TR Forums

I'd say the slowness is more just consistently slow than stuttery. I did used to have a huge stuttery problem, but a clean driver install and game reinstall fixed that, now it just runs slow everywhere almost.

I rounded up to 10% I should say, that's the highest I ever saw it at. Often it was 6% or even 2%.

And yes, Planetside 2 was pretty bad. lol It was super dark, and even looking in keybinds I couldn't find the flashlight key. I kept getting shot by dudes way farther than my rifle could shoot accurately, or got sniped by aircraft. I'm sure if I cared enough to put in some time it'd be ok, but it ain't the original, that's for sure.

Yeah true, Planetside 2 doesn't handle the early region of its learning curve very well. Also on Linux (Wine) my flashlight just straight-up doesn't render, FWIW.

Hmm, so memory is fine and CPU <-> GPU comms are fine. Win10 and thread scheduling?

Thread scheduling has got to be easier to check. Since it's consistently bad, all I can really envision is the game messing something up via core affinity or such. If load looks well-distributed between cores, it's almost certainly fine.

My guess at this point (without much to back it up) would have to be that Win10 + heavy DX9 use (Planetside doesn't abuse DX9 in this way) is extra-demanding on memory bandwidth. GW2 players running Win10 with <2133 RAM on otherwise very fast systems would be a restricted enough set that it could avoid being a widely-known issue. It could well also be a graphics driver interaction, and could be worse for one GPU manufacturer or the other. The easiest way to test this would probably be to drop your RAM to 1333 temporarily and see if the problem gets noticably worse.

So, I went into the BIOS and put the memory frequency to 1333, rebooted the game and no change... to make sure I'd done it right I booted up CPU-Z, and noticed that it claims the max frequency is 800. Weird. Must've set something wrong I though. Went back into the BIOS to set it back to auto, and make sure XMP is on, and it set to 1600. Booted up CPU-Z and same thing, 800. Am I reading CPU-Z wrong? Or could this be the crux of the problem?

Edit= Nevermind, huge brain fart there. Of course you double it. >.< So 800 is normal. Why didn't it change when the motherboard was set otherwise I'll have to look into.

CPU-Z may just be looking at the XMP frequency, in which case it wouldn't see a change. (I wouldn't know.)

I've got one more idea - what's your internet connection like? High latency can make games do otherwise-unnecessary recalculation on networked components. With good architecture, this isn't an issue, but if most stuff is sharing a single thread it could be (especially with a game like this). Trying to ping the server directly might be worthwhile; you can probably figure out its IP by looking in resource monitor at what addresses the game is talking to, if it isn't otherwise well-known.

You are running the 64-bit Guild Wars 2 game client, aren't you?

Yep, 64 bit.

As for my internet... well, wouldn't a tracert be more revealing than a ping? Also, one of my friends and fellow tech dudes has a similar connection to me. He's the one with the 780. He can run the game better than I can.

Something hilarious happened while testing some things earlier. I was dinking around in afterburner, using the stress test in there and my UPS started beeping rather loudly then eventually just shut off. It was really not happy about the load I put on it. Cheap piece of crap, I only have a 600w PSU, can't have been that much of a load. Switched to my rackmount 1200 VA unit to alleviate that, I'll have to get a better one for under my desk. lol

Also while trying to push the bus, I noticed that with shadows on ultra, I can spike the bus load to 14%, but that's the highest I managed to get it. Would dropping to PCIe gen 2 on the firmware of the mobo break anything? I could probably see how bad the issue got if it was indeed a bus throughput issue.

Vinceant wrote:

As for my internet... well, wouldn't a tracert be more revealing than a ping? Also, one of my friends and fellow tech dudes has a similar connection to me. He's the one with the 780. He can run the game better than I can.

Connections of similar spec may still have wildly different pings to a particular location. See ISPs underprovisioning links Netflix needs to go across for an extreme example. I don't know why I was thinking about manual pings though, whatever the game says should be just fine. What are typical/max pings for GW2 on your connection? Tracert is no big deal because we don't much care why the connection is the speed it is.

From personal experience, 100ms is enough to bring this effect mildly into play on many games and pretty strongly on a few. With most architectures it won't affect framerate in such a consistent way, but MMOs like GW2 are potentially pretty vulnerable to it.

