Guild Wars 2

Hi everyone,

now, let’s het to the point!

I have AMD FX8320 CPU, overclocked at 4.5Ghz, VGA is Sapphire HD 7950 vaporx, 8Gb of Kingston beast RAM 1600Mghz

Please, don’t say that it is problem with my hardware, i tried everything, 3 sets of same cpu, ram and vga, and from what i see this game demands intel CPU,s, AMD is just crap here, i tried stock clocks, everything and it still sucks.. in WvW on medium settings its barely 20FPS, lion arch events are unplayeble

I think it’s realy time for u to focus on making this game more compatibile with AMD cpu-s, in other games my CPU is a BEAST, it runs 8 cores on 4.5Mghz, also VGA is BEAST, i have only problems in GW2

I understand that every MMO use more CPU resources, and CPU is more important than VGA, but i mean realy, from what i see in various HW diagnostic tools, this game use only one core???
This is next gen mmo, u advertize that, and it uses only one core, sry but that just not work for me.

Sry for my bad english it’s not my native

I dont mean to offend anyone here, but i realy think it’s time to take all writen here in consideration, not everyone can buy Intel CPUs, they are much more expensive than AMD, personaly i don’t have that problem, but i love AMD, so if game force me to change to intel, il just quit, i DONT want intel in my rig

Thx for reading this, i realy hope it will have effect

You should include what your FPS is.

But I’ll tell you right now, that FX8350 is holding your HD7950 back. And its not really the FX’s fault as much as it is the games. The game uses DX9, which is single threaded, and your FX8350 gets beat by an i3-2120 because of that fact.

There are 3 things you can do to boost, possible fix, some of your issues. But dont expect anything higher then 65FPS and anything lower then 8FPS(Zerg content)

1. Disable the 2nd core per module in the Bios.
2. Disable the Boost option and Peg/OC your CPU at the Boost speed.
3. Disable core parking (locate the Utility via a Google search)

Thats about all that can be done with out further info about your FPS and what you are actually getting in game.

Data from CPU-Z, GPU-Z, and CoreTemp would help.

Thx for the reply,
its realy sad that “next gen mmo” still uses only directx9… i am realy sry but that just sucks
i already did all mentioned above, and i still have low fps
blizzard implemented multicore support in wow few years ago, and its game older than 8 years… so i realy do think that anet need to do something about it…
i totaly agree with u that i3 cpu beats my cpu in single core games, but we are in 2014,
and gw2 only support directx 9(single core), sry arena net but u are slacking behind big time…

An AMD FX-8 series processor cannot be compared between MMOs and non MMOs, especially ones that use DirectX 9. DirectX 9 relies on higher single thread performance than multi threaded performance, something that Intel processors have I think always had over the AMD counterparts.
I guess you could look at the AMD FX-8 series as being like a quad core with each core split into two parts. Two module units and there’s 4 units. In this case, of course Guild Wars 2 is not coded to utilize that. An Intel i7 on the other hand, which in nature could be looked at as being the same in a different way (four cores, eight threads) being clocked at 4.5 GHz blows the FX-8 out of the water (I know this from personal experience after upgrading to 4770K from 8350) and the performance was as some people put it, night and day. However a DX9 game can take advantage of it if it was specifically coded for it, maybe. I’m not sure.

As for Guild wars 2 being coded for DX9, it’s because developers are so used to it by now what with previous generation consoles and majority of PC games using it.

Sure not everyone can buy Intel, but when you look at it, AMD FX-8350 (example) is roughly the same price as the i5 3570K/4670K. so when you weigh those two options, people would likely pick the Intel when you take into account the added features the board could come with, assuming they don’t get a beater board.

In order to achieve similar performance however you’d be looking at clocking the AMD unit to like 5.5 GHz or something.

No I’m not bashing it. I have and used the FX-8350 and loved it.

This “next gen” MMO started being built in 2007, back when gamers detested Vista which was the only way to get Dx10, which still wasn’t thread safe. And the “next gen” aspect was about game design and not the underlying technology.

The game is multithreaded, it’s just that the work load isn’t as balanced as one would like.

As for the high and mighty Blizzard. With a BILLION dollars a year of subscriptions for several years you can afford to recode your entire game engine.

