GTX 480 benchmark far below average, any help would be appreciated

[lotusvibe]

System
MSI P55-GD80
i7 860 at stock 2.8ghz
4GB DDR3 1333
PNY GTX 480

I am new to this forum and to computer gaming in general so please let me know if you need more info to help me.

according to this website http://www.nordichardware.com/compo...force-gtx-480-the-wait-is-over-.html?start=11

I should be getting well above 100fps when benchmarking resident evil 5. However with my GTX 480 I am getting an average of about 72fps
settings are at 1920x1080 highest quality with 8x AA

Before i bought my gtx 480 I had a HD5870 which was benchmarking around 108fps which is matching up with the above website's benchmarks, i got the gtx 480 because It was suppose to be a little better and I wanted to try an NVidia card and see how i liked it. So far well i don't like it cause i cannot get the re5 bencmark about 73fps. I don't have any idea what's wrong I haven't changed any bios settings and everything appears to be running fine, why such low benchmarks?

 

[BFG10K]

Have you patched RE5 to the latest version?

How's the performance of your GTX480 in your other games relative to your 5870?

 

[ModestGamer]

I would say that benchmarks often favor the best paying advertiser on average. Might they have a liberal return policy ?

 

[Janooo]

What's your PSU?

 

[RussianSensation]

Welcome to AnandTech!

My gut feeling is telling me the answer is: Resident Evil 5 benefits a lot from multi-threaded CPU processors and is also VERY cpu frequency dependent.

- Core i9 980X is 3.33ghz vs. yours at 2.8ghz.
- 12mb of L3 cache vs. 8mb with yours
- 12 threaded cpu vs. 8 with yours

Overclock your CPU and you will see your framerates increase. With Core i7 at 3.9ghz on a 4890, I get about 78 fps. This is telling me that your graphics card is majorly held back by your CPU in Resident Evil 5 Fixed benchmark.

This games scales well with CPU speed: http://www.pcgameshardware.com/aid,...nom-strong-Update-Lynnfield-results/Practice/

GTX480 is also a lot more picky about CPU speed than 5870 series. Look here: http://www.xbitlabs.com/articles/cpu/display/cpus-and-games-2010_7.html#sect2

Make sure your CPU's Turbo mode is enabled in the BIOS and that you don't have the more complex anti aliasing mode enabled.

 

[lotusvibe]

my power supply is an corsair hx850 or 850hx can't if the hx comes before or after. I just tried reinstalling my drivers and now my benchmark was at about 108 which seems acceptable. (reinstalling the drivers set my nvidia control panel 3d settings back to default) I wish i knew what benchmarkers used for their nvidia control panel settings. Because i just made some changes to reflect an increase in quality and now my benchmark when back down to 72. So it must be a setting or two in the nvidia control panel that is causing such a drop in performance guess i'll just have to tinker with them and see what it is. And i'm not sure about other games compared to my 5870 I'd rather not mess with putting it in the case again but it may come to that. Do any of you guys have a good free 3d benchmarking program that i should get?

 

[MJinZ]

System
MSI P55-GD80
i7 860 at stock 2.8ghz
4GB DDR3 1333
PNY GTX 480

I am new to this forum and to computer gaming in general so please let me know if you need more info to help me.

according to this website http://www.nordichardware.com/compo...force-gtx-480-the-wait-is-over-.html?start=11

I should be getting well above 100fps when benchmarking resident evil 5. However with my GTX 480 I am getting an average of about 72fps
settings are at 1920x1080 highest quality with 8x AA

Before i bought my gtx 480 I had a HD5870 which was benchmarking around 108fps which is matching up with the above website's benchmarks, i got the gtx 480 because It was suppose to be a little better and I wanted to try an NVidia card and see how i liked it. So far well i don't like it cause i cannot get the re5 bencmark about 73fps. I don't have any idea what's wrong I haven't changed any bios settings and everything appears to be running fine, why such low benchmarks?

Your 5870 was at 108fps? That sounds unlikely to me considering his system is much faster than yours.

