Jump to content

Vista

Members
  • Content Count

    254
  • Joined

  • Last visited

Everything posted by Vista

  1. Sharp, I think your theory is looking good. tRef 2048 appears stable, though Memtest shows a slight bandwidth loss compared to tRef 2560 (2823MHz vs. 2745MHz). Thanks for your help!
  2. I have a Dazzle Hi-Speed USB 2.0 10-in-1 Reader/Writer like the one pictured here. I plugged it in today and moved some images off a Compact Flash card onto my computer. I later rebooted and noticed that after POST and "Updating DMI Pool..." the system would pause with a black screen for almost exactly one minute before starting to load Windows. I unplugged the USB card reader, rebooted, and Windows loaded with no delay. The only other USB device plugged in is a Key Tronic keyboard. I know I tested that USB card reader when I first built the rig and didn't have this problem, so something must have changed. Well, it turned out that going into the BIOS Integrated Peripherals section and setting "USB Mouse Support" to "Enable" resolves the issue. I had recently changed it to Disable when I switched from using a USB mouse to a PS/2. If you are having USB issues it might be worth a try enabling both USB Keyboard and Mouse Support, even if you aren't using a USB keyboard and mouse!
  3. I like your theory. Everest shows the SPD Refresh Rate for my RAM as "Reduced (7.8 us), Self-Refresh". At 260mHz, one clock cycle should be about 3.84 nanoseconds. Working backwards, that means: 7800ns / 3.8ns = 2028 cycles The closest BIOS tRef setting to 2028 is 2048. I've already proven to myself that 3072 and 3120 are slightly unstable while 2560 is solid, so perhaps 2048 would be even better?
  4. You can stick a sensor on them (little thermocouple thing) and read them remotely. You can stick your finger on them and get a rough idea for $0. More expensive options are IR sensors, etc. I don't know how hot RAM can get, I suppose not much more than your CPU is good though.
  5. I just got back from Baton Rouge. Governor Blanco doesn't seem to have many friends at home, but again she was the lesser of many evils as usual (which was what made Edwards so appealing, an honest thief!). I agree we slided down the slippery slope of consumer safety long ago. Has anyone complained about RFI/EMI in the past 10 years? It use to be almost impossible to pass the FCC test for RFI emissions, I remember it cost $10,000 per test back in 1987. Now it seems you can run your case with no shielding at some ungodly gHz and nobody seems to care if it blows up your cellphone/cable TV/microwave/causes brain cancer as a side effect.
  6. Set virtual memory to the same size as your physical ram, with the max at the same limit. So if you have 1GB of RAM, set a 1024MB pagefile size and don't let it grow automatically. This will give you about 1.7GB of RAM playground shared between real and virtual RAM, which works well for most applications. Some people promote putting your pagefile on a different partition, or even a different channel than your Windows and Data drives. I'm giving that a shot for honeypot reasons, but I doubt there is much of a performance difference doing all that versus just formatting and making one big C: partition for everything. Prime95 Blend rags out both RAM and your HD at first, as it allocates everything plus some of your pagefile surplus virtual RAM. This is why it is such a crucial test.
  7. And never rule out, it could just be that your floppy drive is defective. I thought no way until it happened to me. Try swapping out that FDD if it's easy. $15+petrol test?
  8. I just whipped it up in Excel on my Dell while pondering the next BIOS setting attempt for stability on the DFI rig. I'm not 100% sure I got all the setting counts right, you have to include AUTO as a setting though to be honest. But within a few orders of magnitude, this is about right I think, even when you include the rest of the settings which are mostly binary. Feel free to correct my math, update for NF3, etc. LOL :cool:
  9. http://www.memtest.org/download/1.55.1/mem...-1.55.1.iso.zip This is the ISO you need to burn a bootable CD-ROM or DVD image using Nero, etc. It is the preferred Memtest 1.55.1 that is on the 331 and 414 BIOS, versus the 1.51 version that is on the 310 BIOS. If you try MemTest86 3.2 from memtest86.com, be aware it doesn't detect RAM speed and timings, though otherwise seems identical to 1.55 in stability. All that being said, Memtest 1.55.1 is very stable, and a good tool for preflighting your RAM/CPU/PSU before attempting Windows takeoff. Test #8 seems to be the most intense, so I just loop it. There are weird graph bar text overwrite problems on Test #8 using the built-in Memtest 1.51 on BIOS 310, while booting off a CD with Memtest 1.55.1 and BIOS 310 has no graph bar update anomalies. I haven't tried it using 1.51 booting off a CD, the text bar graphic glitches didn't affect the error-detection abilities of Memtest 1.51 on the 310 BIOS, just visual annoyances.
  10. How hard can it be to find stable BIOS tunings for a board? How many settings are possible? What are the chances of randomly encountering the perfect settings? This table is for the 310 BIOS: Well, it seems your chances are very slim to none. Just the top DRAM settings alone give you over a trillion possibilities. This is why auto-gizmos like nTune don't have a chance, and why Oskar Wu loses sleep and has grey hairs.
  11. By the time you finish doing enough research, XP 64 SP4 and drivers and games should be stable and some benchmarks will rock! And maybe the prices will drop on those San Diego's
  12. The pics in the dark are awesome. What digicam are you using? That's almost a good 64 toaster-sig with a tripod