AMD marketing obfuscated the notion of core count with the FX series. While Intel’s method fakes out the OS into believing more cores exist than actually do, AMD’s method was to fuse parts of two cores with the remaining parts of a single core and called it a “module”. While running two threads at the same time, a “module” can be competitive with a HT Intel core but it’s at a severe disadvantage when each are only running only a single thread.

And that’s the rub. This game simply can not provide enough work to maximize the workload that an FX would be competitive to an Intel and isn’t graphically intense enough to make the limiting issue the graphics card. That’s why the HD 7950 doesn’t matter here. In games that are always limited by the graphics card, CPU really doesn’t matter a whole lot.

in 2007 when they started building the game, they also had AMD cpu’s, so that what u are saying is not holding, they could easly optimize game for intel and amd, so this sound’s like favorizing intel
also, yes, blizzard has subscription, and has loads of money, but they keep customer satesfied, here its not the case, i realy dont care about money issue, i just want to play the game i have payed, intel or amd should not matter

maybe they should advertize “gw2, next gen mmo, runs great on intel”, mite even get money from intel for that advertizement…

in last few days, i read lot’s of formus considering this issue, and all say one thing, GW2 is not optimized game, and thats it, it is fun game, but considering hw demands it sucks bigtime, so it’s bad product…

considering my FPS, here’s info

in open world i easly go over 100fps, vsync off ofcourse, but in WwW it, also easly, go below 20fps, i know that cpu need to calculate everything there, but that kind of drops are insane
i have all settings at medium, high char rez and rest like that is turned off, rezolution is 1920×1080

GW2 is DX9 because people hated vista and still used xp, which didnt have dx10 support.

Anet in no way “favored” intel. The performance gap can be found the the decisions each firm made in recent years:

In 2007 AMD was using the K seires (8-10) architecture that was excellent at single thread per core operations. The pinnacle of which were the Phenom chips (Phenom II x4/6 black edition is an excellent chip even today), which were THE chips to get, and highly competitive and arguably better than the intel series at the time. With all this pressure from AMD, intel was further compelled to push the performance of their next chips.

However in a weird move, AMD moved away from the k series excellent architecure and switched to the bulldozer/piledriver/steamroller modules for the new FX series chips and A series APUs.

Both AMD and Intel decided to pursue multi thread performance but their execution was highly divergent. AMD went the hardware route of using combined “modules” with 2 cores sharing a single FPU. This allows for multiple threads but at the costs of each single thread on the cores performing slower. The shared overhead for 2 cores on the modules still affects performance in many applications today mainly because of the way threads are spread by the api and windows. The software doesn’t “know” that it should move threads from core 0 to core 2 then 4 then 8 first, rather than moving from core 0 to 1 to 2 etc. Which is why disabling a core per module increases performance greatly.

Intel however decided to use hyper treading and run 2 threads per core while improving the already higher per clock efficiency of their cores to be able to run both of the threads effectively. This has the added benefit of greatly increased single thread performance per core. With modern software rarely using more than 4 threads (meaning 1 thread per core for an i5), Intel’s choice of improving per clock speed proved to be more effective.

The end result was that AMD sacrificed single thread for mediocre multi thread, while the Intel increased both single and multi thread performance. Which is why consumers have been disappointed with AMD performance lately.

So with DX9 games like gw2, intel has better performance because of the different methods both sides used to improve multi thread performance. With DX11 games multi threading is used, but still not enough to take advantage of the AMD hard ware, but the gap is much much smaller.

This is one of the reasons that AMD developed mantle because they wanted to push the use of multi threading in games and allow their CPU’s to be fully utilized.

I wouldn’t say mediocre performance. You put a FX-8350 up against a i7-2600K in heavily multithreaded applications and it keeps up. Problem is Intel is now 2 generations beyond the -2600K with each generation squeezing a bit more performance each generation.

Overall it’s an improvement over the Phenom II if you compared a 980 quad to an i5-2500 quad (Intel 25% faster whether it’s single or multithreaded). But the trade off is when the CPU isn’t fully used, Intel smashes the FX (single thread FX-8350 vs Intel i7-2600K, the -2600K is 40% faster). So both CPUs running 8 threads, virtual tie but single thread, FX loses badly.

And since GW2 is 1) lightly threaded and 2) not GPU bound except with low end graphics cards, single threaded CPU performance is more important and single threaded performance wasn’t an FX design goal.

The 980 was just a overclocked 940 that was already 2 years old. Comparing it to a 2500 is not exactly a fair comparison. A better choice would be the i7-930 which was a oc’ed i7-920 also 2 years old. If i remember correctly: the i7 was better overall especially multi thread, but the phenom was slightly better single thread.