Vinceant wrote:

Also while trying to push the bus, I noticed that with shadows on ultra, I can spike the bus load to 14%, but that's the highest I managed to get it. Would dropping to PCIe gen 2 on the firmware of the mobo break anything? I could probably see how bad the issue got if it was indeed a bus throughput issue.

Dropping to gen 2 should be fine. It's probably just making a bajillion draw calls (by DX9 standards) - dynamic shadow mapping involves re-rendering the scene from one or more alternate points of view, and can inflate draw call count drastically.

I should specify. He is an IRL friend. Lives in the next town. We have the same ISP. Same hops to the server. He has a slower connection in terms of throughput though, but that doesn't matter in this regard, latency should be comparable.

As for latency, I do pay attention to the in game latency. It doesn't seem to get better or worse with or without latency. I range from 30 all the way to 150 (They've had some DDOS attack issues lately from what I understand). Runs similarly regardless.

I may swap gens down to 2 just to see the effects in the next day or so. What I did consider was a firmware update on the card itself. Unfortunately, while Gigabyte has a newer firmware, it does not seem to have a utility that recognizes my card. So that's a dead end that may have lead somewhere.

I also reseated my GPU earlier. No luck.

And as a total wildcard to eliminate ANY possibility of settings I even made a new free account, deleted the settings file and loaded it up. No dice. My thinking was thus: Some settings are maintained across installs, and while normally I wouldn't think they would effect performance we are quickly running out of good explanations for why this game runs the way it does on my system.

I'm really at a loss here. Between everybody's ideas it seems we've covered things extremely thoroughly. I've tested this system top to bottom and every indication is that not only is my system in perfect repair, it is also running at specc or better. I'm still with you in guessing it's some kind of throughput issue on some memory, but I can't figure out why GW2 would be the only game with this issue, or why I seem to be one of very few that have found it.

I'll keep my eye on this- running a 6700k, 2x16GB DDR4-3000, and a pair of GTX970's with G-Sync.

I've only just updated and popped in with my main, near the bank and the forge/fountain thing with a lot of other characters and the game runs between 25-60FPS per my FRAPS overlay.

Seems pretty crappy for now, but I'll have time to check it out in other areas later.

Hard to say if that's the same as my problem, but sounds about how my performance goes. Perhaps it's a problem with some 900 series cards?

If you could test with a single card and report performance that would help a lot. SLI being SLI means your problem could be completely different from mine.

What I can tell you is that the game was once slower with SLI disabled; most recently I jumped in while monitoring GPU usage, and it was about 50-60% on each card.

I can certainly try disabling SLI, but I don't think that will be a positive long-term solution.

No, but it might tell me if this problem is more than just me. I know people with cards equivalent to a 970 that run this game significantly better than I can.

And if we have the same issue, then it's likely this issue is effecting some series 900 cards at least. It would make my problem less... odd.

There may be more things in Tyria today that hurt performance than there were a few years ago. I have noticed that fires (and the flickering light and shadows that they cast) tend to drag down my fps significantly. I wonder if there are technical as well as artistic reasons that they changed the way that fires are rendered in the new Path of Fire expansion?

Other than that, the typical developer's answer to software coding issues these days is to throw hardware at the problem. Put a Core i7-8700K on your Christmas wishlist.

Vinceant wrote:

I suppose I'll test Planetside 2 here in a bit. These are some great suggestions thus far.

I'd think if WDDM was cutting DX9 performance potentially in half it would be a much wider reported issue, though.

synthel2's thoughts sound like a good track to go down. Any ideas on how one would go about testing or monitoring the PCIe bandwidth? I've never heard of such a thing personally. I know GPU-Z definitely reports gen 3 at 16x, but that's not exactly testing it. Any ideas why GW2 alone would suffer so badly?

I'll see if I can carefully study thread distribution more, maybe that'll give a clue. I want to say that it seemed to be evenly distributed before, but I didn't look TOO carefully at it. Windows does switch threads a lot from what I understand so this may not be easy.

Edit= So as a quick aside, I did some benchmarks in passmark just now. The DX9 test and the memory and processor tests seem to fall in line with similar hardware other users have benchmarked with the software.

