Guild Wars 2

I am formally requesting Arenanet to release a 64 bit client for Guild Wars 2. Far too many AMD users are getting nigh unplayable frame rates in what is considered the two best aspects and two biggest selling points of this game ( WvW and PvE World events) and as such our experience in playing a game we both love and actively support has been diminished.

Better CPU optimization all around should also be a major concern for you as well, as Intel users are experiencing their fair share of issues as well, but these fixes, whatever they may be, need to start somewhere, and I hope the developers will read these forums so they can read and understand the amount of troubleshooting this amazing community has already done for them, and then get to work on delivering the best gaming experience possible.

If Blizzard can release a 64 bit client for an ancient, unoptimized 2005 game engine, so can you Arenanet.

So please, help us out, and help make GW2 an evenkeeled, and wonderful experience for all users. The investment in time and resources to properly optimize this game for modern CPU’s might seem steep now, but I promise you it will pay off in spades later.

Thank you for your time.

See you all in game.

A 64-bit client doesn’t automatically mean better performance. As a matter of fact, last I recalled, I get very similar performance between WoW’s 32-bit and 64-bit clients.

I’m not against a 64-bit client though, but just feel there’s other things needing to be done first. What needs to be done is just overall better optimizations to allow the CPU and GPU to work with each other in a more efficient matter.

For example, Reflections should be both CPU and GPU-intensive, but currently is only CPU-intensive. Spread the load out, and performance should increase in that regard.

I feel introducing a GPGPU language would be a pretty efficient way of doing this. From my understanding, GW2 used to have some CUDA optimizations, but this isn’t the right way to go, especially for a MMORPG. You “want” a larger audience to take advantage of performance improvements, and therefore, OpenCL would likely be the best choice due to it being unbiased on NVIDIA and AMD hardware, and being cross-platform. 2nd best would be DirectCompute, but that’ll limit you to Windows, and probably isn’t a good suggestion in the long run either (OS X and Linux support). Heck, even AMD’s Mantle sounds like a pretty awesome choice, but there’s little details on that currently.

Also, WoW’s engine is not “ancient”. Sure, it’s existed for a while, but it would basically be the same as saying that Windows 8.1 is ancient since Windows existed since like 1985 or so. The engine components are modernized, and I’m almost positive a lot of the core engine parts have different code than original launch.

I get a massive difference between WoW 64 and WoW standard, but part of that might be I’m running an AMD chip.

That being said I do agree with what you said, as the problem is in heavy zergs obviously, and the completely erratic GPU and CPU utilization that plagues other aspects of the game for most users, and therefore better optimization all around would be the ideal route. But the cuda thing is funny to me, because a lot of people have quite a few problems getting any decent GPU utilization in this game ( my GTX 670 included) when running GTX 600 and 700 series cards, though fix for this involves a lot of finger pointing between Arenanet and Nvidia. It could be we need an official optimal profile that Nvidia simply hasn’t gotten to, or it could be poor coding on the Arenanet side of things, or it could of course be both.

Nonetheless there are also a lot of people dealing with memory leaks and the heavy 2.7 GB usage that is found in major zergs, and a 64 bit client would go a long way towards helping those people out.

Basically I’m for anything at this point, 64 bit client not withstanding, that solves the lower fps issues currently plaguing the two best aspects of this game, and I hope we get them.

But since not everyone is experiencing memory leaks then if the cause is the game, it would likely be the game and some other bit of software running that isn’t on everyone’s system. A particular version of a driver. An interaction between a driver and a particular model of video card. Etc.

64 bit client would be a decent start towards better performance I think.

In and of itself 64-bit will do very little.

- Logic and program flow instructions aren’t affected by 64-bit.
- Floating point math isn’t affected by 64-bit.
- Integer math is but the question is how much of the code is being slowed by integer math.
- Address space is but since the game is compiled with large address aware then it has access to 4GB of memory space on a 64-bit OS. Nobody has reported seeing the program go past 3GB much less approach 4GB.
- 64-bit doesn’t impact multi-thread “balance” for lack of a better term. Right now the code is multi-threaded but it appears that the majority of the threads are either very short in execution length or infrequently run and therefore have limited impact when additional cores are present.