I was thinking more of 2007/2008: Phenom X4’s vs (yorkfield/wolfdale)Core 2’s
or even 2008/2009 Phenom II’s vs the Nehalem core i5/i7 (lynnfield/bloomfield). In those days the AMD’s were just as good and people debated about which was a better overclocker and which was a better value, etc.

Then intel took it up a level and released the Sandy Bridge 2500/2600k and blew the competition away. AMD responded with the Bulldozer based FX series and results were…disappointing to say the least. After that there was really no question that intel was better.

The Phenom II X4 980 is a fair comparison because the FX wasn’t out when the i5-2500K was introduced. The 980 was the fastest Phenom II at 3.7GHz. (The i5-2500K came out Jan 2011, the Phenom II X4 980 in May 2011.)

Or would you rather have me compare the -2500K with the 975 which was introduced in the same month as the -2500K? It’s 100MHz slower than the 980. It’s not Intel’s problem that the Phenom II quad core desktop series of CPUs first came out some two years earlier in Jan 2009 and didn’t have a significant upgrade other than clock speed for two years. I picked the fastest 4 thread processor AMD had the year when the -2500K came out. I’m not picking based similar price, where AMD could compete, just simply the fastest available and that was the 980.

Or would you rather have me compare it to the FX-4100 which came out that October 2011? The 980 spanks the -4100.

So lets go back a generation. The i5-750 quad came out in Sept 2009 and the Phenom II X4 965 came out in Nov 2009. Both quad cores that could only run 4 threads and when multithreading they were equivalent. So in 16 months Intel had a completely new architecture that was 45% faster and all AMD did was crank the clock speed up by 9%. And when the first gen FX quad thread CPUs came out they were slower performance wise than the Phenom IIs they were replacing.

Of course I’ve always said comparing a 2 module FX to a 4 core Intel is an unfair comparison because the module was suppose to be compete with an Intel core with HT running 2 threads not without HT or running 1 thread. So the proper comparison with the FX-4100 would be the Intel i3-2130 and in multithreaded tests they are.

And there’s the rub. Every two “cores” of an FX is equivalent to one Intel core with HT, that was the design goal and they made it if they could clock the FX fast enough. However marketing choose to ignore the design requirement, assume the end user is an idiot and therefore number of cores became the new “megahertz myth” (more is better).

I agree with you whole heartly, my whole point was that before sandy bridge was released, amd and intel were very competitive, and that Anet did not favor one over the other. As we both said, it was the decisions made in 2009-2011 that really explains the performance gaps we see today.

For the 2500k, the reason i dont like comparing the i5’s is because they really don’t have a valid competitor from the amd FX series. As you said you either compare the i3 2130 with the fx 4100 or the 2600k with the 8100. The lack of performance in single thread is clearly reflected in the pricing.

As for the marketing, I remember when sandy bridge launched in early 2009, and everyone loved the performance but a lot of the community (including me) was reluctant to jump in because AMD said they would release their new chips (that they over hyped with OCTO CORE and MORE GHZ marketing, without any solid details of course) in the fall of that year. After the excruciating wait, we were all so disappointed in the performance, and immediately decided to build an intel system over the holidays.

I have to admit, I was someone who fell for the 8 core aspect of the FX-8350. But now I look at it as FX-4 being dual core, FX-6 as triple and FX-8 as quad. All of the respective units having two core modules.

FX CPUs have true cores
Problem is that in module they have 128Bit x2 FPU unit = 256bit x1 FPU unit.

FX 4300 is 4 core
FX 6300 is 6 core
FX 8320 is 8 core
Where you people get that idea? I mean if CPU has 6 same cores = 6 core CPU!!

Steamroller already improved P/W and better multithreading while excavator will change whole FPU unit.

The problems is that games will use about 2-3 cores … and AMD is weak or is it?
That means you wont see huge difference between FX 4*** and FX 8*** even if Fx 8*** has 100% more power
http://www.tomshardware.com/reviews/guild-wars-2-performance-benchmark,3268-7.html

FX 4100 4.0Ghz is faster than i3 2*** 3.0Ghz in game that use 2 core.

so if you get 15 FPS with AMD you can say that you will get about 50% more with intel (i5) = 22.5 FPS … FPS still sucks…
But this benchmark show you taht between i5 2500 and FX 4100 at 4.0Ghz is not really
huge difference.