I am certainly not implying that you are not aware of how to install directx 9 but I only suggested it because I remember when I played that game that I had some similar issues and for me it ended up being not having all the directx files that were needed.

A fresh win 10 install still needs close to 100mb of directx 9 files.

credible wrote:

A fresh win 10 install still needs close to 100mb of directx 9 files.

Typically, the game will install whatever dx9 files it actually needs and the others are optional.

JustAnEngineer wrote:

I have noticed that fires (and the flickering light and shadows that they cast) tend to drag down my fps significantly. I wonder if there are technical as well as artistic reasons that they changed the way that fires are rendered in the new Path of Fire expansion?

Usually something of that description is heavy because of the need to render the (moving) shadow map, which in this case should have costs centered on draw call count (single-thread / PCIe / possibly main memory bandwidth). Unless this game has some graphics much better than I'm thinking, they shouldn't pose much GPU load (especially because Maxwell is really good at simple shadows).

Maxwell's drivers being slow on some edge case GW2 hits seems plausible enough.

A friend of mine who plays GW2 on Win10 + a 6600K + fast RAM + either a 390X or dual 480s (dunno what he's got installed right now) reports that it's a pretty stable 70-80 fps except when PvP is dragging it down, FWIW.

JustAnEngineer wrote:

Other than that, the typical developer's answer to software coding issues these days is to throw hardware at the problem. Put a Core i7-8700K on your Christmas wishlist.

Yeah I guess. In all fairness my processor is what, 5+ years old now? I think I've gotten good mileage out of it. Though I've only owned this model for 2.5 years. I originally built this machine with an i3, boy was that upgrade nice.

I've been eyeing the latest processor releases carefully. Threadripper is pretty awesome, but I have almost nothing that would push that very hard. I was thinking about the 1800X and came to the conclusion that covfefe lakes release will make it obvious which top end proc to buy in that range. Either the 1800x or the 8700k. If the 8700k is around the same price, I'll sacrifice 2 cores for the raw per core performance lead that Intel has enjoyed.

In slightly better news I did manage to improve the performance in GW2 decently. In all this faffing about with hardware settings I found that Gigabyte has an OC tool. When I used that I found that my "factory OC" wasn't on this whole time owning the card. lol Thanks Gigabyte. Anyway, boosting the clocks by about 25-30 MHz and putting shadows on high improved performance significantly. Verdant Brink actually runs at 60 FPS in some places. I still get dips quite often to 30, but it's still an improvement.

Putting shadows on Ultra automatically makes the FPS go to 40 or less no matter the view or resolution though. Basically nothing else in the graphics settings seems to have much effect on performance other than shadows at this point.

Vinceant wrote:

Ok, so I have a weird problem and I thought maybe a bunch of tech guys might be able to help me here...

Since the announcement of the new Guild Wars 2 expansion I've been trying to get back into the game. I haven't played in a few years and not since upgrading my hardware significantly and the release of the 64 bit client. Upon my return I was rather disappointed with the performance. Regardless of settings sometimes I don't get more than 30-40 FPS. I have an i7-3770K running at 4.4, and a 980 with 16 gigs of RAM. At first I just assumed the engine was ****, but talking to people with slightly lesser hardware has revealed that no, it's supposed to run at 60 or higher at any decent hardware specc. The question now is, why can't I?

I haven't played the game in a couple of years, but the core engine and graphics should not have changed much. I played PvP quite competitively and easily tuned the settings to 60+ fps with my R9 280X and FX 8350. When I got the Rx480, a quick test showed ~80 fps. I haven't tested since I got the Ryzen 1700X.

I suspect that some settings may have a profound effect on performance (for example, number of visible players or shadow quality, or visible range). Put everything to LOW (yes!) and confirm that you can hit 100+ fps in empty regions, for example when running in the mountains. Then test the same in the big cities to seem what is the impact of having many players. Finally, increase quality until the fps remains solid in the most likely scenario. Medium/high at 1080p should be achievable for most settings, but check all individually.

Also, I haven't had a nvidia card since a few years. Verify that you don't have a snapshot recorder or streaming or something similar running all the time... Disable all unnecessary background tasks.

