That cooler has a 120mm radiator so it will saturate pretty soon but you should get a few runs of CB20 before the clocks drop precipitously. With all that set up you should be seeing it boost single cores to 4.4 Ghz and run all cores closer to 4.2Ghz in heavy loads. You should be running either the Ryzen balanced power plan (installed with the chipset drivers) or the 1Usmus balanced power plan. You want to get updated to v1909 which has a processor aware scheduler. Probably not a problem if you recently installed it but you might want to run Windows UPDATE and let it install any updates for Windows. Leave multiplier and voltage settings on AUTO. And of course, set up your memory and be sure Infinity Fabric is linked to memory clock speed. Once you have that settled Global C-state Control, Processor CPPC, CPPC Preferred Cores, and AMD Cool'n'Quiet should be set to "Enabled" in BIOS. Second but equally important would be to get the latest chipset drivers from AMD's web site and install. CB23 is an all core, heavy threaded load.How have you set up your system? First and foremost you should update to the latest BIOS since early BIOS' did not well support boosting for Ryzen 3000. My CB23 peaks all cores at 4500 or so but it goes to 4750-4790 gaming or doing other lighter threaded apps. CPU-Z is the best, short term test for a stress test of just to see how the boost is working or not when using weak cooling or stock air cooling. So seeing a peak of 4.1-4.5 on all cores maybe less depending on how quickly it senses the temp increasing, is totally normal. When you run CB23 in multi-core, you're putting the CPU into an all core load and it will accordingly add voltage/reduce clock speed to keep you under the 95c max and to keep you stable. If the CPU is boosting at all to 4700 or more it will do so only on a few cores at a time, not all. Your setup is not totally correct, but it's not off enough to even complain. The short answer is ,no your CPU is NOT defective. A few hundred points or even a thousand is no cause for alarm especially when the guy's that run those tests are professionals with better cooling and most likely a better binned CPU. Your scores now aren't too far off from that article anyway. If it crashes, you'll need to back the +100 to say +75, then +50, if it fails to do +50, leave that extra boost off and use regular PBO and the Auto OC alone. You can try enabling PBO and +100 for starters with the SOC manually set to 1.15 and see how that works. Sounds like you're pretty stock though and that type of testing, you'll want a better cooler to hit higher numbers. If you run higher like I run 3933/1966, you need the full 1.2v SOC and some higher other voltages like CCD, VDDP, IOD all at around the 1.05-1.07 range. Set you SOC to around 1.15v for any RAM or FCLK under 3800/1900 respectively. I think you need to stop using Ryzen Master, it's total garbage and go back to old fashioned BIOS. I have a 5600X but he concept is the same. I use a Corsair H110i GTX AIO and my temps stay around the 79-80c range in CB23 multi running PBO +100. I switched back to my old ThermalTake TG-7 paste and temps were back to sanity levels. The CPU was reading ~15-20c higher in general. I found that using Thermal Grizzly Kryonaut, I was spiking quickly to 80-86c in testing. Mine can shutdown at 80c but it depends if it spikes to that or if it gradually gets there. I am using a Scythe Shuriken 2 cooler, which at 58 mm may not provide the best cooling. The idle temperature of the CPU is around 42 Celsius (on Power Saver), but it does seem to heat up rapidly. The computer is in high performance mode. To run these tests, I created a new account and disabled as many applications as possible until the CPU usage was at basically 0%, occasionally fluctuating to 1 or 2% when various Windows processes needed it. If I run the multicore test a dozen or so times in a row, it does approach 90 and start throttling, logically enough given my choice of cooler, but this does not explain the low initial results. It is not thermal throttling during the first part of the multicore test, since the temperature stays below 85 degrees. This is much lower than the 28800 that this chip got in the article. It only boosts to around 3.7 GHz during this test. Similarly, the multicore score is around 25102 points. Still, this does not appear to be the problem with the first run. After the first run, which takes about five minutes, it does seem to start thermal throttling, since the frequency that I see in Task Manager drops a bit, as does the score. The CPU never boosts much above 4.7 GHz in Task Manager during this testing. The single-core score in Cinebench R23 tends to be around 1560, which is much lower than I would have expected given that in this set of test results it got a score of 1694.
0 Comments
Leave a Reply. |