![]() ![]() At least somewhere was tested golden 5950x which showed better results due lower power consumption. In the results (CB20 and membench) all background tasks affect negatively so have you checked active background tasks? Also motherboard vendor settings (default limits etc) and CPU silicon lottery affect as you said. Some tests (hardware unboxed etc) show only slightly bigger numbers. Personally I believe you dont have same issue at least in that extent. No idea how common issue these are (or not), is also discussions about RAM issues as well. Start thinking about more airflow if you hit 50C in HCI. ![]() HCI catches Geardown off instability really quickly, so if it ends up being stable in HCI, you'll know what was up with the instability.Ĥ0C is cool as a cucumber. Like, membench literally includes HCI.exe and uses that in its extremely short and light "stability" test, so you may as well use the actual program instead of a half-assed quickie in membench. Better yet, download HCI Memtest and run a stability test for about an hour - the free version limits you to 3GB of RAM per instance, so just keep opening instances of HCI until you can fill about all of your RAM save for 2GB or so left for Windows, and test all of the instances concurrently. There's plenty of room to tighten timings from what they are right now, after you get rid of the instability. If you really want the 3 seconds, you can try to pump more voltage or mess with Clkdrv strengths, but I recommend neither at this point in time. It costs you about 3sec on Easy and a bit more on Default, but the stability boost is massive. It's not a memory benchmark or stability test and gains nothing from faster for starters, turn Geardown on. Here's some old 4Gb E-die that behaves similarly to B-die:Īnd stop posting your Cinebench results. So get your timings in order now, then wait till the software and firmware are updated, then try again. From the custom latency of 3ns, it's pretty obvious that DRAM calc hasn't been updated to reflect Zen 3's 8-core CCD. Ryzen 5000 does appear to behave a little differently on the memory side than Ryzen 3000, more so than AMD initially let on. ![]() In this case, 277-300+sec is pretty damn terrible for 3800CL14.ĪIDA doesn't care too much about secondaries, but membench certainly does if your timings are that loose.Īfter you properly spend time on the secondaries, I wouldn't be surprised if this all gets better once a more mature AGESA firmware hits the shelves. There are some minute differences in performance between different core count SKUs, but membench results are generally roughly comparable to see where your memory performance lies. At the very least, you should be able to do 4/6/16, 4/12/12 and 350 or so without adding voltage to what you're already using now. Some things here - your RRDS/RRDL/FAW and WTRS/WTRL/WR are looser than a bucket of fishing worms, tRFC is stupidly high for 1T B-die. ![]()
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |