Mass Effect 2: Galactic battle Geforce versus Radeon

[tweakboy]

Not a fair comparison.

ATI is next gen. nvidia doesnt have its next gen yet.

Once Fermi is out then you can go Fermi vs ATI 5890 ... thanks

 

[Occ]

Wait, what.. why would you play at 1366x768 and use AA to emulate higher resolution instead of just playing at a higher resolution?

Not everyone has a monitor with a native resolution of 1900x1200. AA is very useful for making games look smoother at low resolution.

Also, how do you show fps in ME2? I'm running at max details @ 1280x1024 w/ 4xSSAA and 8xAF and would be curious to see what kind of fps I'm getting. It feels ~60fps smooth 95% of the time but would be interesting to double check.

 

[Daedalus685]

Not a fair comparison.

ATI is next gen. nvidia doesnt have its next gen yet.

Once Fermi is out then you can go Fermi vs ATI 5890 ... thanks

You can't be serious...

Where exactly is the universal fair comparison rules book? The game just came out, is one of the best in a long time (if you are into that type of game), and we can't see how well it performs on the best hardware out today because there is new stuff coming?

Last I checked there is always new stuff coming…. Since when is there an obvious generational comparison between any company? Isn’t it far more intelligent to worry about what is out in the now…. We won’t have to listen to this same tired argument from ATI fans in 5 months, and every 6 months after that flipping back and forth, if the releases remain out of phase will we?

Perhaps I missed when I was able to win the Nobel prize for something that was already achieved based on the premise that "Well sure, but it wasn't fair that he won as I had not yet finished."

Companies release new products all the time, am I to believe that if ATI started releasing every 8 months that Nvidia would still win since they are on a lower arbitrary generation now and forever?

 

[Phil1977]

Also, how do you show fps in ME2? I'm running at max details @ 1280x1024 w/ 4xSSAA and 8xAA and would be curious to see what kind of fps I'm getting. It feels ~60fps smooth 95% of the time but would be interesting to double check.

You can use FRAPS. It's a free download. You just start the program before you start the game...

Please note that ME1 had a frame cap of 62fps. ME2 might be similar. You can remove this cap by editing a ini file if you like. You can refer to tweakguides ME article. It has all the details

 

[Daedalus685]

Not everyone has a monitor with a native resolution of 1900x1200. AA is very useful for making games look smoother at low resolution.

Also, how do you show fps in ME2? I'm running at max details @ 1280x1024 w/ 4xSSAA and 8xAA and would be curious to see what kind of fps I'm getting. It feels ~60fps smooth 95% of the time but would be interesting to double check.

How exactly are you running 4xSSAA and 8xAA at the same time? Does nhance allow you to force SSAA on top of forced MSAA?

Or before I dive off the deep end with curiosity.. did you mean 8xAF?

 

[AzN]

How exactly are you running 4xSSAA and 8xAA at the same time? Does nhance allow you to force SSAA on top of forced MSAA?

Or before I dive off the deep end with curiosity.. did you mean 8xAF?

You can do combination of SSAA and MSAA with nhancer.

 

[Occ]

How exactly are you running 4xSSAA and 8xAA at the same time? Does nhance allow you to force SSAA on top of forced MSAA?

Or before I dive off the deep end with curiosity.. did you mean 8xAF?

doh! curse these deliciously greasy lunch burritos! D: Yes, I meant AF. Also, I notice in the CCC, there appear to be options for altering the type of AA algorithm, like you can use box, triangle, and 2 other things that I can't recall at the moment. Not sure what the difference is. Anyone have any ideas? I'll have to check when I get home.

 

[SirPauly]

How exactly are you running 4xSSAA and 8xAA at the same time? Does nhance allow you to force SSAA on top of forced MSAA?

Or before I dive off the deep end with curiosity.. did you mean 8xAF?

They're called hybrid or mixed modes like x4s, x8s, x8sQ ( x8 MSAA + x4 SSAA), x16s.

 

[AzN]

nv

 

[Phil1977]

You obviously have comprehension problems. "These benches" as in Mass Effect 2 benches would make bigger difference with GPU. Guess what? It does! Or did you fail to see linear scaling with GPU?

You told ME that upgrading the GPU on MY system would make an impact because of benchmarks you saw on ANOTHER system. Epic fail.

Athlon x1 2000mhz 20% overclock =2400mhz 1 core
Athlon x2 2000mhz 20% overclock =2400mhz 2 cores

I can't believe I have to spell this out for you. I thought you people where good at maths? Clearly not or you have no education whatsoever.

