WheresWaldo

SN I looked at your results and even tried your settings, but they all come at the expense of memory bandwidth. With the TREF settings you use your memory bandwidth is lower than the bandwidth settings I can get from running at stock settings. So the only thing you are gaining is the CPU clock. Granted the A64 benefits more from CPU clock than memory settings, but your settings put you at the extreme low end of settings. I would think you would want a balance between the two. I'll play around more with this.

Try swapping the slots the RAM is in and see if that changes anything. I found when they are in one particular configuration I get alot of errors in memtest but the other way around I only had 16 errors in test #8 after 65 passes.

SD, Please put your system specs in your signature, and there is a complete thread on the PDP XBLK ram here . Search is your friend.

Rid This is not true for nF4 based boards. The problem was the the SATA speed was not locked on ports 1 and 2, that is why you had to move to 3 and 4. The nF4 chipset does not have this problem.

Demonwolf There is a complete thread on Patriot XBLK along with tons of info on all the BIOS settings, forum search is your friend. http://www.dfi-street.com/forum/showthread.php?t=9451

Jimbo, The numbers you get for 512MB sticks are suprisingly close to how I calculate Tref Your formula = (4/MEM)/(1/FSB)*1000. How was this arrived at and does the XS thread mention anything about refresh rate for particular sticks. The formula that has been discussed on this forum is REFRESH/(1/FSB). Many sticks have and SPD refresh rate of 7.8us so the math with yours worked out for sticks of 512MB @ 200 are (4/512)/(1/200)=1563, the formula I am using taking into account refresh is 7.8/(1/200)=1560, but if the RAM was built to refresh significantly faster, let's say 3.9us then you still come up with 1563 and my results show it should be 780. Just curious.

As CC mentioned some panaflo's only use the third wire to sense on and off, not rpm.

Ok, had another interesting problem with 510 BIOS. Windows hosed so I had to reinstall. No matter what I did I could not reinstall XP SP2, it would always hang at the preparing installation and 39 minutes to go. I reflashed to 414 and the install went fine. I think I am going to stick with 414 since it is the most stable in my other rig (in sig).

OK, did some more testing with TREF settings, you cannot use a figure less than what you calculate and remain stable, you can and should use a higher setting. So if you are running 300 FSB then the calculated TREF is 2340, the closest TREF is 2336 but I have found it is not as stable as 2560. or even 3072. I have seen much better results with 3072 than with 4708 (?).

Okay I have been doing a lot of playing with BIOS settings and was having some issues with MEMTEST #8. I usually had 3 to 5 errors per pass. It was stable in windows, but would crash after 1.5 to 2 hours playing Doom3. So I saw the TCCD guide posted on XtremeSystems and tried their BIOS settings. with it I was able to reduce but not eliminate the MEMTEST #8 errors. so I started looking for other culprits. I stumbled across the TREF thread on this site and took out my calculator to see what TREF should really be. Since PDP RAM has a SPD refresh rate of 7.8us and I was running at 300 Mhz the math was TREF = Refresh / (1/HTT), so in my case it works out as TREF = 7.8 / (1/300) = 2340, since the closest setting is 2336, that is what I set. and now I get 1 error every other pass of MEMTEST #8. At this point I wanted to see if there was any more I could get out of this 3500+ Claw, so I slowly ran the bus up to 312 (note that 312 x 9 = 2.81 Ghz). Still only getting the same 1 error every two passes but I also changed the TREF setting to 2560 this was the closest setting to the 2433 calculated TREF. Then I throttled back the FSB to 310 so I could gain some stability. All this said the RAM settings now are as follows: FSB 310 (highest game stable setting) LDT:FSB x3.0 LDT Bus 16 16 CPU:FSB x9.0 PCIe 101 CnQ Disabled CPU VID Startup CPU VID Control 1.375 CPU VID Special 126% LDT 1.30v nF4 1.60v DRAM 2.90v DRAM plus .03 Enabled DRAM Freq 200 (1:1) CPC Enabled Tcl 2.5 Trcd 04 Tras 09 Trp 04 Trc 08 Trfc 14 Trrd 04 Twr 02 Twtr 02 Trwt 08 Tref 2336 Twcl 01 DRAM Interleave Enabled DQS Control Increase DQS Value 255 DRAM Drive Level 7 DRAM Data Level 2 Max Async 08.0ns Response Fast Read Preamble 05.5ns Idlecycle 256 Dynamic Counter Enabled R/W Queue 16 X Bypass 07 X Granularity Disable This was all with the 510-2 BIOS!

My problem was that 510-1 and 510-3 would reboot in WinXP as soon as all the startup apps finished loading, and there aren't that many. I was testing with Optimized BIOS Defaults loaded, not even overclocked. One of the 1st things I do is disable Windows services after installation and only run with necessary ones. 510-2 does not do this. I am now running 510-2 witrh the settings I previously posted except that I changed DRAM voltage from 2.9 to 2.8 + .03 Enabled. This resulted in a DRAM voltage reported in BIOS as 2.86 instead of the 2.93 I was at before. It is Prime95 stable over 8 hours, I have completed SuperPI up to 16M, but it will only run 14+ minutes in OCCT. I think loosening the RAM timings a bit and I will be able to pull another 50Mhz out of it. Everest 2.x has Read Speed at 8200+ and Write speed 2700+ with a latency just over 34ns, all plenty fast enough.

Back up system is currently under test with 510-2. Will let you know if this is a good combo with PDP XBLK.

I might try 510-3, right now I'm back to 310P on my 2nd system and 414-1 on the system in my sig. everything is very stable that way. Don't see a real need to move away from these BIOSes.

I tried using the 510-1 BIOS on the machine in my sig, and it would spontaneously crash and reboot from WinXP even at stock speeds.

robowang, do you really find that increasing these two voltages are necessary?