PDA

View Full Version : Testing Morbidity and Death and Decay FPS



ttocs
01-19-2009, 08:20 PM
Lately, there has been plenty of discussion on Death Knights and their framerates, particularly when AoEing. I have witnessed this myself, as I am the Lava Blaze/Twilight Whelp tank on Sartharion in my guild. While I'm able to produce a significant amount of DPS during the encounter, as I've come to perfect my AoE tanking Rotation (DnD, IT, PS, Pest, UB, BB, ScS, DnD) and keeping Death and Decay on cooldown, I've noticed my frame rates plummet to the point where I can't move. After doing some reading, many people on the WoW forums and Elitist Jerks have theorized that Morbidity was the cause of the problem.

Some threads on the WoW forums about the issue:

World of Warcraft - English (NA) Forums -> DK's - The lag spikes are due to MORBIDITY! (http://forums.worldofwarcraft.com/thread.html?topicId=14318776458&postId=143174590704&sid=1#0)

World of Warcraft - English (NA) Forums -> Death Knight rune usage - FPS drops (http://forums.worldofwarcraft.com/thread.html?topicId=12891009778&sid=1&pageNo=1)

Death runes are something I'm going to look into another time, though I was spec'd into Reaping throughout these tests.

Now, let's take a look at Morbidity:

Morbidity - Spell - World of Warcraft (http://www.wowhead.com/?spell=49565)
Increases the Damage and healing of Death Coil by 5/10/15% and reduces the cooldown on Death and Decay by 5/10/15 sec.

And the main ability in question, Death and Decay:
Death and Decay - Spell - World of Warcraft (http://www.wowhead.com/?spell=49938)
Corrupts the ground targeted by the Death Knight, causing 62 Shadow damage every sec that targets remain in the area for 10 sec. This ability produces a high amount of threat.

Can a talent that merely buffs damage and lowers cooldown cause a framerate issue? It didn't make much sense, so I wanted to test it.

The Test:

The area I used is the same one shown in Mercfighter's Prot Warrior AoE grinding video (see: http://www.tankspot.com/forums/f64/43068-icecrown-aoe-grinding-area.html). Each test consisted of me pulling all of the mobs, running to a wall near the base of the ramp, waiting for most of them to be in range, then start with my abilities. I would start the Fraps benchmark about 1-2 seconds before I used my first ability. I would stop the benchmark after the last mob was killed. Each pack was anywhere from 15-20 mobs. Maximum of 16 skeletons + 4 Vargul, though some "Evade bugged" on pulls. I can provide a video of a sample run at a later time, if need be.

As was suggested by a blizzard poster in some of these threads, I completely removed ALL of my Addons (keeping just the Blizzard_* folders) and moved them to a temporary folder. I then moved all of my files out of the WTF folder, just as extra assurance that my mods are cleaned out. All of the settings I need are retrieved from the server anyway, so they were instantly restored when I logged back in. I then rebooted my PC, closed all unneded applications, started Fraps, then started WoW.

To ensure all relevant assets are loaded into RAM, I did a "dry run", where I didn't record any benchmark data. I would then round them up again, head to the spot, wait for a group of them to come to me, and did the same test 3 more times. I have graphed the framerate, second-by-second, of each attempt.

When the dry run and 3 attempts were complete, I flew back to Dalaran, changed Video settings if necessary (for the low detail tests), exited the game, and restarted it.

Trying to measure framerate in a controlled environment in WoW is difficult. My setup certainly isn't perfect, and I hope to find a better set of mobs to test with in the future - but I was able to replicate (albeit on a smaller scale) the issues I was having in raids, and that's what mattered here. I'm open to suggestions on how to better test this, not just for this - but there's a lot of possibilities when there's a consistent, repeatable test that can show a computer's performance.

The Computer:

Specs:
Core 2 Duo E6700
2GB Corsair XMS DDR2-800
Radeon HD 4870 X2 2GB
DirectX Version: DirectX 9.0c (Windows XP Professional, 32-bit)

Driver Version: 6.14.0010.6891
Date: 12/1/2008
DDI Version: 9 (or higher)

Resolution: 1680x1050, 1xmultisample, 60Hz
Windowed Mode (Maximized), Vertical Sync off

Detail settings:

http://ttocs7.shackspace.com/morbidity%20test/settings.png

While I normally run my CPU and RAM overclocked, for this test I've run everything at stock speed.

The Character:

I did this in full tanking gear (All lv200/213 epic gear besides Essence of Gossamer, Sigil of the Dark Rider - Weapon is Inevitable Defeat), with a Fallen Crusader Runeforge, and in Frost Presence.

Relevant Glyphs: I used Icy Touch and Scourge Strike as Majors, Pestilence as Minor.

