Guild Wars 2: Your Graphics Card And CPU Performance Guide

[maddogfargo]

After playing for a few weeks I stumbled upon a glitch that may help you guys.

This happens on my AMD based system but it may be worth trying even if you are in nVidia as I am not certain whether this is a driver issue or a renderer issue.

Symptom:

The game performs slowly, responds to mouse movement slower than normal, lower framerate than normal. GPU is not throttling up to 100% of the rated core speed but stays at 500 of 900 MHz. (* Similar to a recent flash bug)

Workaround:

Change fullscreen game resolution, then change it back.


For whatever reason, this seems to be forcing a reset that makes the rendering engine and video card talk to each other and say 'oh hey we should be at full throttle right now not on cruise control'. Could be a driver issue, could be an optimization they need to make, could be my video card is just lazy and doesn't want to work. But When i do this the usage jumps to 100%, the core clock jumps from 500 to 900 where it belongs, and my frames jump from 35-40 up to 50-60.

Hope this helps you folks.

 

[kit_fox]

As a side note to NVIDIA card users it seems that for better performance you may want to move to the beta drivers. upon install the game prompted to go install the latest beta driver and gave the version number. I compared the beta versus the release driver and found that the beta driver did give a performance bump. This was on a GTX 560 Ti

 

[AntiZig]

As a side note to NVIDIA card users it seems that for better performance you may want to move to the beta drivers. upon install the game prompted to go install the latest beta driver and gave the version number. I compared the beta versus the release driver and found that the beta driver did give a performance bump. This was on a GTX 560 Ti

306.02 driver version to be exact

 

[Guest]

Usual tom's test is quite pro intel.
Look at the scalability per core and per frequency.
Ok frequency test al normal, we know that the amd bulldozer based cpu have some "lack" in ipc.
Test per core...1 funny thing i3 = dual core with ht sandy-e dodecacore and not "esacore + ht". Ok but "frequency = 3ghz" i know that you want put llano in this comparison, but you have see the difference in amd cpu with 3 and 4 ghz. So if the gain is linear, seeing the gain for a a 4100 to 4ghz from 3ghz, and 8000 series at 4ghz give about 75 fps and 6000 series 70 fps. So i think that this test is ridicoulos and make only to put some shit on amd. Try to do this test a 4ghz for all core (excluding llano) And we can see a true scalability core to core of all processor.

 

[youssef 2010]

[citation][nom]rdc85[/nom]I'm wonder how my pII x4 955BE will perform, there none in the chart...Anyone know? at stock speed and at 3.8 O.C....[/citation]


I think you're better off upgrading to Ivy bridge..... I never realized how much of a bottleneck it was until I switched to i7-3770K. There're a lot of less expensive Ivy Bridge models out there that perform similarly.Besides, the general belief is that you need to upgrade your PC every three years to preserve the level of detail that you're playing at.Also, you're forgetting that Llano uses the same architecture as the Phenom II minus the L3 cache.

If AMD keeps using this architecture until its time for my next upgrade, I'll have no choice but to go with Intel again. I really respect AMD and what they're trying to do, but their goals and my performance requirements are just too far apart.

 

[mannam]

So are we going to get that updated article since beta build is apparently way different than launch build? They said it was coming middle of September.

 

[cleeve]

So are we going to get that updated article since beta build is apparently way different than launch build? They said it was coming middle of September.

it's still on the to-do list. Some other stuff to get out of the way first, but expect it next week.

 

[mannam]

it's still on the to-do list. Some other stuff to get out of the way first, but expect it next week.

Thanks for the quick reply! Can't wait for the article.

Edit: I would like to add I have an unlocked Phenom II B50 x4 at 3.1 Ghz, Guild Wars people and some Nvidia reps I saw on another forum are saying the game is heavily CPU dependent, so I am really curious of benchmarks that would take a quad core Phenom II and overclock it to see how much of a difference it can really make.

