Poor Low Cinebench R23 results with a Ryzen 7 5700X in stock condition.
Poor Low Cinebench R23 results with a Ryzen 7 5700X in stock condition.
It's worth noting if anyone assumes it's worth it because, haha, I'm going to say vroom for cheaper than the new 3200 C16. Even then, a generic crucial should be around 3600 C16 (16-18-18) with 1.35-1.4v lazy OC. So you don't really need to push crazy speeds or put in a lot of effort to take advantage of these cheaper rams. I'm also thinking about Hynix DJRS since the 5200+ looks really tempting, but it's not as consistent as microns. No guarantee you'll get 4600+ unless your board and CPU are stable.
For smooth benchmarks those velocities won’t stay constant all the time, as long as the board and CPU can handle them. I’d just use a B-Die and skip the complications. B-Die works better and runs consistently. In the “Daily Speed Zone” it’s the top choice because timing will be much tighter. You might think it’s just for bragging, but the results are real, measurable, and repeatable.
I really prefer quick RAM boosts because it feels great when the performance jumps without a big increase in power use. That’s not true for CPUs or GPUs (except VRAM). Also, pushing fast RAM (laughs at 1520 7-9-6-9 DDR2) is something I’m used to doing. As someone who likes saving money, I’ll go with a cheaper chip so I can get similar speed and power efficiency. Just because it’s cheap doesn’t mean I won’t upgrade if it helps me outrun the competition (lucky high bin DJR). Unfortunately, the voltage limit at 1.75V might be tight for something like 5000, but it should still work depending on the model. I probably won’t find any really affordable options with high performance and low cost (most likely just a rev because I’m doubtful about running 5000+ 4x8 on an ASRock X470, and 1.8V feels a bit too much for daily use).
The cost of breaking sync for fast RAM is significant. You should aim to cover the gap, but you’ll likely need to operate near 2500MHz, keeping your FCLK within its limits. Remember this when planning.
Even using Micron Rev. E or DJR, you'll probably get better results by fully utilizing the IF and avoiding desync issues. Memory latency doubles as soon as you lose sync, and aside from a few tests that mainly check bandwidth, running at 5000MT/s still leaves you with performance losses compared to 3800 with decent timing. There are affordable B die options available if you know where to look.
Cezanne chuckles lightly. The main worry might be exceeding 5000 units, as these Cezanne apus usually cap around 2500fclk—especially if the good DJ reaches 5600+ with proper setup and the needed voltage (1.75-1.9v), which itself feels a bit high for everyday use.
The VRM temperatures reached a maximum of 51°C, which makes me question if it’s the main problem.
Implies complete inventory, the VRM could warm up slightly when PBO is activated.