 

[lotusvibe]

It's possible that my ati benchmark was lower than 108 but it was definately definately above 100 i remember cause i was so impressed, so when i was getting 72 with my gtx 480 I was rather alarmed

 

[ModestGamer]

It's possible that my ati benchmark was lower than 108 but it was definately definately above 100 i remember cause i was so impressed, so when i was getting 72 with my gtx 480 I was rather alarmed

my hd5770 at maxxed settings get 60-70 fps in dirt 2 which every bench mark says should be 30-40.YMMV if your video card has onboard audio "it has a HDMI output so most likely" find it in the device manager and siable the cards onboard audio. I found I got a nice performance bump by doing so. I also found that certain cards when tested in certain configurations and setting do things differently. So try experimenting with hardware and software settings.

 

[apoppin]

Welcome to AnandTech!

My gut feeling is telling me the answer is: Resident Evil 5 benefits a lot from multi-threaded CPU processors and is also VERY cpu frequency dependent.

- Core i9 980X is 3.33ghz vs. yours at 2.8ghz.
- 12mb of L3 cache vs. 8mb with yours
- 12 threaded cpu vs. 8 with yours

Overclock your CPU and you will see your framerates increase. With Core i7 at 3.9ghz on a 4890, I get about 78 fps. This is telling me that your graphics card is majorly held back by your CPU in Resident Evil 5 Fixed benchmark.

This games scales well with CPU speed: http://www.pcgameshardware.com/aid,...nom-strong-Update-Lynnfield-results/Practice/

GTX480 is also a lot more picky about CPU speed than 5870 series. Look here: http://www.xbitlabs.com/articles/cpu/display/cpus-and-games-2010_7.html#sect2

Make sure your CPU's Turbo mode is enabled in the BIOS and that you don't have the more complex anti aliasing mode enabled.

How interesting, i am benching RE5 at this very moment (i checked in on PMs )

GTX 480 at stock, 8xAA, maxed details

With i7 920 at 3.8 GHz at 1920x1200 = 111.4 FPS

With Phenom 955 X4 at 3.8 GHz = 84.7
With Phenom 955 X4 at 3.2 GHz = 77.6
With Phenom 955 X4 at 2.6 GHz = 67.8

Resident Evil 5 definitely likes Core i7 over Phenom II architecture (i haven't tested with HT on vs. off yet) and it *does* scale with CPU core speed

As a reference with HD 5870 at 19x12, same settings as above gets 101 FPS average with Core i7 920 at 3.8 GHz

 

[MJinZ]

How interesting, i am benching RE5 at this very moment (i checked in on PMs )

GTX 480 at stock, 8xAA, maxed details

With i7 920 at 3.8 GHz at 1920x1200 = 111.4 FPS

With Phenom 955 X4 at 3.8 GHz = 84.7
With Phenom 955 X4 at 3.2 GHz = 77.6
With Phenom 955 X4 at 2.6 GHz = 67.8

Resident Evil 5 definitely likes Core i7 over Phenom II architecture (i haven't tested with HT on vs. off yet) and it *does* scale with CPU core speed

As a reference with HD 5870 at 19x12, same settings as above gets 101 FPS average with Core i7 920 at 3.8 GHz

Exactly, makes me highly suspect his ability to get 100FPS with stock CPU.

Almost all games these days are heavily heavily bottlenecked by CPU.

 

[apoppin]

Exactly, makes me highly suspect his ability to get 100FPS with stock CPU.

Almost all games these days are heavily heavily bottlenecked by CPU.

Surprisingly, not really. And if you run your same tests at 2560x1600 with maxed out details and 8xAA you get a very different picture. Phenom II is only a couple of FPS slower than the Core i7 when both are clocked at 3.8 GHz and there is also very little penalty for a slower CPU at high resolution:

GTX 480/i7-920@3.8 GHz = 73.2

GTX 480/Ph II 955 X4@3.8 GHz = 71.8
GTX 480/Ph II 955 X4@3.2 GHz = 70.7
GTX 480/Ph II 955 X4@2.6 GHz = 66.8

You just need to shift the burden to the GPU and the picture drastically changes. At 1920x1200 or 1920x1080, the CPU is working much harder relative to the GPU in RE5 than at 2560x1600.