All good thoughts, but alas, they've been done. No, even at low and subsample it does not run at 100+. All the more indication that something is wrong. But with all the tests and ideas that have been thought out and tried in this thread I can't come to a conclusion except that the game is broken in some way that I cannot fix.

I believe you though. My hardware SHOULD run this game amazingly. It just doesn't. At this point I cannot but conclude that this is no fault of my hardware or me.

Vinceant wrote:

All good thoughts, but alas, they've been done. No, even at low and subsample it does not run at 100+. All the more indication that something is wrong. But with all the tests and ideas that have been thought out and tried in this thread I can't come to a conclusion except that the game is broken in some way that I cannot fix.

I believe you though. My hardware SHOULD run this game amazingly. It just doesn't. At this point I cannot but conclude that this is no fault of my hardware or me.

This is sort of bizarre. I play on a I3 3220 with a 2Gb 7850 with 8 Gb of ram. The only slowdowns in FPS I have is when I'm in WVW in big zerg fights or open world boss fights . I have a I5 2500 that I usually game on but its currently boxed up in the closet because Ive had very few issues on the I3. Granted, I dont have all settings cranked up, most are on medium/high. You are correct your hardware should run the game very well. I know that there were some server issues within the last month or so.My system is downright archaic compared to yours, gotta run the Smithsonian is calling.

Vinceant wrote:

I've been eyeing the latest processor releases carefully. Threadripper is pretty awesome, but I have almost nothing that would push that very hard. I was thinking about the 1800X and came to the conclusion that covfefe lakes release will make it obvious which top end proc to buy in that range. Either the 1800x or the 8700k. If the 8700k is around the same price, I'll sacrifice 2 cores for the raw per core performance lead that Intel has enjoyed.

Intel usually makes the top -S part a fine value, but an 1800X is a whole lot more money and barely faster than the two 8C parts below it, especially if you overclock. I'd at least toss a 1700X into that comparison too.

Vinceant wrote:

In slightly better news I did manage to improve the performance in GW2 decently. In all this faffing about with hardware settings I found that Gigabyte has an OC tool. When I used that I found that my "factory OC" wasn't on this whole time owning the card. lol Thanks Gigabyte. Anyway, boosting the clocks by about 25-30 MHz and putting shadows on high improved performance significantly. Verdant Brink actually runs at 60 FPS in some places. I still get dips quite often to 30, but it's still an improvement.

Putting shadows on Ultra automatically makes the FPS go to 40 or less no matter the view or resolution though. Basically nothing else in the graphics settings seems to have much effect on performance other than shadows at this point.

A ~2% clock boost is worth a notable amount of performance? Methinks there's more to that than is obvious.

Maybe GW2's shadows are somehow pretty heavy on GPUs? That could yet be due to driver weirdness. Looking through some screenshots, it looks like they're very low-res and there's a lot of geometry that doesn't apply them, but they look nicely soft and anti-aliased. Random screenshots make it tough to tell what kind of render tech is likely responsible. When you change shadow settings, how does the look of them change?

Lots of shadow tech is heavy on fixed-function GPU hardware and light on shaders, not that that should bother Maxwell. It could explain low GPU utilization figures despite a GPU limit though, depending on how utilization is calculated.

One coffee lake drops I'll look into more comparisons.

So here's the odd thing about shadows, my friend with the 780 says shadows do not load his GPU hardly at all, yet with mine they grind things down significantly. This has got to be something with 900 series and the weird way the shadows are done in this game. For how blurry they are they sure demand a lot. Kinda wish they would update the engine to DX11 at least, it'd probably run 50% better on everyone's machine. I know that's not easy to do, but DX9 is pretty bad.

In-game settings under the graphics options, what have you got Character Model Limit and Character Model Quality set at?

If they aren't already, put them both on "Lowest" and see what that does.

From my own testing, with more players around the game becomes incredibly CPU limited, and by a single thread only - with those at maximum I've seen framerates drop into the 20's with 100% use on a single CPU core and other cores <20%, also GPU usage (monitored using RivaTunerStatisticsServer and its overlay) drops to maybe 30-odd% - it totally chokes on a single CPU thread.

