Joined: 20 Aug 15
Hi. I was recommended to make a thread about this here. I originally made this thread over at overclock.net forums, but some people I was speaking to told me this would be the best place to get data; so I'm going to copy the information from my previous thread below. If this is the wrong place, or it breaks Terms of Service somehow, you can delete it or whatnot, I don't mind. I also don't know how many users here are gamers, so I wouldn't know how hard it is to prove this, so sorry in advance!
Hi. This is going to be a slightly confusing post for many of you, but please, bear with me. I'm going to explain this as best I can.
Windows 8, 8.1 and 10 appear to have CPU utilization problems. I'm sure you've heard of this in the past, but I'm looking toward a very specific approach here. The problem is that Windows Task Manager appears to report incorrect CPU utilization percentages in its "performance" tab, and when WTM hits 100%, even though the actual CPU util is not at 100% (as shown by programs like HWiNFO64 and Throttlestop etc) that some (NOT ALL) programs will simply "stop" using extra power.
"Okay that's great, but what do you need from us?"
I need you to download Throttlestop 8 beta 2 for parity testing. If you want to use other programs in tandem like HWiNFO, that's fine, but mainly WTM and TS8 are the comparisons I want to see. All you need to do is launch Throttlestop. Until you hit "turn on" all it's going to do is monitor your temps/util/speeds. I just want you to use it in monitoring mode, please. C0% is your "util %" for Throttlestop.
I want you to run CPU-heavy, multithreaded games like Dying Light, GTA V and Battlefield 4, using your i5s and i7s, both at stock and with an overclock, to determine how the CPU discrepancy turns out. Lowering GPU-heavy settings to keep a CPU bottleneck would be great. Unlocked FPSes too. PLEASE NOTE: Crysis 3 bypasses this limit. I've seen 95%+ util from Crysis 3 in the past, at which point my GPUs got bottlenecked.
It would also help if you could use Sony Vegas with the "Sony AVC" rendering codec to test. Sony Vegas is one of the rare rendering programs that gets limited by WTM's CPU util % limit. Most of them don't (as I'll show below).
Now in my testing, the discrepancy appears to be ~23% util. I've seen anywhere from "75%" to "80%" actual CPU utilization at 99%-100% WTM util, but usually it sits near 77% for me. The problem is that laptops have quite low base clock speeds, and desktops don't. My 4800MQ also does not have much of an overclock potential, only +400MHz as limited by intel. If I had a 4900MQ, I could go +600MHz, or unlocked slider with a 4930MX, but I don't have those chips. So I want to check with your i5s and i7s.
Here's some examples of it happening:
GTA V giving me a CPU bottleneck with 100% util and only 78% GPU util (some settings are turned down to force a CPU bottleneck most of the time, so that my framerates remain high). Note how both Playclaw's overlay and Throttlestop are well below 100%, but WTM says 100% and GTA says "ok boss".
Vegas 13 Pro rendering at WTM 100% but leaving excess CPU power behind (you can't see vegas in the screenshot, I know, I'm sorry).
Handbrake ignoring that limit and using a cool 99% like it should.
Throttlestop's TSBench using a proper 100% like most benchmarks do.
And please, please, before anybody says "you're on a laptop, you're overheating, you're throttling, blah blah", here is Linpack LinX running at 100% without throttling or overheating. It hammers the system the most early in the bench; later in the bench it uses less power etc, which is why the screenshot is early. This is only here to deter people who will yell at me because I'm on a laptop, because they show up. A lot. I happen to only have a laptop to test, which is why I am asking people with desktops to test more for me.
So the main objective is to see whether your discrepancy is different than mine. I've checked with the author of Throttlestop who confirms that he can see the issue in his lenovo laptop, so it's not a problem that my Clevo specifically has. I don't know if it's a coefficient off the base clock with turbo boost, or if hyperthreading is the reason, or if it's a flat ~23% limiter, etc.
When I find out the issue, I want to bring it and the hard data to news sites, so Microsoft can fix their OSes with some public exposure about this.
Joined: 20 Aug 15
Unfortunately I have no idea what BOINC credit means.
I was truthful when I said I was simply recommended to this particular forum, as there are a lot of users here who are very particular about their system performance, from a public chat room on twitch TV when I was explaining about my wish to test and lack of testers.
I don't see how most of the rest of your post is relevant to what I was asking. What I was trying to get at, is that in the last two Windows operating systems, unlike in previous generations, the operating system seems to limit certain programs' resource pools with respect to the CPU. It seems to be different based on various clockspeeds and whether or not a user is on an i5 or an i7 (I haven't looked at AMD chips as yet). I want data from more users to try and get an accurate... data table? of what's going on.
When I can complete this "data table" and accurately predict what's going on, I would like to present it to news articles so Microsoft can accurately respond to and fix this with their operating systems.
Copyright © 2021 University of California. Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.