You need to take each game on case-by-case. AtM, *most* games are FINE on a Phenom II X2 or C2D at 2.6 GHz if your shift your burden to the GPU at a high resolution. Some games of course - newer games - are much better multi-threaded and will benefit both from more cores and higher clockspeeds.

 

[lotusvibe]

Surprisingly, not really. And if you run your same tests at 2560x1600 with maxed out details and 8xAA you get a very different picture. Phenom II is only a couple of FPS slower than the Core i7 when both are clocked at 3.8 GHz and there is also very little penalty for a slower CPU at high resolution:

GTX 480/i7-920@3.8 GHz = 73.2

GTX 480/Ph II 955 X4@2.6 GHz = 66.8
GTX 480/Ph II 955 X4@3.2 GHz = 70.7
GTX 480/Ph II 955 X4@3.8 GHz = 71.8

You just need to shift the burden to the GPU and the picture drastically changes. At 1920x1200 or 1920x1080, the CPU is working much harder relative to the GPU in RE5 than at 2560x1600.

You need to take each game on case-by-case. AtM, *most* games are FINE on a Phenom II X2 or C2D at 2.6 GHz if your shift your burden to the GPU at a high resolution. Some games of course - newer games - are much better multi-threaded and will benefit both from more cores and higher clockspeeds.

What do you mean by shift burden to your gpu? I know some titles have physx options which i think is along the lines of what your talking about, but how would you shift the burden to your gpu with a game that doesn't have in game settings for physx options

 

[RussianSensation]

MJinZ, that's what happens in the era of console ports.

This is a great article on RE5 development. The game supports 8 threads. http://download.intel.com/products/processor/corei7/ResidentEvil.pdf

Page 6

"Resident Evil 4 did not scale beyond 1 CPU and 1 GPU. Now the team had to handle 1, 2, 4 and 8 threads, and keep their options open for the future. Takeuchi's team assigned the maximum 8 threads in this way: 2 threads starting with 0 went to rendering, with threads 2 through 7 taking care of everything else, including networking, artificial intelligence, physics, animations, thread scheduling, sound, object locations and more."

^^^ I am guessing the game scales a lot better with 6 real cores when handling 8 separate threads than with a 4 core processor with 8 simulated threads. This is because the Core i7 980X can Turbo Mode up to 3.46ghz with up to 6 cores running, while Core i7 860 can only manage 2.93ghz when 4 cores are loaded. That's a frequency difference of 18%.

Even then, on a Fixed Demo DX10 bench, 5870 seems to be the limiting factor as the bench scores are the same beyond 1600x1200 with a Core i7 975 OR a Core i7 980X @ 4.4ghz. The 5870 could not break more than 102 fps at 1920x1080 4AA w/Catalyst 10.2, even when paired with the i7 980x @ 4.4ghz. Surely a case of being GPU limited with the 5870: http://www.guru3d.com/article/core-i7-980x-review/18

What is important to note from the Guru3D article, was that Core i7 980X and i7 975 were 25% faster than the Core i7 940 using the same 5870, even at a high resolution of 1920x1080 4AA. When paired with the i7 940 2.93ghz, the 5870 could only manage 75 fps vs. 102 fps on the 980X.

This is in line with OP's score of 72 fps on a Core i7 860 and a GTX480. The game is CPU limited with a stock Core i7 860. The culprit is CPU frequency.

 

[apoppin]

What do you mean by shift burden to your gpu? I know some titles have physx options which i think is along the lines of what your talking about, but how would you shift the burden to your gpu with a game that doesn't have in game settings for physx options

Let's look at what we have with RE5.

At 1920x1200 resolution, the Core i7 walks all over the Phenom II X4 at the same 3.8 GHz clock speed and when we drop the clocks on the Phenom II (or Core i7) the frame rates in RE5 also drastically drop.

Now when we *shift the burden* to the GPU by upping the resolution to 2560x1600 resolution, we are running almost double the resolution of 1920x1080 and the GPU must work much harder relative to the CPU to push twice the pixels.

Since the GTX 480 is doing much of the work now at the higher resolution, the Phenom II X4 catches the Core i7 920 and when we drop the CPU speeds by 1,200 MHz - from 3.8 to 2.6 GHz we suffer only a small frame rate drop.

