WoW Performance Test Area/Zone?

[fixx]

Perhaps English not your first language, would you prefer I write with smaller words and shorter sentences?

It's not, and neither does being patronizing fucktard doesn't make you smart or right, just annoying. Or do you need that in shorter words, perhaps an explanation of what "patronizing" means in case you missed it?

Likewise, we don't need actual players or NPCs in order to get additional 3D elements into our scene.

Yeah, you do. An addon calculating Pi is usless for benchmarking anything else than either the addon itself, or WoW's LUA interpreter. It has got nothing to do with raid benchmarking when you cant access other critical subsystems like 3D engine or networking engine. When you can simulate the initial lag spike burst with bloodlust, stormlash, three army of the deaths, 20 other pets, skull banner, variable network lag and all trinket procs at the same time you're on to something.

Save for a player multiboxing 25-accounts, you're not going to get any sort of repeatability.

No, and that's why you need to compromise. Synthetic tests like /timetest or calculating Pi touch only minor parts of WoW game engine and can never replicate whole raid environment.

[Noctifer616]

You're said that 'everyone knows' that spell effects and addons are the cause of poor frame rates. Why do you think that forcing the game to render many spell effects, and to perform the same amount of work that addons responding to combat log events would, is a poor way to recreate a raid's performance penalty given that you claimed "everyone knows [those are] the thing that kills WoW performance." ?

Combat log events have a significant effect on the games performance, but they are not the only cause for low FPS, especially on high end systems.

What you are trying to simulate is making the CPU choke by giving it a ton of combat events to handle. While that might be a decent stress test, it is not a realistic one. What a lot of MMO's and RTS games suffer from today is draw call submission limit (CPU can't feed enough data to the GPU, causing low FPS while the GPU is not fully taxed).

What your simulation would do is simply benchmark one of WoW problems, and that is clutter in the combat log. The other issue can only be simulated by actual 3D simulations, like a 25 man raid.

[fixx]

Okay, I will try to write with simple English instead. I have trouble with my second languages too. I also get frustrated when I can't understand what my peers say or write.

So you still are patronizing fucktard? I understood that perfectly well first time around.

You can stress the 3D engine. Network handling has no important impact on frame rate, nor sound.

nope.jpg

3D elements like: spells, players, pets, and NPCs can be created and doing this does cause a measurable decrease in performance as you increase the number of models you would see in a large raid or populated major city.

Some kind of approximation yes, but the created objects arent in the same priority of rendering as the game world. Blizzard explicitly made it impossible to interact with the actual world drawing when they broke the functionality of AVR addon few years back. Again you're benchmarking only WoW's addon space and not the actual game engine.

Networking lag—that is latency—is not a cause of poor frame rates:

Network problems can and will cause latency in framerates when occasional break happens in communication and then you recieve whole pile of packets. Everybody has seen once or twice the rubber banding effect caused by it, and it can and will cause bursts of spell effects much higher than normally possible happening while the game clears backlog of actions you missed.

Sound engine can also cause frameskips when lot of sound effects are queued at the same time and there's a loading bottleneck happening in disc or memory access. Most often this happens as a result of the network burst.

Of course those are rare, but they do happen and they do have direct impact on framerate which you claimed doesnt happen.

Do you mistakenly believe that COMBAT_LOG_EVENT_UNFILTERED differs from UNIT_TARGET or CHAT_MSG when considered by the event-dispatcher? Do you believe that a single event that causes 100 handlers to perform 100 small tasks is significantly different from 1 event causing 1 handler to perform 1 task that is 100 times larger?

You're forgetting the little tidbit that multiple parts of the default UI as well as custom addons need to parse all those COMBAT_LOG_EVENT_UNFILTEREDs and catch the relevant arguments which is significantly slower than any dummy messages you can create. Of course you can just increase the amount of dummy messages and dummy handlers to compensate for it but it'll just be another best guess approximation (like the 3D issues) instead of real simulation.

I think you have not enough time learning how Warcraft works:

I think you might need to heed your own advice.

[tielknight]

Is it just me or is this getting a bit off-topic? Not to mention rather heated.

[Cilraaz]

How about we stop insulting each other and keep this on topic. Further bursts of childishness may result in infractions.

[Instructor Soki]

I can't believe nobody's mentioned LFR.