Firstly: performance = workload = work getting done.
Secondly: assumption: multi threaded application that scales perfectly

Single Core: Gets 100% workload done
Dual Core: Gets 200% workload done
Quad Core: Gets 400% workload done

Still with me? Good keep reading.

Now we are applying a 20% overclock to every CPU:

Single Core: Gets 120% workload done
Dual Core: Gets 240% workload done
Quad Core Gets 480% workload done

See now? 20, 40 and 80 %. WOW. I hope you are having an AHHH moment.

The more cores you have, the more absolute workload gain you will see by overclocking. Overclocking the single core is pretty fruitless. But overclocking the Quad Core gains you almost a the performance equivalent of a full fifth core!

Real option as in 1998 resolution. Talk to me when you can use SSAA @ 1080P.

Why should I. I specifically chose this LCD for gaming. I mentioned this several times. Get over it and move on.

Your amazing IQ. You made a funny!

IQ = Image quality. Might have been funny in kindergarden, but now it just makes you look like a child.

That's before you told me you play 1024x768. Come back to 21st century. We stopped playing in 1024x768 more than 10 years ago.

I play at 1366 x 768. And as above it's a choice I made.

They get fed up because I give them something to think about. You in other hand give 1 big blank. Get an education you.

It's "They got fed up" and "on the other hand". Please learn some basic grammar.

something to think about

Something to think about? Regarding Dot pitch? You basically thought a higher Dot pitch will give you more detail. And everybody was explaining you that this isn't the case. And you went on and on and on like a troll that they must be wrong.

Well what about a 3" LCD with a super high dot pitch? You seriously think you will see more detail compared to a 22" with a lower Dot pitch?

And after you "got it" and realized you where clearly wrong you didn't say thank you to the people helping you out, you didn't apologize for the way you attacked people when clearly you where wrong. Nothing. You just left like a little coward.

Why are you going on about education?

Maybe you should further yourself. Or start by actually listening and thinking and not jumping to conclusions right away.

Rather than starting flame wars, everyone, please just use the report post button & ignore the personal insults & attacks rather than responding to them. (Also, don't post crap like "haha, I reported your post." That just antagonizes people.) Thanks, Admin DrPizza

 

[waffleironhead]

Not a fair comparison.

ATI is next gen. nvidia doesnt have its next gen yet.

Once Fermi is out then you can go Fermi vs ATI 5890 ... thanks

So we are just supposed to wait around for fermi to show up before we can see benchmarks for this game?

 

[evolucion8]

You can use FRAPS. It's a free download. You just start the program before you start the game...

Please note that ME1 had a frame cap of 62fps. ME2 might be similar. You can remove this cap by editing a ini file if you like. You can refer to tweakguides ME article. It has all the details

For some reason editing the config file would make the game unusable, it wouldn't start, even setting the file at default wouldn't work, odd.

Now in a different story, overclocking a Dual Core to 20% will not yield you 40% more performance, why? Because if you overclock both CPU's, each thread will go to its respective core and will get a 20% boost in performance in each respective thread, it isn't like if overclocking the CPU to 20%, would dupplicate to 40% because the thread would split in half and would go to both cores and would duplicate the throughput.

Remember that Dual Cores doesn't work in serial, they work parallel, 20% in each core means 20% in overall boost in performance. Doesn't make sense the 40% result because the threads doesn't work in serial, for that to work, it would mean that CPU 1 and CPU 2 would work together in a single thread which is simply impossible, each CPU gets its thread, period. Sorry for the Redundancy's Deparment of Redundancy.

 

[Phil1977]

For some reason editing the config file would make the game unusable, it wouldn't start, even setting the file at default wouldn't work, odd.

Yea thats a bit odd...


Regarding the over clocking. The main point I was trying to make is that when you are dealing with threaded software, the more cores you have, the more you will get out of over clocking... Or put another way over clocking a single core is pretty fruitless if you are running threaded software...

Check the reply to AnZ. I worked out the figures there...

E.g. if you over clock a single core by 25%, you get a performance gain of a quarter of the equivalent of another core. But if you over clock a dual core, you get a performance gain of half the equivalent of another core.

And if you over clock a quad by 25% you get the performance gain of the equivalent to adding a full core...

You are correct, the relative performance gain remains 25%. But because the dual and quad core are more powerful to begin with, the same relative 25% gain results in a much larger absolute gain (in terms of how much extra workload you can process) ...

That's basically what I was trying to say...

 

[Occ]

I heard the config file was checked via hashing before the game executed, and you have to do some weird thing to keep it the same while making changes. There are threads over at gamefaqs and the steam forums.

