NVidia GTX 670, First Impression

taufikp

Just replaced my Radeon HD5770 with NVidia GTX670 (accidentally, both of them are from Digital Alliance). Here are my impressions of GTX670:

Faster Loading Time
I was pretty surprised actually. Probably it's because the GTX670 has 256-bit interface and 2GB of memory. The HD5770 only has 1GB of memory and 128-bit interface. Can't find another reason except those two.

Smoother Animation
The micro stutter I had with HD5770 is gone now. The overall driving experience is much better. I also think it's related to the fact about how 'jittery' AMD cards are. Read the article about this 'spikey' performance of AMD cards.

Higher FPS
Well, no doubt about it. Higher core clock and wider interface helps a lot. I now can use Environment Reflection (my favorite effect in rF2) and still get smooth animation. Honestly, I think it's unfair to compare the FPS rate between HD5770 and GTX670, since they are not in the same class.

Image Quality
Surprisingly, after 30 minutes of driving and watching replays, I found no difference in terms of image quality. By the way, I don't use driver's AA and AF for both cards. All AA and AF are controlled in-game. I'm using AA Level 1, HDR on, FXAA off, and AnisoFiltering is 8x. Texture Quality Full. Screen resolution is 1600x900 32-bit.

Temperature
Using MSI After Burner, HD5770 tops at 62 degree Celsius, while GTX 670 at 73 degree. The GTX670 fan is less noisy than HD5770, even though the GTX670 Digital Alliance JetStream uses dual fans to cool the GPU. Nice.


Well, that's my first impression about GTX670. I am happy since rF2 is more fun now.

:)
 
Last edited:
taufikp said:
Image Quality
By the way, I don't use driver's AA and AF for both cards. All AA and AF are controlled in-game.

I'm using AA Level 1, HDR on, FXAA off, and AnisoFiltering is 8x. Texture Quality Full. Screen resolution is 1600x900 32-bit.
Click to expand...

Ever heard of Nvidia Inspector. :)

I have GTX570 SOC

I'm using AA Level 8, HDR off, FXAA off, AnisoFiltering is 16x. Texture Quality Full. Screen resolution is 1920x1080 32-bit.
I use Nvidia Inspector to set LOD bias and a few other things.
 
DurgeDriven said:
Ever heard of Nvidia Inspector. :)

I have GTX570 SOC

I'm using AA Level 8, HDR off, FXAA off, AnisoFiltering is 16x. Texture Quality Full. Screen resolution is 1920x1080 32-bit.
I use Nvidia Inspector to set LOD bias and a few other things.
Click to expand...

Yes, I've heard about NVidia Inspector. Haven't try it though. But I will. :)
 
taufikp said:
Yes, I've heard about NVidia Inspector. Haven't try it though. But I will. :)
Click to expand...

I wouldn't worry to much about it unless you like to go crazy changing settings that are not normally there, but it allows for a great level of customization(game settings) another item i always like from time to time is d3doverrider it comes in the riva tuner package, but it allows you to force vsync and it does a better job than forcing in nvidia control panel, plus triple buffering
 
taufikp said:
Just replaced my Radeon HD5770 with NVidia GTX670 (accidentally, both of them are from Digital Alliance)
I am happy since rF2 is more fun now.

:)
Click to expand...

Congrats on your new card.
That extra gig of ram goes a long way.
I'm seeing anywhere from 80-114 fps with my GTX670 in this build...(depending on track) with everything maxed.
I still contend that RF2 does better with large videoram amounts vs raw CPU horsepower.
 
Last edited:
Well, this is my second impression of GTX670 after spending about 6 hours testing it.

Now I can run rF2 smoothly with these settings:
- 1920 x 1080, 60Hz, 32-bit
- HDR on, FXAA off, Anti Aliasing Level 1 (AA is quite good at this level, but the shimmering effect in distance still there)
- Environment Reflection on
- Road Reflection on
- Sun Occlusion on
- 25 AI, 20 visible cars

Wet race is playable now. Previously, with HD5770 my system would hit below 25 FPS in the wet. In terms of image quality, both cards perform really well.
 

This video is about image quality comparison not performance, clearly GTX670 outperforms HD5770 by a large margin.

In my opinion, the image quality differences between both cards are minimal. One thing that I noticed is nVidia has slightly better contrast than AMD.

Sadly in rFactor 2 build 134, the replay module still not at its best. The image quality between racing module and replay module is not the same, hence affecting the image quality in this video. The "real-time quality" actually is much better than this.

rF2 detail settings:
- Vid Res: 1600x900, 60Hz, 32-bit
- HDR on, FXAA off, AA Level 1
- Circuit Detail Low
- Player Detail Full
- Opponent Detail Full
- Texture Quality Full
- Sun Occlusion On
- Reflections Off
- Shadows High
- Shadow Blur Fast
- Number of AI 20
- Visible Vehicles 20
- FOV 35
- HDR profile Realistic Clear
 
Hazi said:
Why do you have circuit detail at low?
Click to expand...

