Видео

I haven't tried it myself but before launch I was told it's a Zen 5 exclusive. However, seems the option is available for Zen 4 on some motherboards and it works too. So, I think it does work with Ryzen 7000.

Thanks for the kind words. I'm happy my video helped! As for CO testing: don't forget to mix workloads when testing. A common error is to test CO only with stress test tools. Those tools tend to run high temperature and moderate frequency. But sometimes CO is unstable at medium temperature and high frequency, or high frequency and high temperature. I found that looping the Y-Cruncher component stress tester can sometimes also uncover CO instabilities.

I run AIDA FPU/Memory/Cache stress test to detect unstable CO. On settings that Cinebench runs for infinite amount of time without any problems, this test crashes my PC in 5-10 seconds. I needed to lower CO by 15 points to make this test run for some extended amount of time.

I too experienced the same problem with ryzen master on my msi board, had to clear bios.

because he have stupid themes about floop cpus which nobody wants

Lol... It does show though that if you did use the most optimal water cooling setup such as direct die that it should definitely improve performance with the amount of voltage that is used which will lower temperatures which will allow higher frequency.

​​@@JMNovak1011i have 5800x 3d that with kombo buster on my msi board the cpu doesn't seem to care about voltage or temps too much. I top out ar 90c, but scored closed to 2,450 in cinebench. Lowering voltage drop 8c to 82c but scroes only 2,325. Think my chip might be golden for LN2 overclocking.

@@kevinerbs2778 yeah it's a very unique give and take between voltage temperature and frequency. There's also quality of the cores which comes into play. We've seen that in a lot of different Zen cores now where one person can hit a certain frequency at a certain voltage and someone else will be 100-200 megahertz under that at the same voltage. I mean you also see it with how CCD 0 usually has the highest binned cores and CCD 1 has lower performing cores because AMD loves to get rid of their trash CCDs by just sticking them on the dual CCD chips

Peltiers are low capacity and grossly inefficient. Probably better to start with an AIO, then duct a window unit air conditioner around the PC, and modify its compressor for variable speed (or start with a model that has that feature). Dry air, no temperature differentials, and you get the extra-low-temperature benefits for RAM and GPU too.

You are right: it's difficult to draw conclusions from just one sample. Also, I am using a new methodology for putting together the Ryzen 9000 V/F curves. So it might not be comparable to my Ryzen 7000 results.

Upvoting you because you are awesome !

Unfortunately not. Maybe with TR9000

@@SkatterBencher When will TR9 be out? 1 year 2 years?

The absolute joys of Asynchronous Eclk! On Ryzen 7000, AMD made Asynchronous ECLK work. On boards with 2 external clock generators and when async eclk is enabled, one clockgen is used exclusively for the CPU cores. So no more data corruption :)

Use it on and off since Ryzen 7000 launch and never had an issue. That said, I'm just using it for tuning and overclocking. Not for a daily system. So I might not be your best source for advise :)

Thanks for the continued support!

That sounds like a BIOS or board issue. I just checked with my 9950X and X670E Hero and can easily go over 190A in CBR23 (46100 points). If I remember correctly, it is possible for the motherboard manufacturer to impose limits to Precision Boost Overdrive settings. So regardless of what's configured in BIOS or Ryzen Master, those limits would be imposed.

​@@SkatterBencher Ok, I wrote the ticket to ASrock support about this TDC limitatiton. This limit is a big joke, X670e Taichi had 24 phase vrm, and if correct there are 105A rated each. And they are limit it to 180A with 24 phase. I think even if it will be halfed it will run pretty well without overheating at 180A.

Strategy 1, 3, and 5 will be the same on the Master (of course your mileage may vary). Strategy 2 is ASUS exclusive. Strategy 4 will depend on whether your board supports asynchronous eclk.

1/2/4/8/16/max are the benchmark tests support.benchmarks.ul.com/support/solutions/articles/44002223746-cpu-profile-tests

In Returnal, sure. But I don't think Final Fantasy XV is GPU-limited with the RTX 2080 Ti. At least not to the extend that there's such little performance scaling. // EDIT: I did actually check the performance scaling with an RTX 4090 in the same benchmarks with the 9700X and the performance improvements were very similar to that of the RTX 2080 Ti.

Curve Optimizer and Curve Shaper is your only option if you want to get the high 1T precision boost frequencies.

Yes! I have another 97X follow up video schedule where I talk about this specific issue a little more in detail. I also already provided the feedback to AMD too (ahead of the Ryzen 9000 launch actually). Long story short, it seems there's some missing guardbands for Curve Shaper allowing the configuration of "illegal" voltages. A voltage is "illegal" when for a given frequency you're requesting a voltage that's lower than for a lower frequency. For example, let's say you're aggressive overvolting the high frequency and undervolting the maximum frequency, the curve might pull the voltage target for "max frequencies" below "high frequencies." It seems the VFT algorithm freaks out and goes into some kind of a safe mode. I've seen this behavior on the 9600X and 9700X but not on the 9950X. The Curve Shaper behavior is also a little different for single vs dual CCD chips. I should maybe make a video about that too. For example, the Max Frequency Curve Shaper setting doesn't do anything for the single-CCD chips. The way to solve what you're seeing is to set all Max Frequency points to +30. This won't affect the curve with the 97X but will ensure that those VFT points stay above the ones you'll undervolt with High Frequency Curve Shaper.

You shouldn't rely exclusively on the OCCT (all-core) stress test for Curve Optimizer, because that tests a specific point of the VFT curve (high temperature at moderate frequency). The likely reason why you see stability in OCCT and not in games is because your game is using a different part of the VFT curve (moderate temperature at high frequency). That's why the OCCT all-core stress test is only one of the points of my Curve Optimizer process, which usually starts with less heavy workloads.