The other way to shift the burden to the GPU would be by maxing out the vendor's control panel for AA - 8X would be a minimum.

 

[lotusvibe]

Let's look at what we have with RE5.

At 1920x1200 resolution, the Core i7 walks all over the Phenom II X4 at the same 3.8 GHz clock speed and when we drop the clocks on the Phenom II (or Core i7) the frame rates in RE5 also drastically drop.

Now when we *shift the burden* to the GPU by upping the resolution to 2560x1600 resolution, we are running almost double the resolution of 1920x1080 and the GPU must work much harder relative to the CPU to push twice the pixels.

Since the GTX 480 is doing much of the work now at the higher resolution, the Phenom II X4 catches the Core i7 920 and when we drop the CPU speeds by 1,200 MHz - from 3.8 to 2.6 GHz we suffer only a small frame rate drop.

The other way to shift the burden to the GPU would be by maxing out the vendor's control panel for AA - 8X would be a minimum.

So your saying that to shift the burden to the GPU i need to crank up the resolution or AA settings? I am using my rig to hook up to a 60" 1080 HDTV so i can't go any higher than 1080. I suppose i could crank up the AA and see what happens but I usually suffer from a lower framerate when i turn up AA. Sorry if I'm just missing your point completely I'm a newb:\

Also (obviously I'll just try it myself and see but i enjoy the options of others) I have heard some people say that Hyper Threading is a gimmick, and that it actually improves your gaming performance to have it off, what do you guys think should HT be on or off for gaming peformance? Or is that also going to vary based on the game.

 

[MJinZ]

Surprisingly, not really. And if you run your same tests at 2560x1600 with maxed out details and 8xAA you get a very different picture. Phenom II is only a couple of FPS slower than the Core i7 when both are clocked at 3.8 GHz and there is also very little penalty for a slower CPU at high resolution:

GTX 480/i7-920@3.8 GHz = 73.2

GTX 480/Ph II 955 X4@3.8 GHz = 71.8
GTX 480/Ph II 955 X4@3.2 GHz = 70.7
GTX 480/Ph II 955 X4@2.6 GHz = 66.8

You just need to shift the burden to the GPU and the picture drastically changes. At 1920x1200 or 1920x1080, the CPU is working much harder relative to the GPU in RE5 than at 2560x1600.

You need to take each game on case-by-case. AtM, *most* games are FINE on a Phenom II X2 or C2D at 2.6 GHz if your shift your burden to the GPU at a high resolution. Some games of course - newer games - are much better multi-threaded and will benefit both from more cores and higher clockspeeds.

GPU just needs to generate the pretty pixels, CPU sometimes needs to do much more.

 

[apoppin]

So your saying that to shift the burden to the GPU i need to crank up the resolution or AA settings? I am using my rig to hook up to a 60" 1080 HDTV so i can't go any higher than 1080. I suppose i could crank up the AA and see what happens but I usually suffer from a lower framerate when i turn up AA. Sorry if I'm just missing your point completely I'm a newb:\

Also (obviously I'll just try it myself and see but i enjoy the options of others) I have heard some people say that Hyper Threading is a gimmick, and that it actually improves your gaming performance to have it off, what do you guys think should HT be on or off for gaming peformance? Or is that also going to vary based on the game.

i am saying you are kinda stuck. You can get the same speeds but with much higher levels of AA (but cannot crank up your resolution)

OR .. if you feel you need more frame rates with GTX 480 at your current level of detail and resolution, you need to overclock your CPU.

Does RE5 feel sluggish to you? If not, don't worry about it. If so, learn to overclock your CPU (or get a higher resolution display to take full advantage of GTX 480).

 

[RussianSensation]

lotusvibe don't be discouraged by your results. RE5 is just not demanding enough to stress a GTX480. In games like STALKER: Call of Pripyat, Metro 2033, Aliens vs. Predator and Dirt 2, GTX480 has a significant advantage over 5870 at 1920x1080 4AA. Play those games and you'll be a lot happier about the power of your card

 

[ModestGamer]

lotusvibe don't be discouraged by your results. RE5 is just not demanding enough to stress a GTX480. In games like STALKER: Call of Pripyat, Metro 2033, Aliens vs. Predator and Dirt 2, GTX480 has a significant advantage over 5870 at 1920x1080 4AA. Play those games and you'll be a lot happier about the power of your card

I don;t get that. I have a 5770 with a lowly2.4ghz 9550 amd with 6gb or ram and it plays Dirt 2 at 60-70 fps. WTF would you need a gtx480 for anyways ?