I’m pretty sure that if the FX-8350 were a true 4 GHz eight core processor, it wouldn’t have been $240 when it was put on the market. More like $700.

You dont understand
it has 8 cores but it has also 4 modules.
WHat module means? 2 cores are sharing same resource! Actually module slows down cores

AMD went modules just to be on same lvl as intel i7 2700K/3770K.
FX 4300 = i3
FX 6300 = i5
FX 8300 = i7
In 100% usage

You’ve just summarized what Behellagh said.

Piledriver has very fast core. Module architecture slows down single core perfromance and multicore performance.

Jaguar cores are much smaller but still faster then piledrivers clock per clock. Is jaguar core really faster? Or is just better architecture?

Module architecture save AMD a lot of place on die. They already put 16 core piledriver on single die.

Because it’s not really 8 cores the same way the i7 isn’t really 8 cores despite what the Task Manager shows.

If you count cores by number of L1 instruction caches, L2 caches, instruction decoders and/or FPUs then the FX-8xxx only has 4 cores. The same methodology also yields a count of four cores on the desktop Intel i7. It’s only when you count ALUs (arithmetic and logic unit) that you get 8 “cores” on an FX. Each one of the ALUs in an AMD FX is of a much simpler design than the ALU that was in the Phenom II which is why it has to be clocked faster than a Phenom II just to match the Phenom II’s performance.

Intel’s core ALU is even larger than the Phenom II’s. It allows the Intel to execute more instructions in the same number of clocks, which is why it has such great single thread performance. But it’s so large that not all of it being used as efficiently as it could be. That led to the notion of Hyper Threading. By decoding two threads of code, the instruction scheduler can improve the utilization of the ALU so even more instructions can be executed in the same number of clocks.

AMD chose dedicated ALUs for each thread per “module”, Intel uses a much larger ALU that can be optionally shared between two threads. “Optionally shared” is the key phrase here because while an AMD module single thread performance does improve a bit when only running one thread (ignoring clock boost) due to not sharing the rest of a traditional core with the overhead of running the second ALU, a single thread on Intel now has access to all of the ALU, even if it’s not using all of it as frequently as when two threads are running.

Analogy time. Intel’s core is a wagon carrying 2 large barrels of ale being pulled by a single large dolyak while AMD’s module is a similar wagon being pulled by two much smaller dolyaks. Both can travel at the same speed. However if each wagon has only a single barrels but one wagon has a large dolyak while the other has only a smaller dolyak, the wagon with the large dolyak can go much faster.

Okay not a great analogy but at least a layperson can visualize it better than abstract elements of a CPU design.

Nope your wrong. You cannot say that FX 6300 has 3 cores and 3 threads.
I can tell you that in crysis 3 you will notice huge difference! with 6 cores is about 90-100% faster

FX 6300 has 6 cores, but it has 3 modules. So 2 cores are using same resource!
FX is 3 core CPu or 6 core CPU!

Why are all people saying that FX 6300 is 3C/6T CPU – stupid. They should say it has 6 cores but they are slower than intel cores!

Why i think so? if you say that FX 6300 has 3C/6T anyone thinks that it has 3 strong cores and 3 slow cores. But if FX 4300 can macth i3 2C/4C, FX 6300 should match intel with 3C/3T.

Compare FX 4300 vs i3 – people say that both are 2C/4T
Problem?
Games that use 1-2-3 cores will run better on i3
While games that use 4 cores will run better on FX 4300

But okay i know what you want to tell me.
Just one last thing.

What about module performance
FX 8350 scores – Cinebench1.06
single – 1.06
4C/4T(1 core per module) – 4.3 (how? 1.06*4 = 4.24) = 405%
2C/4T (2 cores per module) – 3.52

i5
1 core – 1.55
4 cores – 5.74 (1.55*4=6.2) 370%

FX83** has 8 solid cores. Its just that 2 Cores are in a module that shares the same resources.

Same with the FX63** being 6 Cores, and The FX4*** being 4 Cores.

The issue is that the Buss between the cores in the same module is shared, as with Intel it is not. That is despite that the fact remains, AMD’s single core performance is LESS then Intels single core performance. Has been this way since Sandy Bridge and I dont see it changing anytime in the near future.

So even if you disable the 2nd core per module for the FX Series, Intel is going to still be the winner.