 

[jsl]

I have to travel for work at the moment and just purchased an ASUS G75vw and it plays GW2 like a dream.. 60+fps.. gw2 in hotel room FTW!! haha

 

[cleeve]

[citation][nom]mannam[/nom]Thanks for the quick reply! Can't wait for the article. Edit: I would like to add I have an unlocked Phenom II B50 x4 at 3.1 Ghz, Guild Wars people and some Nvidia reps I saw on another forum are saying the game is heavily CPU dependent, so I am really curious of benchmarks that would take a quad core Phenom II and overclock it to see how much of a difference it can really make.[/citation]

The Phenom II will perform right on par with the Llano X4. Same architecture. Yes, the Phenom II has a lot more cache, but in my experience its rare that the cache makes a difference.

spoiler: I've done some re-testing and the llano results haven't changed.

 

[mannam]

The Phenom II will perform right on par with the Llano X4. Same architecture. Yes, the Phenom II has a lot more cache, but in my experience its rare that the cache makes a difference.

spoiler: I've done some re-testing and the llano results haven't changed.

Again, thanks for the quick reply! Shame about the llano results...

How about when overclocked? I've heard overclocking can make a big difference in GW2 since it is CPU intensive, but no one has really shown any real benchmarks. For example, if I overclocked my Phenom II x4 from 3.1 to 3.8 or higher or in your case, the Llano x4 overclocked will we see a big difference?

 

[cleeve]

Again, thanks for the quick reply! Shame about the llano results...

How about when overclocked? I've heard overclocking can make a big difference in GW2 since it is CPU intensive, but no one has really shown any real benchmarks. For example, if I overclocked my Phenom II x4 from 3.1 to 3.8 or higher or in your case, the Llano x4 overclocked will we see a big difference?

Absolutely you should see a big difference!

We did benchmark the impact of clock rates! Note the difference between the 2.0 GHz llano X4 and the 3.0 GHz llano X4 in the "CPU Clock Scaling" chart.
There's almost a linear improvement in Guild Wars 2 frame rates when overclocking AMD CPUs.

 

[mannam]

Absolutely you should see a big difference!

We did benchmark the impact of clock rates! Note the difference between the 2.0 GHz llano X4 and the 3.0 GHz llano X4 in the "CPU Clock Scaling" chart.
There's almost a linear improvement in Guild Wars 2 frame rates when overclocking AMD CPUs.

OH YEAH, I completely forgot about that! I forgot that you said that Llano performance Beta Build vs. Release build is the same, so I kinda forgot the chart.

Even though the performance of the Llano (and my Phenom II x4) is so poor compared to the Sandy Bridges, at least I have confirmation that overclocking will help in my situation. Thanks for the info and can't wait for the redo article!

 

[cleeve]

[citation][nom]mannam[/nom]OH YEAH, I completely forgot about that! I forgot that you said that Llano performance Beta Build vs. Release build is the same, so I kinda forgot the chart. Even though the performance of the Llano (and my Phenom II x4) is so poor compared to the Sandy Bridges, at least I have confirmation that overclocking will help in my situation. Thanks for the info and can't wait for the redo article![/citation]

Hey, I threw a Phenom II X4 just to see what would happen, and it looks like Guild Wars 2 is one of those rare titles that likes the extra cache.

The Phenom II X3 scores the same as the Sandy-Bridge Pentium, and the Phenom II X4 is between the Pentium and the Core i3.

 

[mannam]

Hey, I threw a Phenom II X4 just to see what would happen, and it looks like Guild Wars 2 is one of those rare titles that likes the extra cache.

The Phenom II X3 scores the same as the Sandy-Bridge Pentium, and the Phenom II X4 is between the Pentium and the Core i3.

Hey, a million thanks for doing that! Know you guys at TH have a busy schedule, so it's much appreciated! Interesting that the game benefits from the extra cache. Now I just have to crack my knuckles and get to overclocking, so I can squeeze out some more frames.