64-bit is a buzzword gamers have glomed onto much like Dx11. It’s very much like a first time car buyer wanting alloy rims while ignoring other performance options. It has little impact when the engine is underpowered for the size of the vehicle. Let them fix the engine first before worrying about the 2nd tier performance enhancements.

64-bit and Direct X 11 client would do a lot for this game. Good multi-core support would be appreciated as well.

Signed-

But not just a 64 bit client. Memory leaks, bad multi core optimization, AMD fix, and down the road maybe even DX11 need to be addressed. 1 year + into a game this amount of issues are unacceptable.With most games people scream bloody murder if they can’t hold a steady 60 fps, on this one people brag when they can hold 40 in WvW and the game don’t look “that” great.

kailin.4905:

Memory leaks

Define memory leaks? I’ve had GW2 opened for 12 hours last night (played for 11 of those hours), no crash at all. If the 2.7GB limit thing going around is true (which it isn’t), I would of crashed hours earlier while I was doing a SB event.

Proof of 2.7GB usage: https://forum-en.gw2archive.eu/forum/support/tech/Will-the-Optimization-help-AMD-CPU-s/first#post2974136

Optimizations are nice, but people seem to be asking for the wrong things.

I would but I have no memory leak issues either. I do have a few people in my guild who have been having issues since the teq patch with it though. I don’t want to give details on something I am not getting myself ankitten ot 100% on how is occurring though. Thats a good way to get chewed up on this forum.

This is not a bad idea, but there are plenty of issues that could present a problem. For one, the GW2.DAT file (I believe this is a proprietary database technology) most likely needs to be converted to 64-bit which may require content modifications or conversion.

If it IS possible for a 64-bit client to just allow memory access beyond 4GB, then that could correct some existing issues, but it is doubtful that would do ANYTHING to help general lag / communications issues across the internet.

For those here that have never done any programming; the client would have to be nearly completely rewritten. Yes, it would be nice and could improve the performance a lot. However, I don’t see it happening any time soon.

Furthermore, too many people here (seem) to think that being able to use more memory would help. It won’t. In fact, using excessive amounts of memory (which GW2 doesn’t), is a sign of poor and inefficient programming.

The important thing is getting data from RAM to the GPU. Wasting additional RAM space doesn’t help.

I believe that the main issue here is that the NETWORK portion of the coding is not nearly as efficient as it could be. Most of you aren’t having rendering problems – it’s that the CPU has to process so much information coming from the network. THAT could certainly be more efficient.

We wont get DX11.
We wont get a 64 bit client.
Heck we haven’t gotten an update regarding the “apparent” performance updates that according to the devs we should have already seen ingame.
I would be the 1st to want optimization for the game but I cannot see how they will do that.
In over a year there have been very little to no performance updates and the game still plays like beta. Seriously it does.

I’m not asking for more memory allocation. Nor am I asking for a 64 bit client expecting to fix how inefficiently threaded the game is. I’m asking for one because specifically for AMD and midrange Intel user’s, a 64 bit client in another MMO went a long way towards fixing that games biggest hang ups and lag orgies.

I’m hoping just like you, that any of those fixes “promised” by Arenanet recently will help with the issues a lot of us are experiencing in this games best moments. But you and I know they most likely won’t, so I’m hoping giving us an alternate 64 bit client can pave the way to a more efficiently threaded experience.

I just think it’s ridiculous that quite a few of us are going to have to wait for a better single threaded solution to come along from our manufacturers in the age of multi-core processors to either max out our gfx cards, or get playable frame rates in the best parts of the game. Running 5.0 GHz and not seeing any difference on an FX 8350 is bad enough on the AMD side, but I’m not just thinking of them, i feel just as bad for someone running a massively high IPC single threaded 3770k that they too have to wait to brute force their single threaded performance till Intel releases something better down the road as well.

Whatever the solution is, whatever brings an increase in performance over as many platforms as possible, please optimize your game Arenanet, no matter how much work you have to put in. This problem is a common one and is keeping scores of dedicated gamers from enjoying your game to the fullest extent.

In the end, that’s all that matters.