Now multi-threaded applications, the FX comes in very close to the i-series. I see benches that show the FX being both above and below the Intel by 5-8%.

But this game isnt multi threaded, and that is why AMD cannot hold a candle stick here.

Really, thats the end of this discussion.

XFlyingBeeX.2836:

But okay i know what you want to tell me.
Just one last thing.

What about module performance
FX 8350 scores – Cinebench1.06
single – 1.06
4C/4T(1 core per module) – 4.3 (how? 1.06*4 = 4.24) = 405%
2C/4T (2 cores per module) – 3.52

i5
1 core – 1.55
4 cores – 5.74 (1.55*4=6.2) 370%

Just so everyone is on the same page.

http://www.extremetech.com/computing/138394-amds-fx-8350-analyzed-does-piledriver-deliver-where-bulldozer-fell-short/2

FX-8350
1m/1t – 1.06
2m/4t – 3.53
4m/4t – 4.30

i5-3550
1c/1t – 1.55
4c/4t – 5.74

This shows two things. First, the FX cores work better if threads aren’t doubled up in a module. Second, Intel cores when not using HT murders the FX in terms of performance in 1 thread (+46%) or 4 thread (33%). And that’s not even the fastest i5 at the time of the article.

If you want to point out scaling differences between the i5-3550 and the FX-8350, first you have to realize that both processors will boost clock speed with fewer threads but boost is actually unpredictable. Second compare apples to apples. i5-3550 scaling from 1 to 4 cores, 3.7x or 92% of single core’s performance; FX-8350 scaling from 1 to 8 cores (you insist they are real cores), 6.5x or 81% of single core’s performance.

I don’t give a scritt’s stash about Crysis, while streaming or not, or any other game because I can counter example where AMD FX underperforms compared to an Intel. We are talking about this game and in this game the FX’s design is lacking. It’s not because ANet favored one over another, it’s because the FX was designed for loaded servers running balanced loads and not desktops running an unbalanced load.

yep
intel has faster single thread – this game need good single thread
AMD FX has better multithreading = FX 8350 vs i7 3770K

GW2 doesnt need good optimization.. they need low level API! like MANTLE

Seriously?
Mantle is a pipe dream for guild wars. They’re not going to throw away a currently working API just for some new beta API.
And they’re certainly not going to invest in it when they can just make improvements and better support for the CPU.

MS will release low level API. Even directX 11.2 is much better than directx 9

And Direct X 11.2 requires windows 8.1 otherwise you’re not getting it. And Microsoft’s more than likely Direct X 12 isn’t going to be used for guild wars 2 either.

Recoding a game on a new API requires a large amount of money, something that was easy for Blizzard to do with WoW because they had multiple millions of people paying $15 dollars a month for the past, decade?

As far as I can see, DirectX is made based on the abilities of GPU and CPU at the time so DX9 is single thread intensive, and everyone hated vista which was the only way to get DX 10 as Behellagh stated, so they opted for DX9 so everyone can play the game instead of waiting for Windows 7 which righted what Vista wronged.

GW2 is a directX 9

It will NEVER be upgraded to Mantle BEFORE its upgraded to DX 11.2(or even DX10 at this point).

Windows is the prominent OS on the market, and it comes STOCK with DirectX. 98% of the game makers ONLY code their games for openGL(MacOS Support) and/or DirectX. It will be YEARS before Mantle takes off and becomes the replacement for DirectX.

Games that are made for DX9 RARELY get updated to DX10/11/11.2. And games that get their API updated, it will always be linear in the API that they are already built for. Rarely do Manufactures Jump API’s, so to expect such is just silly.

new directX is comming also check this
http://www.neogaf.com/forum/showthread.php?t=686645

All developer should use DirectX11.2 – better for AMD CPUs
http://www.bf4blog.com/battlefield-4-alpha-gpu-and-cpu-benchmarks/
http://www.bf4blog.com/battlefield-4-retail-gpu-cpu-benchmarks/

Direct X 12 might be coming. But it’s not going to change anything. And it’s certainly not going to change OP’s in game performance, which speaking of, if he’s using an AMD FX-8320, he’s gonna want to look at clocking that sucker to 5.5 GHz before he’ll start seeing intel-like performance in this game.

