More Players = More Lag?

[pollux]

Running as admin gives more priority to the process

What do you mean by "priority"? When I hear this term, I think of Ctrl-Alt-Del -> Task Manager -> Details -> Right-click -> Set Priority.

1. That priority doesn't change regardless of whether you run normally or right-click to "Run as admin"
2. You can adjust that priority without running as admin.
3. I just ran an experiment where I loaded the D&D 5e DMG, and changing that priority made no difference at all, which is what I would expect if the CPU isn't under contention (and if it is, removing the contention is a better approach than changing priority).

... and removes some of the double checking done by the OS.

What specific double-checking does it remove? You certainly have more privilege when running as admin, but that doesn't mean checks don't still happen and don't still have a performance cost. Are there any official sources documenting the performance implications of running as admin vs not?

As an experiment, I tried running the DMG load test as admin as well. I was super surprised to see that the load time went down from 20s to 12s. I was about to come back to the forums and document my result when I decided to try one more time without running as admin. The DMG load times stayed at 12s. What does this mean?

1. It doesn't mean that increased process priority made things run faster as admin. If it did, my performance would have dropped when I went back to running normally.
2. It doesn't mean that reduced runtime checks by the OS made things run faster as admin. If it did, my performance would have dropped when I went back to running normally.
3. Does it mean that running one time as an admin made some persistent change to the filesystem or registry that permanently sped things up even after returning to running as a normal user? Maybe? But if so I haven't yet detected what that change is. If there was such a change, it's hard to predict if it might have a downside for me or for people with other configurations without knowing what it is.
4. Does it mean that the performance change when I switched to admin was a coincidence, and that if my second test was run normally it also would have run in 12s? Maybe? I now can't reproduce my original 20s result so it's hard to know what caused the change. Switching back and forth now causing no change in performance

To me the primary conclusion one can draw from this experiment isn't that running as admin improved performance, it's that knowing why performance changed is complicated and often full of uncertainty. As LordEntrails noted, it's also definitely true that there are novice users participating in the beta at this point and advice with potential downsides should surely be presented with an appropriate caveat.

[4wire]

@pollux thanks for trying it out as well! I'll try to capture timing for different things as well.

[Neovirtus]

I don't have a perspective on the admin issue, but I did notice something when I went to check out the most recent update. Because the patch notes stated that there had been some modest UI loading optimization done I was paying attention to how long it took to load various menus (but not timing it). I noticed on the first load with the new update, many menus took significantly longer than I'd expected. NPCs, Spells, Items were all in the 20-30 second range, if I had to guess. I was used to them taking a bit longer than I'd like, but never more than 3-5 seconds. Frustrated and a bit disappointed by the "step backwards" the update made I shut down FGU. Then a few minutes later, some forum/discord discussion made me want to go back into FGU to check something else. I started it back up, and low and behold, the time to open those menus was still in the 1-5 second range.

I assume this is due to the assets still being in RAM, and I would guess based on Pollux's experiment that that was the case for him as well. I could see how this would be deceiving to other people who are trying to assess the impact of any external factor on FGU performance. (btw I did all this on my gaming laptop which has an NVME SSD drive and decent graphics card - GTX 1050 I believe.

[LordEntrails]

Discussion and ideas and suggestions are great. I hope everyone keeps them coming. And I hope people are just as willing to be open to constructive criticisms

I get where 4Wire is coming from, with expecting people partaking in the open Beta as having some minimum level of knowledge. But, as I've seen in other threads, their are people who want FGU to be so simple that an 70 year old can install and run it just like they can do with Zoom.

So, while I would think people who want to explore a Beta would be more technically inclined, that is certainly not a universal opinion. It seems that lots of people see 'Beta' and 'Early Access' as just another marketing term to hype a new program and expect such a program to run like any 'production' software. And I'm sure their are interpretations of what such means all the way from there to a something akin to an active technology demonstration

Personally I blame it on Agile Software Development Methodology! (No really I don't, but hey, why not?)