[Noctifer616]

I can't believe nobody's mentioned LFR.

LFR isn't as demanding as Heroic.

[chazus]

LFR would be good.. except for a few things.

1) You have to wait in queue (30-45 minutes sometimes) for a single test
2) Each test takes 10-20 minutes, depending on pulling.
3) Group composition and afk-ness varies each time. 15 people fighting is different from 25 people fighting
4) Each boss is different in benchmark, unless you...
5) Only pick one boss and then...
5a) Bail out after a boss
5b) Stay the entire LFR run

[Noctifer616]

I disagree. Create 1000 3D objects, enable ui profiling, notice that UI load remains low yet the frame rate drops.

Heck, if it's just additional 3D work we want then /console spelleffectlevel 5000 gives us an easy way to easily generate additional draws and the sort of performance decrease we'd expect to see too.

Ye, but the OP wants to test 25 man raid performance, unless you can get the exact numbers, you won't get a simulation, just a stress test. And that's where the issue is, not every fight is the same.

Ye, sure, you can cause stress that is on level of 25 man raiding, but how accurate is that going to be?

To be hones, the thing that OP really wants to know is single core performance for his CPU (considering that most modern GPU's today can run WoW just fine).

If you want a realistic scenario, just run any modern 3D game, lower and disable all possible graphics settings, lower resolution. Whatever CPU get's the highest FPS is better at single core performance.

[Faithh]

LFR would be good.. except for a few things.

1) You have to wait in queue (30-45 minutes sometimes) for a single test
2) Each test takes 10-20 minutes, depending on pulling.
3) Group composition and afk-ness varies each time. 15 people fighting is different from 25 people fighting
4) Each boss is different in benchmark, unless you...
5) Only pick one boss and then...
5a) Bail out after a boss
5b) Stay the entire LFR run

Try moltenwow? They use a modded client, wrath of the lich king etc and the cpu bottleneck there seems to be worse. I barely get there just in empty zones above 40 fps

Network problems can and will cause latency in framerates when occasional break happens in communication and then you recieve whole pile of packets. Everybody has seen once or twice the rubber banding effect caused by it, and it can and will cause bursts of spell effects much higher than normally possible happening while the game clears backlog of actions you missed.

No it can't because the gpu keeps rendering whats left in your vram, there's no way network delays can impact your gpu and the gpu has no clue if you're connected or not. He explained it perfectly, plug your connection out and it takes usually like 5-10s before you DC and just compare your frames no difference. I don't really understand why you would argue with evn, you're not on his level he's pretty much pro.

[fixx]

I don't really understand why you would argue with evn, you're not on his level he's pretty much pro.

Pro at writing condescending walls of text, which doesn't mean an addon solution will work any better than the worthless /timetest.

It's not the fact that events exist that cause problem, it's the handlers that run in response to them.

And therein lies the problem why your addon idea is doomed to fail. With all the protected functions you can't model the real game UI accurately enough with your simulated messages.

Yes, it will be a benchmark, but it's benchmarking your addon and not raids.

[Faithh]

Pro at writing condescending walls of text, which doesn't mean an addon solution will work any better than the worthless /timetest.

You're not getting the point, you usually have dunno 100 fps with a 780 there in a empty zone and in 25m raids you get like 30 fps with a 4670k@4.3GHz orsomething, so your goal is applying some extra load to the cpu to reach that 30 fps with ~ 20% gpu load. Now that you have your simulation, then go ahead test different cpu's, same place and you know that the gpu will never exceed 20%.

[tielknight]

Holy walls of text batman o.O



You really should find a dungeon with alot of mobs(not having ones that will cause you to lose aggro would be best) and plan a route that you can easily repeat and get a big group following you. Getting a good route down where you can get a nice big pack going and having alot of stuff going on(spells, particle effects, large group of NPC's smashing your face in, etc) is about as repeatable as you can get in a benchmark-ish kind of environment, at least more reliable than a LFR group as NPC's are alot more predictable(and controllable) than other players. You just get a route, run it, grabbing everything angry at you along the way, and get to the end and round a corner(to get the spellcasters to come to you) and pan out the camera and watch your FPS. Throwing down a spell/ability of your own on such a large group would also be a nice sort of "stress" to see if what framerate drops would be like during intensive phases.

The only hard part is, find a good dungeon to do this in