 

[Phyrexiancure]

[citation][nom]Cleeve[/nom]The Phenom II will perform right on par with the Llano X4. Same architecture. Yes, the Phenom II has a lot more cache, but in my experience its rare that the cache makes a difference.spoiler: I've done some re-testing and the llano results haven't changed.[/citation]

Just to correct you L3 cache does matter. Especially if you overclock it. In games you can get around an 8% to 15% boost in performance by overclocking l3 cache/cpu-nb let alone having it.

 

[Phyrexiancure]

[citation][nom]phyrexiancure[/nom]Just to correct you L3 cache does matter. Especially if you overclock it. In games you can get around an 8% to 15% boost in performance by overclocking l3 cache/cpu-nb let alone having it.[/citation]

I'm referring to the Phenom II architecture. I don't believe this is the case with the FX cpu's.

 

[cleeve]

Just to correct you L3 cache does matter.

After looking at some numbers, you're right, it does make more of a difference in games than I remembered.
Looks like about 10% on average at the same clock between Athlon II and Phenom II at the same clocks.

In games you can get around an 8% to 15% boost in performance by overclocking l3 cache/cpu-nb let alone having it.

I haven't seen much evidence suggesting that a northbridge overclock scales better with more L3 cache though. The northbridge speeds up RAM access, that works as well on Athlon IIs unless you have evidence to the contrary.

 

[jimbaladin]

Cleeve your "FX 4000" at 4 GHz is actually slower than AMD's fastest clocked quad core, the FX 4170 which is 4.2 GHz with 4.3 GHz boost.

Considering that the "FX 4000" at 4 GHz is 3rd fastest of all, you'd expect the 4.2/4.3 GHz version to be faster still and a bit closer to the i5's and maybe even worth it from a price/performance perspective.

 

[jimbaladin]

Ok Cleeve I just realised what you're doing here with certain clock speeds - however if you're going to do this can you also do at least one benchmark of the chips at their stock speeds so we can get an accurate picture of where each cpu stands out of the box?

 

[cleeve]

[citation][nom]jimbaladin[/nom]Ok Cleeve I just realised what you're doing here with certain clock speeds - however if you're going to do this can you also do at least one benchmark of the chips at their stock speeds so we can get an accurate picture of where each cpu stands out of the box?[/citation]

There is only a single FX chip with a base clock faster than 4 GHz, and that's the FX-4170. The FX-4100 is slower, and the 4 GHz I benched is between the two. I think it gives an excellent impression of performance.

Here's another spoiler: FX-6000 and 8000 performance has gone *DOWN* a bit in the new release benchmarks compared to the beta results...

 

[Guest]

Any chance of getting in the new Intel G2120 for testing?

My stock Q6600 is really struggling in GW2 and from your Pentium testing in the Norn starter area it should be a good upgrade for me. I get about 20 fps there. :/

 

[cleeve]

[citation][nom]ChrisSC[/nom]Any chance of getting in the new Intel G2120 for testing?My stock Q6600 is really struggling in GW2 and from your Pentium testing in the Norn starter area it should be a good upgrade for me. I get about 20 fps there.[/citation]

Probably not going to bother in the near future since it's the same clock as the Sandy-Bridge Pentium G870. Ivy Bridge isn't a big bump from Sandy Bridge except when it comes to the graphics chipset...

 

[jimbaladin]

[citation][nom]Cleeve[/nom]There is only a single FX chip with a base clock faster than 4 GHz, and that's the FX-4170. The FX-4100 is slower, and the 4 GHz I benched is between the two. I think it gives an excellent impression of performance.Here's another spoiler: FX-6000 and 8000 performance has gone *DOWN* a bit in the new release benchmarks compared to the beta results...[/citation]

I guess there must be a bit of variation in the benchmarks, which is of course difficult to avoid in an MMO.

 

[cleeve]

[citation][nom]jimbaladin[/nom]I guess there must be a bit of variation in the benchmarks, which is of course difficult to avoid in an MMO.[/citation]

It's been quite consistent for an MMO, actually... very repeatable on the bench method I've used. I can log in at a later time and recreate what I capture with a few percent variance.

I've talked to the devs and they recognize FX performance is a problem, they're working to fix it. But no news other than that I'm afraid.