Because I want to run rF2 at more than 30 FPS :) This was my setting with HD5770 by the way.

Now I can run with Full circuit detail and still get more than 30 FPS all the time.

When I'm recording this video with GTX670 installed, I had to kept all previous settings in order to get equal view.
 
Funny thing happened between GTX670 and rFactor1.

After getting good results in rF2, I turned my attention to rF1. I thought I could make rF1 to look much, much better without sacrificing the oh so important FPS at all.

So, via rF1 Config tool I turned up the AA to Level 6 (that's the highest listed in rF1 Config tool), fired up rF1 and expecting those silky smooth images to show up. But how shocked I was! On the track, the image quality was nowhere near AA Level 1, it was more like running rF1 with no AA at all.

I turned the level down to 3, and the IQ is getting better, but some jaggies are still visible, mainly on the windshield and on track's white side lines.

Why rF1 IQ is so bad at AA Level 6? Do I have to change something in the nVidia's game/app profile?
 
@taufikp: Why rF1 IQ is so bad at AA Level 6? Do I have to change something in the nVidia's game/app profile?
Was reading some time ago about that: FOUND it: http://www.nogripracing.com/forum/showthread.php?p=1449023

Default AA levels in rFactor
I decided to test the AA settings in rFactor's Config utility and I'm surprised by my results. Levels 1/2 and Levels 3/4 appear to be the same. And I'm guessing that Levels 5/6 would have had similar results, but instead L6 looked like L0. And I've seen other people online have problems with L6 so I wonder if the issue is with rFactor or my system.

The frame rate also stayed consistent in spite of the increased AA levels so I'm guessing that just shows how CPU-bound rFactor is.
http://i.imgur.com/HnB53.gif
 
Yarach said:
That is something strange you say there. I have never had anny issues with "spikey" performance when I had the HD5770. It's just strange how to same products behave very different :s"
Click to expand...

If you ever seen a stutter, even the micro/small ones, that's the result of spikes inside one second of gaming. I too rarely seen this spike while gaming. However the latest rF2 build combined with the latest AMD driver (12.11 beta 8) bring back the micro stutters, even worst than before. Enough is enough, and I decided to make a switch to nVidia for the much less pronounced spikes in a second than AMD, and also to take full advantage of PhysX in games that are supporting it.

edit: I know it's wrong to upgrade your system just for the sake of a beta/WIP application, oh well, can't resist the temptation, also it was a good excuse to tell the wife :D
 
Last edited:
Hazi said:
@taufikp: Why rF1 IQ is so bad at AA Level 6? Do I have to change something in the nVidia's game/app profile?
Was reading some time ago about that: FOUND it: http://www.nogripracing.com/forum/showthread.php?p=1449023

Default AA levels in rFactor
I decided to test the AA settings in rFactor's Config utility and I'm surprised by my results. Levels 1/2 and Levels 3/4 appear to be the same. And I'm guessing that Levels 5/6 would have had similar results, but instead L6 looked like L0. And I've seen other people online have problems with L6 so I wonder if the issue is with rFactor or my system.

The frame rate also stayed consistent in spite of the increased AA levels so I'm guessing that just shows how CPU-bound rFactor is.
http://i.imgur.com/HnB53.gif
Click to expand...

Thanks Hazi. That exactly what I was seeing and experiencing with AA Level 6 in rF1. Seems that level 3/4 would be the best choice.
 
Another pleasant surprise from GTX670 over HD5770.

Faster Video Resolution Change Via In-Game Option
With HD5770, usually I have to wait up to apprx. 5 seconds via in-game VidRes option. Now with GTX670 it's less than 2 seconds! Again, I believe this is due to GTX670's 2GB VRAM and 256-bit bus interface.

:)
 
Couldn't be happier with the card. In October last year upgraded my graphics card CPU and motherboard and there's not a game at the moment it wont run on full settings.
The 670 constantly impresses me with performance.

I did read somewhere but can't find it now:confused: that the card performance de-tunes so to speak once the GPU temperature reaches 70° in increments of 5° is that right?
 
Fergy said:
Couldn't be happier with the card. In October last year upgraded my graphics card CPU and motherboard and there's not a game at the moment it wont run on full settings.
The 670 constantly impresses me with performance.

I did read somewhere but can't find it now:confused: that the card performance de-tunes so to speak once the GPU temperature reaches 70° in increments of 5° is that right?
Click to expand...