Lets say that intel has 50% better single thread
i5 haswell = FX 4300
3.2Ghz = 4.8Ghz
3.4Ghz = 5.1Ghz
3.6Ghz = 5.4Ghz
3.8Ghz = 5.7Ghz

SO i5 4430 = FX 4300 4.8Ghz
FX 4300 3.75Ghz = i5 haswell 2.5Ghz

That’s a bit more accurate now.

Avelos.6798:

That’s a bit more accurate now.

Yeah but old benchmarks…
usually i dont see my Fx 6300 4.5Ghz below 30-35 FPS (sometimes 24-26FPS) in LA fights.
I am very happy while a year ago i3 was much better option.
http://www.youtube.com/watch?annotation_id=annotation_113478&feature=iv&src_vid=a8_dh4b0jL8&v=UuCL1gv4vR0

FX cpus works much better now in every game

Because Bulldozer got replaced with Piledriver. That’s why.

But most of the gains from Piledriver was due to clockspeed increases. IPC improvement was relatively minor. Steamroller is better at that but we aren’t going to see an AM3+ socket Steamroller CPU.

Steamroller is good but excavator will feature new FPU unit.

Time for AM4 in my opinion.

Yep.
But am3+ can get steamroller while AM4 should get ddr4 and excavator
Also steamroller on FM2+….

Avelos.6798:

Time for AM4 in my opinion.

I imagine that’s what’ll happen. That and a 20nm feature size.

Page 2 and 3 of this article highlights some of the improvements of Steamroller over Piledriver (APUs so no sluggish L3 cache) when both are clocked at 4.2GHz.

http://www.extremetech.com/computing/177099-secrets-of-steamroller-digging-deep-into-amds-next-gen-core/2

Probably off-topic now, but some things to note about WoW’s multicore and DX11 patch…

Did it help in two zergs (50-60 total) going at in Alterac? No – 28fps.
Did it help in a zone raid (40) on Sha of Anger? No – 18fps.
Did it help in keeping 60fps in org/terrace with an LA-amount of people? No (unfortunately no SS here).
Everywhere else the game would be running >60 so I didn’t bother, most settings ultra-ish.

Those were taken with a 3570k @ 4.6GHz / 3x GTX580. Afterburner monitoring info top-right.

Eh they should do DX 11.2
http://www.pcgameshardware.de/World-of-Warcraft-Cataclysm-PC-232999/Tests/WoW-Cataclysm-Benchmarks-DirectX-11-bis-zu-50-Prozent-schneller-als-DirectX-9-822073/

No, they shouldn’t do DX11.2. Only people with Windows 8.1 would get it, duh.
A new API as Delta has shown with the information provided is pretty much a waste of time.

XFlyingBeeX.2836:

Eh they should do DX 11.2
http://www.pcgameshardware.de/World-of-Warcraft-Cataclysm-PC-232999/Tests/WoW-Cataclysm-Benchmarks-DirectX-11-bis-zu-50-Prozent-schneller-als-DirectX-9-822073/

Don’t confuse Dx11.2 for vanilla Dx11. Blizzard isn’t stupid enough to ignore everyone running Windows 7. Just as nobody is going to do a Mantle version but NOT a Dx11.

Just look at the test set up for a moment before you go all zealot, it’s a Windows 7 platform so it’s not even Dx11.1. Jeez man, there are decaffeinated blends that just as good as the real thing.

Problem is that GW2 is niche game…they’re not as huge as LOTRO, for example, who already recoded engine twice, from dx9 to dx10 to dx11. That cost money

Avelos.6798:

No, they shouldn’t do DX11.2. Only people with Windows 8.1 would get it, duh.
A new API as Delta has shown with the information provided is pretty much a waste of time.

So you wont pay for better performance?

Lets say that MS can do low level API like MANTLE on win 9. Specs: Athlon x4 750K and mid GPU.
Moving on win 9 would bring you better performance than moving on i5 4670K.
Lets look at BF4 and compare FX 4300 Mantle (3.4Ghz) vs i5 4670K directX 11 – clearly win for FX 4300.

Next gen consoles now have 8 jaguar cores!

GW2 is great game – thats why they should optimize it.
- a patch with low level API for 10$ – personally i would buy it

The majority of the performance improvements for D3D11 come from multithreaded rendering (NOT programming). That is, instead of having one CreateDevice() for your application like in D3D9, you can create “virtual” devices in each of your threads – letting you draw from any instead of only the primary.

