Question Weird Ring-Only EDP Throttling

[KSFrosty]

Hi all!

I've seen a couple of similar posts here, but nothing that maps well enough to the problems I've been having, so I figured I'd start a new thread and hope it works out.

Some stats to start off with: i9-10900K with a 360mm AIO cooer on a Z490 Taichi Motherboard running a 4090 GPU on a Corsair RM850X PSU. The screenshot I posted with summary data from HWiNFO64 should speak to the rest of my system specs.

My setup has been going through periods of alternating performance issues, during which nearly any kind of effort leaves it struggling to catch up. For example, mousing over app thumbnails in my Steam library has a significant latency to the zoom-on-hover effect, and games that I've previously had no issues running start running with a near-countable number of frames. Periods of poor performance can last several weeks before returning to normal out of nowhere and running like a dream for another several weeks before dropping precipitously with just as little warning.

I started diagnosing things using Intel XTU, which was the first time I noticed the EDP Throttling flag. During periods of poor performance, it was activated almost immediately upon reboot, but when my computer was working properly, it didn't show up at all. This most recent period of poor performance has started off without the XTU EDP Throttling flag, I started digging into other tools beyond XTU, like ThrottleStop and HWiNFO64 (screenshots from which are attached to this post). I didn't include the TS Limit window in the screenshot, but it's got ONLY a near-constant red EDP Other block in the Ring section, whereas HWiNFO64 shows IA Limits from Max Turbo Limit and Turbo Attenuation. More importantly though, it shows Ring Limits due to EDP-type constraints in line with those shown by TS.

Looking through the other EDP-related issues on this and other forums, I've checked things like the ICCMax Values, etc., and they're already set as high as possible, so I'm running out of steps to take next.

A couple of interesting notes:

• I tried several different Benchmarking approaches, all of which get throttled to just under 800MHz, with core temps not to barely exceeding 30 deg C
• I can physically distinguish between periods of poor performance and acceptable performance by the amount of heat being radiated from my computer. If it's working properly, I can feel the hot air being vented from my PC under stress (while keeping temps below 60-70 deg C in monitoring software)
• I mentioned it before, but these swaps between poor performance and proper performance are infrequent. If my computer is terrible, it's terrible for a long time. If it's good, it's good for a long time, not a lot of switching back and forth.
• I'm not looking to overclock anything. I just want my computer to work right again.

My current guess is there's just a bad sensor somewhere that I need to disable, but I don't want to accidentally fry something. I barely know anything about Computer Hardware/Engineering, so ELI5 responses are welcome!

Thanks,

--KSFrosty

 

[Ralston18]

Try some of the built in Windows tools:

Task Manager, Resource Monitor, and Process Explorer (Microsoft, free). Even Latency Monitor.

Objective being to find some difference between situations and actions when performance is as expected and again when it lags etc..

Could be some background task. Launched at boot time but only causes problems when certain other things are happening.

Look in Task Scheduler as well. Something there could be triggering an app or utility doing (or trying to do) who knows what.

Process Explorer:

https://learn.microsoft.com/en-us/sysinternals/downloads/process-explorer

May take some time and effort to get a sense of the tools and the information they provide. Or work out a methodology to keep an eye on things.

As an example:

"mousing over app thumbnails in my Steam library has a significant latency to the zoom-on-hover effect, and games that I've previously had no issues running start running with a near-countable number of frames."

You can easily see the problem. Open one of tool windows. Drag to one side or to a second monitor - leaving the Window open and observable.

Then mouse over the thumbnails. Look for what changes in the tool window.

 

[KSFrosty]

Try some of the built in Windows tools:

Task Manager, Resource Monitor, and Process Explorer (Microsoft, free). Even Latency Monitor.

Objective being to find some difference between situations and actions when performance is as expected and again when it lags etc..

Could be some background task. Launched at boot time but only causes problems when certain other things are happening.

Look in Task Scheduler as well. Something there could be triggering an app or utility doing (or trying to do) who knows what.

Process Explorer:

https://learn.microsoft.com/en-us/sysinternals/downloads/process-explorer

May take some time and effort to get a sense of the tools and the information they provide. Or work out a methodology to keep an eye on things.