Maybe people just don't know how to optimize there os or something ?

this card renders fine. dirt 2 looks fantastic as very high setting on everything.

Maybe coders need to improve the software. I have come to the conclusion that this generation of consumer parts is good enough.

 

[dookulooku]

So your saying that to shift the burden to the GPU i need to crank up the resolution or AA settings? I am using my rig to hook up to a 60" 1080 HDTV so i can't go any higher than 1080. I suppose i could crank up the AA and see what happens but I usually suffer from a lower framerate when i turn up AA. Sorry if I'm just missing your point completely I'm a newb:\

IMO, "shift the burden to the GPU" is not the best way to describe what goes on. What actually happens is that turning up resolution and AA drastically increases the amount of time the GPU spends rendering a frame without affecting how long the CPU takes to construct a frame. This effectively reduces CPU utilization since the CPU will spend most of the time waiting on the GPU.

Let's assume the CPU needs to wait for the GPU to finish rendering the previous frame before it can construct the next frame. CPU A can construct a frame in 10 ms while CPU B does it in 5 ms.

Let's say that at 1920x1200, the GPU takes 5 ms to render a frame. That means CPU A runs the game at 67 fps, while CPU B runs it at 100 fps.

Now when the game runs at 2560x1600 4xAA/16xAF, the GPU takes 25 ms to render a frame. That means CPU A runs the game at 29 fps, while CPU B runs it at 33 fps.

So in this scenario, increasing resolution and AA/AF settings greatly reduces the advantage seen with CPU B over A.

 

[dookulooku]

Before i bought my gtx 480 I had a HD5870 which was benchmarking around 108fps which is matching up with the above website's benchmarks, i got the gtx 480 because It was suppose to be a little better and I wanted to try an NVidia card and see how i liked it. So far well i don't like it cause i cannot get the re5 bencmark about 73fps. I don't have any idea what's wrong I haven't changed any bios settings and everything appears to be running fine, why such low benchmarks?

The compiler inside a GPU driver is usually one of the most CPU intensive parts of the game.

A possible explanation is that ATI might have a more CPU-friendly compiler than NVIDIA, or that it's instruction architecture allows it to reach peak performance in this game with less software overhead.

Of course, most people with a GTX 480 will have an i7, and in that case, would want a compiler that yields that produces the most optimized GPU code, even it means greater software overhead.

 

[RussianSensation]

I don;t get that. I have a 5770 with a lowly2.4ghz 9550 amd with 6gb or ram and it plays Dirt 2 at 60-70 fps. WTF would you need a gtx480 for anyways ?

You don't need a GTX480 for Dirt 2 per if are playing in DX9 mode. But to say that 5770 is fast enough for Dirt is just a matter of what you call playable. My 4890 can't play Dirt 2 at 8AA 1920x1080 smoothly and it's far faster in DX9 mode than 5770 is in DX11 mode. 4890's minimum framerates in Dirt 2 are higher than 5770's average.

1920x1200 4AA/16AF
5770 = 25 min / 33.8 avg (unplayable)
4890 = 45 min / 52 avg (playable)

It just depends what you want in your racing game. I find that in Dirt 2, anything below 43-47 fps minimum framerates results in choppy gameplay for me. 5770 would not fulfill my Dirt 2 gaming needs regardless of resolution as it can't maintain above 40 fps minimum with 4AA: http://www.xbitlabs.com/articles/video/display/xfx-radeon-hd5830_12.html#sect1

Now if you want to play DX11 in Dirt 2, GTX470 is what you want at min:

1920x1080 4AA - http://www.xbitlabs.com/articles/video/display/gigabyte-gf-gtx400_13.html#sect1
5870 = 38 min (not playable)
GTX470 = 55 min
GTX480 = 74 min

To each his own. I am not picky about 25-30 fps avg in Crysis but I sure want my Dirt 2 min in the 40s.