It's possible. NVidia has GPU Boost feature, so I think it's just logical to have the opposite function as well, to keep GPU temp cool. During watching AI driving the car, I can hear the VGA fan sometimes turning up faster and then slowing down again. Didn't check the temp and speed though. MSI After Burner can be used to confirm this.
 
Quite often AMD cards have a higher peak framerate than Nvidia cards. I believe it's that wide swing in variation from max to min framerate which is probably responsible for the microstutter. It can be quite substantial. That high to low and back to high again is really not ideal. I just wish we could somehow figure out which factors are most responsible for it.
 
Guys, maybe a bit off topic, but I couldn't think of a better place to ask......

My nephew has got a GTX 480 1,5Gb and he is offered a 670 (don't know which kind) and he (and I for that matter) would like to know if that is an improvement.
Now......I do the searching for him (don't ask ;) ) and I do find a lot of charts, but never these two in the same one. It's impossible to make a decent comparison because "the other chart" is always made with different hardware.

Do any of you guys know how these two relate?

Again; sorry for intruding the topic but I didn't want to make a new threat just for this.
 
The extra .5GB of video buffers won't hurt at all on the GTX670. It also has a more efficient core. What is the basis of the offer...card plus money for the swap?
 
As you say you don't often see a 480 v 670 and yes it should be an improvement. One site I did find a comparison was here testing over a variety of games/benchmarks (no rF2 unfortunately), note at the time of this test AMDs 'Never Settle' drivers had not been released.
 
Saabjock said:
Quite often AMD cards have a higher peak framerate than Nvidia cards. I believe it's that wide swing in variation from max to min framerate which is probably responsible for the microstutter. It can be quite substantial. That high to low and back to high again is really not ideal. I just wish we could somehow figure out which factors are most responsible for it.
Click to expand...

According to AMD's David Baumann (as reported in TR's article), AMD is now rewriting the entire driver again since they found out that the memory management in GCN architecture is entirely different to prior GPUs.

AMD plans to release few driver updates in the early part of this year. Can't wait to see the test :)
 
Navigator said:
Guys, maybe a bit off topic, but I couldn't think of a better place to ask......

My nephew has got a GTX 480 1,5Gb and he is offered a 670 (don't know which kind) and he (and I for that matter) would like to know if that is an improvement.
Now......I do the searching for him (don't ask ;) ) and I do find a lot of charts, but never these two in the same one. It's impossible to make a decent comparison because "the other chart" is always made with different hardware.

Do any of you guys know how these two relate?

Again; sorry for intruding the topic but I didn't want to make a new threat just for this.
Click to expand...

This might help: http://www.hwcompare.com/12527/geforce-gtx-470-vs-geforce-gtx-670/ as it gives you theoretical performance of each card based on their basic/reference specification.

I use that site a lot when I was in the market looking for a GeForce card to replace my Radeon HD5770. And in the end I chose GTX670, since it gives me 270% improvement over HD5770 in theoretical performance, and also good 'extras' such as PhysX, and the price is more affordable than GTX680 or GTX690.

EDIT:
Oops, sorry I gave you the wrong comparison! Here's the correct one:
http://www.hwcompare.com/12526/geforce-gtx-480-vs-geforce-gtx-670/
 
Thanks for the response guys; I know enough to tell him.
The guy selling it, wants 275 euro's for it and in my opinion; that's a lot of money for a 30% upgrade; I just tell him that.
 
I just got a 670 today to replace a 6850, went from 1600x900 to 1080p, AA from level 2 FXAA to Level 4 AA, and turned up the rest of the settings to full. The thing is I still get flickering lines and barriers in the distance, can anybody tell me where I'm going wrong?
 
smithaz said:
I just got a 670 today to replace a 6850, went from 1600x900 to 1080p, AA from level 2 FXAA to Level 4 AA, and turned up the rest of the settings to full. The thing is I still get flickering lines and barriers in the distance, can anybody tell me where I'm going wrong?
Click to expand...

You aren't doing anything wrong - it's a common problem with ISI games (from my experience). The only way I managed to get rid of these flickering fences and barriers (and jagged white lines) was to use Nvidia Inspector to apply some Sparse Grid Supersampling.

I use level 3 or 4 AA in game and 4x Sparse Grid Supersampling from Nvidia inspector. You just need to change these two settings from within Inspector - 'AA Transparency Mulitsampling' (change from disabled to enabled) and 'AA Transparency SuperSampling' (set this to 2x or 4x SparseGrid). Obviously the more SparseGrid you use the lower the framerate will be, but your 670 should be able to apply at least 2x which will be an improvement.

Good luck!

Edit - maybe try this without FXAA applied as well. You can still use it with SparseGrid, but I think it's better without FXAA.
 
taufikp said:
However the latest rF2 build combined with the latest AMD driver (12.11 beta 8) bring back the micro stutters, even worst than before.
Click to expand...