Spec: When testing with morbidity, I used my Main Tank Unholy spec, which is 9/11/51: Talent Calculator - World of Warcraft (http://www.wowhead.com/?talent=j0vZhxbzZfghIx0ogc00cout)

When testing without, I also used 9/11/51, but moved the points into two talents that deal with abilities I don't use - Unholy Command, and Corpse Explosion:

Talent Calculator - World of Warcraft (http://www.wowhead.com/?talent=j0vZhxbzZbgGIt0ogc00cout)

The Results:

For the spreadsheet with the graphs and raw data, I have posted it in OpenDocument (.ods) format here. (http://ttocs7.shackspace.com/morbidity%20test/Morbidity%20-%20Framerate%20test.ods)

To set up a baseline, I did a few tests without Death and Decay at all. These test are significantly longer than the other tests, since Death and Decay is a big chunk of my AoE DPS. For the record, I was in fact spec'd 3/3 Morbidity in these tests. I did not use Death Coil for that reason, however.

My rotation was Icy Touch, Plague Strike, Pestilence, Blood Boil, Scourge Strike, and keeping Unholy Blight up.

(Sorry about the different scales - I can't configure OO.org Calc to set a fixed scale...)

http://ttocs7.shackspace.com/morbidity%20test/no%20dnd1.png
http://ttocs7.shackspace.com/morbidity%20test/no%20dnd2.png
http://ttocs7.shackspace.com/morbidity%20test/no%20dnd3.png

Normal Rotation with Morbidity: I opened with a Death and Decay, followed by an Icy Touch, Plague Strike, and a Pestilence. I used Runic Power on Unholy Blight, then followed up with a Blood Boil, Scourge Strike, and then another Death and Decay. The pack died about halfway through the second Death and Decay.

http://ttocs7.shackspace.com/morbidity%20test/normal%20+%20morb1.png
http://ttocs7.shackspace.com/morbidity%20test/normal%20+%20morb2.png
http://ttocs7.shackspace.com/morbidity%20test/normal%20+%20morb3.png

Normal Rotation with Morbidity (Low Spell Detail): To see if it's a Spell Detail issue, I did a few more runs with the Spell Detail slider set all the way to the Left.

http://ttocs7.shackspace.com/morbidity%20test/low%20detail%20morb1.png
http://ttocs7.shackspace.com/morbidity%20test/low%20detail%20morb2.png
http://ttocs7.shackspace.com/morbidity%20test/low%20detail%20morb3.png

Normal Rotation without Morbidity: I opened with a Death and Decay, followed by an Icy Touch, Plague Strike, and a Pestilence. I used Runic Power on Unholy Blight, then followed up with a Blood Boil, Scourge Strike, and another Blood Boil. At that point I tried to do another pestilence if I could get it before the mobs died, if not, I would just tab target and use abilities while refreshing UB.

http://ttocs7.shackspace.com/morbidity%20test/nomorb1.png
http://ttocs7.shackspace.com/morbidity%20test/nomorb2.png
http://ttocs7.shackspace.com/morbidity%20test/nomorb3.png

Normal Rotation without Morbidity (Low Spell Detail):

http://ttocs7.shackspace.com/morbidity%20test/low%20detail%20nomorb1.png
http://ttocs7.shackspace.com/morbidity%20test/low%20detail%20nomorb2.png
http://ttocs7.shackspace.com/morbidity%20test/low%20detail%20nomorb3.png

So, what kind of conclusions can we draw from this? Well, Death and Decay with Morbidity does seem to have a lower framerate than without. Spell detail helps the issue, but how much? Test 1 showed a spike down to 20FPS, and while that could be an outlier, I think it's worth investigating further. Especially since I run with my spell detail (mostly) turned down during Sartharion, with frame rates that were even lower - granted, I also had some mods running, and I'm tanking even more at once.

Either way, looking at the minimums, you're looking at about 15-20FPS difference on my computer, no matter the spell detail. This difference could be even more significant on slower machines - I'm sure there are plenty of people who raid at only 20FPS! Hopefully this will help raise awareness of the problem to get it fixed, because not specing into Morbidity significantly hurts our ability to do damage and generate threat in AoE situations.

Blueduck3285
01-19-2009, 09:50 PM
Great post!

Tarrke
01-20-2009, 01:45 AM
Great post and great analysis.

If I were you I'll be sure to post this on this official forums so noone can argue that you did the tests but never send the results to Blizzard.
Something is bothering me, as I've seen in an other thred, every one is OK to say that morbidity is causing frame rate drops, but some one suggested that this may be linked with the use of the specials voices for DK. Some people reported that once they unchecked this option things were back to normal. If you may test this with the same protocole may be interesteing to have some data on this.

The post I was talking about : World of Warcraft (en) Forums -> Death Knight preformance issue with c. spells (http://forums.wow-europe.com/thread.html?topicId=6915212201&sid=1&pageNo=5)

ttocs
01-20-2009, 07:11 AM
Thanks for the suggestion! I've made a post in the sticky at the tech support forum on the US forums.

DK voices is something I can arrange for another test, though I'm tight on time for the next day or two. I'll see what I can come up with.

Satorri
01-21-2009, 05:14 AM
Very good work ttocs!

Thanks for doing that, it's that sort of thing that can help the devs home in on little quirks like that to fix, and save all of our frame rates who use the tastey DnD in tanking.