tighter than 4-4-4-12, I either hang on reboot before POST, or sometimes (rarely) will POST and then hang.

I know this is a very common, and well liked ram. Can anyone give me some direction? Do i just keep trying random timings that are quicker than what I know works and use varying combinations until I find something that works?

Should I use a program like memset?

Ive gotten very comfortable with OCing intel quad and dual cores, very comortable with OCing Nvidia vid cards (man once I got the driver issues worked out it was easy), and am tinkering with AMD cpu OCs now......but....I still really havnt done much with ram timings.

Thanks!

[Gautam]

At 450 4:5 you're already at 563MHz...you won't get any tighter than 4-4-4 with any ram at such high speeds.

A quick way to increase performance is to lower tRFC in BIOS down to 25, and enabling static read control. (The latter might make your system unstable) Truthfully though the gains from the secondary timings are more or less unnoticeable. Memset really isn't needed as the BIOS has pretty much everything important. (Memset adds RAS to CAS Write Delay which gives a minor boost, but probably isn't worth the hassle)

[Dawgdoc]

Hmm....

Ok, so lets say I seem to be stable at 900mhz 4-4-4-8 2T.

Will I see reasonable performance gains for either regular use (translate=gaming :P) or benching or both by increasing the frequency further and keeping those timings?

Increasing the freq further and loosening timings a bit?

Thanks again!

[inkfx]

In day to day use you wont really see a benefit of higher speed or tighter timings. (maybe a few FPS) With benching, it's another story.

[Dawgdoc]

In day to day use you wont really see a benefit of higher speed or tighter timings. (maybe a few FPS) With benching, it's another story.


How much of a difference in benching marks would you say?

[thideras]

How much of a difference in benching marks would you say?Maybe a 100 points in 3dmark 06....I'd say what? .5 seconds in super pi...not sure.

Maybe a 0-2% increase if you lucky ;)

[Gautam]

Maybe a 100 points in 3dmark 06....I'd say what? .5 seconds in super pi...not sure.


Way less.

[Dawgdoc]

Cool guys thx for the info :)

Ill do a lil testing and see what I can squeak out inching my scores up as best I can!

BTW, since so many Bench team members seem to be responding to this thread......

I have heard/read that serious benchers have a partition that they use exclusively for benching? Does this help?

Also that some benchers boot to safe mode to perform benching?

Are any of these truths or myths?

Thanks!

[Just Learnin']

I use diagnostic mode for SuperPi benching. For 3D benching an IDE drive is helpful because it allows for greater increases in PCI Express Frequency. RAID for PCMark is better I find.

[Dawgdoc]

Cool guys thx for the info :)
I have heard/read that serious benchers have a partition that they use exclusively for benching? Does this help?

Also that some benchers boot to safe mode to perform benching?

Are any of these truths or myths?

Thanks!

Bump plz TY!

[Gautam]

A fresh install is always a good thing to have if you're concerned with getting the highest scores possible.

[Dawgdoc]

A fresh install is always a good thing to have if you're concerned with getting the highest scores possible.


Fresh install of the benchmark program or fresh install of your OS?

The statements regarding benching on its own partition or benching in safemode are accurate, or inaccurate?

TY!

[andyl33t]

Fresh install of the benchmark program or fresh install of your OS?
TY!

A fresh installation of the OS.

[Just Learnin']

Safemode is inaccurate. At least I've never heard of it being used in benching.

Defragging the OS helps tons as well.

Click Start, Run and type msconfig, click OK in windows and choose Diagnostic mode and reboot for SuperPi unless you know how to severely tweak your OS this will gain you some time. While SuperPi is open, ready to run, before you hit ok, open Task Manager and right click SuperPi under the processes tab, under Set Priority choose Real Time for 1M runs helps as well. for 32M runs it may be too much when at the edge of stability FSB wise, experiment with priority high though. Also right click it again and choose Set Affinity, experiment here with running it only on one core at a time to see which is fastest as well as most stable. Usually 1 core will stand out above the rest.

If you want after setting the Priority for SuperPi you may also right click the Explorer process and choose end task. Then run your SuperPI. This gives you one less running process while doing your SuperPi run and can shave a little off your time.Afterwards to get Explorer back press Ctrl + Alt + Delete and click the Applications tab in Task Manager, click New Task at the bottom right and type Explorer and click ok.

Do not use Diagnostic Mode for 3D benching. Usually for 3D and PCMark benching you should reboot for each run as well, Aquamark sometimes will give you a better score if you rerun it without rebooting in my experience.

REMEMBER TO DEFRAG!!!

[jason4207]

Although it may not help your performance very much, here's what I did to test how low I could get my timings:

I used memset. Basically memset will either let you lower the timing or it won't. If it let's you make a change. It will either work, or freeze up. Sometimes it takes a little large FFT to make it lock, but if it doesn't lock under these conditions then you should be good-to-go.

Take notes along the way b/c if your system reboots you'll have to start all over. I like to change 1 timing at a time, and then run a little large FFT to test for stability. Usually I just go for a couple minutes b/c it seems like if it won't work it happens pretty quickly.

Then I run Everest to check bandwidth and latency and take note of everything. Everest's results can vary, though, so sometimes a few runs are in order. Sometimes lowering a timing actually makes your Everest score go down, so keep that in mind. You can also run 3DMark to see if your score is affected. 3DMark takes a while, though, so I tend to save this for the end to see more of an overall result...not perfect I know, but I only have so many hours in the day.

I get everything as tight as possible, and then get some final readings. I then repeat this process w/ a different FSB;DRAM ratio.

Compare the Bandwidth and 3DMark scores from the tightest timings you can achieve at each FSB;DRAM ratio. Then you can decide the winner!

I did this at 1:1 900MHz 4-4-4-4-5 4-25-7-10-10-3-8-3-4-2-2N

and 5:6 1080MHz 5-5-4-4-10 4-25-7-10-11-3-8-3-5-2N

I didn't try very hard at 4:5 1125MHz, but 5-5-5-15 (as set in BIOS) was freezing up w/ large FFT, so I just went w/ what I had.

The difference b/n 1:1 and 5:6 was negligible. Less than 100pts in 3DMark06. I remember getting 12,360 at 5:6, and 1:1 was less, but it was still over 12,300...can't remember off the top of my head. I have notes at home.

I then started figuring out if I could get the timings I wanted w/o using memset, so I wouldn't have to worry about setting it up each time. I was able to do it by making BIOS changes, and then opening up memset and observing what came up initially. I kept making changes in the BIOS until memset opened up w/ my "perfect" settings. I was unable to get the 4:5 settings show up right b/c I didn't have control over that extra main timing (setting 5-5-4-10 in BIOS resulted in 5-5-4-5-10 in memset instead of 5-5-4-4-10 which I wanted). There was 1 sub-timing in the BIOS I had to set to 3 to get to show up as a 10 in memset. There was an obvious "7" skew. I set the tRFC to 25 in the BIOS. All other sub-timings set to auto in BIOS made them show up as my "optimal" settings in memset. I set the main timings in the BIOS to 4-4-4-5, and I had my "optimal" settings upon boot w/o having to load up memset anymore.

I ended up backing it off to 4-4-4-6, and then to 4-4-4-8 later after a few game glitches.

Hope this helps! Good luck Doc!

:beer: