OCZ Prime Stable Database (updated 2/06/2006)

[Miguel]

Frallen, I have a similar setup to you, with the same 0450 winchester. I can acheive about the same maximum clock (2.75GHz with >1.6Vcore), but these 0450's have picky memory controllers. I found that I had trouble getting decent memory clocks and timings at anywhere near 2.7, so I found a bit of a sweet spot at 2.6GHz. (10x260) No matter which way I swing it, I have to have the timings at 2.5-4-4-8 1T or looser for stability. Vdimm does not seem to have a huge impact. I have tried out my memory with a low multiplier, and it is memtest stable for hours at 305MHz 2.5-4-3-7, too bad my chip wont let me do that with a decent clock speed. 9x300 has been a pipe dream for me, and I have never been able to acheive anything longer than 1 minute prime stable at that clock. I'm seriously thinking about picking up one of those new 'AS' core clawhammer 3500+'s

[Frallan]

Well Im game stable now which will last me until the Venice cores come out and then I might even have a chance to move home so I can use my Vapo...

[Guest hazard]

ok lets try this again.. now that my info is at the bottom of the scream im thinking i have one dead ram slot.... being cause when i try to run dual channel it just beeps at me .... any idea or should i just take the board back to the store

[flexy]

Test 8 in memtest is a tougher test 5 basically.

 

I always use the the Toture Blend Test in P95. http://www.mersenne.org/prime.htm has more info.

 

In a nutshell, Tref is the alloted time for the board and chipset to clear data for new data to come in. This is why a bad Tref causes errors everywhere and not only test 5 and 8. It's not specific to memory, and the wrong value hoses everything.

 

Hope this helps.

 

i HARDLY (if ever) run prime anymore..OCCT is so much nicer and AT LEAST as sensitive as prime. I set the tested memory in OCCT to 512MB.

 

Problem with prime blend is that it uses (for my 1gig system) about 900something MB....which is insane...basically page swaws constantly...and no wonder this gives problems....dunno...just like OCCt more.

[ed_jacobson]

Andy, this is a non TCCD issue but involving OCZ memory and could use a hand here. I am getting errors in Memtest 8. I'm running the 3/10/05 bios with this OCZ memory http://www.ocztechnology.com/products/memo...00_dual_channel

 

I am running fine at stock speeds. This is my first computer build, and I am new to overclocking. I mostly want to run at stock speeds, but want to try OCing a bit and try just for a 10 percent OC at this time. I can test the HTT limits of my CPU by increasing the HTT with a variety of memory timings as long as I reduce the DRAM Frequency to 100 1:2. With these settings I can get into Windows and run Prime95.

 

However, whenever I both increase the HTT (for a mild overclock in the range of 210-225) and keep the DRAM Freq. at 200 1:1, I get errors in Memtest 8. I have tried noodling with this thing...I've tried the stock settings, I've tried the settings you mentioned in Post 25 on this thread, including multiple variations on Tref, DRAM voltages ranging from 1.6 to 1.8, and I've tried Rgone's settings from http://www.dfi-street.com/forum/showthread.php?t=5844 and variations on those settings, and all to no avail.

 

I'm new to overclocking, but I've read up on it quite a bit and feel that a mild 10% overclock shouldn't be a problem, and I can't figure out why I'm having such problems.

 

Here are a few examples I've documented:

 

1. Rgone's settings, HTT = 210; Result= memtest 8 errors.

 

2. All SPD settings, except

FSB Bus Frequency 225

DRAM Freq = 200 1:1

LDT/FSB x4

CPU/FSB Freq x6

CPC Enable

3

4

10

4

 

Result = cannot boot into Memtest

 

3. All the same settings as in 2 above, except:

 

FSB Bus Frequency = 210

 

Result = Memtest 8 errors.

 

4. All stock settings except changes from Post 25:

 

2.5

4

6 (also tried 10 here)

3

 

LDT 1.3v

Chipset 1.7v

VID 1.3

Special VID @ 123%

Tref at 2560, 3072, 4708

 

Result Memtest 8 errors.

 

5. I've also tried to isolate the memory by setting

CPU/FSB = x6

LDT/FSB Freq x 3

DRAM Freq Ratio 200 1:1

HTT = 210

DRAM Volt = 2.6 - 2.8

 

Result = Memtest 8 errors.

 

If I try to boot into windows with variations on these settings, one of two things usually happens: (1) Windows cannot load, and I cycle back through the POST; or (2) I get a BIOS ROM checksum error (this usually happens when the timings are tight, such as at the stock ratings).

 

Any ideas on how to resolve this?

[Guest hazard]

ok well i guess ill just call the dfi phone number again and see if they have anymore suggestions . i was thinking my ram is incompatible but i wanted to get some second opinions but no one seems to notice ....

[kuniva]

Not TCCD but a memory problem none the less thats only existent on my DFI SLI-D, PQI Turbomemory PC4000, Hynix D5 chips, capable of 275MHz+, 480w PSU - +12v 34A, board rev A02

[OCZTony]

Guys i feel i can help a little here.

 

TCCD likes a weak drive strength, in bios you have 8 options, what you don't know is the 8 are infact split into 2. All the odd options and all the evens, so 1 3 5 7 and 2 4 6 8..now i don't know what is high and what is low but i do know you need to stick with one or the other...don't go from 5 to 6 as your going low to high etc.

 

I find 5 gives me good results with TCCD and 7 does also... i have a feeling 7 is the highest option on the weak setting with the lowest drive being the lowest # choice.

 

So to move on you also have a data drive option, Ryan our CEO found setting 2 can really help...its worth a shot.

 

Next up TREF, this has to be the MOST important setting there is. Reason for this is if you get the right TREF all the other settings fall into place and the ram clocks well....if your a little ways out your going to struggle.

 

I tend to play with 0648 to start, Oskar has his own option in place and this works well also, thing is play with it...it makes a hige difference as i said.

 

TRC and TRFC....setting 7 for TRC... is NOT a good move. i do this.

I set TRFC at 24 and TRC at 12, i then work TRC down to see the bandwidth changes in memtest. i then set the highest TRC than gives me the most band width, IE if 7 8 9 10 all give the same gain i set 10.

TRFC can dial in instability at high clocks, so work down slow from 24 and test each setting. i usually just leave it at 24 now as im lazy

 

Idle timer....I like 32 and 64 for performance and 256 for high clocks.

 

Read to write and write to read...again huge performance options, keep these as tight as you can but dial in stability by going loser.

 

TRAS...just like TRC set 10 and go down the way, find the spread where you get the best bandwidth and set the highest value.

 

cas3 can really help also, weird but true on the higher clocks.

 

If you are running 2.5-4-4 with TCCD your wasting your time, the memory controller is low latency and you need to feed it nothing higher than 2.5-3-3 or 3-3-3. Now i know this will reduce your clocks but it will speed up your games. I always overclock the LDT bus, so don't be afraid of running the ram in 166 mode with a high 300+HTT so your getting the ram speed up. I like the 183 mode also but its harder to get the LDT over 1K with this.

1:1 is the best and running 4X over 250fsb means your overclocking LDT, ram and CPU

 

As i think of more I will post here

[flexy]

Guys i feel i can help a little here.

 

VERY helpful...thanks a lot !!!

 

room for more experiments

[kuniva]

thanx that will prolly help a good deal of ppl but, omg, is it really so much to ask of the memory to just work when its banged in the slots nowadays? im not blaming the memory manufacturers because its rare something their end goes wrong, im saying board manufacturers should take tme to TEST what memory will WORK in the board from the major manufacturers, i remember my old K6-2 setup, even an unknown brand PC133 module ppl could just bang in back then and the thing would run like a champ still, i mean come on, memory testing is important uknow.

[Frallan]

Well Kuniva thats the price U pay to be on the bleedingedge of things

[kuniva]

i know, i wouldnt be so pissed off if memory compatability tables were easily and readily available, i looked for a good hour on the DFI site and still didnt find that compatability table and google turned up nothing as well.