 

[NoQuarter]

Not everyone has a monitor with a native resolution of 1900x1200. AA is very useful for making games look smoother at low resolution.

Also, how do you show fps in ME2? I'm running at max details @ 1280x1024 w/ 4xSSAA and 8xAF and would be curious to see what kind of fps I'm getting. It feels ~60fps smooth 95% of the time but would be interesting to double check.

Yea I know.. I only just got out of 1280x1024 myself, but he was saying he uses a 1366x768 monitor because it's low resolution (to improve performance and use 4xAA) when AA is obviously a workaround for low resolution instead of a benefit of. If it's better to play with high AA at low res then people would be on 640x480 screens with 16xAA

 

[Lonbjerg]

That is hardly quad optimized. I have single core game that runs between both cores @ 60% and 40%. That doesn't mean it's dual core optimized.

Who gave you the idea that a multicore game must put 100% load on all cores? :\

This is an ARMA2 bench, one of the most multicore optimized games i know of:
http://www.pcgameshardware.com/aid,687620/ArmA-2-tested-Benchmarks-with-18-CPUs/Practice/

Unless you want to move the "goalposts" and redefine what multi-threaded means?

 

[AzN]

Who gave you the idea that a multicore game must put 100% load on all cores? :\

This is an ARMA2 bench, one of the most multicore optimized games i know of:
http://www.pcgameshardware.com/aid,687620/ArmA-2-tested-Benchmarks-with-18-CPUs/Practice/

Unless you want to move the "goalposts" and redefine what multi-threaded means?

Arma 2 isn't truly multi-threaded. Actually there isn't a true multithreaded game in PC. If it was it would double performance in cpu scaling tests but it does not. You get the 20% better frame rates from quads @ same clock as dual core instead of 100%.

 

[evolucion8]

Interesting benchmark, usually very well multi threaded games will put more load in cores than what Arma 2 is doing there which is only scarcing threads across all cores, I think that the developers aren't doing a good job of maximizing the CPU usage, but I prefer to see developers to put most of the work in the GPU instead.

 

[AzN]

I think the only game that almost being a truly multi-threaded game is GTA4. Majority PC games work in a way where threads are shared between the core and not utilizing full potential of quad cores.

 

[Voo]

Firstly: performance = workload = work getting done.
Secondly: assumption: multi threaded application that scales perfectly

Single Core: Gets 100% workload done
Dual Core: Gets 200% workload done
Quad Core: Gets 400% workload done

Still with me? Good keep reading.

Now we are applying a 20% overclock to every CPU:

Single Core: Gets 120% workload done
Dual Core: Gets 240% workload done
Quad Core Gets 480% workload done

See now? 20, 40 and 80 %. WOW. I hope you are having an AHHH moment.

Wow so that's a 20% increase, a 20% increase and a 20% increase from the starting point! I think we all see your point now.. you know that percentages are used for relative comparisons and not absolute ones? Let's just change the workload to 1, 2 and 4, wow now we get only 0.4 more workload done with a dualcore instead of 40.. Oh but still 20% more.
Mixing absolute und relative numbers doesn't get you anywhere, one core more for a single core is 100% more cores, but for a dualcore that's only 50%.


@AzN: If your definition of "truly mulithreaded" is "100% parallel code".. well that's impossible per definition (the best you can do is ld(n) stepps to initialize n processes/threads..), so it's all a question about how the ratio of sequential vs. parallel code is. Getting anywhere near 95% is rather impossible for games I'd think - there are things that are inherently difficult to parallelize. But that doesn't mean a game isn't multithreaded..

 

[Lonbjerg]

Arma 2 isn't truly multi-threaded. Actually there isn't a true multithreaded game in PC. If it was it would double performance in cpu scaling tests but it does not. You get the 20% better frame rates from quads @ same clock as dual core instead of 100%.

As I thought, you are trying to move the goalposts.
Again, even if a game is running 16 threads, that is NO assurance it will max out 16 cores.

The Virtual Reality Engine 3 is multicore:
http://www.arma2.com/supply/presskit/download/20-realvirtualityengine-gc2008.html?lang=en

Dosn't matter if "your" definition of "multicore" means it must max out all availble cores, you stance is flawed.
And to think games scale linear with the number of CPU cores is wrong, by that definition SLI/Crossfire should also scale linear.

ARMA2 is multicore, like it or not...it's your definiton of "multicore" that is incorrect...as I suspected.

 

[AzN]

As I thought, you are trying to move the goalposts.
Again, even if a game is running 16 threads, that is NO assurance it will max out 16 cores.