As an example:

"mousing over app thumbnails in my Steam library has a significant latency to the zoom-on-hover effect, and games that I've previously had no issues running start running with a near-countable number of frames."

You can easily see the problem. Open one of tool windows. Drag to one side or to a second monitor - leaving the Window open and observable.

Then mouse over the thumbnails. Look for what changes in the tool window.

So, this didn't solve my issue, but it has given me an interesting way of characterizing it.

I'm pretty confident at this point that the performance issue is related to the CPU Clock speed being throttled down to 800MHz. Based on your suggestions, I checked the speed under a couple of different actions.

As it turns out, it's running on idle at around 1.3-1.5GHz on average, but as soon as I do almost anything, all of the cores get chunked back down to 800MHz. When I say almost anything, I mean that I was able to trigger the throttle just by moving my mouse quickly. While moving the mouse alone wasn't enough to appreciably increase the Core/Thread Usage levels to the point where I'd expect latency, extra on-hover effects like those on the Steam page were enough to start moving to a significant proportion of overall available usage under the 800MHz restriction.

So the better way to describe my issue is that I'm getting throttled in response to almost any action, with visible impacts (i.e. latency, frame rate drops, etc.) in extremely low-stress situations.

Any other thoughts or next steps?

 

[Ralston18]

PSU: make, model, wattage, age, condition (original to build, new, refurbished, used)?

Throttling may be due to overheating.....

 

[KSFrosty]

PSU: make, model, wattage, age, condition (original to build, new, refurbished, used)?

Throttling may be due to overheating.....

CORSAIR RM850X White, 850W, purchased new in Aug 2020 (~3 years)

Cores are running an average of 30C, with a max of 57C over the last week.

Is it possible that there's a sensor within the PSU causing it to limit output wattage? If so, how would that relate to the EDP Other flags that I'm getting, since most of the associated causes seem to indicate that too much power is being provided, not too little?

 

[Ralston18]

The build will only draw the power required - not more.

Reference thread from this Forum:

https://forums.tomshardware.com/threads/is-buying-an-overpowered-psu-a-bad-idea.3795583/

= = = =

As for the cited problem(s), you need to delve into the details a bit more.

FYI (from Intel):

https://www.intel.com/content/www/u...0039154/processors/intel-core-processors.html

https://www.intel.com/content/www/u...erboard is not able,limits set to the maximum.

What are the "EDP Other flags" that you are getting?

 

[KSFrosty]

The build will only draw the power required - not more.

Reference thread from this Forum:

https://forums.tomshardware.com/threads/is-buying-an-overpowered-psu-a-bad-idea.3795583/

= = = =

As for the cited problem(s), you need to delve into the details a bit more.

FYI (from Intel):

https://www.intel.com/content/www/u...0039154/processors/intel-core-processors.html

https://www.intel.com/content/www/us/en/support/articles/000039154/processors/intel-core-processors.html#:~:text=The motherboard is not able,limits set to the maximum.

What are the "EDP Other flags" that you are getting?

Some of the other utilities like ThrottleStop and HWiNFO64 have slightly more precise versions of the blanket EDP Limit Throttling shown in Intel XTU. TS's version is "EDP Other," which it was indicating as an active throttle for the CPU ring only (as opposed to the general "EDP Limit" indicated in XTU). HWiNFO was showing the same thing, but it spelled it out more directly: "Ring Interconnect Frequency clipped due to electrical design or other constraints (e.g., maximum electrical current consumption, SVID voltage limit, or PL4)".

I checked the Icc Max levels and they were as high as they could go, and I would just chalk it up to the motherboard, but apparently the core is receiving sufficient current. If I were just on the cusp of operating withing EDP for the ring, I'd expect to see much more swapping back and forth between proper performance and throttled performance instead of bad for three weeks/good for three weeks.

 

[Ralston18]

Look in Reliability History/Monitor and Event Viewer.

Either one or both tools may reveal some clues with respect to that "bad for three weeks/good for three weeks" pattern.

Reliability History presents a timeline format that may be helpful.