I tested both - the latest LTSB 1607 and the latest Pro 1703 with LatencyMon. My driver and process latencies are excellent with either edition, BUT LTSB gets hard page faults, even if pagefile is disabled, while Pro 1703 never shows any page faults. This was all tested on non-adjustable and tweaked versions of both. I think Creator's Update forces everything into RAM if there is enough of it, but LTSB does not. In LTSB most page faults, if not all, are caused by Explorer.exe. These page faults, however, never actually cause any problems in games. In fact, my FPS are slightly higher with LTSB. I am curious as to whether the Pro 1607 version gets the same page faults as the LTSB 1607? Could someone with Pro 1607 run a short 30 second test with LatencyMon to see of any page faults occur at all?
Well, with Creator's Update I get zero page faults and I was curious if the same was happening with Pro 1607 version of Windows 10.
Serious question - given that RAM in any system has a physical limit and paging algos can never be 100% perfect in terms of prediction and allocation, how can it be possible for any OS with virtual memory to never have any page faults at all?
Then I guess LatencyMon is lying? I did see something regarding Creator's Update using some "stages" or "levels" of managing the OS within RAM-only. I am searching for it right now to give specifics... I think it was mentioned in one of the Cumulative Updates notes.
I wouldn't call it "lying". LatencyMon 6.51 was released late last year and perhaps just can't handle the changes you indicate have been made in WIn10 CU (do post the link once you find it).
Hmm... Creator's Update can run with UAC disabled and everything will work, but Anniversary Edition cannot... Maybe that's the issue..? EDIT: No, not the case, although by running LatencyMon as Admin, the amount of page faulting BY LatencyMon was drastically reduced. EDIT 2: It appears running just about any program that results in page-faults, as admin, reduces those page-faults. Is there a way to run Explorer.exe as Admin?
There are 2 different types of page faults, namely hard and soft page faults. Soft page faults means that the currently executing (active) application call is not within the currently allocated memory space and has to fetch additional memory. A hard page fault (and this is the one that kills performance bigtime) is that the OS has run out of memory resources and needs to use the pagefile to swap out inactive process memory to disk. TL;DR: soft page faults: no problemo, hard page faults: el problemo grande.
I tested several machines with clean stock install of LTSB 2016 and in each case there was a ton of hard page faults reported by LatencyMon and that is on very high-end rigs with 16+ GB of RAM and SSD's. Maybe LTSB uses some ancient RAM management. I keep asking folks with Pro versions of builds 160714393 to run LatencyMon for 30 seconds and report page faults info, but I guess it's too much to ask these days...
Pro 1607 (14393.1358) LatencyMon run for 30 seconds: 1 page fault caused by chrome.exe Closed chrome LatencyMon run for 30 seconds again Zero page faults
Ha! I got the same results here - 1 or 2 of minor page faults within 60 seconds of use with Pro 14393 and yet heavy high latency page faults with LTSB 14393. They were both stock unedited images, no settings changed, just install OS, install LatencyMon and run it. Don't forget that Pro version has Store Apps running in the background that LTSB does not and yet its the Pro version that perform better. Either LTSB uses some heavily outdated RAM and page file management or ... not sure what, but Pro 14393 is better for gaming, especially if you butcher it with MSMG Toolkit or NTLite. BTW, saw some other reports of similar issue with LTSB.
AHA! The new KB4022723 update confirms that previous update or maybe even the original 14393 had issues with Page Faults, which decreased OS performance! I hope they fixed it for LTSB 2016!
I think MS did something weird with LTSB 1607 because even after the latest update I get a SUPER-mind-blowing amount of Page Faults from all kinds of processes (like Chrome), yet few-to-none with Pro 1607. LTSB = not for gamers and that's for sure!
What does PerfMon shows? Here on Windows 10 LTSB 1607 Build 14393.1358, it shows 0-2 hard page faults/sec, mostly staying at 0 although LatencyMon keep telling i got 1000+ hard page faults (not always). I am doubting LatencyMon since these page faults are not actually causing any issues. Task Manager > Performance > Open Resource Monitor > Overview. Hard Faults/sec. Or right click This PC > Manage > Performance. Cache Faults/sec.
Interesting - will check it out and let you know because as I said earlier somewhere - games run smoothly and game processes never cause Page Faults in LatencyMon. I get this feeling that LatencyMon reports anything that is read off SSD/HDD as Hard Page Fault. EDIT: Yeah, Resource Monitor shows heavy page faults when I launch something like Chrome or Firefox, but the count goes down quickly, while LatencyMon preserves total count (summary) at all times. During use of whichever program, however, Resource Monitor does not show Page Faults - only when something is launched from SSD/HDD or when I open a new Firefox/Chrome window.
If there was actually hard page faults issue, wouldn't you see constant hard page faults whether or not the system is idle or doing something? I think you wont see 0 for the most part and system will actually experience performance issues if there was actual hard page faults issue.