Adjusting performance based on temperature Ram scaling according to ambient conditions
Adjusting performance based on temperature Ram scaling according to ambient conditions
Yesterday I planned to experiment with my extreme dark DDR2 1520 OC. I aimed for a stable 1520 speed without timing issues so I could record a video. But when I adjusted the last settings, it didn’t work even with extra cooling. I kept trying different profiles—1520 7-9-9-12 at 2.26v—but still faced errors in Prime95’s FFTs. Raising the Nb Volt helped a bit, but not enough. It still crashed after just a couple minutes instead of the expected 30 minutes. I kept tweaking Nb and FSB volt, but nothing stuck. By midnight I gave up and went to sleep.
Today I moved the fan closer to cool the RAM more, hoping that would fix things. The Nb Volt stabilized the RAMs, but it still crashed after two minutes instead of the usual 30. I adjusted Nb and FSB volt again, but the problem persisted. Eventually I realized it was the cooling issue. I relocated the fan and added another one to blow over the other side. Suddenly, 7-9-6-10 ran smoothly at 2.34v with the settings I left off. That worked! I used an Intel fan because that’s the only small one I had. The other side stayed cool even without direct airflow. It didn’t matter how much I increased the RAM voltage—it still didn’t help. Eventually, I tried a higher setting (7-9-9-12) but it crashed after just a couple minutes. I kept fine-tuning until I figured it was all about temperature.
The rams didn’t heat up much on the side without airflow, so I didn’t think cooling was the main problem. Maybe some RAMs are more sensitive to heat than others. Even adding more RAM voltage didn’t solve it. I guess if you can’t get higher speeds despite tuning, your temps are likely the bottleneck. I’m planning to try a micron D9 after setting up my X58 rig with a triple-channel RAM for the fan comp. Hopefully I can hit dome D9gkx for better frequency. For now, I might settle for high-frequency overclocking (CL5/6) if I get my X58 ready, but I’m not sure if 2.95v will work.
If the 7-9-6-10 profile holds up longer, I’ll keep tweaking until it runs smoothly or it crashes again. Until then, I’ll focus on getting the timing right so secondary and tertiary timings don’t become limiting factors. I’ll also add a 3000rpm CPU fan to the RAMs for better airflow.
I’m hoping my 8 hours of trial and error won’t waste anyone else’s time. Thanks for reading!
The content lacks depth and focuses on outdated topics. Most discussions revolve around obsolete hardware and we don’t even have clarity on what you’re using. We’re dealing with a DDR2 system, 3GB in dual channel, running Prime95. If you’re trying to boost performance, consider benchmarking with PiMod 32M rather than relying solely on Prime95 for overclocking attempts.
Here are some RAM benchmarks you can use: consider stable options with acceptable timings, even if voltages aren't ideal. At around 1300 MHz, many people on Warp9 report poor performance due to outdated timing data. While your rigs handle speeds like 6-8-5-10 and even 1100/4-4-4, the lack of updated benchmarks makes it hard to judge their real-world stability. A voltage around 2.2–2.3V is typical, but newer models can match or exceed that with better efficiency. At 1520 MHz running at 2.34V still seems reasonable compared to older specs.
I recommended using PiMod 32M for CPU and memory testing. As you improve your performance, you'll notice better results. Experimenting with various processors can help; consider grouping the CPU and memory controller together. Sometimes a chip performs better with memory. I suggest starting around 1200MHz for the CL4-5-4 or aiming higher for memory speed. Those experts at Warp9-systems have solid expertise! I was browsing their forum and found a wealth of useful details.
I own a Blue G.Skill PC8500 with 5-5-5-15 sticks that exceed 600FSB at stock speeds around 2.25-2.35v, using a Rampage Formula with an E8600 E.S. I no longer have the CPU. My machine is an x3360. I don’t focus much on 775 boards anymore; I keep it just because it worked well back then. It’s a great board overall, though it lacks DDR3 support. I have some solid DDR3 options now.
These enthusiast DDR3 775 boards are tough to find, but with a solid DDR2 setup I don’t see much trouble. Theoretically, I could run at 760MHz stable if RAM is set to 1520, though the board itself isn’t built for regular daily use—unless it’s an office or budget gaming machine. With over 4GB RAM it seems unnecessary. A 1200C5 at those voltages is normal for D9GMs/GKX models. Have you ever experimented with higher voltages for better speed or timing? D9s are supposed to handle voltage changes well, but on an ASUS board you won’t push anything extreme even at max voltage with RAM. The safety feature in ASUS boards helps prevent damage if you go beyond the limits.
It can act as a cap for very high RAM usage, but since D9s typically stay under 1400 from what I've observed you probably don't need to boost anything much—just a slight increase or two—to handle 1520. I had to bump my nb to 1.58v instead of the usual 1.56v, which cut my runtime to about 478 fsb.
It seems you're discussing the performance of your x3360 chip. You mentioned it reaches its peak around 4100MHz and that you can push it a bit further without stability.