That;s is simply not true for all systems, my HD6950 Crossfire is as smooth as silk, all details at MAX, 1920x1200x32 res
16xAF AAx3 and its locked at 60fps with V-sync, it never drops below and is real smooth?

with 12.11 beta 8 btw

I simply wouldn't want to swap to anything else just now, as its running sweet

with v-sync off btw it will get as high as 178fps, but picture is smoother with v-sync on, no tearing
 
ForthRight said:
I use level 3 or 4 AA in game and 4x Sparse Grid Supersampling from Nvidia inspector. You just need to change these two settings from within Inspector - 'AA Transparency Mulitsampling' (change from disabled to enabled) and 'AA Transparency SuperSampling' (set this to 2x or 4x SparseGrid). Obviously the more SparseGrid you use the lower the framerate will be, but your 670 should be able to apply at least 2x which will be an improvement.
Click to expand...

Thanks for the advice, I tried 2x and 4x super sampling on transparency from within Nvidia control panel, it does seem to improve it slightly, 4x starts to drop framerate a bit so I'll have to stick with 2x for now. Fingers crossed that this is an area where ISI are able to make improvements in future.
 
smithaz said:
Thanks for the advice, I tried 2x and 4x super sampling on transparency from within Nvidia control panel, it does seem to improve it slightly, 4x starts to drop framerate a bit so I'll have to stick with 2x for now. Fingers crossed that this is an area where ISI are able to make improvements in future.
Click to expand...

Okay, but I should point out that SparseGrid is not the same as the normal SuperSampling you are using from control panel - SparseGrid is better at fixing the problem you have, and in some cases it doesn't cost any more fps than actual SuperSampling (depending on the card). I recommend you try 2x SparseGrid instead.

Nvidia Inspector is a very tiny program and easy to use - it just extracts to a folder and runs from there (it doesn't 'install' itself anywhere). It works in conjunction with the normal Nvidia Control Panel - you can use either as they both do the same thing (so if you set 2xAA in Inspector's global profile you'll notice 2xAA is set in Nvidia Control Panel's global profile too), the only difference is Inspector has more advanced options. It's not like installing some big piece of software - it's very small and you can just delete the folder if you don't want it anymore.

Sorry for going on, it's just I know SparseGrid fixes this problem a lot better than normal SuperSampling.

I'll be quiet now. :p
 
Thanks for the tips ForthRight! I have used level 3 AA and FXAA till now. FXAA for extra AA cause aliasing in stereoscopic 3D is very annoying. Will try SparseGrid without FXAA.
 
You should use 60 FPS as cutoff (aproximately humans' vision "refresh rate"). 30 FPS is too jumpy when you get used with 60 FPS. More than 60 FPS is not really noticeable...

taufikp said:
Because I want to run rF2 at more than 30 FPS :) This was my setting with HD5770 by the way.

Now I can run with Full circuit detail and still get more than 30 FPS all the time.

When I'm recording this video with GTX670 installed, I had to kept all previous settings in order to get equal view.
Click to expand...
 
I am new to nvidia so didn't know about nvidia inspector before but it is a nice little program. I have tried sparse grid super sampling on tranparency briefly and it didnt seem all that different to normal super sampling (which was a great improvement over nothing anyway), I will have to test more when I have more time. Flickering on the white lines in the distance is still not great, maybe it is an area in which not much can be done.
 
Super Sampling (SSAA) is more expensive, in terms of performance. Multi Sampling AA plus Sparse Grid AA gives me an acceptable anti-aliasing, better than FXAA I think, but with better performance than SSAA.
 
Indeed SSAA gives around 100fps where as MSAA will give me around 178fps,
However I prefer the look of SSAA and keep vsync on so only need 60fps anyway lol
 
True. SSAA quality is still the best. Hopefully ISI will implement TXAA as well, since the quality is supposed to be better than FXAA, but the performance is almost equal to FXAA.
 
taufikp said:
Well, this is my second impression of GTX670 after spending about 6 hours testing it.

Now I can run rF2 smoothly with these settings:
- 1920 x 1080, 60Hz, 32-bit
- HDR on, FXAA off, Anti Aliasing Level 1 (AA is quite good at this level, but the shimmering effect in distance still there)
- Environment Reflection on
- Road Reflection on
- Sun Occlusion on
- 25 AI, 20 visible cars

Wet race is playable now. Previously, with HD5770 my system would hit below 25 FPS in the wet. In terms of image quality, both cards perform really well.
Click to expand...


Hi! Im the guy form 660ti 3 screen try out! :p

I see you bought a 670, could you tell me whats the size of you screen/monitor?

I guess Im gonna try a single 40" monitor
 
You must log in or register to reply here.

Back
Top