24 hours prime stable 4.4ghz 965 phenom ii with the nb-cpu at 2800(bigger per mhz performance increase than clock speed itself from a stock 2000 makes a huge difference in pIIs) Shadow Behemoth on High Settings at 1600×900 hits 18 fps lowerging gpu settings does nothing and scaling between 4.0 and 4.4 shows fps gain so i know the issue is cpu limit. I’m using a 7770+7750

My girlfriend is using a dualcore pentium g6950(i3 or i5 based i think) that i’ve pushed from 2.8 to 4.1 and a single 5750 hd. her frame rates beat mine/meet mine depending on location with a gpu equal to my 7750, one half of my gpu equation. and half the available cores. running near the same frequency. my computer pretty much trumps hers in almost every game we play but not gw2….. i know the ipc on intel chips is generally superior but given might overclocker of an intel chip has none of the cache its high end brethren have or my amd chip and its running at a lower clock rate and 2 core disadvantage this is clearly an issue with optimization and threading…..the cheapest proc intel had to offer at the time shouldn’t be beating out my system. not when other cpu bound games perform better on my computer.

CursedShaitan, if you’re running crossfire on your 7700 series cards, disable crossfire for your game and guild wars should give you a couple extra frames. This game hates crossfire.

Yawa what do you mean by an AMD client. One that can support eight cores? Anything that can be done to make it perform better on a midrange AMD CPU will automatically make it better on an Intel CPU. So I’m confused by your request.

Behellagh.1468:

Yawa what do you mean by an AMD client. One that can support eight cores? Anything that can be done to make it perform better on a mid-range AMD CPU will automatically make it better on an Intel CPU. So I’m confused by your request.

an FX 8350 isn’t the usual mid-range CPU… If you compare the full cores with a high end intel one with 4 physical cares with hyper-threading, the amd is more powerful since the cores don’t share the same memory.
Intel is better for you compare just one core. The problem here is more that guild wars 2 uses a ‘updated’ version of the guild wars 1 engine. Maybe they miscalculated the power of it. Maybe they are already making something new. And as some ‘programmers’ here know it can’t be done in 5 minutes, it will take time until we all get a real update for this issue.

Just because game “W” got a 64-bit client that cleared up some issues for AMD users does NOT mean a 64-bit client for GW2 will have similar results. It is SO much more complex than just churning out a client that runs 64-bit.

“…No matter how much work you have to put in…”

Did you seriously type that and believe it? Grow up.

CursedShaitan.9720:

24 hours prime stable 4.4ghz 965 phenom ii with the nb-cpu at 2800(bigger per mhz performance increase than clock speed itself from a stock 2000 makes a huge difference in pIIs) Shadow Behemoth on High Settings at 1600×900 hits 18 fps lowerging gpu settings does nothing and scaling between 4.0 and 4.4 shows fps gain so i know the issue is cpu limit. I’m using a 7770+7750

My girlfriend is using a dualcore pentium g6950(i3 or i5 based i think) that i’ve pushed from 2.8 to 4.1 and a single 5750 hd. her frame rates beat mine/meet mine depending on location with a gpu equal to my 7750, one half of my gpu equation. and half the available cores. running near the same frequency. my computer pretty much trumps hers in almost every game we play but not gw2….. i know the ipc on intel chips is generally superior but given might overclocker of an intel chip has none of the cache its high end brethren have or my amd chip and its running at a lower clock rate and 2 core disadvantage this is clearly an issue with optimization and threading…..the cheapest proc intel had to offer at the time shouldn’t be beating out my system. not when other cpu bound games perform better on my computer.

I have a PhenomII 830 X4 teamed with a HD5770 – Over clocked the CPU to 3.4Ghz and on SB I get 28-30FPS on medium settings. NB/HT are at 2400mhz. If I over clock it to 4.5Ghz (NB/HT 1100, ram at 1100 – cant get stable at 4.5 with higher HT/NB) then I get 30-35FPS with the same settings. (Small boost, would be a lot better if I could get the NB/HT up to 2200-2600mhz)

I upgraded that PhenomII 830 to use a HD7790, repeated the Clock tests (I make sure there are at least 30 Players around when doing these simple tests) and there was absolutely no change in FPS between the 7790 and 5770.