Imagine a simple, theoretical, two-part rendering process. 1-terrain, 2-player. The terrain has an effect on the player. The D3D9 way of doing this is [CPU:terrain][CPU:player][Draw:terrain][Draw:player]. D3D11 let’s you split this to have one thread do [CPU:terrain][Draw:terrain], and another can start [CPU:player][Draw:player] as soon as the first thread is done [CPU:terrain]. If we assume all four pieces are equal weight we just got an fps boost of 33% because player calcs and terrain draw is done in parallel. If the player did not depend on terrain, we can do both tasks completely in parallel (100% boost).

Unfortunately, as the scene becomes more complex, there are more dependencies in place and less things that can be done at the same time, making it a performance boost only in very specific situations.

I have a gut feeling that large fights (model physics, animations, 3D particle effects, projectile physics, and w/e else is there) is not one of them, WoW or GW2.
(edit: that’s not to say there won’t be any improvement from D3D11 in this case, rather it’ll be small enough to not make much of a difference. The only “fix” would be a whole game rewrite, which at this point is likely impossible)

I hate double-posting, but may as well for Mantle; AMD advertised it like D3D11 with some additional low-level control options that expand on multiple render device contexts.

AMD claims it increases BF4 framerate by 45% (https://www.youtube.com/watch?v=Ms16uGxQzSY in the CES teaser) in CPU-limited situations. Even if Arena is able to get the same numbers out of it, 45% over 20fps in wvw = 29fps. People will complain just as hard because it’s still too small a difference to notice.

over 45%…
I did test FX 4300 directX vs Mantle (3.4Ghz).
Same spot, same settings: 32 FPS vs 64FPS

Starswarm benchmark is mostly same as GW2 – D3D11 only 2 cores Mantle up to 8 cores!
FX 4300 3.5Ghz
D3D11 – 21 FPS – 50-60% usage
MANTLE – 61 FPS – 100% usage
i3 4330 3.5Ghz – 60 FPS – 100% usage (mantle)

lets say that you get with FX 4300 3.5Ghz about 10-15 FPS in huge fights.
MANTLE should get you 29-43.5

OC to 5.0Ghz – about 40% FPS boost

personally i want to see same API as Mantle or Mantle running on NVIDIA/AMD systems in GW2

XFlyingBeeX.2836:

Avelos.6798:

No, they shouldn’t do DX11.2. Only people with Windows 8.1 would get it, duh.
A new API as Delta has shown with the information provided is pretty much a waste of time.

So you wont pay for better performance?

I already did by purchasing an i7 4770K. I’m satisfied.

This topic, like a million others, is also starting to turn into another stupid Mantle discussion. Make a thread for it or something because this is not helping OP at all.

XFlyingBeeX.2836:

over 45%…
I did test FX 4300 directX vs Mantle (3.4Ghz).
Same spot, same settings: 32 FPS vs 64FPS

Starswarm benchmark is mostly same as GW2 – D3D11 only 2 cores Mantle up to 8 cores!
FX 4300 3.5Ghz
D3D11 – 21 FPS – 50-60% usage
MANTLE – 61 FPS – 100% usage
i3 4330 3.5Ghz – 60 FPS – 100% usage (mantle)

lets say that you get with FX 4300 3.5Ghz about 10-15 FPS in huge fights.
MANTLE should get you 29-43.5

OC to 5.0Ghz – about 40% FPS boost

personally i want to see same API as Mantle or Mantle running on NVIDIA/AMD systems in GW2

1) Comparing GW2 (an MMO with network, player interaction, animation and other components I can’t think of) to Star Swarm (a demo) is comparing apples to oranges. “If X can do this Y can too” is not how things work. Here’s a quote from the first review, since I assume you didn’t read my example:
Battlefield 4 has multiple CPU tasks going on here, not the least of which is the simulation itself, so in the case of our 4C/4T setups it’s likely we’ve stumbled onto a situation where the game is more strongly CPU-bound by the simulation and other aspects of the game than it is the submission of draw calls.
Replace BF4 with GW2, and simulation with zerg tasks.

2) The many benchmarks suggest a lot less than a 100% gain, and closer to 10-30% for Mantle in BF4.

Avelos.6798:

I already did by purchasing an i7 4770K. I’m satisfied.

+1 for that solution. No more Mantle arguments :p

Avelos.6798:

I already did by purchasing an i7 4770K. I’m satisfied.

I stopped playing, its not bad solution either