[Sulimo]

Discussion and ideas and suggestions are great. I hope everyone keeps them coming. And I hope people are just as willing to be open to constructive criticisms

I get where 4Wire is coming from, with expecting people partaking in the open Beta as having some minimum level of knowledge. But, as I've seen in other threads, their are people who want FGU to be so simple that an 70 year old can install and run it just like they can do with Zoom.

So, while I would think people who want to explore a Beta would be more technically inclined, that is certainly not a universal opinion. It seems that lots of people see 'Beta' and 'Early Access' as just another marketing term to hype a new program and expect such a program to run like any 'production' software. And I'm sure their are interpretations of what such means all the way from there to a something akin to an active technology demonstration

Personally I blame it on Agile Software Development Methodology! (No really I don't, but hey, why not?)

I blame Google , how many of their apps are technically still Beta? They've mangled the term so badly, it has practically lost all meaning.

From the testing Pollux did, I'm suspecting some sort of caching going on (HDD or Windows Prefetcher/SuperFetch).

[LordEntrails]

IME the only way to get reliable performance metrics is to do a full system restart between tests and not have any additional apps etc. And statistically speaking (from a professor I had many years ago!), a sample size of 30 is the minimum for any data condition. So, anyone want to reboot their computer 60 times so we can get 30 of each? *G*

(I hope you all don't think I really expect anyone to actually do that!)

[pollux]

I assume this is due to the assets still being in RAM, and I would guess based on Pollux's experiment that that was the case for him as well. I could see how this would be deceiving to other people who are trying to assess the impact of any external factor on FGU performance

From the testing Pollux did, I'm suspecting some sort of caching going on (HDD or Windows Prefetcher/SuperFetch).

IME the only way to get reliable performance metrics is to do a full system restart between tests and not have any additional apps etc. And statistically speaking (from a professor I had many years ago!), a sample size of 30 is the minimum for any data condition.

FWIW, I did attempt to control for cache effects, but I went the other direction by attempting to measure with warm caches (which are often quicker/easier to get than cold caches when you don't understand the system-under-test very well). You can load and unload the DMG repeatedly and the times don't seem to really change. So I...

1. Unloaded and loaded the DMG 5 or more times from FGU with a normal start and normal priority without restarting FGU between runs. This would be measuring for both a warm FS cache and a warm Lua interpreter (and maybe warm other things that are beyond my ken), but I didn't observe any variance I was able to measure (my accuracy by eye was more or less a second or a little less and it was always pretty much exactly 20s)... so I didn't observe caching effects at this stage.
2. Quit FGU and reran 5x with "High" priority. Again, I observed no measurable variation in timings. 20s pretty exactly in every case.
3. Quit FGU and reran 5x with Run-As-Admin and Normal priority. This time 12s consistently, from the first run.
4. Quit FGU and reran 5x with Normal start and Normal priority (reproducing my first experiment), got 12s consistently again.

So I did measure multiple times and attempt to measure with equally warm caches in all cases. But I nonetheless think you are all correct. Some caching effect that I don't understand and didn't control for is in play, but I don't know what. I don't THINK it's as simple as "stuff loaded in ram" because I think my 5x tests and the high-priority test would have caught that.

Next steps would be to test with reboots between each test, with a new campaign for each test (I reused an existing one), with a fresh data-dir for each test, with a fresh FGU install for each test, and if someone can figure out how with registry diffs or resets of applicable keys between each test. I don't plan on conducting these tests, I feel pretty certain that the run-as-admin thing is a myth. I only went this far because there were specific claims made about process priority and runtime overhead of admin processes that were wrong and I wanted to inject enough data into the discussion to debunk those claims. I'd certainly be happy if someone did further testing though and pinned the cause of this effect down, even if (especially if?) it ends up demonstrating that I'm incorrect.