The Virtual Reality Engine 3 is multicore:
http://www.arma2.com/supply/presskit/download/20-realvirtualityengine-gc2008.html?lang=en

Dosn't matter if "your" definition of "multicore" means it must max out all availble cores, you stance is flawed.
And to think games scale linear with the number of CPU cores is wrong, by that definition SLI/Crossfire should also scale linear.

ARMA2 is multicore, like it or not...it's your definiton of "multicore" that is incorrect...as I suspected.

By your definition of spreading the cores between the CPU automatically becomes multithreaded which is false only to have less than 100% pegged CPU.

In case of Arma you are getting 20% better frame rates between dual and quad. This performance improvement could come from more Cache of the Quad processors over dual core and is not necessarily quad optimized.

Developers claim lot of things but it doesn't mean it's necessarily true. According to Crytek Crysis is quad optimized. Bethesda claimed oblivion was dual core optimized and so on.

SLI and Crossfire does scale linearly 2x fold when it has proper optimization as do multicore optimized games. Go look at some benches.

 

[Lonbjerg]

By your definition of spreading the cores between the CPU automatically becomes multithreaded which is false only to have less than 100% pegged CPU.

In case of Arma you are getting 20% better frame rates between dual and quad. This performance improvement could come from more Cache of the Quad processors over dual core and is not necessarily quad optimized.

Developers claim lot of things but it doesn't mean it's necessarily true. According to Crytek Crysis is quad optimized. Bethesda claimed oblivion was dual core optimized and so on.

SLI and Crossfire does scale linearly 2x fold when it has proper optimization as do multicore optimized games. Go look at some benches.

You are still clining to flawed notion of multicore, please stop.

I am not the only one who has pointed out that you are wrong.
And please whow me these 100% scaling multi-GPU games, I can't see them here:
http://www.anandtech.com/video/showdoc.aspx?i=3643&p=17
http://www.anandtech.com/video/showdoc.aspx?i=3650&p=4
http://www.anandtech.com/video/showdoc.aspx?i=3658&p=5

And even when we look at FarCry 2:
http://www.anandtech.com/showdoc.aspx?i=3505&p=7

FarCry 2 ships with the most impressive benchmark tool we’ve ever seen in a PC game. Part of this is due to the fact that Ubisoft actually tapped a number of hardware sites (AnandTech included) from around the world to aid in the planning for the benchmark.

You won't find you "magical" 100% scaling present.
You overlook the act that threads need to wait for other treads to complete in a game engine.

Only things that scale ~100% are encoding/decoding and other In-Order instructions...game engines dosn't.

But I would love for you to post data in you favour, your word is not enough for me, sorry?

 

[AzN]

You are still clining to flawed notion of multicore, please stop.

I am not the only one who has pointed out that you are wrong.
And please whow me these 100% scaling multi-GPU games, I can't see them here:
http://www.anandtech.com/video/showdoc.aspx?i=3643&p=17
http://www.anandtech.com/video/showdoc.aspx?i=3650&p=4
http://www.anandtech.com/video/showdoc.aspx?i=3658&p=5

Again proper optimization is not at play or it could be that CPU wasn't fast enough to compute both cards. With proper optimization it would closely be doubled in performance.

http://www.guru3d.com/article/radeon-hd-5870-crossfirex-test-review/12

http://www.guru3d.com/article/radeon-hd-5870-crossfirex-test-review/7

http://www.guru3d.com/article/radeon-hd-5870-crossfirex-test-review/10

Notice you are only sticking with Crysis engine to come up with your analysis. Crysis engine had a poor yield with multiple cards from the very start. It was at times 10% faster at game release with dual cards but I see the optimization is creeping up to more than 50%.

And even when we look at FarCry 2:
http://www.anandtech.com/showdoc.aspx?i=3505&p=7

What about it? This doesn't tell much.

You won't find you "magical" 100% scaling present.
You overlook the act that threads need to wait for other treads to complete in a game engine.

Only things that scale ~100% are encoding/decoding and other In-Order instructions...game engines dosn't.

But I would love for you to post data in you favour, your word is not enough for me, sorry?

With proper optimization and fast enough CPU you can get nearly 100% scaling with SLI or crossfire.

The fact that people claim multi-thread optimization only to realized by benchmarking artists that quads have more cache that works in favor of dual core.

This is what happened when real dual core optimized games started to appear.

Go to CPU and memory portion.

http://www.gamespot.com/features/6183967/index.html

Notice from a single core to dual core has nearly doubled in performance within the same architecture. That's no 20% like Arma.