With those on "Lowest" and "Low" respectively, and everything else on maximum (ultra shadows, native render sampling) I usually get 60 fps (using gsync) with occasional drops into the 40's. Similar spec to yours, i7 5930k at 4.4 Ghz with 16 Gb RAM and a GTX 1080ti, running at 3440x1440.

If you are running MSI Afterburner or RivaTunerStatisticsServer, make sure they are on the latest versions, they caused problems for me with GW2 a while back after a Windows 10 update.

Model limit set to lowest, model quality on highest. If you put that on lowest everybody becomes a super generic model that's not animated well. I've never seen a single thread hit 100% on my end. Load seems to be pretty evenly distributed across all 8.

Hmm, based on what people are reporting it might be time to get an AMD card. Depending on how important GW2 performance is.

AMD doesn't have feature parity with Nvidia enough for me to even consider them. The only thing AMD has that Nvidia doesn't is Freesync, which IS nice, but not nice enough for me to drop the rest of the niceties of green team.

Also, this clearly isn't just a green team problem. If all nvidia cards ran at half the speed they should in this game it would be reported everywhere. Whatever problem I have is rather obscure at best.

So playing some more it would seem that speed sweet spot really only applied in parts of VB. Going into the second HoT zone I was basically playing at 30 or less most of the time. It's playable, but hardly ideal.

Vinceant wrote:

AMD doesn't have feature parity with Nvidia enough for me to even consider them. The only thing AMD has that Nvidia doesn't is Freesync, which IS nice, but not nice enough for me to drop the rest of the niceties of green team.

Mainly because they aren't jerks about everything. For example TrueAudio Next is open source. GPUOpen actually helps developers instead of locking them into NV tools which they aren't allowed to optimize for other architectures at all.

I'd be actively avoiding NV still if Vega had come a few months sooner.

Oh, I agree about NV being jerks. It's well known they do such things. I wish it weren't so.

Even still, AMD doesn't have anything like shadowplay streaming, which I use all the time to play coop and other games with my brother over the internet.

For people who don't care about such a feature though, Vega looks to be a nice couple of cards. While AMD doesn't ever hold "top dog" cards I'm not sure they care to, they almost always win price to performance on the midrange to the lower end of the high end market. They are good value for money cards. If you can buy one at MSRP that is.

Running 64 bit GW2 on a i7 4700, 16 gigs RAM @ 1866Mhz, NV 760 w/2 Gigs VRAM on a PCI 3.0 x16 slot; it runs pretty smooth except for huge WvW battles and world train bosses. "Pretty smooth" means 50-60 fps+ in most cases, at 1920x1200 with all options on and up to the max (except depth blur - I just don't like that look) including shadows and character models, and FXAA anti-aliasing on 64-bit Win10. I didn't notice a difference with Win10's "game mode" one way or the other.
My biggest issue with the performance is the map load times (15 seconds+) which is admittedly a first world problem and better than some but it still breaks immersion.
I've read many places that the GW2 engine is single core CPU bound - the game's main thread is single core and not really multi core optimized, so more GPU muscle or mulitcore CPU core muscle won't help. Unless the devs do a huge engine rewrite and move to DX11/12 or Vulkan (native linux without wine would be great - maybe GW3?) we are pretty much single core bound on this game. The 64-bit client helped with some of the memory crashes but it's still a fundamentally single thread engine and game.

I mean, yes, that is true, but Ivy at 4.3 is no slouch in single threaded performance. Either way, it seems nobody knows why a setup like mine wouldn't get the performance of lesser hardware. As I said earlier, I suppose it's time to build a new core of a machine. Coffee lake and Ryzen offer very good platforms in theory, we'll see which stacks up best in a month or so, and in a couple months I should have the dough to replace my board, RAM, and proc.

I should be able to sell my current board and proc for a couple hundred maybe, that'll ease the financial pain of upgrading.

In the meantime I suppose I'll just have to deal with 30 FPS or less. Unless they fix the engine, which I doubt. The engine is really just a modified GW1 engine, and it kinda shows in some places (movement, animation, things like that are really similar between games). I don't think they did too massive of an engine rewrite between games. DX11/12 would make this game run really well though, on most computers.