Upgraded the PhenomII 830 to a i5-4670K, Default clocks (3.4Ghz) Turbo Boost Disabled. SB with the same settings/Players solid 60FPS. If I put the settings on high (have not tried Ultra yet, still needing to run test in Full WvW prime time), 35-45FPS.

Native Sampling
Low Shadows
no post processing
Medium Character Quality
Low Character Limit

Rest of the settings are either High or Medium.

So yea, completely CPU restrained. Though, while so far I am very happy with the i5 Upgrade, I expected more out of it. Can’t wait for my 80mm to 120mm Adapter comes so I can Install my Water Cooling and OC the i5 to 4.5ghz+

And a 64Bit Client (while would be a step in the right direction for GW2 for the future) will not help one bit with the current issues that plague this game.

The main issue is that optimization is so poor, the game runs 60% of its instructions on the CPU. This is the first thing that needs to be addressed before anything else.

The second issue are the lag spikes. Which I still think are Level3/AT&T based, but the last couple of days been seeing weird ‘pause for 5-8 seconds’ and ‘turbo mode go’ latency action (both my wifes PC and mine at the exact same time. However, Last.fm doesn’t drop a beat during these lag spikes, and neither does my Active VPN back to work.). These could be caused by a few things. Server DB Engine Spikes, Cluster resync’s, our Game Client dropping TCP/UDP buffers and refilling them, DDoS Attacks against Anet (this would be behavior one would see if an IP was under attack and a IPS was intercepting the attacks and offloading them to help keep services up), and the Internets backbone (with Verizon/ATT having area outages could very well be a possibility ATM).

Same problems here, Phenom II 965 + Radeon 7850 has less than half the fps than other PC with i3-2120 + radeon 6770 combo, while in many other games its other way around

hoegarden.4287:

Behellagh.1468:

Yawa what do you mean by an AMD client. One that can support eight cores? Anything that can be done to make it perform better on a mid-range AMD CPU will automatically make it better on an Intel CPU. So I’m confused by your request.

an FX 8350 isn’t the usual mid-range CPU… If you compare the full cores with a high end intel one with 4 physical cares with hyper-threading, the amd is more powerful since the cores don’t share the same memory.
Intel is better for you compare just one core. The problem here is more that guild wars 2 uses a ‘updated’ version of the guild wars 1 engine. Maybe they miscalculated the power of it. Maybe they are already making something new. And as some ‘programmers’ here know it can’t be done in 5 minutes, it will take time until we all get a real update for this issue.

What are you talking about? The FX-8350 does to share the same memory. Now if you mean cache the FX series is arranged with 8MB of Level 3 cache, 2MB of Level 2 cache per Module and 64KB Level 1 Instruction cache per Module and 16KB Level 1 Data cache per core. An Intel i7 has 8MB of Level 3 cache, 256KB of Level 2 cache per core and 32KB of Level 1 Instruction and 32KB of Level 1 Data cache per core.

The purpose of cache is to reduce fetches off CPU to system memory which compared to the speed of the CPU is painfully slow. Use of this memory is something you can’t quite program for. Also Intel’s is considerably faster, especially in the Haswell, with noticeably lower latencies. Even in the 512KB and 1MB block size tests, when the AMD has the L2 cache advantage, Intel’s still has a higher bandwidth.

http://techreport.com/review/24879/intel-core-i7-4770k-and-4950hq-haswell-processors-reviewed/8

Then you look at overall memory bandwidth of the entire memory subsystem. Using the same speed DDR3 memory, Intel beats the AMD by 11%. Sure the FX has more L2 cache memory. But that doesn’t fix the problem that it has smaller and poorer performing L1 cache. The L1 instruction cache on the FX is 1-way per core within the module. That’s insane. Yes it’s 2-way overall but 1-way per core means a greater chance of L1 instruction cache misses. I haven’t heard of a 1-way cache since the original Compaq 286 which had the cache built on the motherboard out of discrete parts. I understand why they did it that way, to guarantee balance use of the cache between the two cores in the module but the potential downsize is huge. Which is also why the L2 unified cache per module is so huge.