  13. First off, I must say that link is incredibly helpful: http://www.dfi-street.com/forum/showpost.p...37&postcount=10. The only problem is that 16 out of 32 of the settings don't seem to have any interpretation other than the 100/133/166/200mHz settings. What sort of microsecond timings are these unknown 4 banks set at? I looked at it in base 2 and I still can't really make heads or tails out of the tRef scheme. So I tried each one using the time honored DFI Trial and Error method. At 260x10, every setting posted except for 0016. That crashed so bad I had to reset the CMOS. Looking at the chart, why would tRef 0128 work ([email protected]) and 3072 work ([email protected]), but tRef of 0016 ([email protected]) fail spectacularly? Shouldn't 0016 be more relaxed timings than 3072? Well, let's ignore that for the moment, perhaps Oskar really means Cycles when he says Cycles, so 0016 is super low and 3072 is pretty high. Now we will try the Voodoo tRefs, like 3120 ([email protected]?.?us). Seems pretty stable so far, but I can't see any bandwidth differences. I will just have to try it out to see if it beats the contender for my ram, tRef 3072. RGone posted a hint once to tease us: http://www.dfi-street.com/forum/showpost.p...13&postcount=14 Now that is written in RGonian, so without a new camshaft I can't generalize that into an equation that explains why the 0016 setting is so deadly vs. the other settings. For now it's just a mystery!
  14. Lots of people are trying it for beta testing new software releases, but not for regular daily use! Pro: Just about non-existant at this time since Intel just released the first 64-bit compiler that actually optimizes better than what everyone was futzing about with. So now you should start to see some more benchmarks that show a difference. Con: Same speed or slower than 99% of the things that will actually run. Compatability with 32-bit apps and drivers is chance at best right now. I would give XP 64 a long time to mature and for software developers to catch up before jumping off XP 32 Pro.
  15. Hehe, congratulations! I would have never thought of telling you to enable virtual memory (I was thinking hardware!). Your SmartGuardian volts look absolutely normal, they read a bit lower than a DVM would read directly. I'd run Memtest and Prime longer now and see how it goes. Also check your favorite games or 3D benchmarks.
  16. Just for grins, I artificially raised my ambient room temperature from 68°F (20°C) to 74°F (23.3°C) using the furnace thermostat today. My idea was to simulate worst-case summer temperatures now and preview how my OC stability will cope in a few months when the A/C can't keep up. The results were interesting. I had a hard time causing max-heat CPU temps of 41°C before with the XP-90, now I still can't break 43°C. That tracks nicely with the rise in ambient temperature. PWM IC and Chip Set temperatures rose more than ambient by a few °C but still not alarmingly: PWC IC at 53°C and Chip Set at 47°C max. GPU temperatures also went up about the same, maybe 4-5°C. On the other hand, Memtest 1.55 blew out one error on test #8 after 9 passes, but then settled down and reported no errors after 60 passes! Normally it would take 4-8 hours for this instability to show up, so it gives me a way to fine-tune the timings without waiting as long for the first error to pop up (I've been fighting this error for two weeks now). Still, I think I might pass the summer test with just a few more tweaks! Love this board
  17. I wouldn't ignore crashing in HL2 and failing Prime95 Blend. I think you can rule out the PSU now, and move on to testing your RAM. How about putting just one stick in? How about if you swap slots with the same two sticks? Just some ideas.
  18. When you are trying to eek out just a bit more performance while staying stable, every timing becomes sensitive to small changes. I've spent the last week trying to stabilize 260x10 with my RAM, and some things are just counter-intuitive. For instance, I was stable in OCCT for over 10 hours, but still got some Memtest #8 errors at 2 hours (sheesh!). So I dropped the vDIMM from 2.6v down to 2.5v! This generated a few Memtest #8 errors as early as pass 4. A few tiny DRAM setting changes in BIOS and now have passed 3 hours on Memtest #8 with zero errors and OCCT has been running for over an hour. Theory: TCCD doesn't need all those volts if you loosen up a few key settings, and will run cooler leading to better long-term stability. Another thing I noticed while OC'ing my 6800 GT... I can crank the GPU core up to 440mHz with 1.18gHz memory speed, but then OCCT gave me errors after just a few minutes! I turned it down to 400/1000 and OCCT was stable for over 10 hours, same everything else. So don't expect your GPU overclock to not have any effects on your CPU overclock stability!
  19. If you get a solvent that is too aggressive you can eat through the greenish insulation coating, but modern PCB's seem to have a better coating than in the old days. The PCB itself is epoxy/fiberglass, which is almost impossible to dissolve.
  20. Thanks Shinobi, I tried it out and it is worth adding to my stability test kit. What is especially different about S&M is the way it rapidly cycles CPU load between 50% and 100%. That is a different kind of thermal stress than just going for maximum heat and could uncover problems that just steady heat wouldn't show. It's a bit scary looking at the CPU temp graph, those A64 react so fast to load.
  21. Can you rig a fan in your case to blow on the RAM as a test? If more RAM cooling makes the long-term errors go away, you know it was heat buildup. Otherwise you still need to tune the BIOS.
  22. Exact same board, except the SLI can run two video cards in SLI mode. Same OC potential, etc.
  23. Since you aren't using RAID, you can probably get by without a floppy drive. You can flash your BIOS if needed using a bootable CD-ROM just fine.
  24. Looks sweet but I'd avoid the "NEC FD1231H-302 Black 3.5" Floppy Drive", as the one I got from Newegg was DOA. Get a Mitsumi, Samsung, or Sony instead for about the same price.
  25. Hi mbaylor, Newegg doesn't stock the best fans for the XP-90. I got mine from Jab-tech.com. Here are two models to look at. Panasonic Panaflo FBA09A12L1BX: http://www.jab-tech.com/customer/product.p...8&cat=80&page=1 Nexus DF1209SL-3 (recommended by SilentPC Review and me for low noise): http://www.jab-tech.com/customer/product.p...8&cat=80&page=1
×
×
  • Create New...