Only one core should be at a high frequency - all others should be in Sleep or a few hundred MHz. Post a screenshot of RM during the run - simply drag-n-drop the image into your Reply. Please run a Single core CD R20 after uninstalling all junk applications, including HWinfo, and doing a Clear CMOS. How do know that? Please post a link to a document/site. I do not know that the cores are ranked in the factory. Four cores are running, the fastest running at 815 MHz. Your Ryzen Master (RM) screenshot has no load so shows little. It is hard enough to keep up without your changing tests for no reason. Where did you get these numbers? Please do not run and uninstall HWinfo - it is a bad utility and probably part of the problem, whatever the problem may be. I do not understand what you mean by chiplet (4500-4650MHz) and chiplet (4300-4400MHz). Ltron, I do not know what you are complaining about. I ran Aida 64 Memory Latency test which loads core 0 and gives me 4.6GHz+ so I could show the CPU is capable of these speeds in the screenshot.ģ2GB Crucial Ballistix CAS 16 DDR 4 3600MHz CPU-z is fine at 541 single core (it does its own thing and uses core 0 which is my fastest core so that's why it's not affected by the ranking) and Geekbench 5 is fine probably for the same reason and I get 1315 single core. This CPU core ranking is replicated across both an MSI X570 ACE and a Gigabyte X570 Master and multiple BIOS versions and AGESAs. Is there a plausible explanation for this ranking being correct and does anyone else have this problem? I'm asking for educational purposes, not that I can fix it as that's up to AMD (hopefully they can fix it if I can determine the problem is not unique to me) and I certainly wouldn't RMA a CPU over this. I think someone may have made a mistake in the factory and ranked the cores wrong. I have confirmed that all cores in CCD 0 do at least 4500 MHz by disabling CCD 1 (which resulted in 521 in Cinebench R20 and 210 in R15) whereas all cores in CCD 1 do at most 4425 MHz. #Cinebench r15 vs r20 windowsI have confirmed this ranking in Linux too so it's not a Windows issue and all screenshots I've seen from other people have shown CCD 0 correctly ranked above CCD 1. Can you check this on your CPU, particularly those with low Cinebench scores? Here is a HWInfo screenshot and a Ryzen Master screenshot (attached) so you can see what I mean, in HWInfo the #numbers on the right are how the cores have been ranked in the factory and written into the firmware and the #numbers on the left are what Windows thinks the ranking is. Other people get 520-535 from what I've seen. When applications are not using my fastest core 0 (correctly ranked) they will use a core pair that are ranked #2 and #3 fastest (incorrectly), these are cores 6 and 8 from the slow CCD 1 chiplet and they are only capable of 4.425GHz max which results in low Cinebench R15 and R20 single core scores of 205 and 510 respectively, multi core is fine at 32. I have a strange issue with my 3900X, on average the cores in the slow CCD 1 chiplet (4300-4400MHz) are ranked as faster in the CPU's firmware than the cores in the fast CCD 0 chiplet (4500-4650MHz).
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |